Annual Review of Psychology, Vol 56 2005

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

P1: JRX

December 8, 2004 12:13 Annual Reviews AR231-FM

18 Nov 2004 11:52

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

P1: IKH LaTeX2e(2002/01/18) 10.1146/annurev.psych.56.091103.070239

Annu. Rev. Psychol. 2005. 56:1–23 doi: 10.1146/annurev.psych.56.091103.070239 c 2005 by Annual Reviews. All rights reserved Copyright
First published online as a Review in Advance on June 10, 2004

IN SEARCH OF MEMORY TRACES Richard F. Thompson

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Neuroscience Program, University of Southern California, Los Angeles, California 90089-2520; email: [email protected]

Key Words learning, hippocampus, cerebellum, inactivation, localization ■ Abstract The key issue in analyzing brain substrates of memory is the nature of memory traces, how memories are formed, stored, and retrieved in the brain. In order to analyze mechanisms of memory formation it is first necessary to find the loci of memory storage, the classic problem of localization. Various approaches to this issue are reviewed. A particular strategy is proposed that involves a number of different techniques (electrophysiological recording, lesions, electrical stimulation, pathway tracing) to identify the essential memory trace circuit for a given form of learning and memory. The methods of reversible inactivation can be used to localize the memory traces within this circuit. Using classical conditioning of eye blink and other discrete responses as a model system, the essential memory trace circuit is identified, the basic memory trace is localized (to the cerebellum), and putative higher-order memory traces are characterized in the hippocampus.

CONTENTS INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Simplified Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Strategies to Study Memory Formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 OUR MODEL SYSTEM: EYEBLINK CONDITIONING . . . . . . . . . . . . . . . . . . . . . 6 Motor Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 The Hippocampus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 The Essential Circuitry: The Cerebellum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

INTRODUCTION A major achievement of recent research on brain substrates of learning and memory, to which our work has contributed, is the recognition that there are several different forms or aspects of memory involving different brain systems (Figure 1). Squire (1992) has argued eloquently for the distinction between declarative and nondeclarative forms of memory, as has Schacter (1987). The distinction between episodic and semantic memory—“What did you have for breakfast?” versus “Where is the Eiffel Tower?”—has been stressed by Tulving (1985). 0066-4308/05/0203-0001$14.00

1

2

AR AR231-PS56-01.tex AR231-PS56-01.sgm LaTeX2e(2002/01/18)

Figure 1 A current view of the various forms of learning and memory and their putative brain substrates.

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

18 Nov 2004 11:52 P1: IKH

THOMPSON

18 Nov 2004 11:52

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

LOCALIZATION OF MEMORY TRACES

P1: IKH

3

Implicit or nondeclarative memory is very much a grab bag. In general, explicit memory involves awareness of the memory whereas implicit memory does not necessarily involve being aware of the memory (aware meaning verbal report). The schema of Figure 1 is of course oversimplified. When an organism learns something, a number of brain systems can become engaged. However, in most cases there is one critical brain system, which when damaged causes permanent impairment in the particular form of learning and memory. Many readers will recall Lashley’s (1950) pessimistic conclusion that his series of experiments “has yielded a good bit of information about what and where the memory trace is not” in his famous article, “In Search of the Engram.” In all fairness to Lashley, the existence of several different forms of memory with differing neuronal substrates was not recognized at that time, nor were modern analytic techniques available then. We adopt the following definitions from an earlier review (Lavond et al. 1993). The essential memory trace, together with its associated circuitry, is necessary and sufficient for the basic aspects of a given form of learning (e.g., acquisition and retention). Other brain structures may also form memory traces, defined loosely as learning-induced changes in neuronal activity, but these may not be essential for the basic learning. Thus, in eyeblink conditioning, long-lasting learning-induced increases in neuronal activity develop in the hippocampus (Berger et al. 1976), but the hippocampus is not necessary for basic delay associative learning and memory (Schmaltz & Theios 1972). Aversive classical conditioning can modify the receptive-field properties of neurons in sensory systems. Particularly striking are the studies of Weinberger and associates (Edeline & Weinberger 1991, Weinberger et al. 1984) showing that neurons in secondary areas of auditory thalamus and cortex shift their best frequencies toward the frequency of the conditioning stimulus in aversive Pavlovian conditioning (see also Bao et al. 2003). The motor area of the cerebral cortex provides yet another example. Classical eyeblink conditioning results in marked and persisting increases in excitability of pyramidal neurons in motor cortex (Woody et al. 1984). However, the motor cortex is not necessary for either learning or memory of the conditioned eyeblink response (Ivkovich & Thompson 1997). These “higherorder” memory traces appear to play important roles in the adaptive behavior of the organism. The obvious point here is that the organism can learn, remember, and perform the basic learned response following destruction of nonessential memory trace systems, but destruction of the essential memory trace system abolishes this ability. We draw a distinction between the essential memory trace and the essential memory trace circuit. The latter includes the necessary input and output circuits as well as the essential memory trace itself. These are all identified by lesions. But lesions, per se, cannot distinguish between the essential trace and the essential circuitry; lesions of either completely prevent and abolish the learned response.

18 Nov 2004 11:52

4

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

P1: IKH

THOMPSON

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Simplified Systems How does one go about finding memory traces? One very productive approach is the use of simplified “model” systems, emphasized early by Thompson & Spencer (1966) (see also Kandel & Spencer 1968). Thus, vertebrate spinal reflexes and the gill-withdrawal reflex circuit in Aplysia have both served as very productive models to analyze processes of habituation and sensitization. In the case of habituation, at least, the memory trace is embedded within the reflex pathway under study so lesions abolish the reflex as well as the trace. For monosynaptic pathways in both Aplysia and spinal cord the mechanism of short-term habituation appears to be simply a decrease in the probability of transmitter release as a result of repeated activation, a presynaptic process of synaptic depression (Farel & Thompson 1976, Kandel 1975). But even for a simple behavioral phenomenon like habituation, the neural substrate can prove to be complex, particularly for long-lasting habituation, e.g., between sessions as opposed to within session response decrements. In the Aplysia cellular monosynaptic system, Ezzeddine & Glanzman (2003) have shown that prolonged habituation depends on protein synthesis, protein phosphatase activity, and postsynaptic glutamate receptors. Indeed, their data indicate that stimuli inducing habituation of the gill-withdrawal reflex appear to activate sensitizing processes as well, as proposed in the “dual-process” theory of habituation by Groves & Thompson (1970). Thus, much is still not known about the detailed mechanisms of memory traces in this simplest form of learning. By mechanisms we mean the physical-chemical substrate of memory store. Although there are many candidate mechanisms, as of this writing we do not yet know the detailed physical bases of memory storage for any form of learning and memory in the mammalian brain. But understanding the mechanisms of memory storage will not inform us of what the memories are. The actual memories are coded by the neuronal circuits that code, store, and retrieve the memories, what I term here the “essential circuits.”

Strategies to Study Memory Formation There appear to be two general strategies for the study of memory storage. One approach is to pick a mechanism of neural plasticity such as long-term potentiation (LTP) in a structure known to be important for memory and attempt to show that it is a substrate for memories. Richard Morris and associates published a heroic effort to demonstrate that activity-dependent synaptic plasticity, particularly LTP (and LTD, long-term depression), “is both necessary and sufficient for the information storage underlying the type of memory mediated by the brain area in which that plasticity is observed” (Martin et al. 2000). They established several “formal” criteria by which to judge this hypothesis. We examine these criteria below but here we simply note their conclusion that “synaptic plasticity is necessary for learning and memory, but that little data currently supports the notion of sufficiency” (p. 649).

18 Nov 2004 11:52

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

LOCALIZATION OF MEMORY TRACES

P1: IKH

5

Actually, there are many forms of plasticity in the nervous system, synaptic plasticity being only one. There are also changes in neuron excitability due to factors intrinsic to the neurons, as in decreases in after-hyperpolarization, as well as structural changes in dendrites, formation of new neurons, and actions of glial cells. LTP has been the most intensely studied putative synaptic mechanism of memory storage, particularly in the hippocampus (e.g., Martin et al. 2000), but the evidence is still far from clear. There is a vast literature analyzing mechanisms of LTP but very little work attempting to show that it is a basis for the storage of memories (see Shors & Matzel 1997, Wilson & Tonegawa 1997). The alternative strategy is to begin with a well-characterized form of learning and memory and determine the memory traces involved. In order to do this it is first necessary to find where in the brain the memories are stored, the classical problem of localization. Because learning involves changes in behavior as a result of exposure to stimuli that do not change, there must be alterations at some loci in the brain. A given form of learning may involve one or multiple loci of memory storage, but they must exist. We have used this strategy in our search for memory traces. How does one go about finding a memory trace in the behaving mammal? Olds, Disterhoft, and associates wrestled with this issue in pioneering studies (Olds et al. 1972). Their basic approach was to record extracellular unit cluster discharges in many brain areas in freely moving rats, looking for increases in unit response on the conditioning day that were significantly greater than responses to the stimuli on the preceding pseudoconditioning day, i.e., a learning-induced increase in unit activity. The behavioral situation was simply an auditory conditioned stimulus (CS+) followed after 2 sec by a food pellet unconditioned stimulus (US) versus a CS− with no food pellet. The studies were very well controlled. The initial criterion they set for identification of the memory trace was shortest latency learned responses: The “earliest” conditioned brain responses (i.e., those appearing with the shortest latencies after application of the CS) might thus be considered to be “at the site” of conditioning, and other later conditioned brain responses might be considered to be secondary to them. Similarly (and hopefully validating this supposition) “early” new brain responses might be expected to emerge at the site of old responses indicating this site to be a junction point between old and new (Olds et al. 1972, p. 202). In a most interesting thought piece, Michael Gabriel (1976) argued against the use of the shortest latency response as the criterion for identification of the memory trace. He proposed an alternative “bias” hypothesis. In essence, he suggested that a tonic flow of impulses originating from a distant neuronal locus biased the activity of the neurons being recorded as the shortest latency learned response. So the increase of the shortest latency unit response could be due to tonic actions on these neurons from a distant site where learning has resulted in a tonic increase in activity. Gabriel draws the following rather strong conclusion: “clearly the bias

18 Nov 2004 11:52

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

6

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

P1: IKH

THOMPSON

hypothesis argues against the idea of mapping the brain for short-latency neuronal responses in order to localize engrams” (p. 279). In a companion piece to the Olds et al. (1972) article, Disterhoft & Olds (1972) do consider the possibility of tonic influences, as Gabriel notes. They suggest an additional criterion, namely the earliest increase in neuronal activity that develops over the trials of training. I return to this issue below. In the late 1960s and early 1970s we tried several different types of model systems to analyze neural substrates of mammalian associative learning and memory. Because of our success in using spinal reflexes for analysis of habituation we tackled classical conditioning of the flexion reflex in the acute spinal animal, with some success (Patterson et al. 1973). However, I concluded it was not a good model of associative learning in the behaving mammal.

OUR MODEL SYSTEM: EYEBLINK CONDITIONING Thanks in part to efforts by Michael Patterson, then a postdoc in my lab who had obtained his Ph.D. with Isadore Gormezano, and because of the elegant and comprehensive behavioral studies by Gormezano, we adopted his preparation: classical conditioning of the nictitating membrane (NM) eyeblink response in the restrained rabbit. We have detailed the advantages of this preparation elsewhere (Thompson et al. 1972, Thompson et al. 1976). Perhaps the most important advantages are: (a) The unconditioned response (UR) provides an independent measure of performance against which to compare effects of variables acting on the conditioned response (CR), and (b) the behavioral conditioned response is robust, reliable, and discrete, and the exact amplitude-time course of the response is measurable. There were a number of other advantages as well: no sensitization, pseudoconditioning, or alpha responding; many trials required to learn; and extensive parometric data on the properties of the CR (see Gormezano et al. 1983, and the important theoretical analyses by Wagner and associates, e.g., Wagner & Donegan 1989).

Motor Control We began our analysis with a focus on the motor nerves and nuclei that generated the reflex and learned eyeblink responses (Cegavske et al. 1976, Cegavske et al. 1979, Young et al. 1976). To oversimplify from our work and the work of others, the sixth nerve was critically important for the nictitating membrane (NM) response, but the third and fourth nerves were also involved and the seventh nerve was critical for closure of the external eyelid (obicularis oculi muscles). In terms of motor nuclei, the seventh nucleus controlled external eyelid closure and the accessory portion of the sixth nucleus innervated the retractor bulbous muscle, which retracted the eyeball and forced the nictitating membrane out across the front surface of the eye. The fourth and sixth nuclei acted synergistically with the accessory sixth and seventh. We showed with simultaneous recordings that NM extension and external eyelid closure [actually electromyographic (EMG)

18 Nov 2004 11:52

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

LOCALIZATION OF MEMORY TRACES

P1: IKH

7

activity of the obicularis oculi muscles] were essentially perfectly correlated over the course of learning (McCormick et al. 1982b). Indeed, obicularis oculi EMG appears to be the most robust and sensitive measure of learning (Lavond et al. 1990). Recordings from the relevant motor nuclei, particularly the sixth, accessory sixth, and seventh, showed essentially identical patterns of learning-induced increases in neuronal activity. It is important to stress the fact that the pathways mediating the reflex eyeblink are in the brain stem and do not involve “higher” brain systems such as the cerebellum and hippocampus. Specifically, there are direct projections from the trigeminal nucleus (activated by stimulation of the cornea and surrounding tissues) to the relevant motor nuclei and indirect projections relaying via the brain stem reticular system to the motor nuclei (Hiraoka & Shimamura 1977). The basic logic of our approach seemed reasonable—identify the critical motor neurons involved in control of the learned response and trace the essential circuitry backward to the source, the “engram.” The fact that several motor nuclei showed the same pattern of learning-induced activation argued against any plasticity unique to one motor nucleus (e.g., the accessory sixth) or to processes of plasticity at the level of the motor nuclei. Instead, it seemed most likely that the learninginduced response in the motor nuclei was driven from a common central source. The pattern of learning-induced increase in the activity of neurons in the motor nuclei in the CS period in paired trials and on CS-alone trials (i.e., the neuronal CR) was strikingly similar in form to the amplitude-time course of NM extension, external eyelid closure, and of course the EMG recorded from the obicularis oculi muscles. Indeed, an envelope of increased frequency of discharge of neurons in the motor nuclei preceded in time and closely predicted the form of the behavioral eyeblink-NM response (Figure 2). This close predictive parallel was most evident with unit cluster recordings but was also true for single unit recordings in the motor nuclei. Our initial logic—to work backward from the motor nuclei—was not as simple as it might have seemed because of the very large number of central brain systems that project to the motor and premotor nuclei. So we adopted a different strategy. It was apparent that the amplitudetime course form of the conditioned eyeblink-NM response closely paralleled and followed the pattern of increased unit activity in the motor nuclei. So we focused on this behavioral model of the learning-induced increase in neural activity in the motor nuclei and used it as a template or model (Figure 2). The higher brain systems that acted to generate and drive the learning-induced neuronal CR, the “model” in the motor nuclei, must show the same pattern of learning-induced activity as do the motor nuclei, and hence the amplitude-time course of the learned eyeblink-NM response. Thus, we searched through higher brain systems looking for learning-induced neuronal activity that correlated closely with and preceded in time the form of the learned eyeblink-NM response. In essence, we mapped virtually the entire brain of the rabbit in 1 mm steps, searching for neuronal models

USE OF THE MOTOR NEURON NEURONAL MODEL

18 Nov 2004 11:52

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

8

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

P1: IKH

THOMPSON

Figure 2 Histograms of unit cluster recordings from the ABD and simultaneous recording of the eyeblink response (NM, the “third eyelid”) in classical conditioning in the rabbit summed and averaged over eight trials before (A) and after (B) learning. First cursor onset of tone CS, second cursor onset of corneal airpuff US, trace duration 750 msec, upward movement of the NM trace indicates extension (eyelid closure). Note that the pattern of neural discharges precedes and models the amplitude-time course of the behavioral response. ABD, abducens motor nucleaus; CS, conditioned stimulus; NM, nictitating membrane; US, unconditioned response. (From Cegavske et al. 1979.)

of the learned behavioral response (see, e.g., McCormick et al. 1983, Thompson et al. 1976). We did not search blindly but rather system-by-system. The basic assumption is that the neuronal sites of memory trace formation will show a pattern of increased frequency of unit cluster and single-unit discharges that precedes and closely predicts the pattern of response in the motor nuclei and hence in the amplitude-time course of the behavioral CR. We felt this criterion was preferable to the earlier ideas of shortest latency response, earliest response over trials, or first site to show increased tonic responses. This assumption proved to be the key that permitted us to localize a memory trace.

The Hippocampus Because of its involvement in memory, we began with the hippocampus (Berger et al. 1976). To our delight, both unit cluster recordings and single-unit responses recorded from CA1 and CA3 showed the requisite learning-induced responses, i.e., pattern of increased discharge frequency that preceded in time (within trials) and correlated virtually perfectly with the amplitude-time course of the behavioral CR (Figure 3). We showed that this neuronal model of a memory is generated by identified pyramidal neurons (Berger & Thompson 1978b). Furthermore, the hippocampal unit response began to develop in the US period within just a few trials of training (a criterion by Olds et al. 1972) and moved into the CS period in close association

18 Nov 2004 11:52

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

LOCALIZATION OF MEMORY TRACES

P1: IKH

9

Figure 3 Behavioral eyeblink (nictitating membrane) and hippocampal unit response averaged over 100 trials in a well-trained rabbit. The unit response is of a single, identified pyramidal neuron in CA1. Upper trace, raw single 100-msec sweep to show the neuron response, middle trace averaged NM response, lower trace cumulated histogram of unit activity; cursors and durations in middle and lower traces as in Figure 2. Note that the pattern of response of this neuron both precedes and models the amplitude-time course of the behavioral response. (From Berger & Thompson 1978b.)

with the development of the behavioral CR (Berger & Thompson 1978a). It seemed like a perfect candidate for the engram. In addition, electroencephalogram (EEG) activity recorded from the hippocampus at the beginning of training predicted the rate of learning (Berry & Thompson 1978). We characterized this learning-induced response in the hippocampus in some detail (see Berger et al. 1986). Unfortunately, as we noted above, animals could learn the basic delay-conditioned NM-eyeblink response following hippocampal lesions (Schmaltz & Theios 1972). This apparent enigma illustrates a fundamental limitation of using neuronal recordings to localize memory traces. The fact that a neuronal model of the learned response occurs at a given locus does not necessarily mean it originates there. Indeed, the learning-induced neuronal model in the hippocampus, at least in the CS period, is abolished by appropriate cerebellum lesions (Clark et al. 1984). Actually, a number of loci in the brain exhibit the learning-induced neuronal model of the learned behavioral response. I return to this issue below. The breakthrough insofar as the hippocampus is concerned came in a study in our laboratory by Paul Solomon and Donald Weisz using the trace-conditioning

18 Nov 2004 11:52

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

10

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

P1: IKH

THOMPSON

paradigm, where a 500-msec period of no stimuli separated CS offset from US onset (Solomon et al. 1986). In brief, large bilateral hippocampal lesions made before training markedly impaired subsequent learning of the trace (but not delay) eyeblink CR, a result replicated exactly by Moyer et al. (1990). If the hippocampal lesions are made immediately after learning, the trace CR is abolished (with no effect on the delay CR), but if the lesions are made a month after training the trace CR remains intact (Kim et al. 1995). In sum, hippocampal lesions induced anterograde amnesia (impairment of postlesion learning) and marked but time-limited retrograde amnesia for the trace eyeblink CR. These are precisely the effects that such lesions have on declarative memory in humans (Squire 1987). These results suggest that trace eyeblink conditioning provides a simple model of hippocampal-dependent declarative memory, a possibility strongly supported by studies of humans with hippocampal-medial temporal lobe amnesia. Such patients are markedly impaired on acquisition of trace eyeblink conditioning if the trace interval is sufficiently long but not impaired at all on delay conditioning (see McGlinchey-Berroth et al. 1997). Clark & Squire (1998, 1999) made the striking observation that awareness of the training contingencies in normal human subjects correlated highly with the degree of learning of trace eyeblink conditioning. Recall that awareness is a key defining property of declarative memory. They also showed that awareness played no role in delay conditioning, and that expectancy of US occurrence influenced trace but not delay conditioning (Clark & Squire 2000, Clark et al. 2001). They conclude that delay and trace conditioning are fundamentally different phenomena, with delay conditioning inducing nondeclarative or procedural memory and trace conditioning inducing declarative memory. Hence, trace eyeblink conditioning in rabbits would seem to provide a very elementary instantiation of declarative memory in animals. These strikingly parallel results of hippocampal lesions in rabbits and humans for trace eyeblink conditioning suggest the possibility that a memory trace(s) may develop in the hippocampus in trace conditioning, but do not prove it. Recall that delay eyeblink conditioning results in the development of a pronounced neuronal model of the learned response in the hippocampus (as does trace training). Several lines of evidence support the view that localized changes do develop in the hippocampus in eyeblink conditioning. Thus, the properties of increased neuronal activity are strikingly similar to the properties of LTP (Berger et al. 1986, Thompson 1997, Weisz et al. 1984), and induction of LTP in the hippocampus can enhance eyeblink learning (Berger 1984). There is also a dramatic learninginduced decrease in the after-hyperpolarization in pyramidal neurons in eyeblink conditioning, most evident in trace conditioning, which leads to increased excitability. This alteration is intrinsic to the neurons due to a decrease in one or more calcium-dependent outward potassium currents (Disterhoft et al. 1986, Disterhoft & McEchron 2000). There is also a dramatic increase in dendritic membraneassociated protein kinase C after eyeblink conditioning (Olds et al. 1989). After trace eyeblink conditioning there is a substantial increase in the number of multiple

18 Nov 2004 11:52

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

LOCALIZATION OF MEMORY TRACES

P1: IKH

11

synapse boutons in hippocampal neurons (Geinisman et al. 2001). Finally, there is a dramatic increase in the number of new neurons in the hippocampus in trace, but not delay, eyeblink conditioning (Gould et al. 1999, Shors et al. 2001). These sets of structural and chemical changes are certainly consistent with the formation of higher-order memory traces in the hippocampus. But recall that cerebellar lesions can abolish both the behavioral learned response and the increased neuronal response in the CS period in the hippocampus in both delay and trace learning (see below and Clark et al. 1984, Woodruff-Pak et al. 1985). So doubts remain. Interestingly, the early (over trials)-developing increase in neuronal activity in the hippocampus in the US period is not completely abolished by cerebellar lesions (Clark et al. 1984). In my view, it is necessary to identify the entire essential circuitry from stimulus input to motor output in order to establish definitively that the essential memory trace, e.g., for trace conditioning, develops and is stored, albeit temporarily, in the hippocampus (see below). To date, not all the essential circuitry has been identified (see Thompson & Kim 1996). Yet another remaining puzzle is where the trace memories go after the hippocampus is no longer necessary.

The Essential Circuitry: The Cerebellum So the hippocampus is not the locus of the essential memory trace for standarddelay classical eyeblink conditioning. In a detailed mapping study of the brain stem we identified several loci where neurons exhibited the requisite neuronal model of the CR—increase in frequency of discharge that preceded the onset of the USUR and modeled the amplitude time source of the behavioral CR, just as did the neuronal response in the hippocampus (McCormick et al. 1983). The following brain areas showed this neuronal response predictive of the learned behavioral CR: the relevant motor nuclei (as noted above), a region in the vicinity of the fifth nucleus, various reticular regions, the cerebellar cortex (ansiform and anterior lobes), the cerebellar interpositus nucleus, the pontine nuclei, and the red nucleus. The response in the interpositus nucleus was particularly robust (Figure 4). The fact that a neuronal model of the behavioral CR developed in the cerebellum was suggestive of a memory trace but we knew from our earlier work on the hippocampus that neuronal recordings, per se, could not identify the essential memory trace. We completed a series of lesion studies, initially involving large cerebellar lesions and later lesions limited to the interpositus nucleus (e.g., McCormick & Thompson 1984a,b; Clark et al. 1984), all of which completely prevented learning and completely and permanently abolished the CR with no effect on the UR (thus ruling out performance variables). In a long series of studies completed while we were at Stanford, we identified virtually the entire essential memory trace circuit using electrophysiological recordings, electrical stimulation, lesions (aspiration, electrolytic, and chemical), and anatomical pathway tracing methods (see Chapman et al. 1988; Christian & Thompson 2003; McCormick et al. 1982a, 1985; Lavond et al. 1985; Mauk et al. 1986; Steinmetz et al. 1986, 1987; J.K. Thompson et al. 1985; Woodruff-Pak et al.

18 Nov 2004 11:52

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

12

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

P1: IKH

THOMPSON

Figure 4 Neuronal model of learned eyeblink response recorded from the cerebellar interpositus nucleus. Each graph shows nictitating membrane movement (top trace) and a histogram of multiple unit activity (bottom trace) averaged and summed over 100 trials. Animals that received explicitly unpaired presentations of the conditioning stimuli do not develop altered patterns of neuronal activity (left column). Trained animals develop a neuronal model of the learned behavior. Total trace duration 750 msec. (From McCormick & Thompson 1984b.)

1985). In brief, the efferent CR pathway projects from the interpositus nucleus ipsilateral to the trained eye, via the superior cerebellar peduncle, to the contralateral magnocellular red nucleus and to the relevant motor nuclei ipsilateral to the trained eye. The cerebellar interpositus lesion abolition of the eyeblink CR is strictly ipsilateral and is due to damage to neuron somas, not fibers of passage. The inferior

18 Nov 2004 11:52

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

LOCALIZATION OF MEMORY TRACES

P1: IKH

13

olive-climbing fiber system appears to be the critical US reinforcing or teaching pathway and the pontine nuclei and mossy fiber system appears to convey the necessary CS information to the cerebellum. A particularly satisfying aspect of our discovery of the essential role of the cerebellum in classical conditioning of discrete responses is the relevance of our work to the human condition. Irene Daum and associates, working in Germany, replicated in humans the fact that appropriate cerebellar damage completely prevents learning of the eyeblink CR (Daum et al. 1993), since replicated in many other studies. Christine Logan (a former graduate student of mine) and Scott Grafton, a neurologist then at the University of Southern California, completed an extensive positron emission tomography (PET) study of eyeblink conditioning in humans. They found significant activation in the cerebellar interpositus nucleus and several loci in cerebellar cortex, in close agreement with our recording studies in the rabbit cerebellum (Logan & Grafton 1995), again replicated in many other studies. A point that is obvious from our studies but seems not to be widely understood is that our results on the role of the cerebellum in classical conditioning apply to the learning of any discrete movement: eyeblink, head turn, forelimb flexion, hindlimb flexion, etc. (see Shinkman et al. 1996). Eyeblink conditioning is simply a convenient response to measure. Our findings to date seem to have identified perhaps the critical function of the cerebellum, namely the learning of discrete skilled movements, a basic notion proposed in classic theories of the cerebellum as a learning machine (see Albus 1971, Eccles 1977, Ito 1984, Marr 1969). Indeed, our work constitutes a compelling verification of these theories. Our data to this point strongly supported the hypothesis that the essential memory trace was formed and stored in the cerebellum, but did not prove it. Hence we adapted use of methods of reversible inactivation to localize the memory trace. The logic is straightforward. Having identified the essential memory trace circuit, reversibly inactivate each key locus in the circuit during training. If this completely prevents learning at a given locus, then this locus either conveys essential afferent information to the memory trace or is the site of the memory trace. However, if reversible inactivation of a given locus does not prevent learning at all, then this locus is efferent from the memory trace. But remember that inactivation of all these essential loci will completely prevent expression of the CR. We infused muscimol for inactivation of neuron cell bodies—it acts on gamma amino butyric acid (GABAA) receptors as an agonist and completely shuts down (hyperpolarizes) the neurons for several hours, after which they fully recover. To inactivate axons we infused tetrodotoxin (TTX) (see, e.g., Krupa et al. 1993; Krupa & Thompson 1995, 1997; Krupa et al. 1996; Thompson & Krupa 1994). David Lavond and his students completed parallel studies using reversible cooling (e.g., Clark et al. 1992, Clark & Lavond 1993). Results are completely consistent and have since been replicated in other laboratories. In brief (see Figure 5) inactivation limited to the anterior interpositus nucleus completely prevented learning. After removal of inactivation animals showed no signs of learning; with subsequent training they learned normally as though they

18 Nov 2004 11:52

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

14

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

P1: IKH

THOMPSON

Figure 5 Highly simplified schematic of the essential memory trace circuit for delay classical conditioning of the eyeblink response to illustrate the use of reversible inactivation to localize memory traces. Shading in a, b, and c indicate reversible inactivation of the key region of each structure during training using muscimol; d indicates inactivation of the axonal pathway exiting from the cerebellum, the superior cerebellar peduncle, using tetrodotoxin. Inactivation of the interpositus (c) completely prevents learning, but inactivation of the superior peduncle (d), the red nucleus (b), and the motor nuclei (a) does not prevent learning at all. (Modified from Thompson & Krupa 1994.)

had no prior training; there was no savings. In complete contrast, inactivation of the superior peduncle (the immediate efferent projection from the interpositus/dentate nuclei), the red nucleus, and the relevant motor nuclei, although completely preventing expression of the CR, did not prevent learning at all. After removal of the inactivation the animals had fully learned to asymptote. Thus, inactivation localized to the anterior interpositus nucleus completely prevented learning but inactivation of its immediate output pathway did not prevent learning at all. The fact that after interpositus inactivation there was no savings argues strongly that no part of the memory trace developed in structures afferent to the interpositus. Similarly, the fact that inactivation of structures efferent from the interpositus did not prevent learning at all argues that no part of the memory is formed in these structures. I noted above that several loci in the brain exhibited the neuronal model of the learned behavioral CR, e.g., trigeminal nuclear region, pontine nuclei, etc. In a series of studies Lavond and associates (and we) showed that with inactivation of

18 Nov 2004 11:52

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

LOCALIZATION OF MEMORY TRACES

P1: IKH

15

the interpositus nucleus, all these models disappeared, i.e., they are all driven from the interpositus (Bao et al. 2000, Lavond & Cartford 2000). Because muscimol acts only on neuron cell bodies and not on axons, the inactivation of the interpositus nucleus does not prevent normal activation of cerebellar cortex by the two major afferent systems, climbing fibers from the inferior olive and mossy fibers from many sources. The fact that no learning at all occurred with inactivation limited to the interpositus nucleus would seem to argue against formation of memory traces in the cerebellar cortex. However, this inactivation blocks the direct projections from the interpositus nucleus to the cerebellar cortex. When we first discovered the essential role of the cerebellum in eyeblink conditioning I had hoped the memory traces would be formed and stored in the cerebellar cortex. The cellular machinery in the cortex is vast (each Purkinje neuron receives up to 200,000 synapses from granule neurons). However, we were never able to completely prevent or abolish the CR by lesions limited to the cerebellar cortex. Other workers, particularly Yeo and associates (Yeo et al. 1984), claimed to have done so. The problem is with the lesion method—it is impossible to remove all of the cerebellar cortex without damaging the critical region of the interpositus nucleus (see detailed discussion in Christian & Thompson 2003). However, we consistently found that with large cerebellar cortical lesions or reversible inactivation, learning was much slower and to a lesser degree and adaptive timing of the CR was lost. Normally the eyeblink closure CR peaks at the onset of the US; it is a maximally adaptive response. After large cortical lesions, the CR peaks earlier in time and is no longer adaptive (Garcia et al. 1999, McCormick & Thompson 1984b). So the cortex is critically important for normal adaptive learning and it seemed very likely that higher-order memory traces were established there. But again the lesion method is inconclusive. Fortunately nature provided us with an ideal preparation, the Purkinje cell degeneration (pcd) mutant mouse. The brain of this mutant develops normally until about two weeks after birth. Then over the next several weeks, all Purkinje neurons in the cerebellar cortex die; as a result the animals have no functional cerebellar cortex. But the interpositus nucleus remains functional. Results were clear: The pcd mice were able to learn (L. Chen et al. 1996). They learned much more slowly and to a lesser degree than normal mice of the same strain (littermates), but they still showed significant and substantial learning. The CRs were shorter latency in the pcd mice. They also showed rapid extinction with CS-alone training. The possibility that the memory trace was somehow formed in loci other than the cerebellum was ruled out by lesioning the interpositus nucleus bilaterally in pcd mice before training. The lesioned pcd mice were unable to learn the conditioned eyeblink response (L. Chen et al. 1999). These results were very similar to the effects of large cerebellar cortical lesions we had found earlier in rabbits. So the cortex is critically important for normal learning, and higher-order memory traces are likely formed there. There is considerable evidence supporting the view that a process of long-term potentiation

18 Nov 2004 11:52

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

16

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

P1: IKH

THOMPSON

(LTD) at the parallel fiber-Purkinje neuron dendritic synapses, discovered by Ito (see 1984), plays a key role in initial plasticity (C. Chen & Thompson 1995, C. Chen et al. 1995, Shibuki et al. 1996, Kim & Thompson 1997). Indeed, Purkinje neurons show substantial changes in CS-evoked discharge frequency over the course of training, with many showing learning-induced decrease in discharge frequency, as could be expected if LTD developed (Christian et al. 2002, Christian & Thompson 2003). Other lines of evidence also support the view that memory traces are formed in cerebellar cortex (Cooke et al. 2004). The evidence is now very strong that the basic essential memory trace is formed and stored in the anterior interpositus nucleus for classical conditioning of the eyeblink response. Indeed, the reversible inactivation studies provide the key evidence (Figure 5). It is clear that long-lasting changes in the critical neurons in the anterior interpositus nucleus do develop. Thus, infusion of protein synthesis inhibitors in this locus completely prevents learning (Bracha et al. 1998, G. Chen & Steinmetz 2000, Gomi et al. 1999). Further, training results in selective expression of a protein kinase in this locus, KKIAMRE, an enzyme for the cell division cycle (Gomi et al. 1999). Pharmacological isolation of the interpositus from the cerebellar cortex reveals clear learning-induced increases in excitability of interpositus neurons (Bao et al. 2002, Garcia & Mauk 1998). Perhaps most convincing, eyeblink conditioning results in a dramatic and highly significant increase in the number of excitatory synapses (but not inhibitory synapses) in the interpositus nucleus (Kleim et al. 2002).

CONCLUSION The essential circuitry for classical conditioning of the eyeblink response is shown in Figure 6, along with the site of memory trace formation in the interpositus nucleus and a putative site of plasticity in the cerebellar cortical neurons. The nature of the memory is defined by this circuit; the circuit is the memory. The CS activates the sensory afferent pathways to the site(s) of trace storage in the cerebellum, which activates the efferent pathways to the motor nuclei and the learned behavior. The content of the memory, the conditioned eyeblink response, is completely defined and completely predictable from the essential circuit. The formal criteria developed by Richard Morris and associates to demonstrate that a given set of phenomena establish what they term the “synaptic plasticity and memory” (SPM) hypothesis are as follows (Martin et al. 2000): 1. DETECTABILITY: If an animal displays memory of some previous experience, a change in synaptic efficacy should be detectable somewhere in its nervous system. 2. MIMICRY: Conversely, if it were possible to induce the same spatial pattern of synaptic weight changes artificially, the animal should display “apparent” memory for some past experience which did not in practice occur.

18 Nov 2004 11:52

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

LOCALIZATION OF MEMORY TRACES

P1: IKH

17

Figure 6 Simplified schematic (most interneurons omitted) of the putative essential circuitry for delay classical conditioning of eyeblink (and other discrete responses) learned with an aversive US. (The sensory and motor nuclei activated depend of course on the nature of the CS and US; the more central portions of the circuit appear to be general.) The reflex US-UR pathway involves direct and indirect projection from the trigeminal nucleus to the motor nuclei (for the eyeblink UR and CR, primarily accessory 6 and 7). The tone CS pathway projects from auditory nuclei to the pontine nuclei and to the cerebellum as mossy fibers. The US pathway includes projections from the trigeminal to the inferior olive and to the cerebellum as climbing fibers. The CR pathway projects from the interpositus to the red nucleus and on to premotor and motor nuclei. There is also a direct GABAergic inhibitory projection from the interpositus to the inferior olive. Solid cell bodies and bar terminals indicate inhibitory neurons; open cell bodies and fork terminals indicate excitatory neurons. Stars indicate sites of plasticity based on current evidence. See text for details. (From Christian & Thompson 2003.) CR, conditioned response; CS, conditioned stimulus; UR, unconditioned response; US, unconditioned stimulus.

3. ANTEROGRADE ALTERATION: Interventions that prevent the induction of synaptic weight changes during a learning experience should impair the animal’s memory of that experience. 4. RETROGRADE ALTERATION: Interventions that alter the spatial distribution of synaptic weights induced by a prior learning experience (see detectability) should alter the animal’s memory of that experience (p. 651).

18 Nov 2004 11:52

18

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

P1: IKH

THOMPSON

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

I would expand the SPM notion to include nonsynaptic mechanisms of plasticity and memory as well, referring to all as the memory trace. I would also note that these criteria cannot be met unless the putative memory trace has been localized. Evidence to date on localization of the memory trace(s) for classical conditioning of eyeblink and other discrete responses would seem to satisfy Morris’ criteria. 1. DETECTABILITY: There is a dramatic increase in neuronal/synaptic efficacy in the cerebellum, both in the cortex of lobule HVI and in the anterior interpositus nucleus as a result of training (see, e.g., Shinkman et al. 1996, Bao et al. 2002). Indeed, long-term potentiation (LTP) has been reported to occur in the interpositus nucleus (Racine et al. 1986). 2. MIMICRY: Electrical microstimulation of the locus of the memory trace in the anterior interpositus nucleus can evoke the response to be learned before training (McCormick & Thompson 1984a, Chapman et al. 1988). 3. ANTEROGRADE ALTERATION: Infusion of muscimol into the site of the memory trace in the anterior interpositus nucleus completely prevents learning, i.e., induction of the memory trace. 4. RETROGRADE ALTERATION: Infusion of muscimol into the site of the memory trace in the anterior interpositus nucleus in a well-trained animal alters the synaptic weights (shuts them down) and abolishes the animal’s memory for that experience. These of course are only opinions of what constitutes demonstration of a memory trace. In my view the key is that the essential circuit defines the memory, as I noted above. But we still do not know the detailed nature of the memory trace in the interpositus and how it is formed. It will be necessary to identify all the steps in the causal chain from initial activation of the neurons at the beginning of training to the final form of the memory trace, from the biochemical/genetic processes to the structural changes in the synapses and neurons that code the permanent memory trace. ACKNOWLEDGMENTS Work described in this paper was supported in part by National Science Foundation Grant IBN 92,15069, National Institutes of Aging Grant AG14751, a grant from the Sankyo Company, and funds from the University of Southern California. The Annual Review of Psychology is online at http://psych.annualreviews.org

LITERATURE CITED Albus JS. 1971. A theory of cerebellar function. Math Biosci. 10:25–61 Bao S, Chen L, Kim JJ, Thompson RF. 2002. Cerebellar cortical inhibition and classical

eyeblink conditioning. Proc. Natl. Acad. Sci. USA 99:1592–97 Bao S, Chen L, Thompson RF. 2000. Learningand-cerebellum-dependent neuronal activity

18 Nov 2004 11:52

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

LOCALIZATION OF MEMORY TRACES in the lateral pontine nucleus. Behav. Neurosci. 114:254–61 Bao S, Cheng EF, Davis JD, Gobeske KT, Merzenich MM. 2003. Progressive degradation and subsequent refinement of acoustic representations in the adult auditory cortex. J. Neurosci. 23:10765–75 Berger TW. 1984. Long-term potentiation of hippocampal synaptic transmission affects rate of behavioral learning. Science 224:627– 30 Berger TW, Alger BE, Thompson RF. 1976. Neuronal substrate of classical conditioning in the hippocampus. Science 192:483– 85 Berger TW, Berry SD, Thompson RF. 1986. Role of the hippocampus in classical conditioning of aversive and appetitive behaviors. In The Hippocampus, Vols. III and IV, ed. RL Isaacson, KH Pribram, pp. 203–39. New York: Plenum Berger TW, Thompson RF. 1978a. Neuronal plasticity in the limbic system during classical conditioning of the rabbit nictitating membrane response. I. The hippocampus. Brain Res. 145:323–46 Berger TW, Thompson RF. 1978b. Identification of pyramidal cells as the critical elements in hippocampal neuronal plasticity during learning. Proc. Natl. Acad. Sci. USA 75:1572–76 Berry SD, Thompson RF. 1978. Prediction of learning rate from the hippocampal EEG. Science 200:1298–300 Bracha V, Irwin KB, Webster ML, Wunderlich DA, Stachowiak MK, Bloedel JR. 1998. Microinjections of anisomycin into the intermediate cerebellum during learning affect the acquisition of classically conditioned responses in the rabbit. Brain Res. 788:169–78 Cegavske CF, Patterson MM, Thompson RF. 1979. Neuronal unit activity in the abducens nucleus during classical conditioning of the nictitating membrane response in the rabbit Oryctolagus cuniculus. J. Comp. Physiol. Psychol. 93:595–609 Cegavske CF, Thompson RF, Patterson MM, Gormezano I. 1976. Mechanisms of effer-

P1: IKH

19

ent neuronal control of the reflex nictitating membrane response in the rabbit (Oryctolagus cuniculus). J. Comp. Physiol. Psychol. 90:411–23 Chapman PF, Steinmetz JE, Thompson RF. 1988. Classical conditioning does not occur when direct stimulation of the red nucleus or cerebellar nuclei is the unconditioned stimulus. Brain Res. 442:97–104 Chen C, Kano M, Abeliovich A, Chen L, Bao S, et al. 1995. Impaired motor coordination correlates with persistent multiple climbing fiber innervation in PKCγ mutant mice. Cell 83:1233–42 Chen C, Thompson RF. 1995. Temporal specificity of long-term depression in parallel fiber-Purkinje synapses in rat cerebellar slice. Learn. Mem. 2:185–98 Chen G, Steinmetz JE. 2000. Microinfusion of protein kinase inhibitor H7 in the cerebellum impairs the acquisition but not retention of classical eyeblink conditioning in rabbits. Brain Res. 856:193–201 Chen L, Bao S, Lockard JM, Kim JJ, Thompson RF. 1996. Impaired classical eyeblink conditioning in cerebellar lesioned and Purkinje cell degeneration (pcd) mutant mice. J. Neurosci. 16:2829–38 Chen L, Bao S, Thompson RF. 1999. Bilateral lesions of the interpositus nucleus completely prevent eyeblink conditioning in Purkinje cell degeneration mutant mice. Behav. Neurosci. 113:204–10 Christian KM, Poulos AM, Thompson RF. 2002. Purkinje cell activity during classical conditioning of the eyeblink reflex in rabbits. Soc. Neurosci. Abstr. 79.9 Christian KM, Thompson RF. 2003. Neural substrates of eyeblink conditioning: acquisition and retention. Learn. Mem. 11:427–55 Clark GA, McCormick DA, Lavond DG, Thompson RF. 1984. Effects of lesions of cerebellar nuclei on conditioned behavioral and hippocampal neuronal responses. Brain Res. 291:125–36 Clark RE, Lavond DG. 1993. Reversible lesions of the red nucleus during acquisition and retention of a classically conditioned

18 Nov 2004 11:52

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

20

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

P1: IKH

THOMPSON

behavior in rabbits. Behav. Neurosci. 107: 264–70 Clark RE, Manns JR, Squire LR. 2001. Trace and delay eyeblink conditioning: contrasting phenomena of declarative and nondeclarative memory. Psychol. Sci. 12:304–8 Clark RE, Squire LR. 1998. Classical conditioning and brain systems: the role of awareness. Science 280:77–81 Clark RE, Squire LR. 1999. Human eyeblink classical conditioning: effects of manipulating awareness of the stimulus contingencies. Psychol. Sci. 10:14–18 Clark RE, Squire LR. 2000. Awareness and the conditioned eyeblink response. In Eyeblink Classical Conditioning: Applications in Humans, ed. DS Woodruff-Pak, JE Steinmetz, 1:229–51. Boston, MA: Kluwer Acad. Clark RE, Zhang AA, Lavond DG. 1992. Reversible lesions of the cerebellar interpositus nucleus during acquisition and retention of a classically conditioned behavior. Behav. Neurosci. 106:879–88 Cook SF, Attwell PJE, Yeo CH. 2004. Temporal properties of cerebellar-dependent memory consolidation. J. Neurosci. 24:2934–41 Daum I, Schugens MM, Ackermann H, Lutzenberger W, Dichgans J, Birbaumer N. 1993. Classical conditioning after cerebellar lesions in humans. Behav. Neurosci. 107:748– 56 Disterhoft JF, Coulter DA, Alkon DL. 1986. Conditioning-specific membrane changes of rabbit hippocampal neurons measured in vitro. Proc. Natl. Acad. Sci. USA 83:2733– 37 Disterhoft JF, McEchron MD. 2000. Cellular alterations in hippocampus during acquisition and consolidation of hippocampusdependent trace eyeblink conditioning. In Eyeblink Classical Conditioning: Animal Models, ed. DS Woodruff-Pak, JE Steinmetz, 2:313–34. Boston, MA: Kluwer Acad. Disterhoft JF, Olds J. 1972. Differential development of conditioned unit changes in thalamus and cortex of rat. J. Neurophysiol. 35:665–79 Eccles JC. 1977. An instruction-selection the-

ory of learning in the cerebellar cortex. Brain Res. 127:327–52 Edeline JM, Weinberger NM. 1991. Subcortical adaptive filtering in the auditory system: associative receptive field plasticity in the dorsal medial geniculate body. Behav. Neurosci. 105:154–75 Ezzeddine Y, Glanzman DL. 2003. Prolonged habituation of the gill-withdrawal reflex in Aplysia depends on protein synthesis, protein phosphatase activity, and postsynaptic glutamate receptors. J. Neurosci. 23:9585–94 Farel PB, Thompson RF. 1976. Habituation of a monosynaptic response in frog spinal cord: evidence for a presynaptic mechanism. J. Neurophysiol. 39:661–66 Gabriel M. 1976. Short-latency discriminative unit response: engram or bias? Physiol. Psychol. 4:275–80 Garcia KS, Mauk MD. 1998. Pharmacological analysis of cerebellar contributions to the timing and expression of conditioned eyelid responses. Neuropharmacol. 37:471– 80 Garcia KS, Steele PM, Mauk MD. 1999. Cerebellar cortex lesions prevent acquisition of conditioned eyelid responses. J. Neurosci. 19:10940–47 Geinisman Y, Berry RW, Disterhoft JF, Power JM, Van der Zee EA. 2001. Associative learning elicits the formation of multiple-synapse boutons. J. Neurosci. 21:5568–73 Gomi H, Sun W, Finch CE, Itohara S, Yoshimi K, Thompson RF. 1999. Learning induces a CDC2-related protein kinase, KKIAMRE. J. Neurosci. 19:9530–37 Gormezano I, Kehoe EJ, Marshall BS. 1983. Twenty years of classical conditioning with the rabbit. Prog. Psychobiol. Physiol. Psychol. 10:197–275 Gould E, Beylin A, Tanapat P, Reeves A, Shors TJ. 1999. Learning enhances adult neurogenesis in the hippocampal formation. Nat. Neurosci. 2:260–65 Groves PM, Thompson RF. 1970. Habituation: a dual-process theory. Psychol. Rev. 77:419– 50 Hiraoka M, Shimamura M. 1977. Neural

18 Nov 2004 11:52

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

LOCALIZATION OF MEMORY TRACES mechanisms of the corneal blinking reflex in cats. Brain Res. 125:265–75 Ito M. 1984. The Cerebellum and Neural Control. New York: Raven Ivkovich D, Thompson RF. 1997. Motor cortex lesions do not affect learning or performance of the eyeblink response in rabbits. Behav. Neurosci. 111:727–38 Kandel ER. 1975. The Cellular Basis of Behavior: An Introduction to Behavioral Neurobiology. San Francisco: Freeman Kandel ER, Spencer WA. 1968. Cellular neurophysiological approaches in the study of learning. Physiol. Rev. 48:65–134 Kim JJ, Clark RE, Thompson RF. 1995. Hippocampectomy impairs the memory of recently, but not remotely, acquired trace eyeblink conditioned responses. Behav. Neurosci. 109:195–203 Kim JJ, Thompson RF. 1997. Cerebellar circuits and synaptic mechanisms involved in classical eyeblink conditioning. Trends Neurosci. 20:177–81 Kleim JA, Freeman JH Jr, Bruneau R, Nolan BC, Cooper NR, Zook A, et al. 2002. Synapse formation is associated with memory storage in the cerebellum. Proc. Natl. Acad. Sci. USA 99:13228–31 Krupa DJ, Thompson JK, Thompson RF. 1993. Localization of a memory trace in the mammalian brain. Science 260:989–91 Krupa DJ, Thompson RF. 1995. Inactivation of the superior cerebellar peduncle blocks expression but not acquisition of the rabbit’s classically conditioned eyeblink response. Proc. Natl. Acad. Sci. USA 92:5097–101 Krupa DJ, Thompson RF. 1997. Reversible inactivation of the cerebellar interpositus nucleus completely prevents acquisition of the classically conditioned eyeblink response. Learn. Mem. 3:545–56 Krupa DJ, Weng J, Thompson RF. 1996. Inactivation of brainstem motor nuclei blocks expression but not acquisition of the rabbit’s classically conditioned eyeblink response. Behav. Neurosci. 110:219–27 Lashley KS. 1950. In search of the engram. Soc. Exp. Biol. Symp. 4:454–82

P1: IKH

21

Lavond DG, Cartford MC. 2000. Eyeblink conditioning circuitry: tracing, lesion, and reversible lesion experiments. In Eyeblink Classical Conditioning: Animal Models, ed. DS Woodruff-Pak, JE Steinmetz, 2:51–80. Boston, MA: Kluwer Acad. Lavond DG, Hembree TL, Thompson RF. 1985. Effect of kainic acid lesions of the cerebellar interpositus nucleus on eyelid conditioning in the rabbit. Brain Res. 326:179–83 Lavond DG, Kim JJ, Thompson RF. 1993. Mammalian brain substrates of aversive classical conditioning. Annu. Rev. Psychol. 44:317–42 Lavond DG, Logan CG, Sohn JH, Garner WD, Kanzawa SA. 1990. Lesions of the cerebellar interpositus nucleus abolish both nictitating membrane and eyelid EMG conditioned responses. Brain Res. 514:238–48 Logan CG, Grafton ST. 1995. Functional anatomy of human eyeblink conditioning determined with regional cerebral glucose metabolism and positron-emission tomography. Proc. Natl. Acad. Sci. USA 92:7500–4 Marr D. 1969. A theory of cerebellar cortex. J. Physiol. 202:437–70 Martin SJ, Grimwood PD, Morris RGM. 2000. Synaptic plasticity and memory: an evaluation of the hypothesis. Annu. Rev. Neurosci. 23:649–711 Mauk MD, Steinmetz JE, Thompson RF. 1986. Classical conditioning using stimulation of the inferior olive as the unconditioned stimulus. Proc. Natl. Acad. Sci. USA 83:5349–53 McCormick DA, Clark GA, Lavond DG, Thompson RF. 1982a. Initial localization of the memory trace for a basic form of learning. Proc. Natl. Acad. Sci. USA 79:2731–42 McCormick DA, Lavond DG, Thompson RF. 1982b. Concomitant classical conditioning of the rabbit nictitating membrane and eyelid responses: correlations and implications. Physiol. Behav. 28:769–75 McCormick DA, Lavond DG, Thompson RF. 1983. Neuronal responses of the rabbit brainstem during performance of the classically conditioned nictitating membrane (NM/eyelid response). Brain Res. 271:73–88

18 Nov 2004 11:52

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

22

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

P1: IKH

THOMPSON

McCormick DA, Steinmetz JE, Thompson RF. 1985. Lesions of the inferior olivary complex cause extinction of the classically conditioned eyeblink response. Brain Res. 359: 120–30 McCormick DA, Thompson RF. 1984a. Cerebellum: essential involvement in the classically conditioned eyelid response. Science 223:296–99 McCormick DA, Thompson RF. 1984b. Neuronal responses of the rabbit cerebellum during acquisition and performance of a classically conditioned nictitating membraneeyelid response. J. Neurosci. 4:2811–22 McGlinchey-Berroth R, Carrillo MC, Gabrieli JD, Brawn CM, Disterhoft JF. 1997. Impaired trace eyeblink conditioning in bilateral, medial-temporal lobe amnesia. Behav. Neurosci. 111:873–82 Moyer JR Jr, Deyo RA, Disterhoft JF. 1990. Hippocampectomy disrupts trace eye-blink conditioning in rabbits. Behav. Neurosci. 104:243–52 Olds J, Anderson ML, McPhie DL, Staten LD, Alkon DL. 1989. Imaging of memoryspecific changes in the distribution of protein kinase C in the hippocampus. Science 245:866–69 Olds J, Disterhoft J, Segal M, Kornblith DL, Hirsh R. 1972. Learning centers of rat brain mapped by measuring latencies of conditioned unit responses. J. Neurophysiol. 35: 202–19 Patterson MM, Cegavske CF, Thompson RF. 1973. Effects of classical conditioning paradigm on hindlimb flexor nerve response in immobilized spinal cat. J. Comp. Physiol. Psychol. 84:88–97 Racine RJ, Wilson DA, Gingell R, Sunderland D. 1986. Long-term potentiation in the interpositus and vestibular nuclei in the rat. Exp. Brain Res. 63:158–62 Schacter DL. 1987. Implicit memory: history and current status. Exp. Psychol. Learn. Mem. Cogn. 13:501–18 Schmaltz LW, Theios J. 1972. Acquisition and extinction of a classically conditioned response in hippocampectomized rabbits

(Oryctolagus cuniculus). J. Comp. Physiol. Psychol. 79:328–33 Shibuki K, Gomi H, Chen L, Bao S, Kim JJ, et al. 1996. Deficient cerebellar long-term depression, impaired eyeblink conditioning and normal motor coordination in GFAP mutant mice. Neuron 16:587–99 Shinkman PG, Swain RA, Thompson RF. 1996. Classical conditioning with electrical stimulation of cerebellum as both conditioned and unconditioned stimulus. Behav. Neurosci. 110:914–21 Shors TJ, Matzel LD. 1997. Long-term potentiation: What’s learning got to do with it? Behav. Brain Sci. 20:597–614; discussion 614–55 Shors TJ, Miesegaes G, Beylin A, Zhao M, Rydel T, Gould E. 2001. Neurogenesis in the adult is involved in the formation of trace memories. Nature 410:372–76 Solomon PR, Vander Schaaf ER, Thompson RF, Weisz DJ. 1986. Hippocampus and trace conditioning of the rabbit’s classically conditioned nictitating membrane response. Behav Neurosci. 100:729–44 Squire LR. 1987. Memory and Brain. New York: Oxford Univ. Press Squire LR. 1992. Declarative and nondeclarative memory: multiple brain systems supporting learning and memory. J. Cogn. Neurosci. 4:232–43 Steinmetz JE, Logan CG, Rosen DJ, Thompson JK, Lavond DG, Thompson RF. 1987. Initial localization of the acoustic conditioned stimulus projection system to the cerebellum essential for classical eyelid conditioning. Proc. Natl. Acad. Sci. USA 84:3531–35 Steinmetz JE, Rosen DJ, Chapman PF, Lavond DG, Thompson RF. 1986. Classical conditioning of the rabbit eyelid response with a mossy fiber stimulation CS. I. Pontine nuclei and middle cerebellar peduncle stimulation. Behav. Neurosci. 100:871–80 Thompson JK, Lavond DG, Thompson RF. 1985. Cerebellar interpositus/dentate nuclei afferent seen with retrograde fluorescent tracers in the rabbit. Neurosci. Abstr. 11:1112 Thompson RF. 1997. Classical conditioning

18 Nov 2004 11:52

AR

AR231-PS56-01.tex

AR231-PS56-01.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

LOCALIZATION OF MEMORY TRACES has much to do with LTP. Behav. Brain Sci. 20:632–33 Thompson RF, Berger TW, Cegavske CF, Patterson MM, Roemer RA, et al. 1976. The search for the engram. Am. Psychol. 31:209– 27 Thompson RF, Kim JJ. 1996. Memory systems in the brain and localization of a memory. Proc. Natl. Acad. Sci. USA 93:13438–44 Thompson RF, Krupa DJ. 1994. Organization of memory traces in the mammalian brain. Annu. Rev. Neurosci. 17:519–49 Thompson RF, Patterson MM, Teyler TJ. 1972. Neurophysiology of learning. Annu. Rev. Psychol. 23:73–104 Thompson RF, Spencer WA. 1966. Habituation: a model phenomenon for the study of neuronal substrates of behavior. Psychol. Rev. 73:16–43 Tulving E. 1985. How many memory systems are there? Am. Psychol. 40:385–98 Wagner AR, Donegan NH. 1989. Some relationships between a computational model (SOP) and a neural circuit for Pavlovian (rabbit eyeblink) conditioning. In The Psychology of Learning and Motivation, ed. RD Hawkins, GH Bower, 22:157–203. San Diego: Academic Weinberger NM, Hopkins W, Diamond DM. 1984. Physiological plasticity of single neurons in auditory cortex of the cat during acquisition of the papillary conditioned re-

P1: IKH

23

sponse: I. Primary Field (AI). Behav. Neurosci. 98:171–88 Weisz DJ, Clark GA, Thompson RF. 1984. Increased activity of dentate granule cells during nictitating membrane response conditioning in rabbits. Behav. Brain Res. 12:145– 54 Wilson MA, Tonegawa S. 1997. Synaptic plasticity, place cells and spatial memory: study with second generation knockouts. Trends Neurosci. 20:102–6 Woodruff-Pak DS, Lavond DG, Thompson RF. 1985. Trace conditioning: abolished by cerebellar nuclear lesions but not lateral cerebellar cortex aspirations. Brain Res. 348:249– 60 Woody CD, Alkon DL, Hay B. 1984. Depolarization-induced effects of Ca2+-calmodulindependent protein kinase injection, in vivo, in single neurons of cat motor cortex. Brain Res. 321:192–97 Yeo CH, Hardiman MJ, Glickstein M. 1984. Discrete lesions of the cerebellar cortex abolish the classically conditioned nictitating membrane response of the rabbit. Behav. Brain Res. 13:261–66 Young RA, Cegavske CF, Thompson RF. 1976. Tone-induced charges in excitability of abducens motoneurons and the reflex path of the rabbit nictitating membrane response. J. Comp. Physiol. Psychol. 90:424– 34

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

Annual Review of Psychology Volume 56, 2005

CONTENTS Frontispiece—Richard F. Thompson

xviii

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

PREFATORY In Search of Memory Traces, Richard F. Thompson

1

DECISION MAKING Indeterminacy in Brain and Behavior, Paul W. Glimcher

25

BRAIN IMAGING/COGNITIVE NEUROSCIENCE Models of Brain Function in Neuroimaging, Karl J. Friston

57

MUSIC PERCEPTION Brain Organization for Music Processing, Isabelle Peretz and Robert J. Zatorre

89

SOMESTHETIC AND VESTIBULAR SENSES Vestibular, Proprioceptive, and Haptic Contributions to Spatial Orientation, James R. Lackner and Paul DiZio

115

CONCEPTS AND CATEGORIES Human Category Learning, F. Gregory Ashby and W. Todd Maddox

149

ANIMAL LEARNING AND BEHAVIOR: CLASSICAL Pavlovian Conditioning: A Functional Perspective, Michael Domjan

179

NEUROSCIENCE OF LEARNING The Neuroscience of Mammalian Associative Learning, Michael S. Fanselow and Andrew M. Poulos

207

HUMAN DEVELOPMENT: EMOTIONAL, SOCIAL, AND PERSONALITY Behavioral Inhibition: Linking Biology and Behavior Within a Developmental Framework, Nathan A. Fox, Heather A. Henderson, Peter J. Marshall, Kate E. Nichols, and Melissa A. Ghera

235

BIOLOGICAL AND GENETIC PROCESSES IN DEVELOPMENT Human Development: Biological and Genetic Processes, Irving I. Gottesman and Daniel R. Hanson

263 vii

P1: JRX

December 8, 2004

viii

12:13

Annual Reviews

AR231-FM

CONTENTS

SPECIAL TOPICS IN PSYCHOPATHOLOGY The Psychology and Neurobiology of Suicidal Behavior, Thomas E. Joiner Jr., Jessica S. Brown, and LaRicka R. Wingate

287

DISORDERS OF CHILDHOOD Autism in Infancy and Early Childhood, Fred Volkmar, Kasia Chawarska, and Ami Klin

315

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

CHILD/FAMILY THERAPY Youth Psychotherapy Outcome Research: A Review and Critique of the Evidence Base, John R. Weisz, Amanda Jensen Doss, and Kristin M. Hawley

337

ALTRUISM AND AGGRESSION Prosocial Behavior: Multilevel Perspectives, Louis A. Penner, John F. Dovidio, Jane A. Piliavin, and David A. Schroeder

365

INTERGROUP RELATIONS, STIGMA, STEREOTYPING, PREJUDICE, DISCRIMINATION The Social Psychology of Stigma, Brenda Major and Laurie T. O’Brien

393

PERSONALITY PROCESSES Personality Architecture: Within-Person Structures and Processes, Daniel Cervone

423

PERSONALITY DEVELOPMENT: STABILITY AND CHANGE Personality Development: Stability and Change, Avshalom Caspi, Brent W. Roberts, and Rebecca L. Shiner

453

WORK MOTIVATION Work Motivation Theory and Research at the Dawn of the Twenty-First Century, Gary P. Latham and Craig C. Pinder

485

GROUPS AND TEAMS Teams in Organizations: From Input-Process-Output Models to IMOI Models, Daniel R. Ilgen, John R. Hollenbeck, Michael Johnson, and Dustin Jundt

517

LEADERSHIP Presidential Leadership, George R. Goethals

545

PERSONNEL EVALUATION AND COMPENSATION Personnel Psychology: Performance Evaluation and Pay for Performance, Sara L. Rynes, Barry Gerhart, and Laura Parks

571

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

CONTENTS

ix

PSYCHOPHYSIOLOGICAL DISORDERS AND PSYCHOLOGICAL EFFECTS ON MEDICAL DISORDERS Psychological Approaches to Understanding and Treating Disease-Related Pain, Francis J. Keefe, Amy P. Abernethy, and Lisa C. Campbell

601

TIMELY TOPIC Psychological Evidence at the Dawn of the Law’s Scientific Age, David L. Faigman and John Monahan

631

Annu. Rev. Psychol. 2005.56:1-23. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDEXES Subject Index Cumulative Index of Contributing Authors, Volumes 46–56 Cumulative Index of Chapter Titles, Volumes 46–56

ERRATA An online log of corrections to Annual Review of Psychology chapters may be found at http://psych.annualreviews.org/errata.shtml

661 695 700

18 Nov 2004 10:44

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

P1: IKH LaTeX2e(2002/01/18) 10.1146/annurev.psych.55.090902.141429

Annu. Rev. Psychol. 2005. 56:25–56 doi: 10.1146/annurev.psych.55.090902.141429 c 2005 by Annual Reviews. All rights reserved Copyright
First published online as a Review in Advance on September 10, 2004

INDETERMINACY IN BRAIN AND BEHAVIOR Paul W. Glimcher

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Center for Neural Science, New York University, New York, New York 10003; email: [email protected]

Key Words randomness, decision-making, choice, game theory ■ Abstract The central goal of modern science that evolved during the Enlightenment was the empirical reduction of uncertainty by experimental inquiry. Although there have been challenges to this view in the physical sciences, where profoundly indeterminate events have been identified at the quantum level, the presumption that physical phenomena are fundamentally determinate seems to have defined modern behavioral science. Programs like those of the classical behaviorists, for example, were explicitly anchored to a fully deterministic worldview, and this anchoring clearly influenced the experiments that those scientists chose to perform. Recent advances in the psychological, social, and neural sciences, however, have caused a number of scholars to begin to question the assumption that all of behavior can be regarded as fundamentally deterministic in character. Although it is not yet clear whether the generative mechanisms for human and animal behavior will require a philosophically indeterminate approach, it is clear that behavioral scientists of all kinds are beginning to engage the issues of indeterminacy that plagued physics at the beginning of the twentieth century. CONTENTS INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Determinism and the Philosophy of Science . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Do Indeterminacies in the Physical World Matter for Behavioral Scientists? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . THE RISING TIDE OF APPARENT INDETERMINACY . . . . . . . . . . . . . . . . . . . . . Indeterminacy in the Social Sciences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Empirical Measurements of Behavioral Indeterminacy . . . . . . . . . . . . . . . . . . . . . . Reducing Uncertainty: Looking for Determinacy with Neurophysiology . . . . . . . . Indeterminacy at the Cellular and Subcellular Levels . . . . . . . . . . . . . . . . . . . . . . . . THE CHALLENGE OF INDETERMINACY FOR BEHAVIORAL SCIENCE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

25 26 27 29 29 35 40 46 50

INTRODUCTION Our modern view that the central function of scientific inquiry is to reduce uncertainty emerged early in the scientific revolution that constituted the Enlightenment; by the time of Galileo’s death (cf. Bacon 1620, Descartes 1637, Galilei 1630, 0066-4308/05/0203-0025$14.00

25

18 Nov 2004 10:44

26

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

P1: IKH

GLIMCHER

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Kepler 1618–1621) it was clear that improving the accuracy with which one could predict future events as determinate processes would be a central feature of the scientific method at both theoretical and empirical levels in the physical sciences. Over the course of the eighteenth and nineteenth centuries, the early social sciences emulated this trend, seeking to develop causal relationships in a testable and determinate fashion (cf. Keynes 1936, Smith 1776). By the twentieth century, the notion that scientific inquiry would reduce animal behavior to deterministic certainty had become a mainstream notion in psychological circles as well. Nowhere is this clearer than in the work of Pavlov. As Pavlov put it in Conditioned Reflexes: The physiologist must thus take his own path, where a trail has already been blazed for him. Three hundred years ago Descartes evolved the idea of the reflex. Starting from the assumption that animals behaved simply as machines, he regarded every activity of the organism as a necessary reaction to some external stimulus. . .. A bold attempt to apply the idea of the reflex to the activities of the [cerebral] hemispheres was made by the Russian physiologist I.M. Sechenov, on the basis of the knowledge available in his day of the physiology of the central nervous system. In a pamphlet entitled “Reflexes of the Brain,” published in Russia in 1863, he attempted to represent the activities of the cerebral hemispheres as reflex—that is to say, as determined. (Pavlov 1927) In the period that followed, Skinner and his students (cf. Skinner 1938) strengthened this notion, and psychologists as a group largely embraced the idea that a complete psychological theory would be a determinate one. By studying the causal relationships between environment, organism, and response, these scientists began the process of developing a predictive and testable theory of psychology. The twentieth century witnessed a similar trend in the effective application of the scientific method toward understanding the biological sources of behavior, and as a result, saw the development of a powerful deterministic program for understanding biological systems. Charles Sherrington (1906), for example, applied this programmatic approach to the physiological study of reflexes with great effect.

Determinism and the Philosophy of Science In philosophical circles, the central role of a determinate worldview in the classical scientific method also became a formalized principle in this period. In the early part of the twentieth century, the philosopher Karl Popper (1934) explicitly defined the goal of modern science as the falsification of extant theories through experimental inquiry. For Popper, theories could never be proven in practice, only subjected to the test of falsification. If experimental evidence falsifies a theory then it can be discarded; if experimental evidence corroborates a theory then it can be tentatively retained. What is critical about this logic is what it implies about indeterminacy. Consider the theoretical claim that if I flip a certain coin there is a 50% chance it will land

18 Nov 2004 10:44

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDETERMINATE BEHAVIOR

P1: IKH

27

heads-up. As Popper points out, this is not only an unverifiable theoretical claim, but also an untestable one; my assertion predicts all possible empirical outcomes and is thus unfalsifiable. Even more importantly, my theoretical claim predicts as an experimental result all possible finite series of coin flips that could ever be observed. If the coin is equally likely to land heads-up and tails-up, then any specific series of heads and tails is equally likely, whether that be six heads in a row or six flips that alternate between heads and tails. No formal prediction of any particular outcome is ever possible and for this reason Popper argued that probabilistic claims about indeterminate processes were irremediably problematic. Indeed, in his early writings Popper even used this to argue that the notion of a fundamentally indeterminate universe is at base a nonscientific proposition. In the 1920s and 1930s, however, the emerging discipline of quantum physics raised an important challenge to this notion that had evolved during the Enlightenment and motivated much of Popper’s work. Based initially on the work of Heisenberg and his colleagues (Heisenberg 1930, 1952), strong evidence arose suggesting that several phenomena that occur at the atomic and subatomic scales are, in fact, fundamentally indeterminate and thus could be described only probabilistically. This was a critical challenge to the existing philosophy of science as expressed by Popper because that philosophy argued that a theory of physics built upon probability theory was unfalsifiable, perhaps even unscientific. Nevertheless, the empirical evidence gathered during that period seemed to indicate unambiguously that at a small enough scale of analysis, events occur that are fundamentally indeterminate. This indicated that the philosophy of science, rather than the reality of our physical universe, might have to change.

Do Indeterminacies in the Physical World Matter for Behavioral Scientists? What, if any, are the implications of these issues for the study of behavior? Even if there are fundamental indeterminacies in the physical world, should this matter to behavioral scientists? Many scholars believe that the quantum physicist Edwin Schrodinger provided an answer to that question in his book, What Is Life (1944), in which he argued that for any organism to survive it must operate, in principle, in a fully determinate environment. Indeterminacy, he believed, would be lethal to living systems. Schrodinger’s own work (cf. 1951) had demonstrated that at the atomic and subatomic scales, matter can be described only in probabilistic terms, but it had also shown that large aggregates of these elementary particles behaved in an effectively determinate manner. His argument was that living cells were large enough objects that they would never interact with single atomic or subatomic particles, but only with these larger determinate aggregates. In essence, he argued that cells were large enough that quantum fluctuations in the properties of individual atoms would have no effect on them. Indeed, he went on to argue that biological cells are the size that they are specifically because quantum indeterminacy prevents them from surviving if they become any smaller. Biologists,

18 Nov 2004 10:44

28

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

P1: IKH

GLIMCHER

psychologists, and social scientists, he assured us, need not be concerned with fundamental indeterminacy in the universe:

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

If it were not so, if we were organisms so sensitive that a single atom, or even a few atoms, could make a perceptible impression on our senses—Heavens, what would life be like! To stress one point: an organism of that kind would most certainly not be capable of developing the kind of orderly thought which, after passing through a long sequence of earlier stages, ultimately results in forming, among many other ideas, the idea of an atom. (Schrodinger 1951) Recently, however, evidence has begun to arise in the social, psychological, and neurobiological domains that suggests that, at larger scales of analysis than the one Schrodinger examined in What Is Life, living systems exhibit behavior that is apparently indeterminate (cf. Hastie 2001, Schall 2004, Shafir & LeBoeuf 2002, Staddon & Cerutti 2003). At the largest scale of analysis, social scientists working in areas such as the theory of games have begun to argue that for behavior to be efficient under some circumstances, it must be irreducibly uncertain from the point of view of other organisms and therefore must be studied with the tools of probability theory. In principle, this raises critical problems for game theory. For all of the reasons Popper identified, when game theory makes probabilistic predictions it does so in a manner that is nonfalsifiable. Of course, if Schrodinger was correct, the apparent indeterminacy of game theory presents only a temporary impediment to scientific inquiry. A reductionist approach to human behavior during strategic games would ultimately reveal the mechanisms that give rise to this apparent indeterminacy and thus should ultimately yield a falsifiable determinate theory of human behavior. Although contemporary game theory thus faces indeterminacy, empirical science can hope to resolve this apparent indeterminacy by reduction. Interestingly, however, psychologists working at a lower level of reduction than social scientists have also begun to find evidence of apparent indeterminacy in the systems they study (cf. Staddon & Cerruti 2003). Recently, psychologists have begun the analysis of apparently stochastic patterns of individual responses and have been able to demonstrate classes of individual behavior that appear to be as fully random as can be measured. Indeterminacy, in the hands of these psychologists, seems to be an apparent feature of the behavior of single humans and animals. At a yet deeper level of reduction, neurobiologists have also begun to gather evidence for the existence of apparently indeterminate processes within the architecture of the mammalian brain (cf. Schall 2004). The patterns of action potentials generated by individual neurons, for example, appear highly stochastic for reasons that are not yet well understood. Growing evidence that apparently indeterminate processes operate at social, psychological, and even neurobiological levels are bringing behavioral scientists face-to-face with the same philosophical and scientific issues faced by Popper, Heisenberg, Schrodinger, and others in the last century. Can such theories be scientific, or is calling a neural signal or a behavior a random process only an excuse for ignorance? It may be that behavioral scientists will choose to assert as an

18 Nov 2004 10:44

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

INDETERMINATE BEHAVIOR

P1: IKH

29

axiom that all of the physical phenomena we study are fundamentally determinate in order to avoid these issues, but on the other hand such an assertion may force us to neglect a growing body of compelling evidence.

THE RISING TIDE OF APPARENT INDETERMINACY

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Indeterminacy in the Social Sciences Like scholars in the physical sciences, social scientists in the eighteenth and nineteenth centuries strongly emphasized a determinate scientific approach in their study of human behavior. The classic economic theory of that period, for example, rested on the foundation of a theory of determinate utility developed by Blaise Pascal (1670, Arnauld & Nicole 1662) and Daniel Bernoulli (1738). This utility theory argued that humans act predictably to maximize benefits and to minimize costs, and that the costs and benefits of any action can be reliably computed. Pascal had developed this basic logic in the seventeenth century, arguing that the “expected value” of an action was equal to the product of any possible gain or loss incurred by that action and the likelihood of the gain or loss. Bernoulli had extended this notion with the observation that humans appear at an empirical level to be more averse to risk than Pascal’s formulation predicts. Bernoulli’s conclusion was that humans made decisions based on the product of a subjective estimate of cost or benefit and the likelihood of that gain or loss, rather than based on an objective measure of gains or losses. Because of the precise form of his hypothesis, Bernoulli was able to show that this notion could successfully account for the empirically observed aversion of humans to risk. Thus, the critical idea that utility theory advanced was that one could compute the relative desirabilities of all possible actions to a chooser and, except in the presumably rare case where two actions have identical subjective desirabilities, one could then predict the actions of a chooser with determinate precision. Building on this foundation, Adam Smith (1776) argued that all economic actors could be seen as effectively trading off costs and benefits to maximize gain in a complex marketplace. The prices of goods, for example, were presumed to be set by the determinate interactions of supply and demand curves that represented the aggregate subjective desirabilities and costs of goods to consumers and producers. It was thus a central thesis of eighteenthand nineteenth-century economic theory that the rational process by which desirability was assessed could be accurately modeled and that these models made deterministic predictions about human behavior. Importantly, the incorporation of likelihoods into expected utility theory allowed the approach to make determinate predictions even when the environment in which human decision makers operated was unpredictable. Choosers were assumed to consider risk when they determined the desirability of an action, and the theory explicitly and convincingly predicted that no feature of this environmental uncertainty would be presumed to propagate into the behavior of the choosers. The only time that utility theory predicted indeterminacy in behavior was when

18 Nov 2004 10:44

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

30

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

P1: IKH

GLIMCHER

two or more mutually exclusive actions had precisely equal subjective desirabilities, and the importance of that particular situation seemed limited to the classical economists. In the first half of the twentieth century, the theory of games developed by John VonNeumann, Oskar Morgenstern, and John Nash directly challenged this determinate approach (Nash 1950a,b, 1951; VonNeumann & Morgenstern 1944). Game theory represented a deviation from the classical tradition specifically because it proposed that when a rational chooser faces an intelligent and self-interested opponent, rather than a passive economic environment, situations could easily arise in which the subjective desirabilities of two or more actions are driven toward precise equality. The theory went on to make surprising and fundamentally indeterminate predictions about how rational humans would behave under many conditions that could be well described by game theory. To understand this theoretical insight, consider two opponents repeatedly playing the childhood game of rock-paper-scissors in which the loser pays the winner $2 on a round won by playing paper and $1 on a round won by playing scissors or rock. If the behavior of one’s opponent is unpredictable, any response can win, in principle. Paper will beat a play of rock for $2, scissors will beat a play of paper for $1, and rock will beat a play of scissors, again for $1. Classical utility theory assumes that humans choose between actions by multiplicatively combining the subjective value and likelihood of each outcome and then selecting the action with the outcome that yields the highest expected utility. Assuming naively that one’s opponent is equally likely to play rock, or paper, or scissors, the greater value of winning with paper should lead all players to select paper deterministically on each round. What VonNeumann recognized was that this assumption about the behavior of one’s opponent simply could not be correct. A competitor who simply selected scissors could reliably defeat any player who actually behaved in accord with this strategy. Game theory, as developed by VonNeumann & Morgenstern (1944), addresses this limitation of classical utility theory by making the assumption that both players are aware that they face an intelligent opponent who can anticipate their actions and that both players will shape their behavior accordingly to minimize losses and maximize gains. To accomplish this, players must take into account the potential payoffs associated with each choice, as specified by classical utility theory, but they must also consider how the actions of their opponent will influence those payoffs. Consider again the situation in rock-paper-scissors. Winning with paper yields twice as much money as winning with rock or scissors, but deterministically playing paper leads to certain defeat. What VonNeumann & Morgenstern showed was that under these conditions we can predict that a rational player will titrate risk against gain and play paper two-thirds of the time, scissors one-sixth of the time, and rock one-sixth of the time. Critically, however, he must avoid making his twothirds, one-sixth, one-sixth selections in a determinate fashion; for example, in a repeated version of the game by playing paper, then scissors, then paper, then rock, then paper, and then paper. Were his opponent to divine the determinate nature

18 Nov 2004 10:44

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

INDETERMINATE BEHAVIOR

P1: IKH

31

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

of such a strategy (through observation, for example), then winning would again become trivial for that opponent. He would only have to play scissors, then rock, then scissors, then paper, then scissors, and then scissors to assure a consistent win. The only way to avoid this trap is for a player to incorporate apparent indeterminacy directly into his behavior. He must in essence flip a weighted coin on each round to select between rock and paper and scissors. VonNeumann & Morgenstern were well aware of the implications of this observation. It suggested that under some conditions the study of economic choice would have to become a probabilistic process. As they put it: Consider now a participant in a social exchange economy. His problem has, of course, many elements in common with a maximum problem. [A problem in which a single economic actor seeks to maximize his gain by classically deterministic processes.] But it also contains some, very essential, elements of an entirely different nature. He too tries to obtain an optimum result. But in order to achieve this, he must enter into relations of exchange with others. If two or more persons exchange goods with each other, then the results for each one will depend in general not merely upon his own actions but on those of the others as well. Thus each participant attempts to maximize a function (his above-mentioned “result”) of which he does not control all of the variables. This is certainly no maximization problem, but a peculiar and disconcerting mixture of several conflicting maximum problems. Every participant is guided by another principle and neither determines all of the variables which affect his interest. This kind of problem is nowhere dealt with in classical mathematics. . .. We hope that the reader will be convinced by the above that they face here and now a really conceptual—and not merely technical—difficulty. And it is this problem which the theory of “games of strategy” is mainly devised to meet. (VonNeumann & Morgenstern 1944) VonNeumann & Morgenstern’s critical insight was that under conditions of this type choosers might not be able to identify a single course of action that is deterministically optimal. Instead, they may be forced to select a course of action in as random a fashion as possible. It is this strategy of random selection, known now as a mixed strategy, that distinguishes VonNeumann & Morgenstern’s approach from more classical deterministic approaches to the study of behavior. In sum, VonNeumann & Morgenstern argued that human behavior, under some conditions, must appear indeterminate in order to be efficient. They made this point elegantly when they described, in game theoretic form, a conflict between Sherlock Holmes and his archenemy, Professor Moriarity: Sherlock Holmes desires to proceed from London to Dover and hence to the continent in order to escape from Professor Moriarity who pursues him. Having boarded the train he observes, as the train pulls out, the appearance of Professor Moriarity on the platform. Sherlock Holmes takes it for granted—and

18 Nov 2004 10:44

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

32

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

P1: IKH

GLIMCHER

in this he is assumed to be fully justified—that his adversary, who has seen him, might secure a special train and overtake him. Sherlock Holmes is faced with the alternative of going to Dover or of leaving the train at Canterbury, the only intermediate station. His adversary—whose intelligence is assumed to be fully adequate to visualize these possibilities—has the same choice. Both opponents must choose the place of their detrainment in ignorance of the other’s corresponding decision. If, as a result of these measures, they should find themselves, in fine, on the same platform, Sherlock Holmes may with certainty expect to be killed by Moriarity. If Holmes reaches Dover unharmed he can make good his escape. What are the good strategies, particularly for Sherlock Holmes? [Set the value] to Professor Moriarity [of] catching Sherlock Holmes [at], say 100. [Alternatively, consider what happens if] Sherlock Holmes successfully escaped to Dover, while Moriarity stopped at Canterbury. This is Moriarity’s defeat as far as the present action is concerned, and should be described by a big negative value of the matrix element [for Moriarity]—in the order of magnitude but smaller than the positive value mentioned above—say, −50. [Finally, consider what happens if] Sherlock Holmes escapes Moriarity at the intermediate station, but fails to reach the Continent. This is best viewed as a tie, and assigned the matrix element 0. [From a mathematical analysis of these values one can conclude that] the good strategies (e for Moriarity, n for Sherlock Holmes) [are]: e = {3/5, 2/5}, n = {2/5, 3/5} Thus Moriarity should go to Dover with a probability of 60% while Sherlock Holmes should stop at the intermediate station with a probability of 60%, the remaining 40% being left in each case for the other alternative.1 (VonNeumann & Morgenstern 1944, pp. 177–178) Of course, this theoretical formulation raises critical questions about the scientific nature of game theory. If game theory predicts that Holmes will get off the train at Canterbury with a 60% probability, any action Holmes takes is compatible with the theory. VonNeumann & Morgenstern recognized this but were adamant that this was still the only rational strategy for Holmes to adopt. Holmes must, they argued, be as indeterminate as possible in selecting a course of action. In essence, he must appear to have flipped a weighted coin (weighted 60% for Canterbury and 40% for Dover) in order to maximize his chance of survival. This was true, VonNeumann & Morgenstern suggested, irrespective of whether the theory of games preserved Popperian falsifiability. By the early 1950s, John Nash (1950a,b; 1951) had seen an interesting additional level of structure in game theoretic problems that required a mixed strategy, 1

Our result for e, n yields that Sherlock Holmes is as good as 48% dead when his train pulls out from Victoria Station.

18 Nov 2004 10:44

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDETERMINATE BEHAVIOR

P1: IKH

33

or apparently indeterminate, solution. Building on the work of VonNeumann & Morgenstern, he concluded that stable mixed strategies must in principle reflect an equilibrium point at which the subjective desirabilities of the two or more actions being mixed were precisely equivalent. He argued that it was only this equivalence that could produce the indeterminate behavior that VonNeumann & Morgenstern had predicted. Consider again the situation when a single player must, on repeated rounds, select rock or paper or scissors. If any one of these is truly preferable as a choice, then we can assume that the chooser will always select that option. Mixed strategies should thus emerge, Nash reasoned, only when the two or more actions that are being mixed have identical average desirabilities. Working from VonNeumann & Morgenstern’s insights, Nash argued that these two equivalent desirabilities emerge when the competitive interactions of the two players drive them toward an equilibrium at which the two or more actions being mixed are of equal desirability. What Nash argued was that mixed-strategy equilibriums emerge from dynamic interactions between the players, which yield equal average desirabilities, and thus totally indeterminate patterns of behavioral choice. Classical utility theory had presumed that situations in which two or more actions have precisely equal subjective desirabilities would be encountered only rarely. Nash’s insight was that not only are they encountered, but the dynamic interactions that occur during strategic games actively create these situations of equal subjective desirability. From the point of view of indeterminacy, the critical insight that VonNeumann, Morgenstern, and Nash offered was that indeterminacy is a requisite feature of efficient behavior in a competitive world. That insight means either that humans and animals appear indeterminate to each other under some conditions or they behave inefficiently. In 1982, the evolutionary biologist John Maynard Smith also engaged indeterminacy during strategic games, but from an evolutionary perspective. He argued that any species involved in an internal competition for resources could be described using game theory and that this mathematical formalism predicted that organisms capable of producing apparently indeterminate behavior would be favored by natural selection. Imagine, Maynard Smith proposed, a species of animals in which individuals compete for access to territories that increase the number of young an individual can produce. Individuals without territories produce a small number of young, while individuals with territories produce a large number of young. Obviously, under these conditions it is in the interest of individuals to obtain territories. Now consider a situation in which there are more individuals than territories. In this hypothetical species, territories change hands when an animal without a territory “displays” to an animal with a territory, essentially threatening that individual for control of the territory. In the hawk-dove game, as this competition has come to be known, after such a display each animal must make a decision: whether to escalate the conflict (to fight for the territory) or whether to retreat (give up the territory without a fight). If one of the animals elects to escalate, behaving as a hawk, and

18 Nov 2004 10:44

34

AR

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

P1: IKH

GLIMCHER

TABLE 1

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

AR231-PS56-02.tex

Payoffs for challenger in the hawk-dove game Challenger chooses hawk

Challenger chooses dove

Defender chooses hawk

50% chance of gaining territory 50% chance of injury

Nothing gained

Defender chooses dove

Value of territory gained

50% chance of gaining territory

one decides to retreat, behaving as a dove, then the hawk takes the territory. If both animals elect to be doves, then one of them at random takes the territory. Finally, if both animals elect to be hawks, then they fight, one sustains injuries that reduce the number of young that individual can produce, and the other gains the territory. Table 1 illustrates this simple game as a two-by-two matrix that specifies the costs and benefits of all possible actions to each player. What Maynard Smith realized at a mechanistic level was that each of these values could be expressed in terms of evolutionary fitness, the gain in reproductive success, that an individual achieves with each outcome. Gaining a territory confers an increase in fitness, whereas sustaining an injury confers a decrease. Thus, if the value of a territory is high and the magnitude of injury in a hawk versus hawk fight is low, then animals that behave as hawks are more fit than those that behave as doves. Under these conditions, Maynard Smith reasoned, the population will evolve toward a single pure strategy equilibrium: All animals in the population will always be hawks. Similarly, if the value of a territory is low and the magnitude of injury is high, then all animals that behave as doves will produce more offspring, be more fit, than animals that act as hawks. Under these conditions, the population should converge on a pure strategy equilibrium of dove. If, however, the value of a territory is high and the cost of an injury also is relatively high, then an interesting thing happens. The only reproductively stable strategy for the animal and its offspring is to behave sometimes as a hawk and sometimes as a dove. To be more specific, a single dominant and unbeatable strategy emerges in a population playing the hawk-dove game. The probability that on any given encounter an individual will choose to behave as a hawk must be equal to the value of a territory divided by the magnitude of the injury sustained in a hawk versus hawk conflict. Critically, on each encounter individuals have to behave in an unpredictable fashion, never allowing their opponent to know whether they will be a hawk or dove2. But across many such encounters the only stable and unbeatable solution for the population is for the probability of being a hawk to be equal to the value of a territory divided by the cost of injury. This theoretical analysis suggests that, at least from the point of view of other individuals in this same species, evolution would drive behavior toward 2

Maynard Smith showed mathematically that a population of unpredictable individuals would dominate a population in which separate individuals were committed at birth to playing hawk or dove. For details of that proof, see (Smith 1982).

18 Nov 2004 10:44

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDETERMINATE BEHAVIOR

P1: IKH

35

unpredictability. As in the game theoretic work of VonNeumann, Morgenstern, and Nash, the ability to generate apparently unpredictable behavior seems advantageous. One interesting feature of Maynard Smith’s argument, however, is the mechanism by which this uncertain behavior would be presumed to arise. We have strong reasons to believe that completely novel behaviors arise, at least in part, from genetic mutations. Random changes occur in the genomes of these animals and then selection operates to preserve useful variations in behavioral traits. Atomic-level fluctuations in DNA molecules, induced by quantum-level forces like cosmic radiation, produce unpredictable changes in the genetic make-up of a species. These random changes then influence behavior. We have every reason to believe that the mechanism by which apparently indeterminate behaviors would arise would itself be truly indeterminate. Game theory, whether directed toward the actions of an individual or the evolution of a species, predicts that under some conditions behavior must appear indeterminate in order for it to be efficient. What implications, if any, does this have for the determinate scientific method? Does this mean that social scientists have to abandon Popperian falsifiability? Probably not, for at least two important reasons. First, the theoretical observation that behavior should appear indeterminate does not mean that behavior does appear indeterminate. Physical constraints may make it impossible, or unlikely, for mutations to generate behavior that even appears indeterminate. If this is the case, then at an empirical level we may simply find that apparently indeterminate behavior does not occur. Second, even if apparently indeterminate behavior were to be observed, this would not require that the physical generative process for the behavior itself be indeterminate. The psychological system responsible for decision making during strategic games might operate on totally deterministic grounds. Like a modern digital computer, it may simply generate an appearance of indeterminacy sufficient to defy prediction by the opponent. In summary, there may be reasons why fundamental indeterminacies like those that arise at the quantum level cannot influence the systems that generate behavior. Either of these two observations would rescue Popperian falsifiability in its strongest form.

Empirical Measurements of Behavioral Indeterminacy The theory of games makes it clear that an organism with the ability to produce apparently indeterminate patterns of behavior would have a selective advantage over an animal that lacked this ability. Were apparently indeterminate behavior to have arisen in the evolutionary history of vertebrates, there seems every reason to believe that this behavioral phenotype would be preserved. Do humans have this ability? A common answer to this question, based on studies of humans, is no. Over the course of the last 40 years, a number of psychological studies have suggested that, perhaps because of some fundamental constraint in the human nervous system, humans cannot generate behavior that appears indeterminate (for a review of this literature see Wagenaar 1972). For example, in one of the first of

18 Nov 2004 10:44

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

36

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

P1: IKH

GLIMCHER

these studies, Bakan (1960) asked humans to simulate the action of a random coin flip: subjects were asked to make up a sequence of heads and tails that was fully random in order. When Bakan analyzed the sequences generated by these subjects they were found to be highly nonrandom despite the instructions that the subjects received. Bakan found that the subjects tended to overproduce alternations between heads and tails and to underproduce the occasional long runs of heads or tails that would be predicted from a truly random process. In sum, the humans behaved in a fairly determinate fashion, despite their instructions to do otherwise. Since 1960, dozens of studies have replicated this basic result. When human subjects receive a verbal instruction to produce a random sequence, they reliably fail. On the basis of this evidence, many psychologists have concluded that humans lack the ability, in principle, to generate patterns of behavior that appear indeterminate. According to game theory, however, environmental conditions should arise in which apparently indeterminate behavior would be truly beneficial. Organisms in their natural environment would be reinforced for producing apparently indeterminate behavioral patterns under some conditions. Regardless of these human data, then, can nonhuman animals produce apparently indeterminate behaviors if they are reinforced for doing so? Blough (1966) was one of the first to ask this question directly by specifically reinforcing pigeons for producing behavior that approximated a random process. In that experiment, pigeons were trained to peck a key in a Skinner box, and the amount of grain that they received after each peck was contingent upon the length of time that had intervened since the last peck. The more closely the set of interresponse intervals produced by the pigeon approximated the output of a random Poisson-like process, the more grain the bird earned. Blough found that under these conditions the birds quickly adopted a response strategy that was virtually indistinguishable from the output of a truly random operator. Figure 1 shows the frequency distribution of interresponse intervals Blough obtained from a single pigeon and, plotted as a solid line, the pattern of intervals that would be expected from a fully random process. While Blough’s analysis did not show that the behavior of the pigeons was random by all possible measures, it did demonstrate that when an apparently indeterminate behavior was reinforced, pigeons could produce a behavior of this general type. This study was critical because it provided the first evidence that the ability to produce apparently indeterminate behavior had arisen in the vertebrate line. Since that original study, a voluminous literature has examined the ability of several species of animals to generate apparently indeterminate behavioral sequences when they are specifically reinforced for doing so, and tasks more closely approximating the conditions described by game theorists have also been examined (see Neuringer 2002 for a review of this literature). Shimp (1967) introduced one paradigm that has been particularly widely studied. In that paradigm, pigeons were trained to choose sequentially between left and right response keys for four responses during each of thousands of trials. The behavior of the pigeons on each trial thus produced one of 16 possible patterns, for example, left-right-left-left. The animals then were reinforced for producing the 16 possible patterns with an

18 Nov 2004 10:44

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDETERMINATE BEHAVIOR

P1: IKH

37

Figure 1 Mean interresponse times (IRTs) from two replications of three experimental conditions for a single pigeon (Blough 1966). The graph plots the frequency of each IRT in half-second bins. A fully indeterminate process would produce points that fall along straight lines in this space. (Reproduced with permission from Journal of the Experimental Analysis of Behavior, copyright 1966, Society for the Experimental Analysis of Behavior.)

apparently random frequency distribution. In one important and well-controlled study, Page & Neuringer (1985) employed a strategy of this type to examine indeterminacy in behavior and to see whether the ability of pigeons to produce random sequences depended specifically upon whether or not they were reinforced for apparent randomness. In that experiment, pigeons produced long sets of left and right responses, but under two reinforcement contingencies. Under the first contingency, the animals were specifically reinforced for producing patterns of left and right responses that had a random-like frequency distribution. Under the second contingency, the randomness of the emitted frequency distribution was irrelevant to the reward received. Page & Neuringer found that when reinforcement was contingent on variability, the variability of the pigeons’ responses increased, but when the level of variability was not reinforced directly, the pigeons adopted much more determinate response patterns. More specifically, they found that an information theoretic analysis of the response patterns showed nearly perfect indeterminacy when, but only when, indeterminacy was reinforced. These results suggest two interesting conclusions. First, they suggest that the degree of apparent indeterminacy included in behavior is variable. Animals can be more or less indeterminate based on the requirements of their environment. Second, they suggest that when

18 Nov 2004 10:44

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

38

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

P1: IKH

GLIMCHER

indeterminacy is irrelevant, this species of animal prefers to adopt a fairly determinate response strategy. Machado (1989) employed a similar approach in another important study. In that experiment, pigeons once again emitted four left or right responses in each of thousands of sequential trials, and the variability of the response pattern they produced was assessed statistically to determine the amount of reward that the pigeon would receive. This was accomplished simply by counting the number of trials since that sequence had last been produced and assigning this number, a variability score, to that trial. To determine whether a reinforcement was delivered, the variability scores for the last 20 trials were cumulated and the variability score for the current trial was compared to the variability scores for those last 20 trials. If the percentile rank of the variability score for the current trial exceeded some fixed threshold, for example 50%, then a reinforcement was delivered. If pigeons responded truly randomly, then the probability of emitting all 16 possible sequences within 25 trials would be less than 1%. Accordingly, Machado adjusted the threshold requirement so that it never reinforced patterns of sequences shorter than 25 trials in length. Machado found that under these conditions the frequency distribution of variability scores actually produced by the birds was nearly identical to the frequency distribution that would be expected from a truly random process. Figure 2 plots this relationship for one of Machado’s animals. In this figure, the points plot the frequency with which each possible variability score was observed

Figure 2 Dots are used to plot the distribution of variability scores obtained from a single animal in the Machado (1989) paper. The solid line indicates the frequency distribution that would be expected from a perfectly indeterminate process. (Reproduced with permission from Journal of the Experimental Analysis of Behavior, copyright 1989, Society for the Experimental Analysis of Behavior.)

18 Nov 2004 10:44

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDETERMINATE BEHAVIOR

P1: IKH

39

during a reinforcement contingency that maximized indeterminate behavior. Low variabilities were observed often and high variability scores occurred more rarely. The solid line plots, for comparison, the pattern of variability scores that would be expected from a fully random process measured in this way. Machado’s critical observation is that this particular reinforcement protocol yields a fourth order pattern of responses indistinguishable from the pattern that would be expected from a fully indeterminate process. In sum, many studies (Neuringer 2002) that have yielded data similar to Blough’s, Page & Neuringer’s, and Machado’s suggest that when nonhuman animals are reinforced for producing apparently indeterminate patterns of behavior, they can produce behavior of this type. This set of observations thus led Neuringer (1986) to test the hypothesis that previous studies with human subjects had failed to yield apparently indeterminate behavior because human subjects had not been specifically reinforced for producing apparently indeterminate behaviors. In this study, Neuringer instructed human subjects to produce a random sequence of ones and twos on a computer keyboard. He then analyzed the resulting sequences for nonuniform distributions of ones and twos, and first and second order patterns in the ones and twos. He also analyzed the sequences with a set of statistics related to autocorrelation functions. What he found under these conditions was that the human subjects produced highly nonrandom sequences exactly as had been observed in previous studies. Neuringer provided feedback to these subjects by showing them, after a run of 100 trials, how the distribution they had produced deviated from the distribution that would be expected from a random sequence according to one of the statistical measures of randomness that he employed. In sequence, Neuringer then presented the subjects with each of the additional statistical metrics until they were receiving feedback according to all five metrics at the end of each 100-trial run. Finally, the feedback was terminated and the subjects were told that if they could produce a sequence that could not be discriminated from the product of a computer pseudorandom number generator, they would receive a cash bonus. Under these conditions, Neuringer found that the human subjects essentially all produced sequences that could not be discriminated from random sequences by any of the metrics he employed. From these data, he concluded that like pigeons and other vertebrates, human subjects could produce apparently indeterminate sequences under some conditions. In a similar experiment, Rapoport & Budescu (1992) examined the behavior of humans playing two-person games of the type VonNeumann, Morgenstern, and Nash studied. In Rapoport & Budescu’s experiments, random-like behaviors were reinforced monetarily, and they found that humans could produce behavioral sequences that appeared indeterminate. Under conditions in which game theory predicts that indeterminacy will be reinforced, apparently indeterminate behavior can be produced. Of course, these data do not demonstrate that humans can produce fully indeterminate behavior. All of these data suggest that humans and animals can produce behavior that appears indeterminate, but it seems probable that, like a randomnumber generator in a computer, the generative process for this behavior is likely

18 Nov 2004 10:44

40

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

P1: IKH

GLIMCHER

determinate at a lower level of reduction. To test that hypothesis, however, one would have to turn to a neurophysiological level of analysis.

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Reducing Uncertainty: Looking for Determinacy with Neurophysiology Perhaps the most influential study of choice behavior at the level of interacting neurons has been the work of William Newsome and his colleagues at Stanford University (for a review of Newsome’s work, see Batista & Newsome 2000; for a review of neurobiological choice literature, see Glimcher 2003). Newsome and his colleagues trained rhesus monkeys to monitor a visual display that presents a circular patch of chaotically moving spots of light (Figure 3). Upon viewing a display of this type, human observers report a chaotic blizzard of randomly moving white spots. However, when 15% of the spots move coherently in a single direction, humans subjects report a strong sense that the spots are moving, overall, in that direction. If the fraction of spots moving coherently is reduced, the strength of this perceived motion is reduced. By systematically varying the fraction of spots moving coherently, Newsome and his colleagues could therefore systematically manipulate how difficult it was for observers to determine the average direction in which the spots were moving. In their original experiment (Newsome et al. 1989), monkeys were presented with a display of this type for 2 seconds, after which they had to decide in which of two possible directions the spots, on average, were moving. The animals indicated their decision with an eye movement that shifted the animal’s point-of-gaze in the direction of perceived average motion. If the animals had judged the direction of spot motion correctly, they received a fruit juice reward. While animals made these decisions, the activity of single motion-sensitive neurons in the middle temporal visual cortex (area MT) was monitored. Under these conditions, Newsome and his colleagues found that if 15% of the dots in the display moved to the right, the monkeys always reported that they saw rightward motion, and cells in area MT activated by rightward motion rapidly generated action potentials. As the percentage of rightward dots was systematically decreased, both the probability that the monkey would report that he had seen rightward motion and the probability that the neurons would show an increase in firing rate decreased at roughly the same rate. What can we learn from data of this type about the mechanisms that give rise to apparently indeterminate behaviors? In one interesting study, Britten and colleagues (1996) examined the activity of MT neurons while monkeys viewed a display in which either all of the spots moved in random directions (there was no coherent direction of spot motion) or only a small fraction of the dots moved in a coherent direction. These conditions were selected to examine the relationship between neural activity in area MT and the decisions that an animal made when the visual stimulus was ambiguous. Interestingly, on a subset of the trials in which there was no coherent motion of the spots, the exact same pattern of randomly

18 Nov 2004 10:44

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDETERMINATE BEHAVIOR

P1: IKH

41

Figure 3 The Moving Spot Task. Monkeys fixate a central point while chaotically moving spots of light are presented within a circular aperture. On any given trial, a small fraction of the spots moves in a coherent manner in one of two possible directions. Across trials, the fraction of dots moving in this coherent fashion can be varied systematically to increase or decrease the strength of the perceived motion signal in the two possible directions. After viewing the display for 2 seconds, monkeys indicate the direction of perceived motion with a saccadic eye movement. Correct responses are reinforced with water or fruit juice. (From Shadlen & Newsome 2001. Reproduced with permission from Journal of Neuroscience.)

moving spots was presented. Under these conditions, the animals viewed the exact same stimulus while the activity of MT neurons was monitored. Perhaps surprisingly, even under these conditions the activity of the neurons varied from trial to trial. The precise number of action potentials generated and the precise pattern of action potential generation differed in an apparently random manner from trial to trial, even when the visual stimulus that the animal was evaluating was identical.

18 Nov 2004 10:44

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

42

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

P1: IKH

GLIMCHER

Britten and colleagues also found that the perceptual judgments of the animals were unpredictable on these trials. Like the neurons, the behavior of the animals was variable. Finally, these authors found that the judgments of the animals were always correlated, although only weakly, with the activity of the neurons. In sum, the neurons appeared to be indeterminate with regard to the stimulus, and the decisions that the animals made were correlated with these apparently indeterminate neural events. These data led to the generation of a model (Figure 4) designed to simulate the brain circuits for making this perceptual decision about the direction of spot motion (Shadlen et al. 1996). The model proposed that a group of, for example, rightward motion–sensitive neurons in area MT pooled data according to a fully defined algorithm to yield an instantaneous estimate of the current strength of rightward motion in the moving spot display. In a similar way, a group of leftward motion–sensitive neurons was hypothesized to extract an estimate of the

Figure 4 Shadlen et al.’s (1996) model of a perceptual decision circuit. Pools of neurons in area MT extract the instantaneous strength of visual motion occurring in the display, for motion in all possible directions. The instantaneous pooled estimates of motion strength in each of the two possible directions are passed to elements that compute the time integral of that signal to derive an estimate of the average motion signal over a 2-second display interval. The process of pooling is presumed to involve the addition of a fundamental indeterminacy called the “pooling noise” in the original model and labeled here as the pooling noise generator. These integrative elements project, in turn, to eye movement–producing neurons. The integrative elements are postulated to be mutually inhibitory, assuring that only one eye movement is triggered at a time.

18 Nov 2004 10:44

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDETERMINATE BEHAVIOR

P1: IKH

43

instantaneous strength of leftward motion. Because the monkeys were allowed to view the motion stimulus for 2 seconds, this had allowed them, in principle, to sum 2 seconds of instantaneous motion information before making a choice. Accordingly, Shadlen and his colleagues proposed that the output of each of the neuronal pools of direction-sensitive neurons was summed, or integrated, over the 2-second period, to develop an estimate of the average direction of motion throughout the stimulus interval. They next proposed that the neurons that integrated rightward activity should be able to directly activate circuits that produced rightward eye movements and that leftward integrating neurons should be able to activate circuits for producing leftward eye movements. To make the system capable of decision making, in the sense of making choices, the model employed two inhibitory linkages that allowed the output of each integrator to inhibit the other integrator’s access to the eye movement control circuits. A quantitative analysis of the behavior of the model, however, revealed a surprising result. The behavior of the actual monkeys appeared much more random than would be predicted from the neurobiologically derived model. Moreover, interestingly, this apparent randomness could be accounted for only by assuming neural circuitry that specifically incorporated a degree of intrinsic randomness that they referred to as a neuronal pooling noise. Shadlen and colleagues were forced to incorporate into the model a fully random element in order to account for their results. The Shadlen model was intended to link the activity of neurons in area MT to behavior in as determinate a fashion as possible, but Shadlen and his colleagues concluded that this linkage could only be accomplished if it was presumed that the nervous system incorporated an indeterminate element. Of course, there was no specific claim about the mechanistic nature of this apparently indeterminate neural element. The pooling noise generator could be a determinate device that yields an apparently indeterminate signal, but it is interesting that even at this level of analysis an indeterminate process seemed to operate. Motivated in part by those findings, Dorris & Glimcher (2004) elected to examine the behavior and brains of monkeys employed in a game theoretic conflict that actually required an apparently indeterminate type of behavior, a mixed strategy game of the type VonNeumann and Nash had modeled. In the human version of their inspection game, two opponents face each other, an employer and an employee (Figure 5). In each round of the game, the employee must decide whether to go to work, in which case he earns a fixed wage, or to shirk, in hopes of earning his wage plus a bonus. The goal of the employee is simply to maximize his gains in terms of salary and bonus. The employer, on the other hand, must decide between trusting his employee to arrive for work or spending money to hire an inspector who can check and see whether the employee arrived for work that day. The goal of the employer is to spend as little money as possible on inspections while maximizing the employee’s incentive to work. In this game, both human and monkey contestants played the role of the employee against a standardized, and strategically sophisticated, computer employer. Each round began with the illumination of two lights, one for working and one for

44

AR AR231-PS56-02.tex AR231-PS56-02.sgm

Figure 5 General form of the payoff matrix for the inspection game for both the experimental subjects (employees) and their opponents (employers). The variables in the bottom left of each cell determine the employee’s payoffs and the variables in the top right of each cell determine the employer’s payoffs for each combination of player’s responses. V = value of hypothetical product to the employer, fixed at 4; W = wage paid by employer to employee, fixed at 2; C = cost of working to employee, fixed at 1; I = cost of inspection to the employer, varied from 0.1 to 0.9 in steps of 0.2. Middle and right panels show payoff matrices for 70% and 30% employee-shirk rates. The predicted equilibrium strategy for the employer remains constant at a 50% inspect for all blocks of trials. One unit of payoff is 0.25 mL of water for monkey or 5 cents for human.

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

18 Nov 2004 10:44 LaTeX2e(2002/01/18)

GLIMCHER

P1: IKH

18 Nov 2004 10:44

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDETERMINATE BEHAVIOR

P1: IKH

45

shirking. At the end of each round, players selected one light and the computer employer simultaneously decided whether to pay for an inspection on that round. These responses were then compared by a computer arbiter that paid off players according to a fixed payoff matrix (paying off in juice for monkeys, real currency for humans, and virtual currency for the computer employer). Dorris & Glimcher (2004) found that the overall probability that a human playing the inspection game for money would chose to shirk was reasonably well predicted by the Nash equilibrium computations, but more importantly, they found that human subjects behaved almost perfectly randomly from trial to trial. An analysis of the human data revealed when the Nash solution in the game was for a player to shirk 50% of the time, not only did the players shirk about 50% of the time, but they also showed essentially no patterns in their behavior out to a third-order statistical analysis. As in the experiments of Neuringer (1986) and Rapoport & Budescu (1992), subjects appeared capable of producing largely random patterns of behavior when they were reinforced for doing so. When Dorris & Glimcher analyzed the behavior of their monkeys, they found that the behavior of the monkeys was surprisingly similar, even essentially identical, to the behavior of their human employees. Just like humans, the monkeys seemed to precisely track the Nash equilibrium solutions and to produce those average solutions using largely random sequences of working and shirking. When Dorris & Glimcher (2004) examined the activity of neurons in the posterior parietal cortex while monkeys played the inspection game, they found that the posterior parietal cortex carried a signal essentially identical to one predicted by game theory. The neural activity was correlated with the theoretical quantity economists refer to as expected utility. Importantly, however, this neural encoding of an economic choice variable was not accomplished in a totally deterministic fashion. The cortical neurons responded with an average rate that was correlated with expected utility, but on a moment-by-moment basis, the neurons behaved unpredictably. At a formal level, the neurons behaved roughly like Poisson devices, producing action potentials with random interspike intervals much like the interpeck intervals Blough’s (1966) pigeons produced. So what do we know of the mechanism that generates choice behavior under these conditions? Shadlen’s computational model of the choice process (Shadlen et al. 1996) seems to suggest that at the level of the neural computation we still can see evidence of apparent indeterminacy, and other models loosely related to the original Shadlen model seem to make a similar point (Barraclough et al. 2004, Corrado et al. 2003, Glimcher & Dorris 2005, Lau & Glimcher 2003). The absolute variability of primate behavior seems to be adjustable, and neural models of the machinery that generates this behavior at a neuronal level seem to include apparently random elements. What, then, is the implementation of this random element? One hopeful possibility is that the cellular-level mechanisms that implement this randomness may be, in fact, fully determinate processes. It is at least possible that if we better understood the mechanisms by which cells (for example, cortical neurons) generate action potentials, it would still be possible

18 Nov 2004 10:44

46

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

P1: IKH

GLIMCHER

to reduce this apparent indeterminacy to a determinate process at a subcellular level.

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Indeterminacy at the Cellular and Subcellular Levels Among the first scientists to examine the pattern of cortical neuronal firing rates with regard to indeterminacy were Tolhurst et al. (1981) and Dean (1981), who were extending studies of neuronal variability pioneered by Barlow & Levick (1969; see also Heggelund & Albus 1978). In two landmark papers, Tolhurst et al. (1981) and Dean (1981) examined the firing patterns of neurons in the visual cortices of anesthetized cats viewing visual displays that presented moving bars of light. When a given visual stimulus was presented to the animals, cortical neurons always responded with a fixed average rate of firing. A vertically oriented bar, for example, always produced a fixed rate of average action potential generation. As the bar was rotated toward a horizontal orientation, for example, the cell responded with a different, but also consistent, average rate of action potential production. They found, however, that the exact pattern of firing that gave rise to this average rate seemed to be almost completely unpredictable. Indeed, as the average firing rate increased, the moment-by-moment variability of the spike rate also increased, almost exactly in proportion to this mean rate. Put more formally, Tolhurst et al. and Dean found that the average firing rate was proportional to the square of the variance across a broad range of rates (Figure 6). This was a statistical distribution that would occur if the process of generating an action potential could be described in the following way: Immediately after an action potential is generated there is a 0% chance of generating an action potential for some largely fixed interval. After that interval has elapsed, the probability of generating an action potential in any given instant becomes fixed at a low level until an action potential occurs, after which the probability of action potential generation is again zero and the process repeats. Of course, during the interval when the probability of action potential generation was fixed at this low level, the spike generation could be characterized as, in principle, fully random. The time at which a spike occurred could be described as a fully random process that had all of the hallmarks of a truly stochastic Poisson operator. What Tolhurst et al. (1981) and Dean (1981) found, therefore, was that at the level of action potential generation, cortical neurons could be described as essentially stochastic. This was a surprising result at the time, and it has been widely confirmed (Rieke et al. 1997, Shadlen & Newsome 1998). What then is the source of this apparent stochasticity, and would a more detailed biophysical analysis of the spike generation mechanism reveal an underlying deterministic process that would yield this apparent indeterminacy? To examine one possible answer to that question, Mainen & Sejnowski (1995) sought to determine whether the biophysical process that actually generates action potentials in response to changes in membrane voltage was determinate. They performed intracellular manipulations of single cortical neurons in cortical networks by employing a brain slice preparation, inserting a microelectrode inside a single

18 Nov 2004 10:44

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDETERMINATE BEHAVIOR

P1: IKH

47

Figure 6 Tolhurst et al.’s (1981) plot of variance as a function of the mean firing rates for a cat’s visual-cortical neurons. The different symbols represent different averages of stimulus conditions; the straight lines plot regressions. The graph indicates that the square of variance and mean rate are related by a relatively fixed constant of proportionality. (Reproduced with permission from Vision Research. Copyright Pergamon Press.)

neuron and recording the pattern of membrane voltage produced in the cell by the network in which it was embedded. While membrane voltage was monitored, they also recorded the precise times at which the cell generated action potentials. This allowed them to determine the relationship between membrane voltage and action potential generation under reasonably normal conditions. Next, they disconnected this cell from the rest of the network in which it was embedded and used the microelectrode to reinject exactly the same pattern of membrane voltages that had occurred originally. They found that under these conditions the cell fired action potentials at exactly the same time, with regard to the membrane voltage signal, as it did previously. They found that the spike-generating mechanism was fully

18 Nov 2004 10:44

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

48

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

P1: IKH

GLIMCHER

deterministic. A given pattern of membrane voltage gave rise to exactly the same pattern of action potentials no matter how many times it was injected into the cell. On the one hand, this was a reassuring result. At base, the pattern of action potential generation was found to be governed by a determinate device. However, on the other hand, it was puzzling. Spike rates are not determinate in this sense. Tolhurst et al. and Dean’s work indicates that spike rates are distributed in a Poisson-like fashion, and there clearly is nothing about the spike generator within each cell that produces this pattern. The Mainen & Sejnowski (1995) data indicate that the apparent randomness in spike patterns must be a function of apparent randomness in the underlying membrane voltages. What then are the sources of these Poisson-like fluctuations in membrane voltage? We know that membrane voltages are governed, ultimately, by the pattern of synaptic activations that a cell receives from the neurons that impinge upon it. Each cortical neuron receives about 10,000 synapses from the tissue that surrounds it. The fact that about half of these synapses are excitatory and half are inhibitory is also important. It means that net excitation and inhibition are largely balanced in an active neuron and small shifts in this balance cause the membrane voltage to rise and fall, and thus cause action potentials to be generated. Together, these observations make a clear suggestion. The source of the apparent stochasticity in the membrane voltage either is a determinate pattern of synaptic activations that carefully sculpts the membrane voltage to yield an apparently indeterminate pattern of action potentials for reasons we do not yet understand or the process of synaptic activation is itself apparently indeterminate. A number of groups have investigated this latter possibility by studying the activity of single synapses (see Auger & Marty 2000, Stevens 2003 for reviews of this literature). The basic approach taken by these groups has been to activate a neuron and then monitor the rate at which individual synaptic vesicles are released into the synaptic cleft. Before these experiments were undertaken one could have speculated that synapses were simple determinate mechanisms: When an action potential invades the presynaptic region, it might be presumed that synaptic vesicles of neurotransmitter were deterministically released into the synaptic cleft. Modern studies of this process seem to contradict this view, however. Current evidence indicates that when an action potential invades the presynaptic terminal, the chance that a single synaptic vesicle will be released can be as low as 20%. Examinations of the precise patterns of vesicular release suggest that the likelihood that a vesicle of neurotransmitter will be released in response to a single action potential can be described as a random Poisson-like process. Vesicular release seems to be an apparently indeterminate process. Careful study of other elements in the synapse seems to yield a set of similar, and highly stochastic, results. Postsynaptic membranes, for example, seem to possess only a tiny number of neurotransmitter receptors (cf. Takumi et al. 1999), and during synaptic transmission as few as one or two of a given type of receptor molecules may be activated (Nimchinski et al. 2004). Under these conditions, a single open ion channel may allow a countable number of calcium or sodium ions

18 Nov 2004 10:44

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDETERMINATE BEHAVIOR

P1: IKH

49

to enter the neuron, and there is evidence that the actions of a single receptor and the few ions that it channels into the cell may influence the postsynaptic membrane. Together, all of these data suggest that membrane voltage is the product of interactions at the atomic level, many of which are governed by quantum physics and thus are truly indeterminate events. Because of the tiny scale at which these processes operate, interactions between action potentials and transmitter release as well as interactions between transmitter molecules and postsynaptic receptors may be, and indeed seem likely to be, fundamentally indeterminate. In 1944, Schrodinger argued that the fundamental indeterminacy of the physical universe would have no effect on living systems. He argued that were biological systems to become so small that the actions of single atoms or molecules could influence cells, the resulting organisms would surely perish from the evolutionary landscape. Studies of the mammalian synapse, however, seem to indicate that Schrodinger (1944) was simply wrong in this regard. Single synapses appear to be indeterminate devices; not apparently indeterminate, but fundamentally indeterminate. At base, physical indeterminacy seems to be a fundamental property of the brain. But how sure can we be that this fundamental indeterminacy at the level of the synapse has anything to do with indeterminacy at the level of a single cortical neuron, at the level of a cortical network, at the level of behavior, or at the level of a social theory of behavior? The evidence that we have today suggests that membrane voltage can be influenced by quantum level events, like the random movement of individual calcium ions. So there is every reason to believe that membrane voltage can be viewed, at least under some circumstances, as a formally indeterminate process of the type that precludes Popperian falsifiability. How does this membrane voltage influence action potential generation? Recall that cells receive a mixture of excitation and inhibition from thousands of synapses and that the ratio of this mixture is variable. Imagine that the correlations between the activity of the individual synapses impinging on a given cell were variable. Under conditions in which the activity of many synapses is correlated and the membrane voltage is driven either way above or way below its threshold for action potential generation, the network of neurons itself would maintain a largely determinate characteristic even though the synapses themselves might appear stochastic. Alternatively, when the synaptic activity is uncorrelated and the forces of excitation and inhibition are balanced, small uncorrelated fluctuations in synaptic probabilities drive cells above or below threshold. Under these conditions, indeterminacy in the synapses propagates to the membrane voltage and thence to the pattern of action potential generation. Indeterminacy in the pattern of action potential generation, although variable, would reflect a fundamental indeterminacy in the nervous system. At the level of behavior, apparent indeterminacy is reinforced by the environment and has been observed. Animals can produce behavior that appears to scientists to be indeterminate. How does this apparent indeterminacy arise? Given what we know about the behavior of synapses and action potentials, two possibilities present themselves. The fundamental indeterminacy observed at the cellular

18 Nov 2004 10:44

50

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

P1: IKH

GLIMCHER

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

level could be prevented from influencing higher-level phenomena in the nervous system, rendering these higher-level phenomena determinate. These determinate processes could then instantiate pseudorandom computations that emulate the underlying cellular indeterminacy and yield apparently indeterminate behavior. Alternatively, we can propose the hypothesis that indeterminacy observed at the cellular level could propagate to behavior under some circumstances, yielding truly indeterminate behavior under some conditions and more determinate behaviors under others.

THE CHALLENGE OF INDETERMINACY FOR BEHAVIORAL SCIENCE The traditional scientific method, or at least one interpretation of that method, suggests that the goal of an investigator should be to reduce uncertainty. We make predictions from our data about future states of the world,
i, which have some error, ε. One goal of science is to reduce ε to the smallest possible value and then to use the
i measured under these conditions to falsify incorrect theories. Formally, situations in which behavior appears highly indeterminate are those in which ε is large with regard to
i. The argument that aspects of the world are, however, truly indeterminate necessitates a change in the way measurements of
i are approached. In a fundamentally indeterminate world, ε would have a fixed minimum value beyond which the reduction of uncertainty would be impossible. If this is the case, and ε does have a fixed minimal value, then two critical problems arise for the scientific method. First, a measurement problem arises. If variability is observed during a scientific measurement, does that represent an error on the part of the scientist or variability in the world? Without a reliable technique for specifying the minimum value of ε under a given set of circumstances, there is no way to know if a measurement is accurate. This promotes anarchy in the method by permitting a confusion between error and observation. Second, a falsification problem arises. The existence of a lower limit on ε precludes hard falsification of the type Popper advocated. If a given set of scientific predictions must be couched in probabilistic terms, then—for all the reasons Popper outlined—rigorous falsification is impossible. Good examples of these measurement and falsification problems arise in the contemporary debate about what information is carried in the Poisson-like patterns of action potentials produced by cortical neurons. Cortical neurons produce variable patterns of interspike intervals. All efforts to reduce that variability to a determinate pattern have essentially failed. Some scientists conclude from this failure that spike trains are, at root, indeterminate and that the only information carried by these patterns of action potentials is encoded by the mean rate at which they occur (Shadlen & Newsome 1998). Others propose theories that would yield Poisson-like patterns of interspike intervals, but from underlying determinate processes. Is the first of these hypotheses testable, falsifiable, and scientific? The

18 Nov 2004 10:44

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDETERMINATE BEHAVIOR

P1: IKH

51

answer to that question seems far from clear. What is clear is that two kinds of indeterminacy in principle could occur: a fundamental indeterminacy for which ε cannot be reduced and an apparent indeterminacy for which ε can be reduced. Fundamental indeterminacy challenges the scientific method. Apparent indeterminacy only serves to challenge scientists. In which category does the Poisson-like variability of cortical neurons belong? Popper (1934) argued that science proceeds by falsification. A hypothesis is never proven; it is just discarded when it becomes clearly false. For this reason, Popper was deeply troubled by scientific theories that were fundamentally probabilistic in nature. If a theory proposes that a given neuron will fire an action potential with a probability of 0.2 in the next millisecond, any observation made during the next millisecond is commensurate with the theory. Of course, the longer the neuron is observed, the more robustly the frequency of action potential generation can be described in the past, but the ability of the theory to predict the future remains untestable and perhaps even unscientific, Popper argued. In the end, the theory is untestable because it predicts that given an infinitely long period of observation all possible patterns of action potentials will occur and thus no given pattern can be used to formally falsify the hypothesis. In the behavioral sciences, however, even determinate theories rarely proceed through a process of unambiguous falsification. Nearly all of the measurements made by behavioral scientists are clouded by variability. Variability results from measurement error, from uncontrolled factors that influence the outcome of the experiment, and perhaps even from variability intrinsic to the system under study. As a result, behavioral hypotheses typically are falsified not with unambiguous observations but with statistical generalizations. Further, falsification tends to be iterative. Instead of demonstrating that a single observation is incompatible with a given hypothesis, behavioral scientists gather a distribution of observations and use this distribution to assess the accuracy of the theory. An existing theory is replaced when a new theory can account for a portion of the residual variance unexplained by the old theory. Behavioral scientists accept that measurements are clouded by variance, ε; they work to minimize the magnitude of ε and they make statistical arguments that accommodate ε. However, at a fundamental level the goal of the scientific method remains a reduction in ε. Bacon argued that science must reduce uncertainty, and for working scientists, this usually means reducing ε. For this reason it is difficult to use the formal logical approach embodied by the Popperian scientific method to argue that variance itself, intrinsic indeterminacy, is a fundamental property of a behavioral system. Accepting the level of variance associated with our best theories as the lowest possible variance necessarily forces an abandonment of further inquiry. If some arbitrarily observed variation in a set of measurements is presumed, a priori, to reflect an irreducible feature of the system under study, then there is no reason to engage in further scientific examination. The search for new theories is, in essence, a technique for reducing ε. Over the last century, scholars seeking to understand behavior have struggled with this problem because they have again and again identified systems in which

18 Nov 2004 10:44

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

52

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

P1: IKH

GLIMCHER

variability, ε, seems irreducible. Neuringer (1986), for example, demonstrated that under some conditions human behavior is indistinguishable from a fundamentally indeterminate process. Tolhurst et al. (1981) and Dean (1981) made similar observations in their studies of cortical neurons. How can we ever hope to rigorously test hypotheses that include irreducible variation if the scientific method always seeks to reduce variance to zero? One answer would be to employ a strategy first used by the quantum physicists who encountered evidence of fundamental uncertainty in the physical world. Consider as they did a fundamental process, like their quantum events, that impose a known amount of uncertainty, ω, on a set of scientific measurements. Under such conditions scientists would still make measurements and those measurements would still include an uncertain component, ε total, but under those conditions ε total would be the sum of ω and the variances due to factors like measurement error, ε error. As theories were replaced iteratively by increasingly accurate theories, ε total would begin to approach ω. Under these conditions, knowledge of the value of ω would solve the measurement problem posed by the existence of fundamental indeterminacy. Knowledge of the minimum possible level of indeterminacy would allow one to discriminate between an error on the part of the scientist and variability in the world. One of the two problems posed by uncertainty would become tractable. The existence of a known nonzero ω, however, would do nothing to resolve the falsification problem. Under conditions in which ω has a nonzero value, scientific predictions must always be couched in probabilistic terms and thus rigorous falsification would remain impossible. The two critical issues that would arise were the behavioral world to be indeterminate would therefore be whether ω could be determined, and how one could proceed without rigorous falsification as a scientific goal. Unfortunately, behavioral scientists do not yet have a theory that would allow them to specify the magnitude of ω, and it seems unlikely that such a theory is imminent. This is probably a very important problem, and one with which behavioral scientists are beginning to grapple. The most promising strategy for defining ω today may be to develop converging evidence, from several levels of analysis, for a specific value of ω under a specific set of behavioral conditions. For example, if game theoretic, behavioral, and neurobiological studies all suggested a specific value for ω under some set of conditions, then one could be much more confident that the traditional scientific method could be pursued. The data presented here suggest that the rudiments of just such an approach may be evolving, although it is far too early to suggest that estimates of the behavioral uncertainty intrinsic to any given situation can be made accurately. Research like that of VonNeumann, Neuringer, and Shadlen points to the existence of indeterminate elements that participate in the generation of behavior, and each provides quantitative estimates of that indeterminacy. One goal of these approaches, in the long run, will have to be quantitative convergence around specific predictions for ω. The loss of rigorous falsification may be a more difficult philosophical problem, but may pose fewer difficulties to us as working scientists, especially since quantum

18 Nov 2004 10:44

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDETERMINATE BEHAVIOR

P1: IKH

53

physicists have already begun to engage that problem. In practice, scientists rarely proceed through a process of unambiguous falsification. Instead, we often test theories against each other. We ask which theory provides more explanatory power, which yields a smaller ε, and we then discard the less-efficient theory. Theories are used to falsify each other in an iterative process, and there is nothing about this sequence of events that requires determinacy in the real world. This process of iterative falsification does work, although it is less elegant than the strong falsification Popper advocated. It will probably have to form the philosophical basis on which the study of indeterminate behavior is based, and it likely will form an entirely adequate base. Indeterminacy becomes particularly problematic, however, when we try to ask whether the best currently available theory is a good theory. We traditionally consider a theory good when the predictions it makes are accurate. However, under conditions in which ω is large, good theories—even perfect theories—may not be accurate. An interesting example of this may be contemporary models developed to explain the choice behavior of humans during economic games (Camerer 2003, Erev & Roth 1998). These models seek to explain the play-by-play actions of individuals engaged in repeated rounds of games like rock-paper-scissors. The models seek to explain how human players learn from their experiences, and then use what they have learned to generate actions. Under these game theoretic conditions, however, there may be good reasons to believe that behavior is at least partially indeterminate. If behavior is truly uncertain on a choice-by-choice level, then how much of the behavior should a good theory explain? We can only assess the overall quality of theories like these if we can discover the fraction of the behavior that they seek to model, ω, which is truly indeterminate. The critical point that all of these observations make is that if human behavior is at root indeterminate, we do not need to abandon the scientific method as it is practiced today. The existence of indeterminacy does raise measurement and falsification problems. The measurement problem can be addressed by efforts to bound ω. The falsification problem has to be addressed in a different way. If the behavioral world is indeterminate, we will have to abandon rigorous falsification. That would be a shame, but it is important to remember that as behavioral scientists, we typically rely on an iterative process of theory-by-theory falsification, and there is no compelling reason to believe that this iterative method is challenged by the existence of fundamental indeterminacy in behavior. These considerations suggest that behavioral indeterminacy may be a good deal less threatening to scientists and the scientific method than Popper may have feared originally. At the same time, the empirical observations presented in this review hint that behavioral indeterminacy may be much more likely to occur than Schrodinger imagined. He argued that fundamental indeterminacy would never arise in the living world because If it were not so, if we were organisms so sensitive that a single atom, or even a few atoms, could make a perceptible impression on our senses—Heavens, what would life be like! To stress one point: an organism of that kind would

18 Nov 2004 10:44

54

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

P1: IKH

GLIMCHER

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

most certainly not be capable of developing the kind of orderly thought which, after passing through a long sequence of earlier stages, ultimately results in forming, among many other ideas, the idea of an atom. (Schrodinger 1944) Our existing data, although ambiguous, clearly challenge Schrodinger’s conclusion. The vertebrate nervous system is sensitive to the actions of single quantum particles. At the lowest levels of perceptual threshold, the quantum dynamics of photons, more than anything else, governs whether or not a human observer sees a light (Rieke & Baylor 1998). Synapses and neurotransmission also seem to violate this assumption of Schrodinger’s, and these are the building blocks from which neurocomputation is achieved. In the end, Schrodinger may be right, behavior may be fundamentally determinate, but it would be premature to draw that conclusion now. Behavioral scientists will have to continue to explore apparent indeterminacy in behavior and will have to develop the methodological tools for determining whether this apparent indeterminacy is fundamental. ACKNOWLEDGMENTS The author expresses his profound gratitude to B. Lau and H. Bayer for hours spent in discussion. This work was supported by the James S. McDonnell Foundation and the National Eye Institute. The Annual Review of Psychology is online at http://psych.annualreviews.org

LITERATURE CITED Arnauld A, Nicole P. 1662/1996. Logic or the Art of Thinking. Transl. JV Buroker, ed. London: Cambridge Univ. Press Auger C, Marty A. 2000. Quantal currents at single-site central synapses. J. Physiol. (London) 526:3–11 Bacon F. 1620/1994. Novum Organum. Transl. P Urbach, J Gibson. Chicago: Open Court Bakan P. 1960. Response-tendencies in attempts to generate binary series. Am. J. Psychiatry 73:127–31 Barlow HB, Levick WR. 1969. Three factors limiting the reliable detection of light by retinal ganglion cells of the cat. J. Physiol. (London) 200:1–24 Barraclough DJ, Conroy ML, Lee D. 2004. Prefrontal cortex and decision making in a mixed-strategy game. Nat. Neurosci. 7:404– 10 Batista AP, Newsome WT. 2000. Visuo-motor

control: giving the brain a hand. Curr. Biol. 10:R145–48 Bernoulli D. 1738/1954. Exposition of a new theory on the measurement of risk. Transl. R Sommer. Econometrica 22:23–36 Blough DS. 1966. The reinforcement of least frequent interresponse times. J. Exp. Anal. Behav. 9:581–91 Britten KH, Newsome WT, Shadlen MN, Celebrini S, Movshon JA. 1996. A relationship between behavioral choice and the visual responses of neurons in macaque MT. Vis. Neurosci. 13:87–100 Camerer C. 2003. Behavioral Game Theory. Princeton, NJ: Princeton Univ. Press Corrado GS, Sugrue LP, Newsome WT. 2003. The efficacy of linear integration for reward gathering in a nonstationary foraging task. Soc. Neurosci. Abstr. 385.2 Dean AF. 1981. The variability of discharge

18 Nov 2004 10:44

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDETERMINATE BEHAVIOR of simple cells in the cat striate cortex. Exp. Brain Res. 44:437–40 Descartes R. 1637/2000. Discourse on Method. Transl. DM Clarke. Harmondsworth, UK: Penguin Dorris MC, Glimcher PW. 2004. Lateral intraparietal area activity during manipulations of the expected value and utility of choices. Soc. Neurosci. Abstr. 767.1 Erev I, Roth AE. 1998. Predicting how people play games: reinforcement learning in experimental games with unique, mixed strategy equilibria. Am. Econ. Rev. 88:848–81 Galilei G. 1630/2001. Dialogo Soprai due Massimi Sistemi del Mondo, Tolemaico e Copernicano (Dialogue Concerning the Two Chief World Systems, Ptolemaic and Copernican). Transl. S Drake. New York: Modern Library Glimcher PW. 2003. The neurobiology of visual-saccadic decision making. Annu. Rev. Neurosci. 25:133–79 Glimcher PW, Dorris MC. 2005. Neuronal studies of decision making in the visua-saccadic system. In The Cognitive Neurosciences, ed. MA Gazzaniga. Cambridge, MA: MIT Press. In press Hastie R. 2001. Problems for judgment and decision making. Annu. Rev. Psychol. 52:653– 83 Heggelund P, Albus K. 1978. Response variability and orientation discrimination of single cells in striate cortex of cat. Exp. Brain Res. 32:197–211 Heisenberg W. 1930. Physical Principles of Quantum Theory. New York: Dover Heisenberg W. 1952/1979. Philosophic Problems of Quantum Physics. Woodbridge, CT: Ox Bow Kepler J. 1618–21/1995. Epitome of Copernican Astronomy. Transl. CG Wallis. New York: Prometheus Keynes JM. 1936/1997. The General Theory of Employment, Interest, and Money. New York: Prometheus Lau B, Glimcher PW. 2003. How monkeys learn from their decisons: behavior and models for physiology. Soc. Neurosci. Abstr. 518.16 Machado A. 1989. Operant conditioning of be-

P1: IKH

55

havioral variability using a percentile reinforcement schedule. J. Exp. Anal. Behav. 52: 155–66 Mainen ZF, Sejnowski TJ. 1995. Reliability of spike timing in neocortical neurons. Science 268:1503–6 Maynard Smith J. 1982. Evolution and the Theory of Games. London: Cambridge Univ. Press Nash JF. 1950a. Equilibrium points in n-person games. Proc. Natl. Acad. Sci. USA 36:48–49 Nash JF. 1950b. The bargaining problem. Econometrica 18:155–62 Nash JF. 1951. Non-cooperative games. Ann. Math. 54:286–95 Neuringer A. 1986. Can people behave “randomly?” The role of feedback. J. Exp. Psychol.: Gen. 115:62–75 Neuringer A. 2002. Operant variability: evidence, functions, theory. Psychon. Bull. Rev. 9:672–705 Newsome WT, Britten KH, Movshon JA. 1989. Neuronal correlates of a perceptual decision. Nature 341:52–54 Nimchinski EA, Yasuda R, Oertner TG, Svoboda K. 2004. The number of glutamate receptors opened by synaptic stimulation in single sippocampal spines. J. Neurosci. 24: 2054–64 Page S, Neuringer A. 1985. Variability is an operant. J. Exp. Psychol.: Anim. Behav. Process. 11:429–52 Pascal B. 1670/1966. Pensees. Transl. AJ Krailsheimer. Harmondsworth, UK: Penguin Pavlov IP. 1927. Conditioned Reflexes: An Investigation of the Physiological Activity of the Cerebral Cortex. London: Oxford Univ. Press Popper K. 1934/1992. The Logic of Scientific Discovery. Transl. K Popper. London: Routledge Rapoport A, Budescu DV. 1992. Generation of random series in two-person strictly competitive games. J. Exp. Psychol.: Gen. 121:352– 62 Rieke F, Baylor D. 1998. Origin of reproducibility in the responses of retinal rods to single photons. Biophys. J. 75:1836–57

18 Nov 2004 10:44

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

56

AR

AR231-PS56-02.tex

AR231-PS56-02.sgm

LaTeX2e(2002/01/18)

P1: IKH

GLIMCHER

Rieke F, Warland D, de Reuter van Steveninck R, Bialek W. 1997. Spikes. Cambridge, MA: MIT Press Schall JD. 2004. On building a bridge between brain and behavior. Annu. Rev. Psychol. 55:23–50 Schrodinger E. 1944/1967. What is Life? London: Cambridge Univ. Press Schrodinger E. 1951. Science and Humanism: Physics in Our Time. London: Cambridge Univ. Press Shadlen MN, Britten KH, Newsome WT, Movshon JA. 1996. A computational analysis of the relationship between neuronal and behavioral responses to visual motion. J. Neurosci. 16:1486–510 Shadlen MN, Newsome WT. 1998. The variable discharge of cortical neurons: implications for connectivity, computation, and information coding. J. Neurosci. 18:3870– 96 Shadlen MN, Newsome WT. 2001. Neural basis of a perceptual decision in the parietal cortex (area LIP) of the rhesus monkey. J. Neurophysiol. 86:1916–36 Shafir E, LeBoeuf RA. 2002. Rationality. Annu. Rev. Psychol. 53:491–517 Sherrington CS. 1906. The Integrative Ac-

tion of the Nervous System. New York: Scribner’s Shimp CP. 1967. Reinforcement of leastfrequent sequences of choices. J. Exp. Anal. Behav. 10:57–65 Skinner BF. 1938. The Behavior of Organisms. East Norwalk, CT: Appleton-Century-Crofts Smith A. 1776/1991. Wealth of Nations. New York: Prometheus Staddon JER, Cerutti DT. 2003. Operant conditioning. Annu. Rev. Psychol. 54:115–44 Stevens CF. 2003. Neurotransmitter release at central synapses. Neuron 40:381–88 Takumi Y, Ramirez-Leon V, Laake P, Rinvik E, Ottersen OP. 1999. Differential modes of expression for AMPA and NMDA receptors in hippocampal synapses. Nat. Neurosci. 2:618–24 Tolhurst DJ, Movshon JA, Dean AF. 1981. The statistical reliability of signals in single neurons in cat and monkey visual cortex. Vision. Res. 23:775–85 VonNeumann JV, Morgenstern O. 1944. Theory of Games and Economic Behavior. Princeton, NJ: Princeton Univ. Press Wagenaar WA. 1972. Generation of random sequences by human subjects: a critical survey of literature. Psychol. Bull. 77:65–72

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

Annual Review of Psychology Volume 56, 2005

CONTENTS Frontispiece—Richard F. Thompson

xviii

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

PREFATORY In Search of Memory Traces, Richard F. Thompson

1

DECISION MAKING Indeterminacy in Brain and Behavior, Paul W. Glimcher

25

BRAIN IMAGING/COGNITIVE NEUROSCIENCE Models of Brain Function in Neuroimaging, Karl J. Friston

57

MUSIC PERCEPTION Brain Organization for Music Processing, Isabelle Peretz and Robert J. Zatorre

89

SOMESTHETIC AND VESTIBULAR SENSES Vestibular, Proprioceptive, and Haptic Contributions to Spatial Orientation, James R. Lackner and Paul DiZio

115

CONCEPTS AND CATEGORIES Human Category Learning, F. Gregory Ashby and W. Todd Maddox

149

ANIMAL LEARNING AND BEHAVIOR: CLASSICAL Pavlovian Conditioning: A Functional Perspective, Michael Domjan

179

NEUROSCIENCE OF LEARNING The Neuroscience of Mammalian Associative Learning, Michael S. Fanselow and Andrew M. Poulos

207

HUMAN DEVELOPMENT: EMOTIONAL, SOCIAL, AND PERSONALITY Behavioral Inhibition: Linking Biology and Behavior Within a Developmental Framework, Nathan A. Fox, Heather A. Henderson, Peter J. Marshall, Kate E. Nichols, and Melissa A. Ghera

235

BIOLOGICAL AND GENETIC PROCESSES IN DEVELOPMENT Human Development: Biological and Genetic Processes, Irving I. Gottesman and Daniel R. Hanson

263 vii

P1: JRX

December 8, 2004

viii

12:13

Annual Reviews

AR231-FM

CONTENTS

SPECIAL TOPICS IN PSYCHOPATHOLOGY The Psychology and Neurobiology of Suicidal Behavior, Thomas E. Joiner Jr., Jessica S. Brown, and LaRicka R. Wingate

287

DISORDERS OF CHILDHOOD Autism in Infancy and Early Childhood, Fred Volkmar, Kasia Chawarska, and Ami Klin

315

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

CHILD/FAMILY THERAPY Youth Psychotherapy Outcome Research: A Review and Critique of the Evidence Base, John R. Weisz, Amanda Jensen Doss, and Kristin M. Hawley

337

ALTRUISM AND AGGRESSION Prosocial Behavior: Multilevel Perspectives, Louis A. Penner, John F. Dovidio, Jane A. Piliavin, and David A. Schroeder

365

INTERGROUP RELATIONS, STIGMA, STEREOTYPING, PREJUDICE, DISCRIMINATION The Social Psychology of Stigma, Brenda Major and Laurie T. O’Brien

393

PERSONALITY PROCESSES Personality Architecture: Within-Person Structures and Processes, Daniel Cervone

423

PERSONALITY DEVELOPMENT: STABILITY AND CHANGE Personality Development: Stability and Change, Avshalom Caspi, Brent W. Roberts, and Rebecca L. Shiner

453

WORK MOTIVATION Work Motivation Theory and Research at the Dawn of the Twenty-First Century, Gary P. Latham and Craig C. Pinder

485

GROUPS AND TEAMS Teams in Organizations: From Input-Process-Output Models to IMOI Models, Daniel R. Ilgen, John R. Hollenbeck, Michael Johnson, and Dustin Jundt

517

LEADERSHIP Presidential Leadership, George R. Goethals

545

PERSONNEL EVALUATION AND COMPENSATION Personnel Psychology: Performance Evaluation and Pay for Performance, Sara L. Rynes, Barry Gerhart, and Laura Parks

571

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

CONTENTS

ix

PSYCHOPHYSIOLOGICAL DISORDERS AND PSYCHOLOGICAL EFFECTS ON MEDICAL DISORDERS Psychological Approaches to Understanding and Treating Disease-Related Pain, Francis J. Keefe, Amy P. Abernethy, and Lisa C. Campbell

601

TIMELY TOPIC Psychological Evidence at the Dawn of the Law’s Scientific Age, David L. Faigman and John Monahan

631

Annu. Rev. Psychol. 2005.56:25-56. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDEXES Subject Index Cumulative Index of Contributing Authors, Volumes 46–56 Cumulative Index of Chapter Titles, Volumes 46–56

ERRATA An online log of corrections to Annual Review of Psychology chapters may be found at http://psych.annualreviews.org/errata.shtml

661 695 700

18 Nov 2004 12:34

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

P1: IKH LaTeX2e(2002/01/18) 10.1146/annurev.psych.56.091103.070311

Annu. Rev. Psychol. 2005. 56:57–87 doi: 10.1146/annurev.psych.56.091103.070311 c 2005 by Annual Reviews. All rights reserved Copyright
First published online as a Review in Advance on August 24, 2004

MODELS OF BRAIN FUNCTION IN NEUROIMAGING Karl J. Friston

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Wellcome Department of Cognitive Neurology, University College London, London, WC1N 3BG, United Kingdom; email: [email protected]

Key Words dynamic, inference, causal, Bayesian, fMRI, hemodynamics, connectivity ■ Abstract Inferences about brain function, using neuroimaging data, rest on models of how the data were caused. These models can be quite diverse, ranging from conceptual models of functional anatomy to nonlinear mathematical models of hemodynamics. However, they all have to be internally consistent because they model the same thing. This consistency encompasses many levels of description and places constraints on the statistical models, adopted for data analysis, and the experimental designs they embody. The aim of this review is to introduce the key models used in imaging neuroscience and how they relate to each other. We start with anatomical models of functional brain architectures, which motivate some of the fundaments of neuroimaging. We then turn to basic statistical models (e.g., the general linear model) used for making classical and Bayesian inferences about where neuronal responses are expressed. By incorporating biophysical constraints, these basic models can be finessed and, in a dynamic setting, rendered causal. This allows us to infer how interactions among brain regions are mediated.

CONTENTS INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ANATOMIC MODELS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Functional Specialization and Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Functional Specialization and Segregation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . STATISTICAL MODELS OF REGIONAL RESPONSES . . . . . . . . . . . . . . . . . . . . . Statistical Parametric Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The General Linear Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Experimental Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Classical and Bayesian Inference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dynamic Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Biophysical Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . MODELS OF FUNCTIONAL INTEGRATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Functional and Effective Connectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dynamic Causal Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0066-4308/05/0203-0057$14.00

58 58 58 59 60 60 61 64 67 69 71 77 77 79 84

57

18 Nov 2004 12:34

58

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

P1: IKH

FRISTON

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INTRODUCTION Understanding the brain depends on conceptual, anatomical, statistical, and causal models that link ideas about how it works to observations and experimental data. The aim of this review is to highlight the relationships among the sorts of models that are employed in imaging neuroscience. These relationships reflect an increasing mechanistic finesse as one moves from simple statistical models used to identify where evoked brain responses are expressed (cf. neo-phrenology) to models of how neuronal responses are caused (e.g., causal modeling). In the near future, models of representational inference and learning may be used as observation models to confirm some fundamental hypotheses about what the brain is doing (e.g., predictive coding). We review a series of exemplar models that cover conceptual models, motivating experimental design, to detailed biophysical models of coupled neuronal ensembles that enable questions to be asked at a physiological and computational level. Anatomical models of functional brain architectures motivate the fundaments of neuroimaging. We start by reviewing the distinction between functional specialization and integration and how these principles serve as the basis for most analyses of neuroimaging data. We then turn to simple statistical models (e.g., the general linear model) used for making classical and Bayesian inferences about functional specialization in terms of where neuronal responses are expressed. Characterizing a region-specific effect rests on estimation and inference. Inferences in neuroimaging may be about differences expressed when comparing one group of subjects to another or, within subjects, inferences may be about changes over a sequence of observations. They may pertain to structural differences (e.g., in voxel-based morphometry; Ashburner & Friston 2000) or neurophysiological indices of brain functions (e.g., functional magnetic resonance imaging, or fMRI). The principles of data analysis are very similar for all these applications. We focus on the analysis of fMRI time-series because this covers most of the issues encountered in other modalities. By incorporating biophysical constraints, simple observation models can be rendered biologically more realistic and, in a dynamic framework, causal. This allows us to infer how interactions among brain regions are mediated. This sort of characterization speaks to functional integration and relies on the notions of functional and effective connectivity.

ANATOMIC MODELS Functional Specialization and Integration The brain appears to adhere to two fundamental principles of functional organization, functional specialization and functional integration, where the integration within and among specialized areas is mediated by effective connectivity. The distinction relates to that between “localizationism” and “[dis]connectionism” that dominated thinking about cortical function in the nineteenth century. Since the

18 Nov 2004 12:34

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN FUNCTION MODELING

P1: IKH

59

early anatomic theories of Gall, the identification of a particular brain region with a specific function has become a central theme in neuroscience. However, functional localization per se was not easy to demonstrate: For example, a meeting that took place on August 4, 1881, addressed the difficulties of attributing function to a cortical area, given the dependence of cerebral activity on underlying connections (Phillips et al. 1984). This meeting was entitled “Localization of Function in the Cortex Cerebri.” Goltz (1881), although accepting the results of electrical stimulation in dog and monkey cortex, considered that the excitation method was inconclusive, in that movements elicited might have originated in related pathways, or current could have spread to distant centers. In short, the excitation method could not be used to infer functional localization because localizationism discounted interactions, or functional integration among different brain areas. It was proposed that lesion studies could supplement excitation experiments. Ironically, it was observations on patients with brain lesions some years later (see Absher & Benson 1993) that led to the concept of “disconnection syndromes” and the refutation of localizationism as a complete or sufficient explanation of cortical organization. Functional localization implies that a function can be localized in a cortical area, whereas specialization suggests that a cortical area is specialized for some aspects of perceptual or motor processing, and that this specialization is anatomically segregated within the cortex. The cortical infrastructure supporting a single function may then involve many specialized areas whose union is mediated by the functional integration among them. In this view, functional specialization is only meaningful in the context of functional integration and vice versa.

Functional Specialization and Segregation The functional role of any component (e.g., cortical area, subarea, or neuronal population) of the brain is defined largely by its connections (Passingham et al. 2002). Certain patterns of cortical projections are so common that they could amount to rules of cortical connectivity. “These rules revolve around one, apparently, overriding strategy that the cerebral cortex uses—that of functional segregation” (Zeki 1990). Functional segregation demands that cells with common functional properties be grouped together. This architectural constraint necessitates both convergence and divergence of cortical connections. Extrinsic connections among cortical regions are not continuous but occur in patches or clusters. This patchiness has, in some instances, a clear relationship to functional segregation. For example, V2 has a distinctive cytochrome oxidase architecture consisting of thick stripes, thin stripes, and interstripes. When recordings are made in V2, directionally selective (but not wavelength or color-selective) cells are found exclusively in the thick stripes. Retrograde (i.e., backward) labeling of cells in V5 is limited to these thick stripes. All the available physiological evidence suggests that V5 is a functionally homogeneous area that is specialized for visual motion. Evidence of this nature supports the notion that patchy connectivity is the anatomical infrastructure that mediates functional segregation and specialization. If it is the case that neurons

18 Nov 2004 12:34

60

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

P1: IKH

FRISTON

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

in a given cortical area share a common responsiveness, by virtue of their extrinsic connectivity, to some sensorimotor or cognitive attribute, then this functional segregation is also an anatomical one. In summary, functional specialization suggests that challenging a subject with the appropriate sensorimotor attribute or cognitive process should lead to activity changes in, and only in, the specialized areas. This is the anatomical and physiological model upon which the search for regionally specific effects is based. We deal first with models of regionally specific responses and later return to models of functional integration.

STATISTICAL MODELS OF REGIONAL RESPONSES Statistical Parametric Mapping Functional mapping studies are usually analyzed with some form of statistical parametric mapping. Statistical parametric mapping entails the construction of spatially extended statistical processes to test hypotheses about regionally specific effects (Friston et al. 1991). Statistical parametric maps (SPMs) are image processes with voxel values that, under the null hypothesis, are distributed according to a known probability density function, usually the student’s T or F distributions. These are known colloquially as T- or F-maps. The success of statistical parametric mapping is due largely to the simplicity of the idea. Namely, one analyzes every voxel using any standard (univariate) statistical test, usually one testing for activation or regression on some explanatory variable. The resulting statistical parameters are assembled into an image—the SPM. SPMs are interpreted as spatially extended statistical processes by referring to the probabilistic behavior of random fields (Adler 1981; Friston 1991; Worsley et al. 1992, 1996). Random fields model both the univariate probabilistic characteristics of an SPM and any nonstationary spatial covariance structure under the null hypothesis. “Unlikely” excursions of the SPM are interpreted as regionally specific effects, attributable to the sensorimotor or cognitive process that has been manipulated experimentally. Over the years, statistical parametric mapping (Friston et al. 1995b) has come to refer to the conjoint use of the general linear model (GLM) and Gaussian random field (GRF) theory to analyze and make classical inferences about spatially extended data through statistical parametric maps. The GLM is used to estimate some parameters that could explain the spatially continuous data in exactly the same way as in conventional analysis of discrete data. GRF theory is used to resolve the multiple-comparisons problem that ensues when making inferences over a volume of the brain. GRF theory provides a method for adjusting p values for the search volume of an SPM to control false positive rates. It plays the same role for continuous data (i.e., images or time-series) as the Bonferonni correction for a family of discontinuous or discrete statistical tests. Below we consider the Bayesian alternative to classical inference with SPMs. This rests on conditional inferences about an effect, given the data, as opposed to

18 Nov 2004 12:34

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

BRAIN FUNCTION MODELING

P1: IKH

61

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

classical inferences about the data, given the effect is zero. Bayesian inferences about spatially extended effects use posterior probability maps (PPMs). Although less established than SPMs, PPMs are potentially very useful, not least because they do not have to contend with the multiple-comparisons problem induced by classical inference (see Berry & Hochberg 1999). In contradistinction to SPM, this means that inferences about a given regional response do not depend on inferences about responses elsewhere. Before looking at the models underlying Bayesian inference, we first consider estimation and classical inference in the context of the GLM.

The General Linear Model Statistical analysis of imaging data corresponds to (a) modeling the data to partition observed neurophysiological responses into components of interest, confounds, and error, and (b) making inferences about interesting effects using the variances of the partitions. A brief review of the literature may give the impression that there are numerous ways to analyze positron emission tomography (PET) and fMRI time-series, with a diversity of statistical and conceptual approaches. This is not the case. With very few exceptions (see Nichols & Holmes 2002 for an overview), every analysis is a variant of the general linear model. These include (a) simple T-tests on scans assigned to one condition or another, (b) correlation coefficients between observed responses and boxcar stimulus functions in fMRI (Bandettini et al. 1993), (c) inferences made using multiple linear regression, (d) evoked responses estimated using linear time invariant models, and (e) selective averaging to estimate event-related responses. Mathematically, all are identical and can be implemented with the same equations and algorithms. The only thing that distinguishes among them is the design matrix encoding the experimental design. The general linear model is an equation, y = Xβ + ε,

(1)

expressing the observed response y in terms of a linear combination of explanatory variables in the matrix X plus a well-behaved error term (i.e., an independently and identically distributed Gaussian random variable). The general linear model is variously known as “analysis of [co]variance” or “multiple regression” and subsumes simpler variants, like the T-test for a difference in means, to more elaborate linear convolution models such as finite impulse response (FIR) models. The matrix X that contains the explanatory variables (e.g., designed effects or confounds) is called the design matrix. Each column of the design matrix corresponds to some effect one has built into the experiment or that may confound the results. These are referred to as explanatory variables, covariates, or regressors. The example in Figure 1 relates to an fMRI study of visual stimulation under four conditions. The effects on the response variable are modeled in terms of functions of the presence of these conditions (i.e., box or stick functions smoothed with components of a hemodynamic response function; in Figure 1, two components were used for each

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

18 Nov 2004 12:34

62

AR

FRISTON

AR231-PS56-03.tex AR231-PS56-03.sgm LaTeX2e(2002/01/18) P1: IKH

18 Nov 2004 12:34

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN FUNCTION MODELING

P1: IKH

63

of the four conditions, giving eight regressors in total). The relative contribution of each of these columns to the response is controlled by the parameters β. These are estimated using standard least squares, and inferences about the parameter estimates are made using T or F statistics, depending upon whether one is looking at a particular linear combination (e.g., a subtraction), or all of them together. The design matrix can contain covariates or indicator variables that take values of 0 or 1, to indicate the presence of a particular level of an experimental factor. Each column of X has an associated but unknown parameter. Some of these parameters will be of interest (e.g., the effect of a sensorimotor or cognitive condition or the regression coefficient of hemodynamic responses on reaction time). The remaining parameters will be of no interest and pertain to nuisance or confounding effects (e.g., the effect of being a particular subject or the regression slope of regional activity on global activity). Inferences about the parameter estimates are made using their estimated variance. This allows one to test the null hypothesis that some particular linear combination (e.g., a subtraction) of the estimates is zero using an SPM{T}. The T statistic obtains by dividing a contrast or compound (specified by contrast weights) of the parameter estimates by the standard error of that compound. Sometimes, several contrasts of parameter estimates are jointly interesting; for example, when using polynomial (B¨uchel et al. 1996) or basis function expansions of some experimental factor. In these instances, the SPM{F} is used and is specified with a matrix of contrast weights that can be thought of as a collection of “T contrasts” that one wants to test together. In most analyses, the design matrix contains indicator variables or parametric variables encoding the experimental manipulations. These are formally identical to classical analysis of [co]variance (i.e., AnCova) models. An important instance of the GLM, from the perspective of fMRI, is the linear time invariant model. Mathematically this is no different from any other GLM. However, it explicitly ←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− Figure 1 The general linear model. This model is an equation expressing the response variable Y in terms of a linear combination of explanatory variables in a design matrix X and an error term with assumed or known autocorrelation . In fMRI, the data can be filtered with a convolution or residual-forming matrix (or a combination) S, leading to a generalized linear model that includes (intrinsic) serial correlations and applied (extrinsic) filtering. Different choices of S correspond to different estimation schemes as indicated on the upper left. The parameter estimates obtain in a least squares sense using the pseudoinverse (denoted by +) of the filtered design matrix. Generally, an effect of interest is specified by a vector of contrast weights c that give a weighted sum or compound of parameter estimates referred to as a contrast. The T statistic is simply this contrast divided by the estimated standard error (i.e., the square root of its estimated variance). The ensuing T statistic is distributed with v degrees of freedom. The equations for estimating the variance of the contrast and the degrees of freedom are provided in the right-hand panel and accommodate the nonsphericity implied by .

18 Nov 2004 12:34

64

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

P1: IKH

FRISTON

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

treats the data sequence as an ordered time-series and enables a signal processing perspective that can be useful. The equations summarized in Figure 1 can be used to implement a vast range of statistical analyses. The issue is therefore not so much the mathematics but the formulation of a design matrix X appropriate to the study design and inferences that are sought. Before considering general linear models as biophysical or causal models of brain responses, we focus on the design matrix as a device to specify experimental design. Here the explanatory variables encode treatment effects that we assume are expressed in a linear and instantaneous fashion in the data, without reference to any particular mechanism.

Experimental Design This section considers the different sorts of designs employed in neuroimaging studies. Experimental designs can be classified as single factor or multifactorial designs; within this classification, the levels of each factor can be categorical or parametric. The tenet of cognitive subtraction is that the difference between two tasks can be formulated as a separable cognitive or sensorimotor component and that regionally specific differences in hemodynamic responses, evoked by the two tasks, identify the corresponding functionally specialized area. Early applications of subtraction range from the functional anatomy of word processing (Petersen et al. 1989) to functional specialization in extrastriate cortex (Lueck et al. 1989). The latter studies involved presenting visual stimuli with and without some sensory attribute (e.g., color and motion). The areas highlighted by subtraction were identified with homologous areas in monkeys that showed selective electrophysiological responses to equivalent visual stimuli. Cognitive conjunctions (Price & Friston 1997) can be thought of as an extension of the subtraction technique, in the sense that they combine a series of subtractions. In subtraction, one tests a single hypothesis pertaining to the activation in one task relative to another. In conjunction analyses, several hypotheses are tested to determine whether all the activations, in a series of task pairs, are expressed conjointly. Consider the problem of identifying regionally specific activations due to a particular cognitive component (e.g., object recognition). If one can identify a series of task pairs whose differences have only that component in common, then the region that activates, in all the corresponding subtractions, can be associated with the common component. In short, conjunction analyses allow one to disclose context-invariant regional responses.

CATEGORICAL DESIGNS, COGNITIVE SUBTRACTION, AND CONJUNCTIONS

The premise behind parametric designs is that regional physiology will vary systematically with the degree of cognitive or sensorimotor processing or deficits thereof. Examples of this approach include the PET

PARAMETRIC DESIGNS

18 Nov 2004 12:34

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN FUNCTION MODELING

P1: IKH

65

experiments of Grafton et al. (1992) that demonstrated significant correlations between hemodynamic responses and the performance of a visually guided motor tracking task. On the sensory side, Price et al. (1992) demonstrated a remarkable linear relationship between perfusion in periauditory regions and frequency of aural word presentation. This correlation was not observed in Wernicke’s area, where perfusion appeared to correlate, not with the discriminative attributes of the stimulus, but with the presence or absence of semantic content. These relationships or neurometric functions may be linear or nonlinear. Using polynomial regression, in the context of the GLM, one can identify nonlinear relationships between stimulus parameters (e.g., stimulus duration or presentation rate) and evoked responses. To do this one usually uses a SPM{F} (see B¨uchel et al. 1996). The example provided in Figure 2 illustrates both categorical and parametric aspects of design and analysis. These data were obtained from an fMRI study of visual motion processing using radially moving dots (Chawla et al. 1999). The stimuli were presented over a range of speeds using isoluminant and isochromatic stimuli. To identify areas involved in visual motion, a stationary dots condition was subtracted from a moving dots conditions (see the contrast weights, upper right). To ensure significant motion-sensitive responses, using color and luminance stimuli, a conjunction of the equivalent subtractions was assessed under both viewing contexts. Areas V5 and V3a are seen in the ensuing SPM{T}. The T values in this SPM are simply the minimum of the T values for each subtraction. Thresholding this SPM ensures that all voxels survive a threshold in each subtraction separately. This conjunction SPM has an equivalent interpretation; it represents the intersection of the excursion sets, defined by the threshold of each component SPM. This intersection is the essence of a conjunction. The responses in left V5 are shown in the lower panel of Figure 2 and speak to a compelling inverted “U” relationship between speed and evoked response that peaks at around 8◦ per second. It is this sort of relationship that parametric designs try to characterize. Interestingly, the form of these speed-dependent responses was similar using both stimulus types, although luminance cues elicit a greater response. From the point of view of a factorial design, there is a main effect of stimulus (isoluminant versus isochromatic), a main (nonlinear) effect of speed, but no speed by stimulus interaction. Factorial designs are more prevalent than single-factor designs because they enable inferences about interactions. At its simplest, an interaction represents a change in a change. Interactions are associated with factorial designs where two or more factors are combined in the same experiment. The interaction term assesses the effect of one factor on the effect of the other. Factorial designs have a wide range of applications. An early application, in neuroimaging, examined physiological adaptation and plasticity during motor performance by assessing time by condition interactions (Friston et al. 1992). Factorial designs have an important role in the context of cognitive subtraction and additive factors logic by virtue of being able to test for interactions, or context-sensitive activations

MULTIFACTORIAL DESIGNS

18 Nov 2004 12:34

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

66

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

P1: IKH

FRISTON

Figure 2 (Top right): An image representation of the design matrix. The vectors of contrast weights define the linear compounds of parameters tested. The contrast weights are displayed over the column of the design matrix that corresponds to the effects in question. The design matrix here includes condition-specific effects (boxcars convolved with a hemodynamic response function). Odd columns correspond to stimuli shown under isochromatic conditions and even columns model responses to isoluminant stimuli. The first two columns are for stationary stimuli and the remaining columns are for stimuli of increasing speed. The final column is a constant term. (Top left) A maximum intensity projection of the SPM{T} conforming to the standard anatomical space of Talairach & Tournoux (1988). The T values here are the minimum T values from both contrasts, thresholded at p = 0.001 uncorrected. The most significant conjunction is seen in left V5. (Lower panel) Plot of the condition-specific parameter estimates for this voxel. The T value was 9.25 (p < 0.001, corrected according to GRF theory). This example is based on data from Chawla et al. 1999.

18 Nov 2004 12:34

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN FUNCTION MODELING

P1: IKH

67

(i.e., to demonstrate the fallacy of “pure insertion”; see Friston et al. 1996c). These interaction effects can sometimes be interpreted as (a) the integration of the two or more (cognitive) processes or (b) the modulation of one (perceptual) process by another. In summary, the design matrix encodes the causes of observed data and, in particular, treatment effects caused by changes in the level of various experimental factors. These factors can have categorical or parametric levels and most experiments nowadays use multiple factors to test for both main effects and interactions. Before turning to mechanistically more informed formulations of the general linear model we consider briefly the two sorts of inferences that can be made about the parameter estimates.

Classical and Bayesian Inference To date, inference in neuroimaging has been restricted largely to classical inferences based upon statistical parametric maps. The statistics that comprise these SPMs are essentially functions of the data. The probability distribution of the chosen statistic, under the null hypothesis (i.e., the null distribution), is used to compute a p value. This p value is the probability of obtaining the statistic, or the data, given that the null hypothesis is true. If sufficiently small, the null hypothesis can be rejected and an inference is made. The alternative approach is to use Bayesian or conditional inference based upon the posterior distribution of the activation given the data (Holmes & Ford 1993). This necessitates the specification of priors (i.e., the probability distribution of the activation). Bayesian inference requires the posterior distribution and therefore rests upon a posterior density analysis. A useful way to summarize this posterior density is to compute the probability that the activation exceeds some threshold. This computation represents a Bayesian inference about the effect, in relation to the specified threshold. By computing a posterior probability for each voxel, we can construct posterior probability maps or PPMs that are a useful complement to classical SPMs (Friston et al. 2002, Friston & Penny 2003). The motivation for using conditional or Bayesian inference is that it has high face validity. This is because the inference is about an effect, or activation, being greater than some specified size that has some meaning in relation to underlying neurophysiology. This contrasts with classical inference, in which the inference is about the effect being significantly different from zero. The problem for classical inference is that trivial departures from the null hypothesis can be declared significant, with sufficient data or sensitivity. From the point of view of neuroimaging, posterior inference is especially useful because it eschews the multiple-comparisons problem. In classical inference, one tries to ensure that the probability of rejecting the null hypothesis incorrectly is maintained at a small rate, despite making inferences over large volumes of the brain. This induces a multiple-comparisons problem that, for continuous spatially extended data, requires an adjustment or correction to the p value using GRF theory as mentioned above. This correction means that classical inference becomes less sensitive or powerful with large search

18 Nov 2004 12:34

68

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

P1: IKH

FRISTON

volumes. In contradistinction, posterior inference does not have to contend with the multiple-comparisons problem because there are no false-positives. The probability that activation has occurred, given the data, at any particular voxel is the same, irrespective of whether one has analyzed that voxel or the entire brain. For this reason, posterior inference using PPMs represents a relatively more powerful approach than classical inference in neuroimaging. PPMs require the posterior distribution or conditional distribution of the activation (a contrast of conditional parameter estimates) given the data. This posterior density can be computed, under Gaussian assumptions, using Bayes’ rule. Bayes’ rule requires the specification of a likelihood function and the prior density of the model’s parameters. The models used to form PPMs and the likelihood functions are the same as in classical SPM analyses, namely the GLM. The only extra bit of information that is required is the prior probability distribution of the parameters. Although it would be possible to specify them using independent data or some plausible physiological constraints, there is an alternative to this fully Bayesian approach. The alternative is empirical Bayes, in which the variances of the prior distributions are estimated directly from the data. Empirical Bayes requires a hierarchical observation model where the parameters and hyperparameters at any particular level can be treated as priors on the level below. There are numerous examples of hierarchical observation models in neuroimaging. For example, the distinction between fixed- and random/mixedeffects analyses of multisubject studies relies upon a two-level hierarchical model (Friston et al. 2002). However, in neuroimaging there is a natural hierarchical observation model that is common to all brain mapping experiments. This is the hierarchy induced by looking for the same effects at every voxel within the brain (or gray matter). The first level of the hierarchy corresponds to the experimental effects at any particular voxel and the second level of the hierarchy comprises the effects over voxels. Put simply, the variation in a particular contrast, over voxels, can be used as the prior variance of that contrast at any particular voxel. Generally, hierarchical linear models have the following form:

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HIERARCHICAL MODELS AND EMPIRICAL BAYES

y = X (1) β (1) + ε (1) β (1) = X (2) β (2) + ε (2) . β (2) = . . .

(2)

This is the same as Equation 1, but now the parameters of the first level are generated by a supraordinate linear model and so on to any hierarchical depth required. These hierarchical observation models are an important extension of the GLM and are usually estimated using Expectation Maximization (Dempster et al. 1977). In the present context, the response variables comprise the responses at all voxels and β (1) are the treatment effects about which we want to make an inference. Because we have invoked a second level, the first-level parameters embody random effects and are generated by a second-level linear model. At the second level, β (2) is the

18 Nov 2004 12:34

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN FUNCTION MODELING

P1: IKH

69

average effect over voxels and ε (2) is voxel-to-voxel variation. By estimating the variance of ε(2), one is implicitly estimating an empirical prior on the first-level parameters at each voxel. This prior can then be used to estimate the posterior probability of β (1) being greater than some threshold at each voxel. An example of the ensuing PPM is provided in Figure 3 along with the classical SPM. In this section we have seen how the GLM can be used to test hypotheses about brain responses and how, in a hierarchical form, it enables empirical Bayesian or conditional inference. In the next section, we deal with dynamic systems and how they can be formulated as GLMs. These dynamic models take us closer to how experimental manipulations actually cause brain responses, and represent the next step in working toward causal models of brain responses.

Dynamic Models In Friston et al. (1994), the form of the hemodynamic impulse response function (HRF) was estimated using a least squares deconvolution and a time-invariant model, where evoked neuronal responses are convolved or smoothed with an HRF to give the measured hemodynamic response (see also Boynton et al. 1996). This simple linear convolution model is the cornerstone for making statistical inferences about activations in fMRI with the GLM. An impulse response function is the response to a single impulse, measured at a series of times after the input. It characterizes the inputoutput behavior of the system (i.e., voxel) and places important constraints on the sorts of inputs that will excite a response. Knowing the forms that the HRF can take is important for several reasons, not least because it allows for better statistical models of the data. The HRF may vary from voxel to voxel and this has to be accommodated in the GLM. To allow for different HRFs in different brain regions, the notion of temporal basis functions, to model evoked responses in fMRI, was introduced (Friston et al. 1995a) and applied to event-related responses in Josephs et al. (1997) (see also Lange & Zeger 1997). The basic idea behind temporal basis functions is that the hemodynamic response, induced by any given trial type, can be expressed as the linear combination of several (basis) functions of peristimulus time. The convolution model for fMRI responses takes a stimulus function encoding the neuronal responses and convolves it with an HRF to give a regressor that enters the design matrix. When using basis functions, the stimulus function is convolved with all the basis functions to give a series of regressors (in Figure 1 we used four stimulus functions and two basis functions to give eight regressors). Mathematically we can express this model as

CONVOLUTION MODELS AND TEMPORAL BASIS FUNCTIONS

y(t) = Xβ + ε X i = Ti (t) ⊗ u(t)

⇔

y(t) = u(t) ⊗ h(t) + ε , h(t) = β1 T1 (t) + β2 T2 (t) + . . .

(3)

where ⊗ means convolution. This equivalence illustrates how temporal basis functions allow one to take any convolution model (right) and convert it into a GLM (left). The parameter estimates β i are the coefficients or weights that determine the

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

18 Nov 2004 12:34

70

AR

FRISTON

AR231-PS56-03.tex AR231-PS56-03.sgm LaTeX2e(2002/01/18) P1: IKH

18 Nov 2004 12:34

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

BRAIN FUNCTION MODELING

P1: IKH

71

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

mixture of basis functions Ti(t) that best models h(t), the HRF for the trial type and voxel in question. We find the most useful basis set to be a canonical HRF and its derivatives with respect to the key parameters that determine its form (see below). Temporal basis functions are important because they enable a graceful transition between conventional multilinear regression models with one stimulus function per condition and FIR models with a parameter for each time point following the onset of a condition or trial type. Figure 4 illustrates this graphically. In short, temporal basis functions offer useful constraints on the form of the estimated response that retain the flexibility of FIR models and the efficiency of single regressor models.

Biophysical Models By adopting a convolution model for brain responses in fMRI, we are implicitly positing some underlying dynamic system that converts neuronal responses into observed hemodynamic responses. Our understanding of the biophysical and physiological mechanisms that underpin the HRF has grown considerably in the past few years (e.g., Buxton & Frank 1997, Hoge et al. 1999, Mandeville et al. 1999). Figure 5 shows some simulations based on the hemodynamic model described in Friston et al. (2000). Here, neuronal activity induces some autoregulated vasoactive signal that causes transient increases in regional cerebral blood flow (rCBF). The resulting flow increases lead to the dilation of a venous balloon, increasing its volume (v) and diluting venous blood to decrease deoxyhemoglobin content (q). The blood oxygenation level–dependent (BOLD) signal is roughly proportional to the concentration of deoxyhemoglobin (q/v) and follows the rCBF response with about a one-second delay. The model is framed in terms of differential equations, examples of which are provided in Figure 5. Notice that we have introduced variables, like volume and deoxyhemoglobin concentrations, that are not actually observed. These are referred to as the hidden states of input-state-output models. The state and output equations of any analytic dynamical system are

INPUT-STATE-OUTPUT SYSTEMS

←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− Figure 3 Statistical parametric map (SPM) and posterior probability map (PPM) for a functional magnetic resonance imaging study of attention to visual motion (B¨uchel & Friston 1997). The display format in the lower panel uses an axial slice through extrastriate regions but the thresholds are the same as employed in maximum-intensity projections (upper panels). (Upper right) The activation threshold for the PPM was 0.7 a.u., meaning that all voxels shown had a 90% chance of an activation of 0.7% or more. (Upper left) The corresponding SPM using a corrected threshold at p = 0.05. Note the bilateral foci of motion-related responses in the PPM that are not seen in the SPM (gray arrows). As can be imputed from the design matrix (upper middle panel), the statistical model of evoked responses comprised boxcar regressors convolved with a canonical hemodynamic response function. The middle column corresponds to the presentation of moving dots and was the stimulus property tested by the contrast.

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

18 Nov 2004 12:34

72

AR

FRISTON

AR231-PS56-03.tex AR231-PS56-03.sgm LaTeX2e(2002/01/18) P1: IKH

18 Nov 2004 12:34

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

P1: IKH

BRAIN FUNCTION MODELING

73

x˙ (t) = f (x, u, θ)

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

y(t) = g(x, u, θ) + ε.

(4)

The first line is an ordinary differential equation and expresses the rate of change of the states as a parameterized function of the states and input. Typically, the inputs u(t) correspond to designed experimental effects (e.g., the stimulus function in fMRI). There is a fundamental and causal relationship (Fliess et al. 1983) between the outputs and the history of the inputs in Equation 4. This relationship conforms to a Volterra series, which expresses the output y(t) as a generalized convolution of the input u(t), critically without reference to the hidden states x(t). This series is simply a functional Taylor expansion of the outputs with respect to the inputs (Bendat 1990). The reason it is a functional expansion is that the inputs are a function of time1, y(t) =


 i

0

t

t

...

κi (σ1 , . . . , σi )u(t − σ1 ), . . . , u(t − σi )dσ1 , . . . , dσi 0

κi (σ1 , . . . , σi ) =

∂ i y(t) , ∂u(t − σ1 ), . . . , ∂u(t − σi )

(5)

where κ i(σ 1, . . ., σ i) is the ith-order kernel. In Equation 5, the integrals are restricted to the past. This renders Equation 5 causal. The key thing here is that Equation 5 is simply a convolution and can be expressed as a GLM as in Equation 3. This means that we can take a neurophysiologically realistic model of hemodynamic responses and use it as an observation model to estimate parameters using observed data. Here the model is parameterized in terms of kernels that have a direct analytic 1

For simplicity, here and in Equation 7, we deal with only one experimental input.

←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− Figure 4 Temporal basis functions offer useful constraints on the form of the estimated response that retain (a) the flexibility of finite impulse response (FIR) models and (b) the efficiency of single regressor models. The specification of these constrained FIR models involves setting up stimulus functions u(t) that model expected neuronal changes [e.g., boxcars of epoch-related responses or spikes (delta functions) at the onset of specific events or trials]. These stimulus functions are then convolved with a set of basis functions Ti(t) of peristimulus time that, in some linear combination, model the hemodynamic impulse response function (HRF). The ensuing regressors are assembled into the design matrix. The basis functions can be as simple as a single canonical HRF (middle), through to a series of delayed delta functions (bottom). The latter case corresponds to an FIR model and the coefficients constitute estimates of the impulse response function at a finite number of discrete sampling times. Selective averaging in event-related functional magnetic resonance imaging (Dale & Buckner 1997) is mathematically equivalent to this limiting case.

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

18 Nov 2004 12:34

74

AR

FRISTON

AR231-PS56-03.tex AR231-PS56-03.sgm LaTeX2e(2002/01/18) P1: IKH

18 Nov 2004 12:34

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

BRAIN FUNCTION MODELING

P1: IKH

75

relation to the original parameters θ of the biophysical system (through Equation 5). The first-order kernel is simply the conventional HRF. High-order kernels correspond to high-order HRFs and can be estimated using basis functions as described above. In fact, by choosing basis functions according to

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

T (σ )i =

∂κ(σ )1 , ∂θi

(6)

one can estimate the biophysical parameters because, to a first-order approximation, β i = θ i. The critical step we have taken here is to start with a causal dynamic model of how responses are generated and construct a general linear observation model that allows us to estimate and infer things about the parameters of that model. This is in contrast to the conventional use of the GLM with design matrices that are not informed by a forward model of how data are caused. This approach to modeling brain responses has a much more direct connection with underlying physiology and rests upon an understanding of the underlying system. Once a suitable causal model has been established (e.g., Figure 5), we can estimate second-order kernels. These kernels represent a nonlinear characterization of the HRF that can model interactions among stimuli in causing responses. One important manifestation of the nonlinear effects, captured by the second-order kernels, is a modulation of stimulus-specific responses by preceding stimuli that are proximate in time. This means that responses at high stimulus presentation rates saturate and, in some instances, show an inverted U behavior. This behavior appears to be specific to BOLD effects (as distinct from evoked changes in cerebral blood flow) and may represent a hemodynamic refractoriness. This effect has important implications for event-related fMRI, where one may want to present trials in quick succession. The results of a typical nonlinear analysis are given in Figure 6. The results in the right panel represent the average response, integrated over a 32-second train of stimuli as a function of stimulus onset asynchrony (SOA). These responses are based on the kernel estimates (left panels) using data from a voxel in the left posterior temporal region of a subject obtained during the presentation of single words at different rates. The solid line represents the estimated response and shows

NONLINEAR SYSTEM IDENTIFICATION

←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− Figure 5 (Right) Hemodynamics elicited by an impulse of neuronal activity as predicted by a dynamical biophysical model (left). A burst of neuronal activity causes an increase in flow-inducing signal that decays with first-order kinetics and is downregulated by local flow. This signal increases regional cerebral blood flow (rCBF), which dilates the venous capillaries, increasing volume (v). Concurrently, venous blood is expelled from the venous pool, decreasing deoxyhemoglobin content (q). The resulting fall in deoxyhemoglobin concentration leads to a transient increase in the blood oxygenation level–dependent (BOLD) signal and a subsequent undershoot. (Left) Hemodynamic model on which these simulations were based (see Friston et al. 2000 for details).

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

18 Nov 2004 12:34

76

AR

FRISTON

AR231-PS56-03.tex AR231-PS56-03.sgm LaTeX2e(2002/01/18) P1: IKH

18 Nov 2004 12:34

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN FUNCTION MODELING

P1: IKH

77

a clear maximum at just less than one second. The dots are responses based on empirical data from the same experiment. The broken line shows the expected response in the absence of nonlinear effects (i.e., that predicted by setting the second-order kernel to zero). It is clear that nonlinearities become important at around two seconds, leading to an actual diminution of the integrated response at subsecond SOAs. The implication of this sort of result is that the assumptions of the linear convolution models discussed above are violated with subsecond SOAs (see also Buckner et al. 1996, Burock et al. 1998). In summary, we started with models of regionally specific responses, framed in terms of the general linear model, in which responses were modeled as linear mixtures of designed changes in explanatory variables. Hierarchical extensions to linear observation models enable random-effects analyses and, in particular, an empirical Bayesian approach. The mechanistic utility of these models is realized through forward models that embody causal dynamics. Simple variants of these are the linear convolution models used to construct explanatory variables in conventional analyses of fMRI data. These are a special case of generalized convolution models that are mathematically equivalent to input-state-output systems comprising hidden states. Estimation and inference with these dynamic models tells us something about how the response was caused, but only at the level of single voxels. In the next section, we adopt the same perspective on models, but in the context of distributed responses and functional integration.

MODELS OF FUNCTIONAL INTEGRATION Functional and Effective Connectivity Imaging neuroscience has firmly established functional specialization as a principle of brain organization in man. The integration of specialized areas has proven ←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− Figure 6 (Left panels) Volterra kernels from a voxel in the left superior temporal gyrus at −56, −28, and 12 mm. These kernel estimates were based on a single-subject study of aural word presentation at different rates (from 0 to 90 words per minute) using a second-order approximation to a Volterra series expansion modeling the observed hemodynamic response to stimulus input (a delta function for each word). These kernels can be thought of as a characterization of the second-order hemodynamic response function. The first-order kernel κ 1 (upper panel) represents the (first-order) component usually presented in linear analyses. The second-order kernel (lower panel) is presented in image format. The color scale is arbitrary; white is positive and black is negative. The insert on the right represents κ1 κ1T , the second-order kernel predicted by a simple model that involves a linear convolution with κ1 followed by some static nonlinearity. (Right panel) Integrated responses over a 32-second stimulus train as a function of stimulus onset asynchrony. Solid line: estimates based on the nonlinear convolution model parameterized by the kernels on the left. Broken line: the responses expected in the absence of second-order effects (i.e., in a truly linear system). Dots: empirical averages based on the presentation of actual stimulus trains.

18 Nov 2004 12:34

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

78

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

P1: IKH

FRISTON

more difficult to assess. Functional integration is usually inferred on the basis of correlations among measurements of neuronal activity. Functional connectivity has been defined as statistical dependencies or correlations among remote neurophysiological events. However, correlations can arise in a variety of ways; for example, in multiunit electrode recordings they can result from stimulus-locked transients evoked by a common input or reflect stimulus-induced oscillations mediated by synaptic connections (Gerstein & Perkel 1969). Integration within a distributed system is usually better understood in terms of effective connectivity: Effective connectivity refers explicitly to the influence that one neural system exerts over another, either at a synaptic (i.e., synaptic efficacy) or population level. It has been proposed that “the [electrophysiological] notion of effective connectivity should be understood as the experiment- and time-dependent, simplest possible circuit diagram that would replicate the observed timing relationships between the recorded neurons” (Aertsen & Preißl 1991). This speaks to two important points: (a) Effective connectivity is dynamic, i.e., activity- and time-dependent and (b) it depends upon a model of the interactions. The estimation procedures employed in functional neuroimaging can be divided into those based on (a) linear regression models (e.g., McIntosh & Gonzalez-Lima 1994) or (b) nonlinear dynamic causal models. There is a necessary link between functional integration and multivariate analyses because the latter are necessary to model interactions among brain regions. Multivariate approaches can be divided into those that are inferential in nature and those that are data-led or exploratory. We first consider multivariate approaches that generally are based on functional connectivity or covariance patterns (and generally are exploratory) and then turn to models of effective connectivity (that allow for some form of inference). In Friston et al. (1993), we introduced voxel-based principal component analysis (PCA) of neuroimaging time-series to characterize distributed brain systems implicated in sensorimotor, perceptual, or cognitive processes. These distributed systems are identified with principal components or eigenimages that correspond to spatial modes of coherent brain activity. This approach represents one of the simplest multivariate characterizations of functional neuroimaging time-series and falls into the class of exploratory analyses. Principal component or eigenimage analysis generally uses singular value decomposition to identify a set of orthogonal spatial modes that capture the greatest amount of variance expressed over time. As such, the ensuing modes embody the most prominent aspects of the variance-covariance structure of a given time-series. Noting that covariance among brain regions is equivalent to functional connectivity renders eigenimage analysis particularly interesting because it was among the first ways of addressing functional integration (i.e., connectivity) with neuroimaging data. Subsequently, eigenimage analysis has been elaborated in a number of ways. Notable among these is canonical variate analysis and multidimensional scaling (Friston et al. 1996a,b). Canonical variate

EIGENIMAGE ANALYSIS AND RELATED APPROACHES

18 Nov 2004 12:34

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN FUNCTION MODELING

P1: IKH

79

analysis was introduced in the context of ManCova (multiple analysis of covariance) and uses the generalized eigenvector solution to maximize the variance that can be explained by some explanatory variables relative to error. Canonical variate analysis can be thought of as an extension of eigenimage analysis that refers explicitly to some explanatory variables and allows for statistical inference. In fMRI, eigenimage analysis (e.g., Sychra et al. 1994) generally is used as an exploratory device to characterize coherent brain activity. These variance components may or may not be related to experimental design and endogenous coherent dynamics that have been observed in the motor system (Biswal et al. 1995). Despite its exploratory power, eigenimage analysis is fundamentally limited for two reasons. First, it offers only a linear decomposition of any set of neurophysiological measurements, and second, the particular set of eigenimages or spatial modes obtained is uniquely determined by constraints that are biologically implausible. These aspects of PCA confer inherent limitations on the interpretability and usefulness of eigenimage analysis of biological time-series and have motivated the exploration of nonlinear PCA and neural network approaches (e.g., Mørch et al. 1995). Two other important approaches deserve mention here. The first is independent component analysis (ICA). ICA uses entropy maximization to find, using iterative schemes, spatial modes or their dynamics that are approximately independent. This is a stronger requirement than orthogonality in PCA and involves removing highorder correlations among the modes (or dynamics). It was initially introduced as spatial ICA (McKeown et al. 1998) in which the independence constraint was applied to the modes (with no constraints on their temporal expression). Approaches that are more recent use, by analogy with magneto- and electrophysiological timeseries analysis, temporal ICA where the dynamics are enforced to be independent (e.g., Calhoun et al. 2001). This requires an initial dimension reduction (usually using conventional eigenimage analysis). Finally, there has been an interest in cluster analysis (Baumgartner et al. 1997). Conceptually, this can be related to eigenimage analysis through multidimensional scaling and principal coordinate analysis. All these approaches are interesting, but hardly anyone uses them. This is largely because the approaches tell you nothing about how the brain works and don’t allow one to ask specific questions. Simply demonstrating statistical dependencies among regional brain responses (i.e., demonstrating functional connectivity) does not address how these responses were caused. To address this, one needs explicit models of integration, or more precisely, effective connectivity.

Dynamic Causal Modeling This final section examines the modeling of interactions among neuronal populations, at a cortical level, using neuroimaging time-series. The aim of these models is to estimate, and make inferences about, the coupling among brain areas and how that coupling is influenced by changes in experimental context (e.g., time or

18 Nov 2004 12:34

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

80

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

P1: IKH

FRISTON

cognitive set). The basic idea is to construct a reasonably realistic neuronal model of interacting cortical regions or nodes. This model is then supplemented with a forward model of how neuronal or synaptic activity translates into a measured response (see previous section). This enables the parameters of the neuronal model (i.e., effective connectivity) to be estimated from observed data. Intuitively, this approach regards an experiment as a designed perturbation of neuronal dynamics that are promulgated and distributed throughout a system of coupled anatomical nodes to change region-specific neuronal activity. These changes engender, through a measurement-specific forward model, responses that are used to identify the architecture and time constants of the system at a neuronal level. This represents a departure from conventional approaches (e.g., structural equation modeling and autoregression models; B¨uchel & Friston 1997, Harrison et al. 2003, McIntosh & Gonzalez-Lima 1994), in which one assumes the observed responses are driven by endogenous or intrinsic noise (i.e., innovations). In contradistinction, dynamic causal models assume the responses are driven by designed changes in inputs. An important conceptual aspect of dynamic causal models pertains to how the experimental inputs enter the model and cause neuronal responses. Experimental variables can illicit responses in one of two ways. First, they can elicit responses through direct influences on specific anatomical nodes. This would be appropriate, for example, in modeling sensory evoked responses in early visual cortices. The second class of input exerts its effect vicariously, through a modulation of the coupling among nodes. These sorts of experimental variables would normally be more enduring; for example, attention to a particular attribute or the maintenance of some cognitive set. These distinctions are seen most clearly in relation to particular forms of causal models used for estimation; for example, the bilinear approximation x˙ (t) = f (x, u) = Ax + u Bx + Cu y = g(x) + ε A=

∂f ∂x

B=

∂2 f ∂ x∂u

C=

∂f . ∂u

(7)

This is an approximation to any model of how changes in neuronal activity in one region, x(t)i, are caused by activity in the other regions. Here the output function g(x) embodies a hemodynamic model, linking neuronal activity to BOLD, for each region (e.g., that in Figure 5). The matrix A represents the connectivity among the regions in the absence of input u(t). Effective connectivity is the influence that one neuronal system exerts over another in terms of inducing a response ∂ x˙ /∂ x. This latent connectivity can be thought of as the intrinsic coupling in the absence of experimental perturbations. The matrix B is effectively the change in intrinsic coupling induced by the input. It encodes the input-sensitive changes in A or, equivalently, the modulation of effective connectivity by experimental manipulations. Because B is a second-order derivative, it is referred to as bilinear.

18 Nov 2004 12:34

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN FUNCTION MODELING

P1: IKH

81

Finally, the matrix C embodies the extrinsic influences of inputs on neuronal activity. The parameters θ = {A, B, C} are the connectivity or coupling matrices that we wish to identify and that define the functional architecture and interactions among brain regions at a neuronal level. Because Equation 7 has exactly the same form as Equation 4, we can express it as a GLM and estimate the parameters. Generally, estimation in the context of highly parameterized models like DCMs requires constraints in the form of priors. These priors enable conditional inference about the connectivity estimates. The sorts of questions that can be addressed with DCMs are now illustrated by looking at how attentional modulation might be mediated in sensory processing hierarchies in the brain. It has been established that the superior posterior parietal cortex (SPC) exerts a modulatory role on V5 responses using Volterra-based regression models (Friston & B¨uchel 2000), and that the inferior frontal gyrus (IFG) exerts a similar influence on SPC using structural equation modeling (B¨uchel & Friston 1997). The example here shows that DCM leads to the same conclusions but starting from a completely different construct (see Friston et al. 2003 for details). The experimental paradigm and data acquisition parameters are described in the Figure 7 legend. Figure 7 also shows the location of the regions that entered into the DCM. These regions were based on maxima from conventional SPMs testing for the effects of photic stimulation, motion, and attention. Regional time courses were taken as the first eigenvariate of 8 mm spherical volumes of interest centered on the maxima shown in the figure. The inputs, in this example, comprise one sensory perturbation and two contextual inputs. The sensory input was simply the presence of photic stimulation, and the first contextual input was presence of motion in the visual field. The second contextual input, encoding attentional set, was unity during attention to speed changes and zero otherwise. The model parameters were fitted such that the regional outputs from the model corresponded as closely as possible to the four regional eigenvariates (Figure 7, left panel). The intrinsic connections were constrained to conform to a hierarchical pattern in which each area was reciprocally connected to its supraordinate area. Photic stimulation entered at, and only at, V1. The effect of motion in the visual field was modeled as a bilinear modulation of the V1 to V5 connectivity and attention was allowed to modulate the backward connections from IFG and SPC. The results of the DCM are shown in Figure 7 (right panel). Of primary interest is the modulatory effect of attention that is expressed in terms of the bilinear coupling parameters for this input. Analysis of the posterior densities of the bilinear parameters shows that we can be highly confident that attention modulates the backward connections from IFG to SPC and from SPC to V5. Indeed, the influences of IFG on SPC are negligible in the absence of attention (dotted connection). It is important to note that the only way attentional manipulation could effect brain responses was through this bilinear effect. Attention-related responses are seen throughout the system (attention epochs are marked with arrows in the plot of IFG

DCM AND ATTENTIONAL MODULATION

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

18 Nov 2004 12:34

82

AR

FRISTON

AR231-PS56-03.tex AR231-PS56-03.sgm LaTeX2e(2002/01/18) P1: IKH

18 Nov 2004 12:34

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN FUNCTION MODELING

P1: IKH

83

responses in Figure 7). This attentional modulation is accounted for, sufficiently, by changing just two connections. This change is, presumably, instantiated by an instructional set at the beginning of each epoch. The second idea this analysis illustrates is how DCM models functional segregation. Here one can regard V1 as “segregating” motion from other visual information and distributing it to the motion-sensitive area V5. This segregation is modeled as a bilinear “enabling” of V1 to V5 connections when, and only when, motion is present. Note that in the absence of motion the intrinsic V1 to V5 connection was trivially small (in fact the estimate was −0.04). The key advantage of entering motion through a bilinear effect, as opposed to a direct effect on V5, is that we can finesse the inference that V5 shows motion-selective responses with the assertion that these responses are mediated by afferents from V1. The two bilinear effects above represent two important aspects of functional integration that DCM is able characterize. ←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− Figure 7 Results of a dynamic causal modeling (DCM) analysis of attention to visual motion with functional magnetic resonance imaging. (Right panel) Functional architecture based upon the conditional estimates shown alongside their connections, with the percent confidence that they exceeded threshold in brackets. The most interesting aspects of this architecture involve the role of motion and attention in exerting bilinear effects. Critically, the influence of motion is to enable connections from V1 to the motion-sensitive area V5. The influence of attention is to enable backward connections from the inferior frontal gyrus (IFG) to the superior parietal cortex (SPC). Furthermore, attention increases the influence of SPC on the V5. Dotted arrows connecting regions represent significant bilinear affects in the absence of a significant intrinsic coupling. (Left panel) Fitted responses based upon the conditional estimates and the adjusted data are shown for each region in the DCM. The insert (upper left) shows the location of the regions. The fMRI data were from a study in which subjects viewed identical stimuli (visual motion subtended by radially moving dots) under different attentional manipulations of the task (detection of velocity changes) (B¨uchel & Friston 1997). The data were acquired from a normal subject at 2 Tesla, using a whole-body MRI system, equipped with a head volume coil. Contiguous multislice T2∗ -weighted fMRI images were obtained with a gradient echo-planar sequence (TE = 40 ms, TR = 3.22 s, matrix size = 64 × 64 × 32, voxel size 3 × 3 × 3 mm). Each subject had four consecutive 100-scan sessions comprising a series of 10-scan blocks under five different conditions D F A F N F A F N S. The first condition (D) was a dummy condition to allow for magnetic saturation effects. F (fixation) corresponds to a low-level baseline where the subjects viewed a fixation point at the center of a screen. In condition A (attention), subjects viewed 250 dots moving radially from the center at 4.7◦ per second and were asked to detect changes in radial velocity. In condition N (no attention), the subjects were asked simply to view the moving dots. In condition S (stationary), subjects viewed stationary dots. The order of A and N was swapped for the last two sessions. In all conditions, subjects fixated the center of the screen. There were no speed changes during scanning. No overt response was required in any condition.

18 Nov 2004 12:34

84

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

P1: IKH

FRISTON

The central ideal behind dynamic causal modeling is to treat the brain as a deterministic nonlinear dynamic system that is subject to inputs and produces outputs. Effective connectivity is parameterized in terms of coupling among unobserved brain states (e.g., neuronal activity in different regions). The objective is to estimate these parameters by perturbing the system and measuring the response. This is in contradistinction to established methods for estimating effective connectivity from neurophysiological time-series, which include structural equation modeling (B¨uchel & Friston 1997, McIntosh & Gonzalez-Lima 1994) and models based on multivariate auto-regressive processes (Goebel et al. 2003, Harrison et al. 2003). In these models, there is no designed perturbation and the inputs are treated as unknown and stochastic. Furthermore, the inputs are assumed to express themselves instantaneously such that, at the point of observation, the change in states will be zero. From Equation 7, in the absence of bilinear effects, we have

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

STRUCTURAL EQUATION MODELING

x˙ = 0 = Ax + Cu x = −A−1 Cu.

(8)

This is the regression equation used in SEM where A = A − I and A contains the off-diagonal connections among regions. The key point here is that A is estimated by assuming u is some random innovation with known covariance. This is not tenable for designed experiments when u represents carefully structured experimental inputs. Although SEM and related autoregressive techniques are useful for establishing dependence among regional responses, these techniques are not a surrogate for informed causal models based on the underlying dynamics of these responses.

CONCLUSION In this article, we reviewed the main models that underpin image analysis and touched briefly on ways of assessing specialization and integration in the brain. The key principles of functional brain architectures were used to motivate the various models considered. These can be regarded as a succession of modeling endeavors, drawing more and more on our understanding of how brain-imaging signals are generated, both in terms of biophysics and the underlying neuronal interactions. We have seen how hierarchical linear observation models encode the treatments effects elicited by experimental design. General linear models based on convolution models imply an underlying dynamic input-state-output system. The form of these systems can be used to constrain convolution models and explore some of their simpler nonlinear properties. By creating observation models based on an explicit forward model of neuronal interactions, one can now start to model and assess interactions among distributed cortical areas and make inferences about

18 Nov 2004 12:34

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

BRAIN FUNCTION MODELING

P1: IKH

85

coupling at the neuronal level. The next few years will probably see an increasing realism in the dynamic causal models introduced above (see Horwitz et al. 2001). Attempts already have been made to use plausible models of neuronal ensemble to estimate network parameters of evoked responses in EEG (David & Friston 2003). These endeavors are likely to encompass fMRI signals in the near future, enabling the conjoint modeling, or fusion, of different modalities and the marriage of computational neuroscience with the modeling of brain responses.

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

ACKNOWLEDGMENTS The Wellcome Trust funded this work. I would like to thank all of my colleagues at the Wellcome Department of Imaging Neuroscience for their conceptual input. The Annual Review of Psychology is online at http://psych.annualreviews.org

LITERATURE CITED Absher JR, Benson DF. 1993. Disconnection syndromes: an overview of Geschwind’s contributions. Neurology 43:862–67 Adler RJ. 1981. The Geometry of Random Fields. New York: Wiley Aertsen A, Preißl H. 1991. Dynamics of activity and connectivity in physiological neuronal networks. In Nonlinear Dynamics and Neuronal Networks, ed. HG Schuster, pp. 281– 302. New York: VCH Publ. Ashburner J, Friston KJ. 2000. Voxel-based morphometry—the methods. NeuroImage 11:805–21 Bandettini PA, Jesmanowicz A, Wong EC, Hyde JS. 1993. Processing strategies for time course data sets in functional MRI of the human brain. Magn. Reson. Med. 30:161–73 Baumgartner R, Scarth G, Teichtmeister C, Somorjai R, Moser E. 1997. Fuzzy clustering of gradient-echo functional MRI in the human visual cortex. Part 1: reproducibility. J. Magn. Reson. Imaging 7:1094–101 Bendat JS. 1990. Nonlinear System Analysis and Identification from Random Data. New York: Wiley Berry DA, Hochberg Y. 1999. Bayesian perspectives on multiple comparisons. J. Stat. Plan. Infer. 82:215–27 Biswal B, Yetkin FZ, Haughton VM, Hyde JS.

1995. Functional connectivity in the motor cortex of resting human brain using echoplanar MRI. Magn. Reson. Med. 34:537–41 Boynton GM, Engel SA, Glover GH, Heeger DJ. 1996. Linear systems analysis of functional magnetic resonance imaging in human V1. J. Neurosci. 16:4207–21 B¨uchel C, Friston KJ. 1997. Modulation of connectivity in visual pathways by attention: cortical interactions evaluated with structural equation modeling and fMRI. Cereb. Cortex 7:768–78 B¨uchel C, Wise RJS, Mummery CJ, Poline JB, Friston KJ. 1996. Nonlinear regression in parametric activation studies. NeuroImage 4:60–66 Buckner R, Bandettini P, O’Craven K, Savoy R, Petersen S, et al. 1996. Detection of cortical activation during averaged single trials of a cognitive task using functional magnetic resonance imaging. Proc. Natl. Acad. Sci. USA 93:14878–83 Burock MA, Buckner RL, Woldorff MG, Rosen BR, Dale AM. 1998. Randomized eventrelated experimental designs allow for extremely rapid presentation rates using functional MRI. NeuroReport 9:3735–39 Buxton RB, Frank LR. 1997. A model for the coupling between cerebral blood flow and

18 Nov 2004 12:34

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

86

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

P1: IKH

FRISTON

oxygen metabolism during neural stimulation. J. Cereb. Blood Flow Metab. 17:64–72 Calhoun VD, Adali T, Pearlson GD, Pekar JJ. 2001. Spatial and temporal independent component analysis of functional MRI data containing a pair of task-related waveforms. Hum. Brain Mapp. 13:43–53 Chawla D, Buechel C, Edwards R, Howseman A, Josephs O, et al. 1999. Speed-dependent responses in V5: a replication study. NeuroImage 9:508–15 Dale A, Buckner R. 1997. Selective averaging of rapidly presented individual trials using fMRI. Hum. Brain Mapp. 5:329–40 David O, Friston KJ. 2003. A neural mass model for MEG/EEG: coupling and neuronal dynamics. NeuroImage 20:1743–55 Dempster AP, Laird NM, Rubin DB. 1977. Maximum likelihood from incomplete data via the EM algorithm. J. R. Stat. Soc. Ser. B 39:1–38 Fliess M, Lamnabhi M, Lamnabhi-Lagarrigue F. 1983. An algebraic approach to nonlinear functional expansions. IEEE Trans. Circuits Syst. 30:554–70 Friston KJ, B¨uchel C. 2000. Attentional modulation of effective connectivity from V2 to V5/MT in humans. Proc. Natl. Acad. Sci. USA 97:7591–96 Friston KJ, Frith CD, Fletcher P, Liddle PF, Frackowiak RSJ. 1996a. Functional topography: multidimensional scaling and functional connectivity in the brain. Cereb. Cortex 6:156–64 Friston KJ, Frith CD, Liddle PF, Frackowiak RSJ. 1991. Comparing functional (PET) images: the assessment of significant change. J. Cereb. Blood Flow Metab. 11:690–99 Friston KJ, Frith CD, Liddle PF, Frackowiak RSJ. 1993. Functional connectivity: the principal component analysis of large data sets. J. Cereb. Blood Flow Metab. 13:5–14 Friston KJ, Frith CD, Passingham RE, Liddle PF, Frackowiak RSJ. 1992. Motor practice and neurophysiological adaptation in the cerebellum: a positron tomography study. Proc. R. Soc. London Ser. B 248:223– 28

Friston KJ, Frith CD, Turner R, Frackowiak RSJ. 1995a. Characterizing evoked hemodynamics with fMRI. NeuroImage 2:157–65 Friston KJ, Harrison L, Penny W. 2003. Dynamic causal modeling. NeuroImage 19: 1273–302 Friston KJ, Holmes AP, Worsley KJ, Poline JB, Frith CD, Frackowiak RSJ. 1995b. Statistical parametric maps in functional imaging: a general linear approach. Hum. Brain Mapp. 2:189–210 Friston KJ, Jezzard PJ, Turner R. 1994. Analysis of functional MRI time-series. Hum. Brain Mapp. 1:153–71 Friston KJ, Mechelli A, Turner R, Price CJ. 2000. Nonlinear responses in fMRI: the Balloon model, Volterra kernels, and other hemodynamics. NeuroImage 12:466–77 Friston KJ, Penny W. 2003. Posterior probability maps and SPMs. NeuroImage 19:1240– 49 Friston KJ, Penny W, Phillips C, Kiebel S, Hinton G, Ashburner J. 2002. Classical and Bayesian inference in neuroimaging: theory. NeuroImage 16:465–83 Friston KJ, Poline J-B, Holmes AP, Frith CD, Frackowiak RSJ. 1996b. A multivariate analysis of PET activation studies. Hum. Brain Mapp. 4:140–51 Friston KJ, Price CJ, Fletcher P, Moore C, Frackowiak RSJ, Dolan RJ. 1996c. The trouble with cognitive subtraction. NeuroImage 4:97–104 Gerstein GL, Perkel DH. 1969. Simultaneously recorded trains of action potentials: analysis and functional interpretation. Science 164:828–30 Goebel R, Roebroeck A, Kim DS, Formisano E. 2003. Investigating directed cortical interactions in time-resolved fMRI data using vector autoregressive modeling and Granger causality mapping. Magn. Reson. Imaging 21:1251–61 Goltz F. 1881. In Transactions of the 7th International Medical Congress, ed. W MacCormac, Vol. I, pp. 218–28. London: Kolkmann Grafton S, Mazziotta J, Presty S, Friston KJ, Frackowiak RSJ, Phelps M. 1992. Functional

18 Nov 2004 12:34

AR

AR231-PS56-03.tex

AR231-PS56-03.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN FUNCTION MODELING anatomy of human procedural learning determined with regional cerebral blood flow and PET. J. Neurosci. 12:2542–48 Harrison LM, Penny W, Friston KJ. 2003. Multivariate autoregressive modeling of fMRI time series. NeuroImage 19:1477–91 Hoge RD, Atkinson J, Gill B, Crelier GR, Marrett S, Pike GB. 1999. Linear coupling between cerebral blood flow and oxygen consumption in activated human cortex. Proc. Natl. Acad. Sci. USA 96:9403–8 Holmes A, Ford I. 1993. A Bayesian approach to significance testing for statistic images from PET. In Quantification of Brain Function, Tracer Kinetics and Image Analysis in Brain PET, ed. K Uemura, NA Lassen, T Jones, I. Kanno, Ser. 1030, pp. 521–34. Amsterdam: Excerpta Medica Horwitz B, Friston KJ, Taylor JG. 2001. Neural modeling and functional brain imaging: an overview. Neural Netw. 13:829–46 Josephs O, Turner R, Friston KJ. 1997. Eventrelated fMRI. Hum. Brain Mapp. 5:243– 48 Lange N, Zeger SL. 1997. Non-linear Fourier time series analysis for human brain mapping by functional magnetic resonance imaging (with discussion) J. Roy. Stat. Soc. Ser. C 46:1–29 Lueck CJ, Zeki S, Friston KJ, Deiber MP, Cope NO, et al. 1989. The color centre in the cerebral cortex of man. Nature 340:386–89 Mandeville JB, Marota JJ, Ayata C, Zararchuk G, Moskowitz MA, et al. 1999. Evidence of a cerebrovascular postarteriole Windkessel with delayed compliance. J. Cereb. Blood Flow Metab. 19:679–89 McIntosh AR, Gonzalez-Lima F. 1994. Structural equation modeling and its application to network analysis in functional brain imaging. Hum. Brain Mapp. 2:2–22 McKeown M, Jung T-P, Makeig S, Brown G, Kinderman S, et al. 1998. Spatially independent activity patterns in functional MRI data during the Stroop color naming task. Proc. Natl. Acad. Sci. USA 95:803–10 Mørch N, Kjems U, Hansen LK, Svarer C, Law

P1: IKH

87

I, et al. 1995. Visualization of neural networks using saliency maps. IEEE Int. Conf. Neural Netw., pp. 2085–90. Perth, Austr. Nichols TE, Holmes AP. 2002. Nonparametric permutation tests for functional neuroimaging: a primer with examples. Hum. Brain Mapp. 15:1–25 Passingham RE, Stephan KE, Kotter R. 2002. The anatomical basis of functional localization in the cortex. Nat. Rev. Neurosci. 3:606– 16 Petersen SE, Fox PT, Posner MI, Mintun M, Raichle ME. 1989. Positron emission tomographic studies of the processing of single words. J. Cogn. Neurosci. 1:153–70 Phillips CG, Zeki S, Barlow HB. 1984. Localisation of function in the cerebral cortex: past, present, and future. Brain 107:327–61 Price CJ, Friston KJ. 1997. Cognitive conjunction: a new approach to brain activation experiments. NeuroImage 5:261–70 Price CJ, Wise RJS, Ramsay S, Friston KJ, Howard D, et al. 1992. Regional response differences within the human auditory cortex when listening to words. Neurosci. Lett. 146:179–82 Sychra JJ, Bandettini PA, Bhattacharya N, Lin Q. 1994. Synthetic images by subspace transforms. I. Principal component images and related filters. Med. Phys. 21:193–201 Talairach P, Tournoux J. 1988. A Stereotactic Coplanar Atlas of the Human Brain. Stuttgart: Thieme Worsley KJ, Evans AC, Marrett S, Neelin P. 1992. A three-dimensional statistical analysis for rCBF activation studies in human brain. J. Cereb. Blood Flow Metab. 12:900– 18 Worsley KJ, Marrett S, Neelin P, Vandal AC, Friston KJ, Evans AC. 1996. A unified statistical approach or determining significant signals in images of cerebral activation. Hum. Brain Mapp. 4:58–73 Zeki S. 1990. The motion pathways of the visual cortex. In Vision: Coding and Efficiency, ed. C Blakemore, pp. 321–45. London: Cambridge Univ. Press

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

Annual Review of Psychology Volume 56, 2005

CONTENTS Frontispiece—Richard F. Thompson

xviii

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

PREFATORY In Search of Memory Traces, Richard F. Thompson

1

DECISION MAKING Indeterminacy in Brain and Behavior, Paul W. Glimcher

25

BRAIN IMAGING/COGNITIVE NEUROSCIENCE Models of Brain Function in Neuroimaging, Karl J. Friston

57

MUSIC PERCEPTION Brain Organization for Music Processing, Isabelle Peretz and Robert J. Zatorre

89

SOMESTHETIC AND VESTIBULAR SENSES Vestibular, Proprioceptive, and Haptic Contributions to Spatial Orientation, James R. Lackner and Paul DiZio

115

CONCEPTS AND CATEGORIES Human Category Learning, F. Gregory Ashby and W. Todd Maddox

149

ANIMAL LEARNING AND BEHAVIOR: CLASSICAL Pavlovian Conditioning: A Functional Perspective, Michael Domjan

179

NEUROSCIENCE OF LEARNING The Neuroscience of Mammalian Associative Learning, Michael S. Fanselow and Andrew M. Poulos

207

HUMAN DEVELOPMENT: EMOTIONAL, SOCIAL, AND PERSONALITY Behavioral Inhibition: Linking Biology and Behavior Within a Developmental Framework, Nathan A. Fox, Heather A. Henderson, Peter J. Marshall, Kate E. Nichols, and Melissa A. Ghera

235

BIOLOGICAL AND GENETIC PROCESSES IN DEVELOPMENT Human Development: Biological and Genetic Processes, Irving I. Gottesman and Daniel R. Hanson

263 vii

P1: JRX

December 8, 2004

viii

12:13

Annual Reviews

AR231-FM

CONTENTS

SPECIAL TOPICS IN PSYCHOPATHOLOGY The Psychology and Neurobiology of Suicidal Behavior, Thomas E. Joiner Jr., Jessica S. Brown, and LaRicka R. Wingate

287

DISORDERS OF CHILDHOOD Autism in Infancy and Early Childhood, Fred Volkmar, Kasia Chawarska, and Ami Klin

315

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

CHILD/FAMILY THERAPY Youth Psychotherapy Outcome Research: A Review and Critique of the Evidence Base, John R. Weisz, Amanda Jensen Doss, and Kristin M. Hawley

337

ALTRUISM AND AGGRESSION Prosocial Behavior: Multilevel Perspectives, Louis A. Penner, John F. Dovidio, Jane A. Piliavin, and David A. Schroeder

365

INTERGROUP RELATIONS, STIGMA, STEREOTYPING, PREJUDICE, DISCRIMINATION The Social Psychology of Stigma, Brenda Major and Laurie T. O’Brien

393

PERSONALITY PROCESSES Personality Architecture: Within-Person Structures and Processes, Daniel Cervone

423

PERSONALITY DEVELOPMENT: STABILITY AND CHANGE Personality Development: Stability and Change, Avshalom Caspi, Brent W. Roberts, and Rebecca L. Shiner

453

WORK MOTIVATION Work Motivation Theory and Research at the Dawn of the Twenty-First Century, Gary P. Latham and Craig C. Pinder

485

GROUPS AND TEAMS Teams in Organizations: From Input-Process-Output Models to IMOI Models, Daniel R. Ilgen, John R. Hollenbeck, Michael Johnson, and Dustin Jundt

517

LEADERSHIP Presidential Leadership, George R. Goethals

545

PERSONNEL EVALUATION AND COMPENSATION Personnel Psychology: Performance Evaluation and Pay for Performance, Sara L. Rynes, Barry Gerhart, and Laura Parks

571

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

CONTENTS

ix

PSYCHOPHYSIOLOGICAL DISORDERS AND PSYCHOLOGICAL EFFECTS ON MEDICAL DISORDERS Psychological Approaches to Understanding and Treating Disease-Related Pain, Francis J. Keefe, Amy P. Abernethy, and Lisa C. Campbell

601

TIMELY TOPIC Psychological Evidence at the Dawn of the Law’s Scientific Age, David L. Faigman and John Monahan

631

Annu. Rev. Psychol. 2005.56:57-87. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDEXES Subject Index Cumulative Index of Contributing Authors, Volumes 46–56 Cumulative Index of Chapter Titles, Volumes 46–56

ERRATA An online log of corrections to Annual Review of Psychology chapters may be found at http://psych.annualreviews.org/errata.shtml

661 695 700

18 Nov 2004 10:54

AR

AR231-PS56-04.tex

AR231-PS56-04.sgm

P1: IKH LaTeX2e(2002/01/18) 10.1146/annurev.psych.56.091103.070225

Annu. Rev. Psychol. 2005. 56:89–114 doi: 10.1146/annurev.psych.56.091103.070225 c 2005 by Annual Reviews. All rights reserved Copyright
First published online as a Review in Advance on June 21, 2004

BRAIN ORGANIZATION FOR MUSIC PROCESSING Isabelle Peretz

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Department of Psychology, University of Montreal, Montreal, Quebec H3C 3J7; email: [email protected]

Robert J. Zatorre Montreal Neurological Institute, McGill University, Montreal, Quebec H3A 2B4; email: [email protected]

Key Words auditory cognitive neuroscience, neural correlates, musical disorders, neuroimaging ■ Abstract Research on how the brain processes music is emerging as a rich and stimulating area of investigation of perception, memory, emotion, and performance. Results emanating from both lesion studies and neuroimaging techniques are reviewed and integrated for each of these musical functions. We focus our attention on the common core of musical abilities shared by musicians and nonmusicians alike. Hence, the effect of musical training on brain plasticity is examined in a separate section, after a review of the available data regarding music playing and reading skills that are typically cultivated by musicians. Finally, we address a currently debated issue regarding the putative existence of music-specific neural networks. Unfortunately, due to scarcity of research on the macrostructure of music organization and on cultural differences, the musical material under focus is at the level of the musical phrase, as typically used in Western popular music.

CONTENTS INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . MUSIC PERCEPTION AND RECOGNITION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pitch and Time Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pitch Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Time Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Emotion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . MUSIC PERFORMANCE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Singing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Music Playing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sight-Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Training . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . MUSIC SPECIFICITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0066-4308/05/0203-0089$14.00

90 90 91 91 94 95 98 99 99 100 101 102 105 106

89

18 Nov 2004 10:54

90

AR

PERETZ

AR231-PS56-04.tex




AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

P1: IKH

ZATORRE

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INTRODUCTION Music processing has fascinated neuroscientists for more than a century (Critchley & Henson 1977). Yet it is only over the last decade that music processing has become an area of intense and systematic study, as illustrated by three special issues recently published on the cognitive neuroscience of music (Annals of the New York Academy of Sciences issues 930 in 2001 and 999 in 2003, and Nature Neuroscience issue 6, 2003). Two forces have driven this scientific effort. First, music offers a unique opportunity to better understand the organization of the human brain. For example, only a minority of individuals become proficient musicians through explicit tutoring. This particularity in the distribution of acquired skills confers to music a privileged role in the exploration of the nature and extent of brain plasticity. The other major motivation is that the study of brain organization provides a unique tool to reveal the inner working of music processing. For instance, brain anomalies may reveal to what extent and at what level music processing recruits neural networks that are distinct from those involved in other auditory-vocal functions, such as language. Although these two major approaches for exploring the brain principles that underlie music processing may not always converge, it is hoped that the combined effort gives rise to a better understanding of the neurobiological roots of one of the most characteristic traits of humans, namely music.

MUSIC PERCEPTION AND RECOGNITION A sound reaching the eardrum sets into motion a complex cascade of mechanical, chemical, and neural events in the cochlea, brain stem, midbrain nuclei, and cortex that eventually—but rapidly—results in a percept. The task of auditory cognitive neuroscience is to figure out how this happens. Musical sounds and all other sounds share most of the processing stages throughout the auditory neuraxis. However, as we discuss below, evidence points to a degree of functional segregation in the processing of music, which may in part be related to the important role played by pitch processing within music. In a broader context, the functional system that handles music processing must solve a similar computational problem to that faced by any perceptual system: It must generate internal representations of any given input, permitting the stimulus to be segregated from its background, analyzed along several dimensions, recognized, and possibly acted upon. Importantly, the nature of the representations eventually generated by this system need to be relatively abstract, in the sense that they must be insensitive to superficial variations in stimulus features (loudness, reverberation, spectral filtering, etc.). In other words, perceptual constancy must be maintained. Indeed, music is an excellent medium to study this process, as the Gestaltists realized long ago, because music relies on relations between elements, rather than on absolute values of elements (the most obvious example being a tune, which is defined not by the pitches of its constituent tones, but by the arrangement of the intervals between the pitches).

18 Nov 2004 10:54

AR

AR231-PS56-04.tex

AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

BRAIN ORGANIZATION FOR MUSIC PROCESSING

P1: IKH

91

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Pitch and Time Relations Musical pitch-based (melodic) and time-based (temporal) relations traditionally are treated separately, both empirically and theoretically (Justus & Bharucha 2002, Krumhansl 2000). Their processing independence has been questioned, however. For example, Jones & Boltz (1989) have argued that perception, attention, and memory for pitch relations are inherently rhythmical. By this view, listeners treat melody and rhythm as a unified dimension. Nevertheless, the neuropsychological literature is more consistent with the traditional view, by which melodic and temporal structures are processed independently. Brain damage can interfere with the discrimination of pitch relations while sparing the accurate interpretation of time relations (Ayotte et al. 2000, Li´egeois-Chauvel et al. 1998, Peretz 1990, Peretz & Kolinsky 1993, Piccirilli et al. 2000, Vignolo 2003). Conversely, rhythmic discrimination of musical events can be impaired while extraction of pitch content is spared (Di Pietro et al. 2004, Peretz 1990). This double dissociation between melodic and temporal processing has also been observed in singing and sight-reading (see below). Hence, the evidence suggests that these two dimensions involve the operation of separable neural subsystems. To our knowledge, neuroimaging techniques have not yet corroborated this dissociation. On the contrary, Griffiths et al. (1999) obtained similar activation patterns for pitch and time discrimination tasks. However, similarity of activation does not entail that the underlying operations are similar, or that the recruited brain areas are critical to performing a given task. This general point always must be taken into account in integrating lesion and neuroimaging findings.

Pitch Relations The right temporal neocortex plays a particularly important role in the computation of pitch relations. The initial evidence came from studies of patients with focal brain excisions (Milner 1962), which showed that (a) right temporal damage entailed a greater deficit than left, and (b) such impairment could not be explained by a deficit in simple pitch discrimination. Subsequent studies have generally supported these findings (Li´egeois-Chauvel et al. 1998, Zatorre 1985). In addition, however, deficits in certain aspects of pitch processing have been noted specifically after damage to the right anterolateral part of Heschl’s gyrus. Patients with such damage have difficulty perceiving the pitch of the missing fundamental (Zatorre 1988) and have an increased threshold for determining the direction of pitch change (Johnsrude et al. 2000). Thus, this area of right auditory cortex seems to play a particular role in the analysis of pitch information (Tramo et al. 2002). Converging evidence is now beginning to accumulate from neuroimaging studies to support the view that analysis of pitch changes may involve areas of posterior secondary cortex. Several investigators have found activation in this location when comparing complex tones containing frequency or amplitude modulation, or spectral changes, to static tones (Hall et al. 2002, Hart et al. 2003, Thivard et al. 2000). In many cases, manipulations of fine-grained pitch lead to greater response from right

18 Nov 2004 10:54

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

92

AR

PERETZ

AR231-PS56-04.tex




AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

P1: IKH

ZATORRE

auditory regions (Zatorre & Belin 2001, Zatorre et al. 2002). Using isochronous melodies made with iterated noise bursts, Patterson et al. (2002) observed similar results, but activity increases in anterior areas were also found (Griffiths et al. 1998). Warren et al. (2003) additionally report that posterior regions are involved in processing pitch height whereas anterior regions are important for pitch chroma. In summary, the data are quite consistent in implicating the right secondary auditory cortex region in operations related to processing relationships between pitch elements as they change over time, especially if the pitch changes are small. One plausible interpretation of these findings is that they represent an early stage of melodic analysis, beyond simple pitch extraction, where interval and/or contour information is processed. This interpretation is consistent with data from electroencephalography and magnetoencephalography (MEG) recordings indicating that the auditory cortex responds to such pitch relations even in the absence of attention (Tervaniemi 2003). The contribution of contour (defined by pitch directions) and intervals (defined by frequency ratios between successive notes) to melodic perception and recognition also can be distinguished in brain-damaged patients (Peretz 2001). Taken together, the results converge to show that when listeners rely on contour representation to discriminate melodies, the right superior temporal gyrus plays a critical role. When contour cues are not available and interval information is required, both the right and left temporal structures appear to be involved (Ayotte et al. 2000, Li´egeois-Chauvel et al. 1998, Peretz 1990, Vignolo 2003). This cooperation of the hemispheres in the integration of contour with interval information in melodies is echoed by the pattern of coherence observed dynamically with MEG while neurologically intact people listen to melodylike sequences (Patel & Balaban 2000). Such integration may occur very early in development. Balaban et al. (1998) have shown that the same left-right brain asymmetry in the processing of interval and contour cues in melodies is already present in infancy. In the adult, extraction of both contour and intervals does not require attention either, as electrical brain responses suggest (Tervaniemi 2003, Trainor et al. 2002). Pitch relations do not define only direction and interval sizes in tonal music; they also evoke a particular scale. A musical scale refers to the use of a small subset of pitches (typically between five and seven) in a given piece. The scale tones are not equivalent and are organized around a central tone, called the tonic. A hierarchy of importance or stability exists among the other scale tones, with the fifth and the third scale tones more closely related to the tonic than to the other scale tones. Substantial empirical evidence demonstrates that listeners use scale structure in melodies for perception and memory, albeit in an implicit manner (Tillmann et al. 2000). Despite the central importance of this tonal knowledge to the encoding of pitch relations in a melodic context, it has been little explored with neuropsychological methods. Yet, this processing component appears isolable both functionally and neuroanatomically. Brain damage can disrupt the normal intervention of tonal knowledge in melodic processing (Fran¸ce` s et al. 1973), while sparing perception

18 Nov 2004 10:54

AR

AR231-PS56-04.tex

AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN ORGANIZATION FOR MUSIC PROCESSING

P1: IKH

93

of intervals and contour (Peretz 1993). In such cases, the patient is no longer able to distinguish tonal from atonal music nor to judge melodic closure properly and suffers from a severe reduction in pitch memory. The use of scale structure in melody processing can also be indexed by electrical brain responses such as event-related potentials time-locked to tones that deviate from the expected scale degree (Besson & Fa¨ıta 1995, Besson et al. 1998) and, more surprisingly, by model-fitting of the haemodynamic responses obtained with functional magnetic resonance imaging (fMRI; Janata et al. 2002). Hence, the evidence points to the existence of neural networks that are specialized for the processing of scale structure in melodies. Their localization, however, remains to be determined. Up to this point, we have considered pitch relations between sequentially presented tones, but pitch relations also exist in simultaneously presented tones, as in chords. The pitch relations in chord sequences are governed by harmonic principles that are similar to those described earlier for the tonal hierarchy of scale tones. That is, different chords are differently related to one another, with one chord acting as the reference: the triadic chord that builds on the first degree of the scale (the tonic). Listeners have assimilated these principles via passive exposure to samples of Western music (Bigand 2003). These harmonic principles influence chord processing automatically even when more veridical information about the chord progression has been made explicitly available (Justus & Bharucha 2001). Hence, perception of pitch relations in chords seems to operate like the perception of the tonal hierarchy among successive tones, although the acquisition of harmonic hierarchy appears to emerge later in development (Trainor & Trehub 1992). By the age of five, the degree of harmonic appropriateness of chord progression appears assimilated, as electric brain responses indicate (Koelsch et al. 2003). Processing of harmonic relations rarely has been studied in brain-damaged populations. The only study, to our knowledge, in which chord processing has been examined showed sparing of the ability to generate expectancies from chord processing after bilateral lesion of the auditory cortex (Tramo et al. 1990). However, the patient suffered from a deficit in pitch processing, with a bias to judge welltuned chords as out of tune (Tramo et al. 2003). In contrast, detection of harmonic violations has been studied with diverse neuroimaging techniques. Deviations from harmonic expectancies elicit robust event-related potentials (Regnault et al. 2001). Their neural generators appear to be located in the inferior frontal areas (the frontal operculum) on both sides of the brain (which corresponds to Broca’s area on the left side; Maess et al. 2001). Involvement of the same areas has also been found in two recent fMRI studies of harmonic expectancies (Koelsch et al. 2002, Tillmann et al. 2003). Thus, the data point to bilateral involvement of the inferior frontal regions in detecting deviations from harmonic expectancies. Finally, it should be pointed out that both harmonic relations between chords and tonal relations between tones might be rooted in pitch consonance (or dissonance). Consonant intervals typically are expressed in terms of simple frequency ratios, such as the octave (2:1) and the perfect fifth (3:2), whereas dissonant intervals are related by complex ratios, such as the minor second (16:15). Despite the

18 Nov 2004 10:54

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

94

AR

PERETZ

AR231-PS56-04.tex




AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

P1: IKH

ZATORRE

saliency of dissonance and its initial account in terms of the poor spatial resolution of the basilar membrane (Plomp & Levelt 1965), its functional origins and neural consequences are still open questions (Tramo et al. 2003). Current evidence suggests that dissonance might be further computed bilaterally in the superior temporal gyri by specialized mechanisms. Assemblies of auditory neuron populations in Heschl’s gyrus exhibit phase-locked activity for dissonant chords but not for consonant chords, as measured with implanted electrodes in both humans and monkeys (Fishman et al. 2001). Such cortical responses to dissonance can be disrupted by bilateral lesion of the auditory cortex, resulting in a loss of sensitivity to dissonance (Peretz et al. 2001). A remaining question for the future is to determine at what stage in auditory processing (e.g., at the level of acoustical analysis and/or of tonal encoding of pitch) the computation of the perceptual attribute of dissonance is critical to the perceptual organization of music. This question often is framed in terms of a distinction between sensory and musical dissonance that would reflect built-in auditory constrains and learned associations, respectively (Krumhansl 2000, Schellenberg & Trehub 1994).

Time Relations Two types of time relations are fundamental to the temporal organization, or “rhythm,” of musical sequences: the segmentation of an ongoing sequence into temporal groups of events based on their durational values, and the extraction of an underlying temporal regularity or beat (Fraisse 1982). Beat perception leads to the perception of a metrical organization corresponding to periodic alternation between strong and weak beats (the strong beats generally correspond to the spontaneous tapping of the foot). Grouping and regularity are conceived as hierarchically organized by certain researchers (Povel & Essens 1985), while others (Drake 1998, Lerdahl & Jackendoff 1983) conceive these relations as the result of distinct processing components. The available neuropsychological evidence supports the latter view, in showing functional dissociations between grouping and regularity. Ibbotson & Morton (1981) provided initial support by showing that subjects more easily tapped a rhythmic pattern with their right hand and the beat with their left hand than the other way around. These findings suggest that the right hemisphere better handles meter, whereas grouping would rely essentially on the left. Further support for the separation of the two types of organization mediating temporal pattern processing has been provided by the study of two brain-damaged patients who, after lesions of the right temporal auditory cortex, could no longer tap the beat or generate a steady pulse (Fries & Swihart 1990, Wilson et al. 2002). In contrast, both patients were able to discriminate or reproduce irregular temporal sequences. Neuroimaging data also suggest that metrical rhythms may be processed differently than nonmetrical rhythms, engaging different frontal and cerebellar mechanisms (Sakai et al. 1999). All of these studies have used tapping tasks. In studies that assessed more perceptual aspects of temporal processing, convergent

18 Nov 2004 10:54

AR

AR231-PS56-04.tex

AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN ORGANIZATION FOR MUSIC PROCESSING

P1: IKH

95

evidence for dissociating meter from grouping was also found (Li´egeois-Chauvel et al. 1998, Peretz 1990). Patients who have sustained an excision in either their left or right temporal lobe have been found to discriminate rhythmic patterns normally, but fail on meter evaluation after a right-sided lesion of the anterior part of the superior temporal gyrus (Li´egeois-Chauvel et al. 1998). Conversely, impairments in rhythmic discrimination that spare meter evaluation can also be observed in brain-damaged patients after a left-sided lesion (Di Pietro et al. 2004). Hence, the data are generally consistent with the manual asymmetries observed in normal tapping (Ibbotson & Morton 1981) in pointing to the left hemisphere for temporal grouping (Di Pietro et al. 2004, Vignolo 2003) and to the right temporal auditory cortex for meter (Li´egeois-Chauvel et al. 1998, Penhune et al. 1999, Wilson et al. 2002). The fact that essentially all the results coming from production tasks have parallels in perceptual tasks suggests a strong motor component to the mental representation of musical rhythm. Indeed, data from both lesion and neuroimaging studies have shown the participation of the cerebellum and/or basal ganglia as a possible central mechanism controlling motor and perceptual timing (Janata & Grafton 2003). Patients with cerebellar damage have increased variability in motor timing, as well as decrements in discrimination of auditory intervals (Ivry & Keele 1989). Further, patients with damage to the lateral cerebellar hemispheres showed increased variability in the timing component of the tapping task, in contrast to patients with damage to medial cerebellar regions who showed increased variability on the motor implementation component of the task (Ivry et al. 1988). The contribution of the lateral cerebellar hemispheres to timing has been corroborated by recent neuroimaging studies examining reproduction of rhythmic sequences (Penhune & Doyon 2002, Penhune et al. 1998) and perceptual monitoring of auditory and visual rhythmic sequences (Schubotz et al. 2000). These studies converge on the conclusion that a supramodal cerebellar timing system is involved in processing temporally organized events. Other fMRI studies have produced evidence for the possible involvement of the basal ganglia in both motor and perceptual timing (Harrington et al. 1998; Rao et al. 1997, 2001). Finally, several studies have pointed to the involvement of motor cortical areas in rhythm perception and production, including the supplementary motor area, premotor cortex, and parietal cortex (Halsband et al. 1993).

Memory Adding to the complexity—and interest—of studying the neural correlates of musical processing is the fact that music, like all sounds, unfolds over time. Thus, the auditory cognitive system must depend to a large degree on mechanisms that allow a stimulus to be maintained on-line to be able to relate one element in a sequence to another that occurs later. These working memory mechanisms apply broadly to many types of processes. As applied to pitch processing, cognitive studies of working memory suggest that there may be dissociable systems for maintenance of pitch information over short periods as compared to speech information (Deutsch 1970, Semal et al. 1996). Similarly, lesion studies focusing on working memory

18 Nov 2004 10:54

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

96

AR

PERETZ

AR231-PS56-04.tex




AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

P1: IKH

ZATORRE

for pitch materials have implicated the right auditory cortex (Zatorre & Samson 1991), which is not surprising because a disruption in perceptual processing would likely lead to difficulty in maintaining a perceptual trace over a length of time. In addition to the participation of auditory cortex to working memory, both lesion and neuroimaging studies have found a role for frontal cortical areas (Gaab et al. 2003, Holcomb et al. 1998, Zatorre et al. 1994). In particular, dorsolateral and inferior frontal areas are most often recruited when working memory load is high (Griffiths et al. 1999, Zatorre et al. 1994). These studies largely converge on the idea that working memory for tones engages interactions between frontal cortical and posterior temporal areas, as is true for other domains. In general, auditory cortical regions are more strongly recruited during high-load conditions that require active rehearsal (Gaab et al. 2003, Zatorre et al. 1994), such as retaining a pitch in memory while other tones are presented, and may also be related to auditory-motor interactions (Hickok et al. 2003). These findings fit within a broader context in which working memory for pitch may be seen as a specialized subsystem within the framework of general working memory (Marin & Perry 1999). The contribution of memory to music processing is crucial not only because music unfolds over long periods of time but also because music is highly structured along multiple principles that require the contribution of different sources of knowledge. In the present review, however, we limit our attention to the memory component that enables recognition and mental representation (imagery) of a familiar tune. To enable recognition of a given tune, melodic and time relations must be mapped onto a stored long-term representation that contains invariant properties of the musical selection. As for words in language, the process of music recognition requires access and selection of potential candidates in a perceptual memory system (Dalla Bella et al. 2003). This musical memory is a perceptual representation system that is conceived as representing information about the form and structure of events, and not the meaning or other associative properties. Music is by essence perceptually driven. Unlike speech, music is not associated with a fixed semantic system, although it may convey meaning through other systems, such as emotional analysis (reviewed below) and associative memories (to retrieve contextual information, such as the title of a piece, the name of the singer, or its genre). Associative memories are probably the locus of the semantic priming effects recently observed between music and words with event-related brain potentials (Koelsch et al. 2004). Perceptual memories of familiar tunes must be relatively abstract in order to allow recognition despite transposition to a different register (Dowling & Fujitani 1971), change in instrumentation (Radvansky et al. 1995), and change in tempo (Warren et al. 1991). The stored representations can nonetheless preserve some surface features, such as absolute pitch and precise tempo (Halpern 1988, 1989; Levitin 1994; Levitin & Cook 1996). This duality between abstract and surface representations is congruent with our intuitions as to the role of memory in music listening. On the one hand, most listeners will not remember every detail of a musical segment but instead will follow the piece by a process of abstraction

18 Nov 2004 10:54

AR

AR231-PS56-04.tex

AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN ORGANIZATION FOR MUSIC PROCESSING

P1: IKH

97

and organization, remembering its “gist” (Dowling & Harwood 1986, Large et al. 1995). On the other hand, appreciation of interpretation requires the consideration of surface characteristics that are unique to a particular rendition (Raffman 1993). Thus, both surface and structural features may be contained in the stored representations and fit with the role and definition of the perceptual representation systems that are posited in other perceptual domains. Neuropsychological support for a perceptually based memory system for music derives mainly from lesion studies. In effect, persistent loss of recognition abilities for music can occur despite normal or near-normal perceptual processing of musical input (Eustache et al. 1990, Peretz 1996). Above all, such memory loss can be limited to music. For instance, an amusic patient with bilateral damage to the auditory cortex was normal at recognizing and memorizing spoken lyrics, whereas she performed at chance when required to recognize or to relearn the corresponding melody (played without lyrics). The deficit was selective because the patient had no difficulties with other nonmusical auditory materials, such as voices and animal cries, and had no memory impairment for visual stimuli (Peretz 1996). Hence, the memory failures were not only modality-specific but also musicspecific. A milder dissociation between melodies and speech sounds has also been reported in patients with focal lesions of the medial temporal lobe (Samson & Zatorre 1992). A lesion to either medial temporal region led to initial difficulties in learning the melodies; after right-sided lesions, retention of melodies was affected more severely and selectively over time. The right temporal structures appear less critically involved when recognition of highly familiar tunes is considered. Difficulties in recognizing familiar melodies tend to occur after a surgery to either superior temporal region (Ayotte et al. 2000). Moreover, the participation of left inferior temporal and frontal areas for recognizing familiar music has been pointed out in neuroimaging studies (Platel et al. 1997, 2003). These findings do not imply that familiar tunes are processed differently from novel melodies. Familiar melodies are associated to a series of extramusical and extraexperimental events that may contribute to recognition. For instance, song melodies (played without lyrics) automatically trigger the lyrics with which they are typically paired (Peretz et al. 2004b). The presence of these associate memories may even confer an advantage in the case of lesion. As uncovered by Steinke et al. (2001), brain damage can impair recognition of instrumental music but spare recognition of song melodies. The principles along which the stored representations of music can be impaired, when partially damaged, are potentially instructive with respect to the structure of their internal organization. Such degradation remains, however, to be observed and systematically studied. One way to probe the nature of the representations that are stored in memory is to study musical imagery. Imagery refers here to the subjective experience of being able to imagine music or musical attributes in the absence of real sound input. By applying behavioral methods developed by Halpern (1988) to patients with focal auditory cortex lesions, it was possible to demonstrate that perceptual deficits and imagery deficits are found in common, following damage to right auditory

18 Nov 2004 10:54

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

98

AR

PERETZ

AR231-PS56-04.tex




AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

P1: IKH

ZATORRE

cortical areas (Zatorre & Halpern 1993), suggesting that imagery requires access to perceptual mechanisms that are involved in processing melodies. This general conclusion has been supported in subsequent work using functional neuroimaging, which consistently indicates that secondary auditory cortices are recruited during a variety of tasks that require imagery or rehearsal of melodies (Halpern & Zatorre 1999, Zatorre et al. 1996), including sequences of tones (Penhune et al. 1998, Rao et al. 1997, Yoo et al. 2001) and isolated tones (Halpern et al. 2004). Further evidence comes from electrophysiological measures showing that the scalp topography elicited by imagining the continuation of a melody is similar to the electrical activity elicited by a real tone (Janata 2001). These demonstrations of auditory cortex activity in the absence of an acoustical stimulus, or at least not driven solely by external input, support the contention that perceptual mechanisms within auditory areas are responsible for the subjective experience of imagery. Retrieval processes from long-term representations, such as might occur when generating a familiar tune, tend to engage inferior frontal regions (Halpern & Zatorre 1999, Zatorre et al. 1996) in accord with many studies showing the importance of these mechanisms for memory retrieval in general (Nyberg & Tulving 1996). Additional neural mechanisms related to motor processes may also be involved in musical imagery. In particular, the supplementary motor area has often been identified in musical imagery studies (Halpern & Zatorre 1999, Halpern et al. 2004, Zatorre et al. 1996), which may relate to subvocalization and motor imagery (Lotze et al. 1999).

Emotion Music experience is not limited to perception and memory. Music experience is intimately related to its emotional appeal (Juslin & Sloboda 2001). Although the study of music as a means to express and induce emotions is a recent endeavor in neurosciences, a few studies (Blood et al. 1999, Peretz et al. 1998, Schmidt & Trainor 2001) have already highlighted important facts about music, emotions, and the brain. More specifically, these studies have suggested that the system for the analysis of emotional expression is neurally isolable (Peretz & Gagnon 1999, Peretz et al. 1998). Recognition of the emotional tone in music can be spared by brain damage while recognition of music identity is impaired (Peretz & Gagnon 1999), indicating that emotion recognition may rely on perceptual determinants that play little role in identity recognition. For example, the mode (major or minor) in which the music is written and the tempo (fast or slow) at which it is played can convey the happy-sad distinction (e.g., Peretz et al. 1998). In contrast, mode and tempo are not perceptually relevant for recognition; a piece of music can be easily recognized despite changes in mode and tempo (Halpern et al. 1998). Thus, the analysis of emotion expression may take as input emotion-specific musical features, as derived from pitch and time relation computations. In other words, emotional analysis could be mediated by a common (perceptual) cortical relay, suggesting no direct access to subcortical, limbic structures (Peretz et al. 2001). It would make sense for affective responses elicited by music to be cortically mediated, since there is a large cultural learning component to musically induced

18 Nov 2004 10:54

AR

AR231-PS56-04.tex

AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN ORGANIZATION FOR MUSIC PROCESSING

P1: IKH

99

emotion. Yet it is possible to elicit consistent affective responses, albeit not necessarily to the same stimuli across people. One such response that has been studied with neuroimaging techniques is the “chills” effect, which many people report as a particularly pleasant or even euphoric musical experience. This effect has been documented behaviorally (Panksepp 1995) and can be elicited relatively reliably. Moreover, this and other emotions are associated with objective physiological markers (Goldstein 1980, Krumhansl 1997). Blood & Zatorre (2001) reported that while people experienced musical chills, cerebral blood flow changes occurred in several brain areas, including the dorsal midbrain, ventral striatum (which contains the nucleus accumbens), insula, and orbitofrontal cortex. Some of these regions have previously been implicated in response to highly rewarding or motivationally important stimuli, including food (Small 2001) and drugs of abuse (Breiter et al. 1997). Thus, under certain circumstances, music can access neural substrates that are primarily associated with biologically significant stimuli. Whether music is unique in this respect remains to be seen; it may be one of a class of human constructs that elicit pleasure by co-opting ancient neural systems via inputs from neocortex. In this respect, music may serve as an excellent paradigm to explore the interactions between neocortically mediated cognitive processes and subcortically mediated affective responses.

MUSIC PERFORMANCE Music performance includes a variety of tasks, such as playing or singing welllearned pieces from memory, sight-reading, and improvisation. They all combine rapid motor skills and relatively elaborate cognitive operations in addition to the perceptual, memory, and emotion components described above. In order to account for production, one can add to this core system a visual-to-musical conversion subsystem for sight-reading (Cappelletti et al. 2000) and a motor planning component for singing and for playing (Peretz & Coltheart 2003). Such a tight coupling between input and output processes is reflected in the involuntary motor activity observed in pianists as they listen to well-trained music (Haueisen 2001), which may reflect a so-called mirror system important for imitation and learning (Rizzolatti & Arbib 1998). However, there is very little neuropsychological research bearing on this issue. Part of this situation is due to the limited number of proficient musicians who have sustained focal brain damage and to methodological difficulties in studying motor behavior with neuroimaging techniques. Yet, singing is not the privilege of musicians. Singing is the most widespread mode of musical expression; hence, it has the potential to shed light on how the brain orchestrates music production in general.

Singing Infants spontaneously sing around the age of one. By the age of five, children have a large repertoire of songs of their own culture, and they display singing

18 Nov 2004 10:54

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

100

AR

PERETZ

AR231-PS56-04.tex




AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

P1: IKH

ZATORRE

abilities that will remain qualitatively unchanged in adulthood (Dowling 1999). Thus, even without much practice, the ordinary adult seems to be endowed with the basic abilities that are necessary to sing simple songs. This conclusion has a long history in clinical neurology because researchers frequently observe nonfluent aphasics who can hardly speak intelligibly but who can produce recognizable songs (Yamadori et al. 1977). Similarly, transcranial magnetic stimulation to the left frontal cortex can temporarily produce speech arrest without interfering with singing in normal subjects (Stewart et al. 2001). These observations point to a neural dissociation between speech and music performance that is supported by the detailed analysis of aphasics’ production errors. H´ebert and colleagues (H´ebert et al. 2003, Peretz et al. 2004a) have shown that patients produce as few intelligible words in speaking as in singing, when tested in comparable conditions. Hence, the results indicate that verbal production, be it sung or spoken, is mediated by the same (impaired) language output system and that this speech route is distinct from the (spared) melodic output system. Conversely, amusic individuals can be normal at speaking while being unable to sing (Ayotte et al. 2002, Peretz et al. 1994). This distinction between singing and speaking is consistent with recent PET studies (Jeffries et al. 2003, Perry et al. 1999) that observed relative increases in activity during singing (versus speaking or listening) in bilateral motor structures, with predominance in the right hemisphere, particularly for premotor, insular, and auditory regions. Thus, in the vocal mode of expression, music can be dissociated from speech, as is also true in perception (see Music Specificity below). Similarly, as in perception studies, it has been found that singing pitch relations can be maintained while temporal relations are lost and vice versa; the melody can be impaired while the rhythm is intact (Alcock et al. 2000). Loss of melody, in both perception and production, is associated with a right-hemisphere lesion, whereas the loss of rhythm, in both perception and production, is related to a lesion of left-hemispheric structures. These findings are consistent with the notion that the melodic and temporal processing subsystems can function relatively independently of one another in music processing and that these support both perception and production.

Music Playing The study of musical playing can also yield insights into the mental representations used to plan the execution of music. Generally, a planned sequence is segmented in small units such as phrases. With practice, these units become larger, future events are more strongly anticipated, and expressive features of the performance are enhanced. The performer’s task is to highlight the structure of the musical piece and its emotional content through the complex programming of finely coordinated motor movements (Gabrielson 1999, Palmer 1997). Unfortunately, the contribution of cognitive neuroscience to the understanding of musical playing skills lags behind its cognitive study. Lesion studies are either anecdotal or sketchy regarding the nature of the expressive disorder (McFarland & Fortin 1982). Several neuroimaging

18 Nov 2004 10:54

AR

AR231-PS56-04.tex

AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

BRAIN ORGANIZATION FOR MUSIC PROCESSING

P1: IKH

101

studies (Kristeva et al. 2003, Lotze et al. 2003) have attempted to identify the neural correlates of musical performance but this has proven to be a difficult task due to its complex nature. Some of these findings highlight brain activity in certain areas (such as bilateral frontal opercular regions) when both executing and imagining music, which is consistent with the perceptual imagery research reviewed above.

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Sight-Reading The cognitive neuroscience of sight-reading is more informative. It has been known for some time that sight-reading of musical notation is a distinct ability because it is neurally and functionally separable from reading words and Arabic numerals. Brain damage can selectively impair reading of musical notation while sparing reading of other kinds of symbols (Brust 1980, Cappelletti et al. 2000, Horikoshi et al. 1997). Moreover, the musical reading disorder can be observed in relative isolation because playing, singing, and musical memory can be well preserved (e.g., Cappelletti et al. 2000). The reverse situation has also been described. Brain damage can impair the ability to read letters and words while sparing music reading (Signoret et al. 1987). The lesions responsible for music alexia are located in the left hemispheric structures (Midorikawa et al. 2003, Sch¨on et al. 2001). However, results of a recent fMRI study indicate the right occipital-temporal region is critically involved in deciphering pitch notation on a keyboard (Sch¨on et al. 2002). The fact that sight-reading of musical notation has been examined as a unitary function might explain in part this conflicting information regarding the neural correlates of music reading. Music sight-reading calls for the simultaneous and sequential processing of a vast amount of information in a very brief time for immediate use. This task requires, at the very least, interpretation of the pitch and duration of the notes (written on the two staves of a piano score) in the context of the prespecified key signature and meter, detection of familiar patterns, anticipation of what the music should sound like, and generation of a performance plan suited for motor translation. This sketchy componential task analysis illustrates the number of operations that are involved in music sight-reading and that in principle can be distinguished neuropsychologically. The need for further fractionation of sight-reading skills finds support in the recurrent observation that after brain damage pitch and time relations are dissociable in reading musical notation, as seen in both perception and singing. Some patients are still able to convert printed information into a rhythmical pattern but are no longer able to decipher pitch-related information (see, e.g., Brust 1980, Fasanaro et al. 1990) and vice versa: Pitch reading can be preserved while rhythm reading is impaired (Midorikawa et al. 2003). This dissociation needs to be studied in more detail because it may arise as a result of damage at different levels in the decoding of musical notation. It can occur as early as when the music score is visually inspected because pitch and duration are represented differently in print (duration is determined by the shape of the note itself, whereas pitch is indicated by the spatial position on the stave). This dissociation may also result from a difficulty occurring

18 Nov 2004 10:54

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

102

AR

PERETZ

AR231-PS56-04.tex




AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

P1: IKH

ZATORRE

at a later stage, when the reader imagines what the written code sounds like (Sch¨on & Besson 2002). Moreover, there are different routes to achieve a proper interpretation of the pitch code. One route involves naming, which is learned by association between a spatial location on the stave and a name (Do, r´e, mi or A, B, C; see below). Another involves some form of procedural memory, which is the consequence of learning to associate a notated pitch to a motor gesture (key press). However, most musicians are able to sight-read the same score in different modalities by singing and playing different musical instruments, and hence are likely to use a more abstract, modality-independent representation system of written scores. In sum, there are different ways to read musical notation. Moreover, this is a skill that is highly automated and specialized compared to what novices may learn (Stewart et al. 2003), an issue we discuss next.

Training Even if everyone engages in some sort of musical activity in everyday life, it is generally limited in time and effort. In contrast, a few individuals become proficient musicians through extensive practice from an early age and often with the help of explicit tutoring. The fact that musical training is not uniformly and systematically imposed in current educational curricula makes this natural variety of musically acquired skills a formidable laboratory in which to study the effects of training on brain functioning. Musicians represent a unique model in which to study plastic changes in the human brain (M¨unte et al. 2002). As suggested by animal research, experience can shape the size of cortical networks either by expansion or by reduction, depending on stimuli and on the structural levels examined (i.e., synaptic or macroscopic; see M¨unte et al. 2002 for a review). Hence, one would expect to find evidence of size differences in certain regions of the musician’s brain compared to the brain of an untrained person. The prime areas to look for differences are the motor areas. Indeed, there is clear evidence that the motor cortex of musicians is enhanced structurally (Gaser & Schlaug 2003) and functionally (Krings et al. 2000). Anatomical changes also have been seen in other motor-related structures, including cerebellum and corpus callosum (Schlaug 2003). In a seminal study, Elbert et al. (1995) investigated somatosensory-evoked magnetic fields in string players. Source analysis revealed that the cortical representation of the digits of the left hand (the fingering hand, especially for its fifth digit) was larger in musicians than in nonmusicians. In the case of the right hand, in which no independent movements of the fingers are required in string players, no differences were found between musicians and nonmusicians. Moreover, the cortical reorganization of the representation of the fingers was more pronounced in musicians who had begun their musical training at an early age. However, an investigation of sequential finger movements with fMRI found a reduction rather than an increase in activity in motor areas when comparing professional pianists to nonmusicians (Hund-Georgiadis & von Cramon 1999).

18 Nov 2004 10:54

AR

AR231-PS56-04.tex

AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN ORGANIZATION FOR MUSIC PROCESSING

P1: IKH

103

In general, there is no doubt that a complex interplay exists between structural changes that may accompany prolonged behavioral performance and neural responses that underlie that performance. A greater volume of tissue may reflect a reorganization, which in turn may manifest itself as recruitment of fewer neurons, or differently synchronized firing patterns, or different effective connectivity with other regions under different circumstances. Each of these possibilities would have quite different consequences for functional measures; different functional measures are sensitive to different parameters of neuronal function. Thus, we are far from understanding in detail the nature of the reorganization associated with musical training. Yet, the study of musical training effects is a unique paradigm to achieve this understanding. A few fascinating training effects are worth reporting in some detail. One of these is the fact that functional and morphological brain changes as a function of musical training are not limited to motor control. Researchers have identified several auditory brain areas that differ between musicians and nonmusicians. For example, Pantev and colleagues (Pantev et al. 1989), using MEG, have shown that brain responses to piano tones were 25% larger in musicians than in nonmusicians. This effect appears to be more pronounced for tones from each musician’s own type of instrument (Pantev et al. 2003, Tervaniemi 2003), strongly implying a use-dependent plasticity phenomenon. However, using a similar MEG technique, Schneider and collaborators (Schneider et al. 2002) found that both the early activity evoked by pure tones and the gray matter volume of the anteromedial portion of Heschl’s gyrus were more than 100% larger in professional musicians than in nonmusicians. Pure tones do not exist in the environment and hence may not account for the observed plasticity effects. In fact, the functional and the morphological differences were related to musical aptitude, suggesting influence of innate determinants. These findings reopen the debate about whether the observed brain differences between musicians and nonmusicians arise from genetic or other predispositions (or talent) as well as from practice and experience. Absolute pitch (AP) is another domain in which the issues of training or exposure, development, and innate predispositions arise (Takeuchi & Hulse 1993, Ward 1999, Zatorre 2003). The principal feature that distinguishes an AP possessor from anyone else is that he or she can identify any pitch by its musical note name without reference to any other sound. In other words, with AP, one has a fixed associative memory for a large number of pitch categories (as many as 70, according to Ward 1999); nonpossessors of AP—nonmusicians and most musicians—also have fixed categories available (Levitin 1994), but the precision is an order of magnitude lower. Because AP is not a universally expressed trait, other predispositions must be at play. One possibility is that genetic factors may be involved. The evidence for this idea is growing (Baharloo et al. 2000, Gregersen et al. 2000). The sibling recurrence rate, for example, is quite high (on the order of 8% to 15%). Thus, AP likely depends on some innate neural substrate (genetically determined or not) in interaction with a certain environmental input at a certain time during neural development.

18 Nov 2004 10:54

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

104

AR

PERETZ

AR231-PS56-04.tex




AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

P1: IKH

ZATORRE

Several studies have probed the differences in patterns of brain activity in AP possessors and musicians who do not have AP. One of the earliest demonstrations reported that AP listeners showed an absent or reduced electrical-evoked potential component thought to index updating of working memory (Klein et al. 1984, Wayman et al. 1992). Thus, a typical listener will show an electrophysiological response to a pitch change, indicating that some type of “on-line” memory system has been refreshed, but AP possessors do not show this effect. A related phenomenon is that an area of the right frontal cortex believed to be important for monitoring pitch information in working memory is more active in musicians lacking AP than in those who have AP (Zatorre et al. 1998). Thus, instead of requiring continuous maintenance of a sensory trace, AP possessors appear to use their categorical representation of tones in such tasks. The posterior dorsolateral frontal cortex also responds differentially in AP possessors as compared to control musicians (Zatorre et al. 1998). This area, concerned with establishing and maintaining conditional associations in memory (Petrides 1995), is a logical candidate area for the link between a pitch and its label in AP subjects. These findings help to demystify AP somewhat, by demonstrating that working memory and associative learning aspects of AP draw on well-understood neural resources. However, this still does not explain why some people are able to form the pitch categories in the first place and others are not. Part of the answer to this issue has been proposed to lie in the function of subcortical nuclei that are important for periodicity coding, such as the inferior colliculus (Langner et al. 2002) The presence of functional differences between AP and non-AP people leads to the question of whether brain anatomy might not also differ. Several studies have now shown that there are indeed significant differences in the degree of lateral asymmetry of auditory cortical areas (Schlaug et al. 1995, Zatorre et al. 1998). However, it remains uncertain how to interpret this effect: Whereas initial indications were that AP subjects showed relative enhancement of left-hemisphere structures, the latest evidence suggests that instead it may be a relative reduction in volume of right-hemisphere structures that accounts for the difference (Keenan et al. 2001). A simple explanation based on size may not be sufficient because there are likely complex interactions between gross morphological features and the underlying cortical function. One way to study the neural correlates of musical training while controlling for complex interactions between innate and environmental factors is to train novices to learn to play music. For example, training on the piano for two hours a day over five days leads to increased excitability of the cortical motor areas that control the contralateral finger muscles of the hand, as measured by applying transcranial magnetic stimulation. Interestingly, mental practice of the motor tasks leads to similar but less-pronounced effects (Pascual-Leone 2001). This type of research allows control for the short- and long-term effects of training on the nature of the plastic changes in the brain. However, this sort of study should be undertaken longitudinally to explore the interaction between plasticity and brain development.

18 Nov 2004 10:54

AR

AR231-PS56-04.tex

AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN ORGANIZATION FOR MUSIC PROCESSING

P1: IKH

105

All of these studies demonstrate quantitative neural changes associated with musical training. Musicians appear to recruit more neural tissue or to use it more efficiently than do nonmusicians. An important question is to what extent training also modifies the way music is processed and organized in the brain. For instance, it has been widely argued that musicians might be more prone to adopt an analytical mode of processing music (Bever & Chiarello 1974). However, the use of such a strategy is not exclusive to musicians. As seen earlier, nonmusicians are able to consider local interval sizes when required by the task (Peretz & Morais 1987), and they do so automatically (Trainor et al. 2002). Moreover, musicians’ approach of musical material is not confined to an analytic mode; they can flexibly use either contour or interval information, depending on the structure of the stimuli (Peretz & Babai 1992). More generally, when musical abilities are assessed indirectly, more similarities than differences are noted between musicians and nonmusicians (Bigand 2003).

MUSIC SPECIFICITY As we have seen, a vast network of regions located in both the left and right hemispheres of the brain, with an overall right-sided asymmetry for pitch-based processing, is recruited by musical activities. This should not come as a surprise since musical activities are numerous, complex, and diverse (Tramo 2001). However, finding such a wide distributed network of brain regions raises the issue of which of these brain areas is dedicated to music processing. Some of these regions might not only overlap but might also share processing components with other functions, such as those involved in language. This issue of music and language specificity has a long history in neurology (Henschen 1924) and has been recently reopened by the observation that the detection of harmonic deviations activates Broca’s area (Koelsch et al. 2002). This research suggests that the mechanisms underlying syntactic processing are shared between music and language (Levitin & Menon 2003, Patel 2003). However, there are problems with this conclusion that we address in the present review, because the question of domain-specificity is important from both an evolutionary (Hauser et al. 2002) and an empirical perspective. First, we should keep in mind that Broca’s area is a vast brain region that can easily accommodate more than one distinct processing network (Marcus et al. 2003). Second, the degree of anatomical proximity of activation maxima that should count as reactivation of the same region in another task is not straightforward, particularly when activation is not limited to Broca’s area but involves the right hemisphere homologous region (Maess et al. 2001) and when the cognitive domains (music and language) are studied separately in different laboratories. Clearly, further comparison is needed between music and language in the same experimental setting, using similar tasks that are matched (or manipulated) for attentional resources (Shallice 2003).

18 Nov 2004 10:54

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

106

AR

PERETZ

AR231-PS56-04.tex




AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

P1: IKH

ZATORRE

Indeed, comparisons across domains have generally been more consistent with the idea that musical abilities are subserved, at least in part, by neural networks that are dedicated to their respective processing. As seen previously, singing, memory, and sight-reading are all musical activities that are functionally and neuroanatomically dissociable from analogous activities that involve speech. The evidence rests on both lesion and neuroimaging studies. The strongest support for the existence of processing components whose operation is specific to the processing of music comes from the study of auditory disorders. Selective impairments and sparing of music recognition abilities can occur after brain damage. As recently reviewed in Peretz & Coltheart (2003), patients may suffer from recognition failures that affect the musical domain exclusively. Such patients can no longer recognize melodies (presented without words) that were highly familiar to them prior to the onset of their brain damage. In contrast, they are normal at recognizing spoken lyrics (and spoken words, in general), familiar voices, and other environmental sounds (such as animal cries). Similarly, there are individuals who suffer from lifelong difficulties with music but can recognize the lyrics of familiar songs even though they are unable to recognize the accompanying tune. Conversely, one can find people in whom brain damage has impaired the ability to recognize spoken words while sparing the ability to recognize music. Hence, the evidence points to the existence of distinct processing modules for music and speech. Music specificity may concern different processing components. Some components might be perceptual prerequisites for music processing (Peretz & Hyde 2003), while some processing components appear to be genuinely specialized for music (Peretz & Morais 1989). Still other components conceivably could be involved in the processing not only of music but also of speech (Patel et al. 1998). One important question for the future is to determine which other processing components are uniquely involved in music and which components are not. This would provide clues as to the roots of brain specialization for music processing and hence the roots of musicality in general.

CONCLUSIONS The work presented in this review constitutes a selection of what currently is known (and is not known) about how musical functions are organized in the brain. As pointed out, a number of musical processes have not yet been considered in neuropsychology. Yet, we trust that the accumulated knowledge on how and where short musical phrases are perceived, recognized, imagined, appreciated, and sung will provide the building blocks of a broader and more complete neurofunctional account of music processing. One such fundamental principle concerns the distinction between pitch-based and time-based mechanisms. As seen in both perception and performance, the musical processing of pitch and rhythm appears to result from the operation of largely distinct neural mechanisms. Extraction of musical pitch relations seems to depend on a series of operations that predominantly involve the right auditory

18 Nov 2004 10:54

AR

AR231-PS56-04.tex

AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN ORGANIZATION FOR MUSIC PROCESSING

P1: IKH

107

cortex, whereas extraction of musical time relations recruits more widespread and bilateral neural networks. Obviously, music processing cannot be ascribed wholly to one cerebral hemisphere. However, as we have argued elsewhere (Peretz & Hyde 2003, Zatorre et al. 2002), it might be the case that right-sided neural mechanisms play a determinant role. Pitch-based mechanisms might be rooted in the specialization of the right auditory cortex for spectral resolution (Zatorre et al. 2002) and be instrumental in the normal development of musical competence (Peretz et al. 2002). These conclusions also assume that music processing is mapped onto the human brain with a certain consistency. If there were no consistency, understanding the relations between music and its neural correlates would be impossible. However, as discussed above, people differ in their musical training or experience, and perhaps also in aptitude or talent; these differences manifest themselves at various levels, and understanding the interplay between brain processes, environmental inputs, innate factors, and development will be one of the major challenges for our field. It should also be considered that the mapping between processing components and neural networks might not be one-to-one. The fact that the processing framework can account for the musical disorders exhibited by brain-damaged patients suggests that this mapping is feasible. Indeed, it is conceivable that some music processing components lack neuroanatomical separability. In that case, the neural substrates of the components would be intermingled with the neural systems devoted to the processing of other complex patterns. If so, brain damage could never affect just one processing component while sparing all other aspects. If the process under study possesses the property of neural separability, as we have described here for most processing components, then we expect to find corresponding isolable neural areas in the brain. Thus, we believe that the evidence is solid enough to envisage a proper mapping between processes and neural networks. Above all, we believe that convergence between lesion and neuroimaging data is the optimal strategy to build a sound model of how the brain processes music.

ACKNOWLEDGMENTS This chapter was written while both authors were supported by grants from the Canadian Institute of Health Research and the Natural Science and Engineering Research Council of Canada. The Annual Review of Psychology is online at http://psych.annualreviews.org

LITERATURE CITED Alcock KJ, Wade D, Anslow P, Passingham RE. 2000. Pitch and timing abilities in adult lefthemisphere-dysphasic and right-hemisphere subjects. Brain Lang. 75:47–65

Ayotte J, Peretz I, Rousseau I, Bard C, Bojanowski M. 2000. Patterns of music agnosia associated with middle cerebral artery infarcts. Brain 123:1926–38

18 Nov 2004 10:54

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

108

AR

PERETZ

AR231-PS56-04.tex




AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

P1: IKH

ZATORRE

Baharloo S, Service E, Risch N, Gitschier J, Freimer NB. 2000. Familial agregation of absolute pitch. Am. J. Hum. Genet. 67:755–58 Balaban MT, Anderson LM, Wisniewski AB. 1998. Lateral asymmetries in infant melody perception. Dev. Psychol. 34:39–48 Besson M, Fa¨ıta F. 1995. An event-related potential (ERP) study of musical expectancy: comparison of musicians with nonmusicians. J. Exp. Psychol.: Hum. Percept. Perform. 21:1278–96 Besson M, Fa¨ıta F, Peretz I, Bonnel A-M, Requin J. 1998. Singing in the brain: independence of lyrics and tunes. Psychol. Sci. 9:494–98 Bever TG, Chiarello RJ. 1974. Cerebral dominance in musicians and nonmusicians. Science 185:537–39 Bigand E. 2003. More about the musical expertise of musically untrained listeners. Ann. NY Acad. Sci. 999:304–12 Blood AJ, Zatorre RJ. 2001. Intensely pleasurable responses to music correlates with activity in brain regions implicated in reward and emotion. Proc. Natl. Acad. Sci. USA 98:11818–23 Blood AJ, Zatorre RJ, Bermudez P, Evans AC. 1999. Emotional responses to pleasant and unpleasant music correlate with activity in paralimbic brain regions. Nat. Neurosci. 2:382–87 Breiter HC, Gollub RL, Weisskoff RM, Kennedy D, Makris N, et al. 1997. Acute effects of cocaine on human brain activity. Neuron 19:591–611 Brust JC. 1980. Music and language: musical alexia and agraphia. Brain 103:367–92 Cappelletti M, Waley-Cohen H, Butterworth B, Kopelman M. 2000. A selective loss of the ability to read and write music. Neurocase 6:321–32 Critchley M, Henson R, eds. 1977. Music and the Brain. Studies in the Neurology of Music. London: Heinemann Dalla Bella S, Peretz I, Aronoff N. 2003. Time course of melody recognition: a gating paradigm study. Percept. Psychophys. 65:1019–28

Deutsch D. 1970. Tones and numbers: specificity of interference in immediate memory. Science 168:1604–5 Deutsch D. 1999. The Psychology of Music. San Diego: Academic Di Pietro M, Laganaro M, Leeman B, Schnider A. 2004. Receptive amusia: temporal auditory deficit in a professional musician following a left temporo-parietal lesion. Neuropsychologia 42:868–977 Dowling WJ. 1999. The development of music perception and cognition. See Deutsch 1999, pp. 603–25 Dowling WJ, Fujitani DS. 1971. Contour, interval, and pitch recognition in memory for melodies. J. Acoust. Soc. Am. 49:524–31 Dowling WJ, Harwood D. 1986. Music Cognition. New York: Academic Drake C. 1998. Psychological processes involved in the temporal organization of complex auditory sequences: universal and acquired processes. Music Percept. 16:11– 26 Elbert T, Pantev C, Wienbruch C, Rockstroh B, Taub E. 1995. Increased cortical representation of the fingers of the left hand in string players. Science 270:305–7 Eustache F, Lechevalier B, Viader F, Lambert J. 1990. Identification and discrimination disorders in auditory perception: a report on two cases. Neuropsychologia 28:257–70 Fasanaro AM, Spitaleri DLA, Valiani R. 1990. Dissociation in musical reading: a musician affected by alexia without agraphia. Music Percept. 7:259–72 Fishman YI, Volkov IO, Noh MD, Garell PC, Bakken H, et al. 2001. Consonance and dissonance of musical chords: neural correlates in auditory cortex of monkeys and humans. J. Neurophysiol. 86:2761–88 Fraisse P. 1982. Rhythm and tempo. See Deutsch 1999, pp. 149–80 Fran¸ce` s R, Lhermitte F, Verdy M. 1973. Le d´eficit musical des aphasiques. Rev. Int. Psychol. Appl. 22:117–35 Fries W, Swihart AA. 1990. Disturbance of rhythm sense following right hemisphere damage. Neuropsychologia 28:1317–23

18 Nov 2004 10:54

AR

AR231-PS56-04.tex

AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN ORGANIZATION FOR MUSIC PROCESSING Gaab N, Gaser C, Zaehle T, J¨ancke L, Schlaug G. 2003. Functional anatomy of pitch memory—an fMRI study with sparse temporal sampling. NeuroImage 19:1417–26 Gabrielson A. 1999. The performance of music. See Deutsch 1999, pp. 501–602 Gaser C, Schlaug G. 2003. Brain structures differ between musicians and non-musicians. J. Neurosci. 23:9240–45 Goldstein A. 1980. Thrills in response to music and other stimuli. Physiol. Psychol. 8:126– 29 Gregersen PK, Kowalsky E, Kohn N, Marvin EW. 2000. Early childhood music education and predisposition to absolute pitch: teasing apart genes and environment. Am. J. Hum. Genet. 98:280–82 Griffiths TD, Beuchel C, Frackowiak RSJ, Patterson RD. 1998. Analysis of temporal structure in sound by the human brain. Nat. Neurosci. 1:422–27 Griffiths TD, Johnsrude I, Dean JL, Green GGR. 1999. A common neural substrate for the analysis of pitch and duration pattern in segmented sound? NeuroReport 10:3825–30 Hall DA, Johnsrude I, Haggard MP, Palmer AR, Akeroyd MA, Summerfield AQ. 2002. Spectral and temporal processing in human auditory cortex. Cereb. Cortex 12:140–49 Halpern AR. 1988. Mental scanning in auditory imagery for songs. J. Exp. Psychol.: Learn. Mem. Cogn. 14:434–43 Halpern AR. 1989. Memory for the absolute pitch of familiar songs. Mem. Cogn. 17:572– 81 Halpern AR, Bartlett D, Dowling W. 1998. Perception of mode, rhythm, and contour in unfamiliar melodies: effects of age and experience. Music Percept. 15:335–56 Halpern AR, Zatorre R. 1999. When that tune runs through your head: a PET investigation of auditory imagery for familiar melodies. Cereb. Cortex 9:697–704 Halpern AR, Zatorre R, Bouffard M, Johnson JA. 2004. Behavioral and neural correlates of perceived and imagined musical timbre. Neuropsychologia. In press Halsband U, Tanji J, Freund H-J. 1993. The

P1: IKH

109

role of premotor cortex and the supplementary motor area in the temporal control of movement in man. Brain 116:243–46 Harrington DL, Haaland K, Hermanowicz N. 1998. Temporal processing in the basal ganglia. Neuropsychology 12:3–12 Hart HC, Palmer AR, Hall DA. 2003. Amplitude and frequency-modulated stimuli activate common regions of human auditory cortex. Cereb. Cortex 13:773–81 Haueisen JK, Kn¨osche T. 2001. Involuntary motor activity in pianists evoked by music perception. J. Cogn. Neurosci. 13:786–92 Hauser M, Chomsky N, Fitch WT. 2002. The faculty of language: What is it, who has it, and how did it evolve? Science 298:1569–79 H´ebert S, Racette A, Gagnon L, Peretz I. 2003. Revisiting the dissociation between singing and speaking in expressive aphasia. Brain 126:1838–50 Henschen SE. 1924. On the function of the right hemisphere of the brain in relation to the left in speech, music and calculation. Brain 44:110–23 Hickok G, Buchsbaum B, Humphries C, Muftuler T. 2003. Auditory-motor interaction revealed by fMRI: speech, music, and working memory in area SPT. J. Cogn. Neurosci. 15:673–82 Holcomb H, Medoff D, Caudill P, Zhao Z, Lahti A, et al. 1998. Cerebral blood flow relationships associated with difficult tone recognition task in trained normal volunteers. Cereb. Cortex 8:534–42 Horikoshi T, Asari Y, Watanabe A, Nagaseki Y, Nukui H, et al. 1997. Music alexia in a patient with mild pure alexia: disturbed visual perception of nonverbal meaningful figures. Cortex 33:187–94 Hund-Georgiadis M, von Cramon DY. 1999. Motor-learning-related changes in piano players and non-musicians revealed by functional magnetic-resonance signals. Exp. Brain Res. 125:417–25 Ibbotson NR, Morton J. 1981. Rhythm and dominance. Cognition 9:125–38 Ivry RB, Keele S. 1989. Timing functions of the cerebellum. J. Cogn. Neurosci. 1:136–52

18 Nov 2004 10:54

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

110

AR

PERETZ

AR231-PS56-04.tex




AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

P1: IKH

ZATORRE

Ivry RB, Keele S, Diener H. 1988. Dissociation of the lateral and medial cerebellum in movement timing and movement execution. Exp. Brain Res. 73:167–80 Janata P. 2001. Brain electrical activity evoked by mental formation of auditory expectations and images. Brain Topogr. 13:169–93 Janata P, Birk J, Van Horn J, Leman M, Tillmann B, Bharucha J. 2002. The cortical topography of tonal structures underlying Western music. Science 298:2167–70 Janata P, Grafton S. 2003. Swinging in the brain: shared neural substrates for behaviors related to sequencing and music. Nat. Neurosci. 6:682–87 Jeffries KJ, Fritz JB, Braun AR. 2003. Words in melody: an H2 15O PET study of brain activation during singing and speaking. NeuroReport 14:749–54 Johnsrude I, Penhune V, Zatorre R. 2000. Functional specificity in the right auditory cortex for perceiving pitch direction. Brain 123:155–63 Jones MR, Boltz M. 1989. Dynamic attending and responses to time. Psychol. Rev. 96:459– 91 Juslin P, Sloboda J, eds. 2001. Music and Emotion: Theory and Research. Oxford: Oxford Univ. Press Justus T, Bharucha J. 2001. Modularity in musical processing: the automaticity of harmonic priming. J. Exp. Psychol.: Hum. Percept. Perform. 27:1000–11 Justus T, Bharucha J. 2002. Music perception and cognition. In Stevens Handbook of Experimental Psychology, Volume I: Sensation and Perception, ed. S Yantis, H Pasler, pp. 453–92. New York: Wiley Keenan J, Thangaraj V, Halpern A, Schlaug G. 2001. Absolute pitch and planum temporale. NeuroImage 14:1402–8 Klein M, Coles MGH, Donchin E. 1984. People with absolute pitch process tones without producing a P300. Science 223:1306– 9 Koelsch S, Grossmann T, Gunter T, Hahne A, Schr¨oger E, Friederici AD. 2003. Children processing music: electric brain responses re-

veal musical competence and gender differences. J. Cogn. Neurosci. 15:683–93 Koelsch S, Gunter TC, van Cramon DY, Zysset S, Lohmann G, Friederici AD. 2002. Bach speaks: a cortical “language-network” serves the processing of music. NeuroImage 17:956–66 Koelsch S, Kasper E, Sammler D, Schulze K, Gunter T, Friederici AD. 2004. Music, language and meaning: brain signatures of semantic processing. Nat. Neurosci. 7:302– 7 Krings T, T¨opper R, Foltys H, Erbich S, Sparing R, et al. 2000. Cortical activation patterns during complex motor tasks in piano players and control subjects. A functional magnetic resonance imaging study. Neurosci. Lett. 278:189–93 Kristeva R, Chakarov V, Schulte-M¨onting J, Spreer J. 2003. Activation of cortical areas in music execution and imagining: a highresolution EEG study. NeuroImage 20:1872– 83 Krumhansl CL. 1997. An exploratory study of musical emotions and psychophysiology. Can. J. Exp. Psychol. 51:336–52 Krumhansl CL. 2000. Rhythm and pitch in music cognition. Psychol. Bull. 126:159–79 Langner G, Albert M, Briede T. 2002. Temporal and spatial coding of periodicity information in the inferior colliculus of awake chinchilla (Chinchilla laniger). Hear. Res. 168:110–30 Large EW, Palmer C, Pollack JB. 1995. Reduced memory representations for music. Cogn. Sci. 19:53–96 Lerdahl F, Jackendoff R. 1983. A Generative Theory of Tonal Music. Cambridge, MA: MIT Press Levitin DJ. 1994. Absolute memory for musical pitch: evidence from the production of learned melodies. Percept. Psychophys. 56:414–23 Levitin DJ, Cook PR. 1996. Memory for musical tempo: additional evidence that auditory memory is absolute. Percept. Psychophys. 58:927–35 Levitin DJ, Menon V. 2003. Musical structure is processed in “language” areas of the brain:

18 Nov 2004 10:54

AR

AR231-PS56-04.tex

AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN ORGANIZATION FOR MUSIC PROCESSING a possible role for Brodmann area 47 in temporal coherence. Neuroimage 20:2142–52 Li´egeois-Chauvel C, Peretz I, Babai M, Laguitton V, Chauvel P. 1998. Contribution of different cortical areas in the temporal lobes to music processing. Brain 121:1853–67 Lotze M, Montoya P, Erb M, Hulsmann E, Flor H, et al. 1999. Activation of cortical and cerebellar motor areas during executed and imagined hand movements: an fMRI study. J. Cogn. Neurosci. 11:491–501 Lotze M, Scheler G, Tan H-R, Braun C, Birbaumer N. 2003. The musician’s brain: functional imaging of amateurs and professionals during performance and imagery. Neuroimage 20:1817–29 Maess B, Koelsch S, Gunter TC, Friederici AD. 2001. Musical syntax is processed in Broca’s area: an MEG study. Nat. Neurosci. 4:540–45 Marcus GF, Vouloumanos A, Sag IA. 2003. Does Broca’s play by the rules? Nat. Neurosci. 6:651–52 Marin OSM, Perry DW. 1999. Neurological aspects of music perception and performance. See Deutsch 1999, pp. 653–724 McFarland HR, Fortin D. 1982. Amusia due to right temporoparietal infarct. Arch. Neurol. 39:725–27 Midorikawa A, Kawamura M, Kezuka M. 2003. Musical alexia for rhythm notation: a discrepancy between pitch and rhythm. Neurocase 9:232–38 Milner B. 1962. Laterality effects in audition. In Interhemispheric Relations and Cerebral Dominance, ed. VB Mountcastle, pp. 177– 95. Baltimore, MD: Johns Hopkins Press M¨unte TF, Altenm¨uller E, J¨ancke L. 2002. The musician’s brain as a model of neuroplasticity. Nat. Rev. 3:473–78 Nyberg L, Tulving E. 1996. Classifying human long-term memory: evidence from converging dissociations. Eur. J. Cogn. Psychol. 8:163–83 Palmer C. 1997. Music performance. Annu. Rev. Psychol. 48:115–38 Panksepp J. 1995. The emotional sources of “chills” induced by music. Music Percept. 13:171–207

P1: IKH

111

Pantev C, Engelien A, Candia V, Elbert T. 2003. Representational cortex in musicians. See Peretz & Zatorre 2003, pp. 382–95 Pantev C, Hoke M, Lutkenhoner B, Lehnertz K. 1989. Tonotopic organization of the auditory cortex: pitch versus frequency representation. Science 246:486–88 Pascual-Leone A. 2001. The brain that plays music and is changed by it. See Peretz & Zatorre 2003, pp. 396–409 Patel A. 2003. Language, music, syntax and the brain. Nat. Neurosci. 6:674–81 Patel AD, Balaban E. 2000. Temporal patterns of human cortical activity reflect tone sequence structure. Nature 404:80–84 Patel AD, Peretz I, Tramo M, Labrecque R. 1998. Processing prosodic and musical patterns: a neuropsychological investigation. Brain Lang. 61:123–44 Patterson R, Uppenkamp S, Johnsrude I, Griffiths T. 2002. The processing of temporal pitch and melody information in auditory cortex. Neuron 36:767–76 Penhune V, Doyon J. 2002. Dynamic cortical and subcortical networks in learning and delayed recall of timed motor sequences. J. Neurosci. 22:1397–406 Penhune V, Zatorre R, Evans AC. 1998. Cerebellar contributions to motor timing: a PET study of auditory and visual rhythm reproduction. J. Cogn. Neurosci. 10:752–65 Penhune VB, Zatorre RJ, Feindel W. 1999. The role of auditory cortex in retention of rhythmic patterns in patients with temporal-lobe removals including Heschl’s gyrus. Neuropsychologia 37:315–31 Peretz I. 1990. Processing of local and global musical information by unilateral brain-damaged patients. Brain 113(Pt. 4):1185–205 Peretz I. 1993. Auditory atonalia for melodies. Cogn. Neuropsychol. 10:21–56 Peretz I. 1996. Can we lose memories for music? The case of music agnosia in a nonmusician. J. Cogn. Neurosci. 8:481–96 Peretz I. 2001. Music perception and recognition. In The Handbook of Cognitive Neuropsychology, ed. B Rapp, pp. 519–40. Hove, UK: Psychology Press

18 Nov 2004 10:54

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

112

AR

PERETZ

AR231-PS56-04.tex




AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

P1: IKH

ZATORRE

Peretz I, Ayotte J, Zatorre RJ, Mehler J, Ahad P, et al. 2002. Congenital amusia: a disorder of fine-grained pitch discrimination. Neuron 33:185–91 Peretz I, Babai M. 1992. The role of contour and intervals in the recognition of melody parts: evidence from cerebral asymmetries in musicians. Neuropsychologia 30:277–92 Peretz I, Blood AJ, Penhune V, Zatorre R. 2001. Cortical deafness to dissonance. Brain 124:928–40 Peretz I, Coltheart M. 2003. Modularity of music processing. Nat. Neurosci. 6:688–91 Peretz I, Gagnon L. 1999. Dissociation between recognition and emotional judgment for melodies. Neurocase 5:21–30 Peretz I, Gagnon L, Bouchard B. 1998. Music and emotion: perceptual determinants, immediacy, and isolation after brain damage. Cognition 68:111–41 Peretz I, Gagnon L, Macoir J, H´ebert S. 2004a. Singing in the brain: insights from cognitive neuropsychology. Music Percept. 21:373–90 Peretz I, Hyde K. 2003. What is specific to music processing? Insights from congenital amusia. Trends Cogn. Sci. 7:362–67 Peretz I, Kolinsky R. 1993. Boundaries of separability between melody and rhythm in music discrimination: a neuropsychological perspective. Q. J. Exp. Psychol. A 46:301–25 Peretz I, Kolinsky R, Tramo M, Labrecque R, Hublet C, et al. 1994. Functional dissociations following bilateral lesions of auditory cortex. Brain 117:1283–301 Peretz I, Morais J. 1987. Analytic processing in the classification of melodies as same or different. Neuropsychologia 25:645–52 Peretz I, Morais J. 1989. Music and modularity. Contemp. Music Rev. 4:277–91 Peretz I, Radeau M, Arguin M. 2004b. Two-way interactions between music and language: evidence from priming recognition of tune and lyrics in familiar songs. Mem. Cogn. 32:142– 52 Peretz I, Zatorre R. 2003. The Cognitive Neuroscience of Music. Oxford: Oxford Univ. Press Perry DW, Zatorre RJ, Petrides M, Alivisatos B, Meyer E, Evans AC. 1999. Localization

of cerebral activity during simple singing. Neuroreport 10:3979–84 Petrides M. 1995. Functional organization of the human frontal cortex for mnemonic processing: evidence from neuroimaging studies. Ann. NY Acad. Sci. 769:85–96 Piccirilli M, Sciarma T, Luzzi S. 2000. Modularity of music: evidence from a case of pure amusia. J. Neurol. Neurosurg. Psychiatry 69:541–45 Platel H, Baron JC, Desgranges B, Bernard F, Eustache F. 2003. Semantic and episodic memory for music are subserved by distinct neural networks. Neuroimage 20:244–56 Platel H, Price C, Baron JC, Wise R, Lambert J, et al. 1997. The structural components of music perception. A functional anatomical study. Brain 120:229–43 Plomp R, Levelt WJ. 1965. Tonal consonance and critical bandwidth. J. Acoust. Soc. Am. 38:548–60 Povel DJ, Essens P. 1985. Perception of temporal patterns. Music Percept. 2:411–40 Radvansky GA, Flemming KJ, Simmons JA. 1995. Timbre reliance in nonmusicians’ memory for melodies. Music Percept. 13: 127–40 Raffman D. 1993. Language, Music and Mind. Cambridge, MA: MIT Press Rao SM, Harrington DL, Haaland K, Bobholz JA, Cox RW, Binder J. 1997. Distributed neural systems underlying the timing of movements. J. Neurosci. 17:5528–35 Rao SM, Mayer AR, Harrington DL. 2001. The evolution of brain activation during temporal processing. Nat. Neurosci. 4:317–23 Regnault P, Bigand E, Besson M. 2001. Eventrelated brain potentials show top-down and bottom-up modulations of musical expectations. J. Cogn. Neurosci. 13:241–55 Sakai K, Hikosaka O, Miyauchi S, Takino R, Tamada T, et al. 1999. Neural representation of a rhythm depends on its interval ratio. J. Neurosci. 19:10074–81 Samson S, Zatorre RJ. 1992. Learning and retention of melodic and verbal information after unilateral temporal lobectomy. Neuropsychologia 30:815–26

18 Nov 2004 10:54

AR

AR231-PS56-04.tex

AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BRAIN ORGANIZATION FOR MUSIC PROCESSING Schellenberg E, Trehub SE. 1994. Frequency ratios and the perception of tone patterns. Psychon. Bull. Rev. 1:191–201 Schlaug G. 2003. The brain of musicians. See Peretz & Zatorre 2003, pp. 366–81 Schlaug G, Jancke L, Huang Y, Steinmetz H. 1995. In vivo evidence of structural brain asymmetry in musicians. Science 267:699– 701 Schmidt LA, Trainor LJ. 2001. Frontal brain electrical activity (EEG) distinguishes valence and intensity of musical emotions. Cogn. Emot. 15:487–500 Schneider P, Scherg M, Dosch G, Specht H, Gutschalk A. 2002. Morphology of Heschl’s gyrus reflects enhanced activation in the auditory cortex of musicians. Nat. Neurosci. 5:688–94 Sch¨on D, Anton J-L, Roth M, Besson M. 2002. An fMRI study of music sight-reading. NeuroReport 13:2285–89 Sch¨on D, Besson M. 2002. Processing pitch and duration in music reading: a RT-ERP study. Neuropsychologia 40:868–78 Sch¨on D, Semenza C, Denes G. 2001. Naming of musical notes: a selective deficit in one musical clef. Cortex 37:407–21 Schubotz RI, Friederici AD, von Cramon DY. 2000. Time perception and motor timing: a common cortical and subcortical basis revealed by fMRI. Neuroimage 11:1–12 Semal C, Demany L, Ueda K, Halle P-A. 1996. Speech versus nonspeech in pitch memory. J. Acoust. Soc. Am. 100:1132–40 Shallice T. 2003. Functional imaging and neuropsychology findings: How can they be linked? NeuroImage 20:S146–54 Signoret JL, van Eeckhout P, Poncet M, Castaigne P. 1987. Aphasia without amusia in a blind organist. Verbal alexia-agraphia without musical alexia-agraphia in braille. Rev. Neurol. (Paris) 143:172–81 Small DM, Zatorre R, Dagher A, Jones-Gotman M. 2001. Brain activity related to eating chocolate: from pleasure to aversion. Brain 124:1720–33 Steinke WR, Cuddy LL, Jakobson LS. 2001. Dissociations among functional subsystems

P1: IKH

113

governing melody recognition after righthemisphere damage. Cogn. Neuropsychol. 18:411–37 Stewart L, Henson R, Kamp K, Walsh V, Turner R, Frith U. 2003. Brain changes after learning to read and play music. Neuroimage 20:71– 83 Stewart L, Walsh V, Frith U, Rothwell J. 2001. Transcranial magnetic stimulation produces speech arrest but not song arrest. Ann. NY Acad. Sci. 930:433–35 Takeuchi AH, Hulse SH. 1993. Absolute pitch. Psychol. Bull. 113:345–61 Tervaniemi M. 2003. Musical sound processing: EEG and MEG evidence. See Peretz & Zatorre 2003, pp. 294–309 Thivard L, Belin P, Zibovicius M, Poline J, Samson Y. 2000. A cortical region sensitive to auditory spectral motion. NeuroReport 11:2969–72 Tillmann B, Bharucha JJ, Bigand E. 2000. Implicit learning of tonality: a self-organizing approach. Psychol. Rev. 107:885–913 Tillmann B, Janata P, Bharucha J. 2003. Activation of the inferior frontal cortex in musical priming. Cogn. Brain Res. 16:145–61 Trainor LJ, McDonald KL, Alain C. 2002. Automatic and controlled processing of melodic contour and interval information measured by electrical brain activity. J. Cogn. Neurosci. 14:430–42 Trainor LJ, Trehub SE. 1992. A comparison of infants’ and adults’ sensitivity to Western musical structure. J. Exp. Psychol.: Hum. Percept. Perform. 18:394–402 Tramo M, Cariani PA, Delgutte B, Braida LD. 2003. Neurobiology of harmony perception. See Peretz & Zatorre 2003, pp. 127–51 Tramo M, Shah GD, Braida LD. 2002. Functional role of auditory cortex in frequency processing and pitch perception. J. Neurophysiol. 87:122–39 Tramo MJ. 2001. Biology and music. Music of the hemispheres. Science 291:54–56 Tramo MJ, Bharucha JJ, Musiek FE. 1990. Music perception and cognition following bilateral lesions of auditory cortex. J. Cogn. Neurosci. 2:195–212

18 Nov 2004 10:54

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

114

AR

PERETZ

AR231-PS56-04.tex




AR231-PS56-04.sgm

LaTeX2e(2002/01/18)

P1: IKH

ZATORRE

Vignolo LA. 2003. Music agnosia and auditory agnosia. Dissociations in stroke patients. Ann. NY Acad. Sci. 999:50–57 Ward WD. 1999. Absolute pitch. See Deutsch 1999, pp. 265–99 Warren J, Uppenkamp S, Patterson R, Griffiths T. 2003. Separating pitch chroma and pitch height in the human brain. Proc. Natl. Acad. Sci. USA 100:10038–42 Warren RM, Gardner DA, Brubaker BS, Bashford JA. 1991. Melodic and nonmelodic sequences of tones: effects of duration on perception. Music Percept. 8:277–90 Wayman JW, Frisina RD, Walton JP. 1992. Effects of musical training and absolute pitch ability on event-related activity in response to sine tones. J. Acoust. Soc. Am. 91:3527– 31 Wilson SJ, Pressing J, Wales RJ. 2002. Modelling rhythmic function in a musician poststroke. Neuropsychologia 40:1494–505 Yamadori A, Osumi Y, Masuhara S, Okubo M. 1977. Preservation of singing in Broca’s aphasia. J. Neurol. Neurosurg. Psychiatry 40: 221–24 Yoo S-S, Lee CU, Choi BG. 2001. Human brain mapping of auditory imagery: eventrelated functional MRI study. NeuroReport 12:3045–49 Zatorre RJ. 1985. Discrimination and recognition of tonal melodies after unilateral cerebral excisions. Neuropsychologia 23:31–41 Zatorre RJ. 1988. Pitch perception of complex

tones and human temporal-lobe function. J. Acoust. Soc. Am. 84:566–72 Zatorre RJ. 2003. Absolute pitch: a paradigm for understanding the influence of genes and development on cognitive function. Nat. Neurosci. 6:692–95 Zatorre RJ, Belin P. 2001. Spectral and temporal processing in human auditory cortex. Cereb. Cortex 11:946–53 Zatorre RJ, Belin P, Penhune VB. 2002. Structure and function of auditory cortex: music and speech. Trends Cogn. Sci. 6:37–46 Zatorre RJ, Evans AC, Meyer E. 1994. Neural mechanisms underlying melodic perception and memory for pitch. J. Neurosci. 14:1908– 19 Zatorre RJ, Halpern AR. 1993. Effect of unilateral temporal-lobe excision on perception and imagery of songs. Neuropsychologia 31:221–32 Zatorre RJ, Halpern AR, Perry DW, Meyer E, Evans AC. 1996. Hearing in the mind’s ear: a pet investigation of musical imagery and perception. J. Cogn. Neurosci. 8:29–46 Zatorre RJ, Perry DW, Beckett CA, Westbury CF, Evans AC. 1998. Functional anatomy of musical processing in listeners with absolute pitch and relative pitch. Proc. Natl. Acad. Sci. USA 95:3172–77 Zatorre RJ, Samson S. 1991. Role of the right temporal neocortex in retention of pitch in auditory short-term memory. Brain 114 (Pt. 6):2403–17

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

Annual Review of Psychology Volume 56, 2005

CONTENTS

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Frontispiece—Richard F. Thompson

xviii

PREFATORY In Search of Memory Traces, Richard F. Thompson

1

DECISION MAKING Indeterminacy in Brain and Behavior, Paul W. Glimcher

25

BRAIN IMAGING/COGNITIVE NEUROSCIENCE Models of Brain Function in Neuroimaging, Karl J. Friston

57

MUSIC PERCEPTION Brain Organization for Music Processing, Isabelle Peretz and Robert J. Zatorre

89

SOMESTHETIC AND VESTIBULAR SENSES Vestibular, Proprioceptive, and Haptic Contributions to Spatial Orientation, James R. Lackner and Paul DiZio

115

CONCEPTS AND CATEGORIES Human Category Learning, F. Gregory Ashby and W. Todd Maddox

149

ANIMAL LEARNING AND BEHAVIOR: CLASSICAL Pavlovian Conditioning: A Functional Perspective, Michael Domjan

179

NEUROSCIENCE OF LEARNING The Neuroscience of Mammalian Associative Learning, Michael S. Fanselow and Andrew M. Poulos

207

HUMAN DEVELOPMENT: EMOTIONAL, SOCIAL, AND PERSONALITY Behavioral Inhibition: Linking Biology and Behavior Within a Developmental Framework, Nathan A. Fox, Heather A. Henderson, Peter J. Marshall, Kate E. Nichols, and Melissa A. Ghera

235

BIOLOGICAL AND GENETIC PROCESSES IN DEVELOPMENT Human Development: Biological and Genetic Processes, Irving I. Gottesman and Daniel R. Hanson

263 vii

P1: JRX

December 8, 2004

viii

12:13

Annual Reviews

AR231-FM

CONTENTS

SPECIAL TOPICS IN PSYCHOPATHOLOGY The Psychology and Neurobiology of Suicidal Behavior, Thomas E. Joiner Jr., Jessica S. Brown, and LaRicka R. Wingate

287

DISORDERS OF CHILDHOOD Autism in Infancy and Early Childhood, Fred Volkmar, Kasia Chawarska, and Ami Klin

315

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

CHILD/FAMILY THERAPY Youth Psychotherapy Outcome Research: A Review and Critique of the Evidence Base, John R. Weisz, Amanda Jensen Doss, and Kristin M. Hawley

337

ALTRUISM AND AGGRESSION Prosocial Behavior: Multilevel Perspectives, Louis A. Penner, John F. Dovidio, Jane A. Piliavin, and David A. Schroeder

365

INTERGROUP RELATIONS, STIGMA, STEREOTYPING, PREJUDICE, DISCRIMINATION The Social Psychology of Stigma, Brenda Major and Laurie T. O’Brien

393

PERSONALITY PROCESSES Personality Architecture: Within-Person Structures and Processes, Daniel Cervone

423

PERSONALITY DEVELOPMENT: STABILITY AND CHANGE Personality Development: Stability and Change, Avshalom Caspi, Brent W. Roberts, and Rebecca L. Shiner

453

WORK MOTIVATION Work Motivation Theory and Research at the Dawn of the Twenty-First Century, Gary P. Latham and Craig C. Pinder

485

GROUPS AND TEAMS Teams in Organizations: From Input-Process-Output Models to IMOI Models, Daniel R. Ilgen, John R. Hollenbeck, Michael Johnson, and Dustin Jundt

517

LEADERSHIP Presidential Leadership, George R. Goethals

545

PERSONNEL EVALUATION AND COMPENSATION Personnel Psychology: Performance Evaluation and Pay for Performance, Sara L. Rynes, Barry Gerhart, and Laura Parks

571

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

CONTENTS

ix

PSYCHOPHYSIOLOGICAL DISORDERS AND PSYCHOLOGICAL EFFECTS ON MEDICAL DISORDERS Psychological Approaches to Understanding and Treating Disease-Related Pain, Francis J. Keefe, Amy P. Abernethy, and Lisa C. Campbell

601

TIMELY TOPIC Psychological Evidence at the Dawn of the Law’s Scientific Age, David L. Faigman and John Monahan

631

Annu. Rev. Psychol. 2005.56:89-114. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INDEXES Subject Index Cumulative Index of Contributing Authors, Volumes 46–56 Cumulative Index of Chapter Titles, Volumes 46–56

ERRATA An online log of corrections to Annual Review of Psychology chapters may be found at http://psych.annualreviews.org/errata.shtml

661 695 700

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

P1: IKH LaTeX2e(2002/01/18) 10.1146/annurev.psych.55.090902.142023

Annu. Rev. Psychol. 2005. 56:115–47 doi: 10.1146/annurev.psych.55.090902.142023 c 2005 by Annual Reviews. All rights reserved Copyright
First published online as a Review in Advance on September 10, 2004

VESTIBULAR, PROPRIOCEPTIVE, AND HAPTIC CONTRIBUTIONS TO SPATIAL ORIENTATION Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

James R. Lackner and Paul DiZio Ashton Graybiel Spatial Orientation Laboratory, Brandeis University, Waltham, Massachusetts 02454; email: [email protected]

Key Words semicircular canals, otoliths, muscle spindles, somatosensation, artificial gravity ■ Abstract The control and perception of body orientation and motion are subserved by multiple sensory and motor mechanisms ranging from relatively simple, peripheral mechanisms to complex ones involving the highest levels of cognitive function and sensory-motor integration. Vestibular contributions to body orientation and to spatial localization of auditory and visual stimuli have long been recognized. These contributions are reviewed here along with new insights relating to sensory-motor calibration of the body gained from space flight, parabolic flight, and artificial gravity environments. Recently recognized contributions of proprioceptive and somatosensory signals to the appreciation of body orientation and configuration are described. New techniques for stabilizing posture by means of haptic touch and for studying and modeling postural mechanisms are reviewed. Path integration, place cells, and head direction cells are described along with implications for using immersive virtual environments for training geographic spatial knowledge of real environments.

CONTENTS INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VESTIBULAR RECEPTORS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VESTIBULAR REFLEXES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Vestibulo-Ocular Reflexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Vestibulo-Spinal Reflexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Loss of Vestibular Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . EFFECT OF UNUSUAL PATTERNS OF ACCELERATION ON ORIENTATION AND SENSORY LOCALIZATION . . . . . . . . . . . . . . . . . . . . . Oculogyral and Audiogyral Illusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Oculogravic, Audiogravic, and Somatogravic Illusions . . . . . . . . . . . . . . . . . . . . . . PROPRIOCEPTIVE AND SOMATOSENSORY CONTRIBUTIONS TO ORIENTATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Muscle Spindles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Somatosensation and Orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Haptic Stabilization of Posture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0066-4308/05/0203-0115$14.00

116 116 119 119 121 121 121 122 122 124 124 126 127

115

6 Dec 2004 10:37

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

116

AR

AR231-PS56-05.tex

LACKNER




AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

DIZIO

MODELS OF ORIENTATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Internal Models and Cognitive Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tri-Axis Model of Spatial Orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . FALLING IN THE ELDERLY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Experimental Tests and Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . MOTION SICKNESS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Incidence and Eliciting Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Space Motion Sickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ARTIFICIAL GRAVITY ENVIRONMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Coriolis Forces and Disruption of Movement Control . . . . . . . . . . . . . . . . . . . . . . . Adaptive Accommodations and Aftereffects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Self-Generated Coriolis Forces in Everyday Behavior . . . . . . . . . . . . . . . . . . . . . . . Coriolis Forces and Motion Sickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . PATH INTEGRATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Place Cells and Head Direction Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . COMPLEXITY OF SPATIAL REPRESENTATIONS: IMPLICATIONS FOR IMMERSIVE VIRTUAL ENVIRONMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . GLOSSARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

128 128 128 129 129 130 130 131 132 132 135 136 136 137 137 138 139

INTRODUCTION Thirty years have elapsed since the Annual Review of Psychology last included a chapter on the vestibular system. In that period, enormous progress has been made. Space flight became commonplace with the development of the U.S. space shuttle program, the Russian Mir space station, and the International Space Station. Now planning is taking place for a manned mission to Mars. A new panorama of information concerning human spatial orientation and performance in weightless and unusual force environments has emerged. In addition, important contributions of somatosensation and proprioception to orientation have been recognized as well as a contribution of cognitive factors. Falling in the elderly has become a major national and international health problem with the burgeoning numbers of older individuals. The present review provides a basic summary of vestibular function and then focuses on these new discoveries and problems.

VESTIBULAR RECEPTORS The bilaterally symmetric labyrinths of the vertebrate consist on either side of three orthogonally oriented semicircular canals and two otolith organs.1 Each semicircular canal is a skull-fixed cartilaginous ring filled with a fluid called endolymph. 1

For definitions of terms, see Glossary at end of text.

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

SPATIAL ORIENTATION MECHANISMS

117

A gelatinous, elastic membrane, the cupula, spans the cross-section of each canal within an enlargement known as the ampulla. The cupulae encapsulate the cilia of hair cell receptors whose somas are embedded in the canal walls. Rotary acceleration of the head in the plane of a semicircular canal causes the endolymph to lag relative to the canal wall because of its inertia and low viscosity, and the resulting endolymphatic pressure deflects the cupula–hair cell complex. Bilateral parallel pairs of semicircular canals are symmetric with respect to ampulla location and are activated in push-pull fashion. Head angular acceleration causes deflection either toward or away from the ampulla, increasing or decreasing the afferent discharge rate, respectively, relative to the high tonic discharge rate when the head is stationary. Over the frequency range of natural head movements, the outputs of the canals are proportional to head angular velocity owing to the viscoelastic properties of the cupula-endolymph system. Some of the experimental procedures described below involve use of a rotating chair to create a step-like onset of head rotation followed by prolonged, constant velocity motion. Cupula-endolymph dynamics are such that the rotating chair evokes canal discharge that shows a direction and velocity-dependent sudden rise followed by a decay back to resting level. The otolith organs by contrast are responsive to linear acceleration, be it of gravitational or inertial origin. These organs consist of aggregates of high specificgravity crystals, or otoconia, embedded in a gelatinous membrane that also encapsulates the cilia of many hair cells whose bodies are fixed relative to the skull. Linear acceleration displaces the roughly planar otoconial masses relative to the skull, thereby bending the embedded hair cells. Hair cells collectively have a wide distribution of morphological orientations, resulting in their receptor potentials being tuned to different directions of acceleration in a plane. The utricular otolith organs are oriented roughly in the head’s coronal plane and the saccular otoliths in the sagittal plane. Different classes of hair cells and their associated primary afferent fibers of the semicircular canals and otoliths segregate angular and linear accelerations into low- and high-frequency bands. There are also efferent pathways from the CNS to the end organs, but the function of these pathways remains largely unknown. The vestibular system is illustrated in Figure 1. For a detailed review of vestibular anatomy and physiology, see Wilson & Melvill Jones (1979). These specialized receptor systems participate in a variety of postural and oculomotor reflexes, and in the perception of body motion and orientation. Direct and indirect projections of the canal and otolith receptors to the vestibular nuclei, the “vomiting centers” in the pontine reticular formation (Miller & Wilson 1983), the cerebellum, the oculomotor nuclei, and the spinal cord have been systematically charted. Recently, vestibular projections to parietal and temporal cortical areas and hippocampus (Guldin & Grusser 1998, Kahane et al. 2003, Wenzel et al. 1996) have been identified that may underlie vestibular influences on perceived orientation and self-motion.

6 Dec 2004 10:37

AR231-PS56-05.tex

LACKNER




AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

DIZIO

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

118

AR

Figure 1 (A) Schematic illustration of the vestibular end organs. For clarity, the semicircular canals and the otolithic maculae are shown larger than life size and unilaterally, although both sets of organs are actually present on each side. The arrows associated with the semicircular canals indicate the head rotation vectors for which individual canals are selective; the arrows on the otolith organs are a reminder that individual hair cells on each macula are attuned to different linear acceleration directions. (B) A block diagram of connections from the vestibular organs to some important brain regions: PCRF is the parvicellular reticular formation, the electrical stimulation of which triggers vomiting; PIVC is the parieto-insular vestibular cortex; TPSVC is the temporo-perisylvian vestibular cortex.

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

SPATIAL ORIENTATION MECHANISMS

119

VESTIBULAR REFLEXES

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Vestibulo-Ocular Reflexes Vestibulo-ocular reflexes and vestibulocollic reflexes are elicited both by angular and linear acceleration. During gaze shifts in which the eyes and head are oriented to a new target position, the eyes generally reach the target by a rapid saccadic movement followed by a head turn toward the target. The eyes are maintained on the target position during the head turn by a vestibulo-ocular reflex that drives the eyes in the opposite direction to the head at the same rate (see Figure 2). The vestibulocollic reflex stabilizes the head in the final position if additional passive whole-body motions are imposed. In the dark, during exposure to passive angular acceleration of the whole body, the vestibulo-ocular reflex produces a nystagmoid pattern of eye movements with a slow phase component compensatory for the direction of acceleration (“field-holding reflex”) interrupted by rapid saccades that reposition the eyes. When vision is allowed, the resulting optokinetic stimulation reinforces the vestibular compensation. During constant velocity rotation in the dark, the semicircular canal signals decay to baseline and the nystagmoid eye movements will gradually abate, but if vision is allowed the resulting optokinetic stimulation will continue to drive the eyes. The vestibulo-ocular reflex has a three-dimensional structure (Cohen et al. 1999) that can be adaptively remapped (Dulac et al. 1995). Vestibulo-ocular reflex adaptation can be measured by comparing ocular responses to body oscillation in darkness before and after exposure to body oscillation relative to a rich visual scene. For example, a few minutes exposure to left-right rotation in synchrony with a head-fixed visual scene results in a change in gain of the vestibulo-ocular reflex such that the magnitude of compensatory eye movements in darkness is reduced. Furthermore, left-right rotation in synchrony with a visual scene oscillating up and down ultimately leads to a geometrical reorganization of the vestibulo-ocular reflex such that left-right oscillation of the body in darkness elicits vertical eye movements (Schultheis & Robinson 1981). Physiological pathways with as few as three neurons can mediate these reflexes, but higher-level inputs modulate them according to current goals. For example, a subclass of neurons in the vestibular nucleus faithfully conveys the primary afferent signals they receive about head motion in space during passive but not active head rotations (Gdowski & McCrea 1999, Roy & Cullen 1998). This task-dependent processing is based on an efferent copy of the head-to-torso rotation command (Roy & Cullen 2001). Compensatory eye movements induced by linear acceleration include ocular counterrolling of the eyes during static roll tilt of the body (Miller & Graybiel 1962), the doll’s eye reflex (Ebenholtz & Shebilske 1975) during head pitch, and elevations and depressions of the eyes during exposure to decreases and increases in background gravitoinertial acceleration (GIA) level in the z axis of the upright head (Whiteside 1961) and to linear translations of the body (Jongkees & Philipszoon 1962). The otolith organs are responsive to GIA, the resultant of gravitational and inertial accelerations acting on the body. Under normal static conditions, GIA

6 Dec 2004 10:37

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

120

AR

AR231-PS56-05.tex

LACKNER




AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

DIZIO

Figure 2 Schematic illustration of the circuit for an idealized unidimensional angular vestibulo-ocular reflex. A leftward head turn produces ampulopetal flow of endolymph in the ipsilateral semicircular canal (SCC) and increases the receptor potential (+sign); ampulofugal flow and depolarization (−sign) occur on the right side. Primary afferent fibers whose cell bodies are in Scarpa’s ganglia (SG) have high tonic firing rates that are increased on the ipsilateral side and reduced on the contralateral side. Excitatory synapses (filled circles) cause parallel increases and decreases in the ipsilateral and contralateral vestibular nuclei (VN) and para-pontine reticular formation (PPRF), respectively. Rightward eye movements result from a combination of ipsilateral connections to the oculomotor nuclei (III) that innervate the medial rectus muscles and contralateral excitatory input to the abducens nuclei (VI) that innervate the lateral rectus muscles. Circuitry involved in adaptation is not included.

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

SPATIAL ORIENTATION MECHANISMS

121

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

magnitude simply corresponds to 1 g (9.8 m/s2), the magnitude of earth gravity. Eye movements evoked by linear acceleration are rarely fully compensatory, e.g., the gain of ocular counterrolling and of the linear vestibulo-ocular reflex is about 0.1 to 0.15 (Paige & Tomko 1991). However, the linear vestibulo-ocular reflex shows sophisticated coordination with other orienting movements. For example, when an individual is undergoing left-right translation in darkness, the amplitude of lateral eye movements rapidly increases when the eyes converge to a nearer distance, as would be required to prevent visual slip of near objects if any were visible (Telford et al. 1997).

Vestibulo-Spinal Reflexes Vestibular influences on postural control include modulation of the postural tone of the body and the antigravity reflexes that are key to maintaining without conscious effort the configuration and stance of the body vis-`a-vis gravity (see Wilson & Peterson 1978, Wilson et al. 1995 for reviews). Such reflexes related to standing and balance interact in a synergistic fashion with reflexes related to the neck (tonic neck reflexes), the limbs, and the overall disposition of forces on the body (see Roberts 1978 for a review).

Loss of Vestibular Function Progressive loss of vestibular function owing to a degenerative process may be largely unnoticed by an individual, unlike loss of other sensory functions such as hearing or vision, because of adaptive changes that occur to maintain function. The neck proprioceptive contribution to compensatory eye movements during gaze shifts, for example, increases and largely compensates for the loss of the vestibuloocular reflex (Bronstein & Hood 1986). Individuals with total loss of function who under normal circumstances are “fully compensated” in terms of balance and locomotion, nevertheless experience difficulty on uneven terrain or soft surfaces, e.g., sand, and under conditions of low light levels or darkness. Thresholds for perception of angular or linear acceleration are greatly increased but perception of body orientation relative to gravity may be largely intact, especially for conditions of self-support and active maintenance of posture. The preservation of accurate perception of body orientation despite loss of vestibular function is based on somatosensory, proprioceptive, efferent, and visual signals (Graybiel et al. 1968, Niven & Graybiel 1953). Loss of labyrinthine function confers total immunity to motion sickness.

EFFECT OF UNUSUAL PATTERNS OF ACCELERATION ON ORIENTATION AND SENSORY LOCALIZATION Changes in visual and auditory localization and in perceived self-orientation are elicited by unusual patterns of vestibular stimulation. Many of these effects were initially described by Mach (1875).

6 Dec 2004 10:37

122

AR

AR231-PS56-05.tex

LACKNER




AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

DIZIO

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Oculogyral and Audiogyral Illusions Exposure to constant angular acceleration in a rotating chair produces changes in both visual and auditory localization (Clark & Graybiel 1949a, Graybiel & Hupp 1946). If a head-fixed visual target is being viewed in an otherwise dark chamber, it will be seen as moving with the body’s changing apparent position in space but leading the body as well in the direction of acceleration. The time course and magnitude of this illusion, known as the oculogyral illusion, is related to the discharge dynamics of the semicircular canals, as described in the Vestibular Receptors section. The occurrence of the oculogyral illusion is related to the partial or complete suppression by visual fixation of the involuntary eye movements that otherwise would occur. The fixation “hold signal” to override the covert vestibular nystagmus and average shift in eye position (Schlagfeld displacement) is misrepresented as an eye deviation, thus causing an error in body relative visual localization (Whiteside et al. 1965). In the dark, a head-fixed auditory target will be heard to displace relative to the body in the direction opposite angular acceleration. The contrasting directions of the audiogyral illusion and of the oculogyral illusion have yet to be adequately explained. Figure 3 illustrates the oculogyral and audiogyral illusions.

Oculogravic, Audiogravic, and Somatogravic Illusions Exposure to unusual patterns of GIA also leads to changes in sensory localization, in apparent self-orientation, and in apparent vehicle orientation. The oculogravic illusion is illustrated in Figure 4. During exposure to an increase in magnitude and a tilt of the GIA vector, a subject seated in a slow rotation room facing the direction of rotation, with left ear toward the center of the room, will experience rightward body tilt and see a luminous line as having displaced clockwise. If the room lights are on, the magnitude of the apparent body tilt will be decreased but a clockwise displacement of the entire room will be seen. Similar mislocalizations

Figure 3 Illustration of the oculogyral and audiogyral illusions. The large arrow indicates the direction of head acceleration. The unfilled icons represent the actual locations of a loudspeaker and a target light fixed in the head midline (dotted line). The gray icons and arrows represent the perceived locations and motions, respectively, of the sound and light source.

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

123

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

SPATIAL ORIENTATION MECHANISMS

Figure 4 (A) Illustration of the gravitoinertial acceleration (GIA) experienced on a rotating room by a subject who is facing the direction of rotation, with left ear toward the center of the room, when the centrifuge acceleration is 1 g. The GIA magnitude and direction are determined by the vector sum of gravity and centrifugal force (ω2r). (B) The somatogravic illusion refers to the apparent body tilt the subject experiences. (C) The oculogravic illusion. The subject perceives an externally presented, earthhorizontal line as tilted. If the room lights are on, the magnitude of the apparent body tilt will be decreased but a clockwise displacement of the entire room will be seen.

6 Dec 2004 10:37

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

124

AR

AR231-PS56-05.tex

LACKNER




AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

DIZIO

of auditory targets occur. The apparent body tilt is known as the somatogravic illusion (Clark & Graybiel 1949b), and the visual and auditory mislocalizations are referred to as the oculogravic (Clark & Graybiel 1949c, Graybiel 1952) and audiogravic illusions (Graybiel & Niven 1951). These effects have sometimes been described as not illusions at all but simply accurate reflections of body and visual orientation in relation to a new vertical corresponding to the GIA orientation as sensed by the otolith organs, which cannot distinguish between inertial and gravitational accelerations (Howard & Templeton 1966). Recent experiments indicate a more complex mechanism (DiZio et al. 2001). Studies of visual and auditory localization have been carried out in which, during exposure to altered GIA direction and magnitude, subjects were required to adjust a visual or an auditory target so that it was perceived to remain in their body midline. In these circumstances, target position had to be displaced by about 15% of the GIA displacement and in the same direction in order to still be seen or heard in the midline of the head. Localization of somatosensory stimulation on the forehead was similarly affected (unpublished observation). Coincident with the altered sensory spatial localizations was a reorientation of the apparent median plane of the head. A joystick that was kept aligned with the apparent midline of the body was displaced in the direction opposite the rotation of the GIA. The time course and magnitude of this shift exactly paralleled the shifts in visual, auditory, and somatosensory localization but were of opposite sign. These localization changes are embedded in the framework of simultaneous changes in apparent orientation of the body with respect to external space. The overall pattern of remapping of sensory localization coupled with a shift in the apparent body midline suggests that a common reference frame for bodyrelative localization and for localization of the body with respect to external space has been modified (see also Lewald & Karnath 2000). A potentially related finding is the observation that vestibular stimulation can alter the pattern of egocentric, unilateral spatial neglect experienced by some patients with brain injury (Vallar et al. 1990). The underlying physiological bases for these modulations remain uncertain, but temporal and parietal cortices contain projections from the relevant sensory modalities, receive vestibular projections, and are implicated in the perception of body orientation in space (Bottini et al. 2001). Thus, they represent natural sites for participation in the observed changes.

PROPRIOCEPTIVE AND SOMATOSENSORY CONTRIBUTIONS TO ORIENTATION Muscle Spindles The otolith organs of the inner ear provide information about the orientation of the head to the GIA. However, the ongoing sense of orientation includes the orientation of the entire body, and all of its links and segments, relative to gravity. The sense of the relative configuration of the body is commonly referred to as the body

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

SPATIAL ORIENTATION MECHANISMS

125

schema. Joint receptors within the capsules of the joints were long thought to serve as potentiometers providing a fiducial representation of joint position that could contribute to an overall mapping of body configuration. However, with the advent of joint replacement surgery, it was found that accurate position sense was retained even in the absence of joint receptors (Grigg et al. 1973). About the same time, it was found that position sense was influenced by the muscle spindle fibers that are interspersed in parallel with the extrafusal muscle fibers that do the actual work of muscle contraction (cf. Matthews 1972). This arrangement means the total length of the muscle spindles is a function of overall muscle length; however, the length of their sensory region innervated by primary and secondary receptor endings is modulated by intrafusal muscle fibers innervated by gamma motoneurons in the spinal cord. The extrafusal muscle fibers are activated by the alpha motoneurons of the spinal cord. The spindle sensory signal is interrelated to the ongoing pattern of gamma and alpha activation and other signals about body loading to compute the angle and rate of change of the joint controlled by the muscle. This normal relationship can be distorted by vibrating a muscle mechanically, circa 100–120 Hz, to activate artificially its muscle spindle primary and secondary receptor endings (Hagbarth & Eklund 1966). In this circumstance, the heightened discharge is centrally interpreted as lengthening of the vibrated muscle and is referred to the joint controlled by the muscle. For example, vibration of the biceps brachii of the arm leads to the forearm feeling more extended than it actually is (Goodwin et al. 1972). When postural muscles are vibrated, various illusions of body motion can be elicited. For example, simultaneous vibration of the Achilles tendons of a standing subject restrained in position will cause the subject to experience forward pivoting in pitch about the ankles. The subject, if in total darkness, will exhibit nystagmoid eye movements with the slow phase compensatory for the direction of apparent self-displacement. If a visual target is presented for the subject to fixate, it will be seen to move in the direction of apparent self-motion and to displace ahead of the subject in the same direction (Lackner & Levine 1979). Thus, the visual target motion has the same characteristics as the oculogyral illusion described above, and a similar physiological explanation in terms of suppression of involuntary eye movements can account for its properties. Vibration of neck muscles leads to illusions of head rotation and displacement (Karnath et al. 1994, Lackner & Graybiel 1974, Lackner & Levine 1979). In fact, with vibration of the appropriate skeletal muscles, apparent motion and displacement of the body or its segments can be elicited in virtually any desired configuration (Lackner 1988). If visual or auditory targets are present, their positions also are remapped in the direction of apparent body motion and displacement. For example, when a small target light is attached to the hand and illusory motion of the restrained forearm is elicited by vibration of the biceps brachii or triceps muscles, the target light will be seen to displace physically in the direction of the apparent motion of the hand. This phenomenon is known as the oculobrachial illusion (Lackner & Levine 1978).

6 Dec 2004 10:37

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

126

AR

AR231-PS56-05.tex

LACKNER




AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

DIZIO

Muscle spindle signals can be more important than vestibular cues in their influence on perceived orientation in altered gravitoinertial force environments. Subjects who move about during exposure to increases or decreases in background force level, for example during parabolic flight, experience instability of the aircraft deck and misperceive their own motion (Lackner & Graybiel 1981). When a deep knee bend is made in a greater than 1 g force level, the weight of the body is more than normal and the antigravity muscles that are undergoing eccentric contraction lengthen slightly more than for a lowering movement of the body in 1 g; in addition, an unusually high level of spindle discharge for the level of alphagamma motoneuronal activity likely is present. The nervous system “attributes” this unexpected lengthening to the floor moving up under the feet. With repeated deep knee bends, self-motion perception gradually becomes more accurate, and the aircraft again is perceived as stable under the feet. On return to a normal 1 g environment, aftereffects are experienced, with opposite signs. Astronauts on return to earth experience the same effects as individuals adapted to 1 g who are exposed to 1.8 g. The ground seems to move upward under their feet when they lower their body in a deep knee bend.

Somatosensation and Orientation Proprioceptive information about limb configuration combined with somatosensory information about hand or limb contact with the body itself and with external objects is a key factor in calibration of the apparent dimensions of the body and of its relationship to external space. For example, if a subject is grasping his nose with his fingers and the biceps brachii muscle of the arm is vibrated, an illusion of arm extension will be elicited and the subject may feel his nose elongate in Pinocchio fashion (Lackner 1988). If the subject is seated with hands on the waist and arms akimbo, bilateral vibration of the triceps brachii muscles will elicit the illusion of the hands approaching one another and of the waist shrinking to become wasp-like. Such interactions indicate that the calibration of the body, in terms of its spatial dimensions, may be achieved by interaction with the hands. Control of the hands and their represented size can be calibrated by interacting with the external environment. Visual direction and the direction of regard also can be calibrated through sight of the hands. Somatosensory stimulation influences perception of the upright and the control of posture. Receptors in the soles of the feet are important in the control of posture (Kavounoudias et al. 1998). For example, as the body leans forward, regions of the feet that are more anterior receive the most mechanical pressure in supporting the body against gravity. During passive stance, this provides a mapping of body orientation to the upright. Diabetic patients with peripheral sensory neuropathies have increased postural sway in part because of degradation of these foot-related signals. Postural control can be enhanced by the use of shoe insoles that vibrate to stimulate the plantar receptors, making it easier for the central nervous system to detect sway changes (Collins et al. 2003). Vests with tactor vibrators also have been employed with mild success to enhance balance by providing tactile stimulation

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

SPATIAL ORIENTATION MECHANISMS

127

of the torso dependent on body orientation to the vertical (Kentala et al. 2003). Such systems also have been proposed for use in signaling body orientation in helicopters and high-performance aircraft (Rupert 2000).

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Haptic Stabilization of Posture Light touch cues from the hand have a powerful stabilizing influence on posture when a surface is touched with the index finger at mechanically nonsupportive force levels. In blindfolded individuals, finger contact reduces mean sway amplitude by about 50%. Even when normal sight of the surroundings is present, balance is further stabilized by light fingertip contact (Holden et al. 1994, Jeka & Lackner 1994). Labyrinthine-defective patients who cannot stand heel-to-toe with eyes closed for more than a few seconds can stand stably indefinitely when allowed light touch (Lackner et al. 1999). Finger stabilization of posture occurs in patients with diabetic neuropathy, cerebellar disease, alcoholism, and in fact, in all patient groups tested to date. If unbeknownst to an individual, the touched surface is oscillated, postural entrainment occurs. The person will be made to sway at frequencies and amplitudes above normal detection thresholds and not be aware of it while the oscillating surface will be perceived as stationary (Jeka et al. 1997). Even when the test subject is aware that the surface can oscillate, postural entrainment will occur, albeit of lesser amplitude. However, in this circumstance, the subject will perceive the touched surface as moving and in trials in which the surface is stationary, it nevertheless may be perceived to be moving. This latter fact emphasizes the importance of cognitive knowledge and assumptions (i.e., “top-down effects”) about the environment for whether unseen touched objects will be perceived to be stationary or not. The time course of postural stabilization by fingertip contact is remarkably rapid. When the finger is dropped to contact a surface, it will be stabilized on the surface within 100 milliseconds and a downward trend in body mean sway amplitude will be detectible within the next 100 milliseconds. By contrast, visual stabilization of posture by turning on the lights takes three or four times longer to begin and still longer to complete (JR Lackner, E Rabin, & P DiZio, submitted; Rabin et al. 2004). Touch signals from the hand also completely override the otherwise destabilizing effects of tonic vibration reflexes in leg muscles (Lackner et al. 2000). In addition, the illusions of self-displacement and aircraft displacement that occur with whole-body movements during exposure to increased or decreased background force levels in parabolic flight (see above) are virtually abolished by light contact of the hand with a stationary surface (Lackner & DiZio 1993). The postural stabilization conferred by light touch seems to reflect a form of “precision touch” analogous to precision grip. When an object is held between the index finger and thumb and lifted, the forces at the fingertip are “automatically adjusted” to generate appropriate grip and load forces. The latency of grip adjustments, about 85 ms, is well below conscious reaction times, and reflects the

6 Dec 2004 10:37

128

AR

AR231-PS56-05.tex

LACKNER




AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

DIZIO

activity of a somatosensory-motor cortical control loop (Johansson & Westling 1984). During stance with maintained finger contact, the changes in the body’s force at the fingertip phase lead changes in the body center of pressure and center of mass by about 300 ms. With fingertip contact, inverted pendulum sway of the body is maintained.

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

MODELS OF ORIENTATION Physical sensors such as the otolith organs cannot distinguish gravitational from inertial accelerations. However, for humans, other forms of information that can disambiguate situations generally are available. One important factor is the control of the body itself in terms of the forces necessary to maintain balance or body support, and whether the body is actively moving. For example, when a person walks in a tight circle or makes a sharp turn, quite high centripetal forces are generated but body orientation is not misperceived (Imai et al. 2001). When the same forces are passively applied to the body, the same tilt of the GIA will lead to apparent roll tilt of the body (Miller & Graybiel 1966).

Internal Models and Cognitive Factors The otolith organs have “regular” and “irregular” receptor units with different firing patterns (Fernandez et al. 1972). The regular fibers have discharges that are tuned to different static orientations of the body relative to gravity and that maintain a steady-state discharge virtually indefinitely. By contrast, irregular fibers respond to changes in acceleration and in position, but adapt over time. Regular units show a less-prominent dynamic modulation. Such frequency-dependent segregation of otolith afferent firing has been proposed as a way to distinguish gravity from inertial accelerations, especially for frequency domains that exceed those of voluntary head movements (Mayne 1974). Investigators have also proposed that the CNS uses internal models to discriminate inertia and gravity based on representations of the body’s dynamics and on available semicircular canal as well as otolith input (Merfeld et al. 1999). Cognitive information also usually is available about the environment and its properties, such as whether the body is in a vehicle of some sort.

Tri-Axis Model of Spatial Orientation A recent tri-axis spatial orientation model has been developed based on the hypothesis that the vestibular system evolved in a 1 g, Earth gravity context and that 1 g is taken as a standard relative to which all shear and deformation patterns of the otolith membranes are interpreted (Bortolami et al. 2004). The model incorporates the known structural features of the otolith organs and includes “cross talk” between the pitch and roll axes. This cross talk occurs because of the complex three-dimensional organization of the otolith membranes and their resulting displacements to shear forces (Grant & Best 1987). It accurately predicts both the

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

SPATIAL ORIENTATION MECHANISMS

129

errors in spatial localization of the vertical that occur for different static orientations of the body in yaw, pitch, and roll relative to gravity, and the potential elicitation of inversion illusions in weightless conditions. A surprising prediction of the model is that when the recumbent body is in different static yaw orientations with respect to gravity, the level of resultant GIA will have negligible influences on perceived body orientation to the vertical. This prediction results from calculating yaw orientation from signals related to the pitch and roll axes and using a 1 g gravity standard to interpret these shear signals. Parabolic flight experiments have recently validated the model’s predictions. The model uses somatosensory and proprioceptive cues to distinguish up and down, which is consistent with the important influence of touch and contact cues on spatial orientation in weightless conditions.

FALLING IN THE ELDERLY Falls in the elderly have become a major health problem as longevity has increased (Maki et al. 1987). In the United States, the average survival time for a faller who has fractured a hip is about two years. Many factors contribute to falls, including reduced visual acuity, diminished vestibular function, lessened somatosensory and proprioceptive acuity, reduced muscle strength, and medications (Brown et al. 1999, Chen et al. 1996, Draganich et al. 2001, Klein et al. 1998, Lajoie et al. 1993, Leipzig et al. 1999, McIllroy & Maki 1996, Shumway-Cook et al. 1997, Thelen et al. 1996). Tripping on obstacles and when changing direction is also commonplace (Blake et al. 1988, Campbell et al. 1990, Overstall et al. 1977, Ruberstein et al. 1988). Individuals who are in the weightless phase of parabolic flight maneuvers and astronauts who are in orbital flight are actually in a state of continuous free fall; nevertheless, they do not perceive themselves to be falling (Lackner 1992). This means that otolith signals per se do not determine whether one experiences falling but that additional sensory and cognitive factors must contribute.

Experimental Tests and Models A number of studies have investigated the role of surface features, obstacles in the walking path (Persad et al. 1995), sudden changes in direction (Mbourou et al. 2003), and visual factors (Judge et al. 1995) in predisposing individuals to lose balance during locomotion. These studies highlight the relative importance of various external factors in triggering loss of balance. Other studies have evaluated how various perturbations of the stance surface and of the visual surround affect postural recovery (Luchies et al. 1994, Pai & Patton 1997). Still others have explored the ability to recover from artificially induced forward leans of the body (Thelen et al. 1997). From these investigations, age- and gender-related changes in somatosensory, proprioceptive, and vestibular function, and muscular strength have been identified that may contribute to falling in the elderly.

6 Dec 2004 10:37

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

130

AR

AR231-PS56-05.tex

LACKNER




AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

DIZIO

Mathematical models of falling have been developed using an inverted pendulum representation of the body. The most advanced of these accurately reproduces the body’s response to a Hold and Release perturbation (Bortolami et al. 2003). The individual being tested actively resists a force applied to his or her chest that is unpredictably released. For some tens of milliseconds, the body is physically in a state analogous to free fall, pivoting about the ankles. After a latency of about 50 milliseconds, compensatory postural adjustments are made to arrest falling and the body undergoes damped oscillations over 10 seconds before settling. Video recordings of the body multilink response and recovery from the perturbation, and electromyographic recordings from the postural muscles involved in the recovery, serve as observational input providing a description of the body’s response. Fitting these data to a multilink inverted pendulum model allows computation of the stiffness and damping at the ankles, knees, hips, and neck, along with the recovery time to resettling. The patterning and magnitudes of muscle activation allow a snapshot of the underlying neurophysiological processes mediating recovery. It appears that postural responses to Hold and Release involve one-time, stable adjustment of stiffness and damping after a variable latency. The identified stiffness and damping parameters, as well as the latency, vary with visual and tactile sensory cues and with background GIA. The Hold and Release paradigm and model provide a useful clinical tool for experimentally studying falling and for characterizing the deficits in motor control of people prone to falling. A linear, four-link inverted pendulum model is sufficient to characterize postural recovery in patients with labyrinthine loss, those with somatosensory loss, and those with cerebellar lesions. For normal, healthy subjects only a two-link model is necessary to adequately describe recovery of balance.

MOTION SICKNESS Motion sickness has been a persistent accompaniment of exposure to vehicles of virtually any sort: cars, trains, boats, aircraft, camels, swings, and spacecraft. The acceleration patterns in terms of amplitudes, frequencies, and durations that are most provocative are relatively well known. Many theories have been proposed to account for motion sickness, including overstimulation of the vestibular system, reaction to a perceived poison, and sensory conflict (cf. Reason 1970). None is fully satisfactory nor has adequate predictive power (e.g., the ability to distinguish which sensory conflict situations will be nauseogenic and which will not). The most certain factor known with regard to motion sickness under terrestrial conditions is that to date subjects without functioning labyrinths have not been made motion sick despite sometimes heroic efforts to do so (Graybiel 1969, Johnson et al. 1962).

Incidence and Eliciting Factors It initially was thought that there were two general categories of responders to motion sickness, those expressing primarily “head symptoms,” and those expressing

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

SPATIAL ORIENTATION MECHANISMS

131

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

primarily “gut symptoms.” Head symptoms include headache, drowsiness, and eyestrain, whereas gut symptoms include stomach awareness, discomfort, nausea, and vomiting. It now is recognized that a person’s response to a particular form of motion depends on his or her individual susceptibility to that motion as well as to the relative intensity of the provocative stimulation. For example, making tilting head movements during constant velocity, passive body rotation at 3 rpm will elicit mainly head symptoms, whereas making tilting head movement at 30 rpm will rapidly induce nausea and vomiting for most subjects.

Space Motion Sickness Space motion sickness, now often referred to as the space adaptation syndrome to recognize the complex pattern of changes associated with exposure to weightlessness, is an operational problem in space travel. Symptoms may develop in some astronauts and cosmonauts as early as the first hour in flight, but generally within the first day. Nearly 70% of all space travelers are affected to some extent during the first 72 hours (Jennings 1997). Pitch and roll head movements are significant etiological factors in eliciting the space adaptation syndrome (Lackner & Graybiel 1984, 1986, 1987). After return to earth, a variety of disturbances of posture, gait, and motor control are present until readaptation to earth gravity occurs. Symptoms of motion sickness may redevelop postflight as well. Astronauts generally are less susceptible in subsequent space flights and experience less severe re-entry disturbances and motion sickness as well. Severe cases of space motion sickness can be treated with anti-motion-sickness drug injections of promethazine (Graybiel & Lackner 1987, Lackner & Graybiel 1994). It is unknown whether labyrinthine-defective subjects would be insusceptible to space motion sickness. However, they would not be immune to the postural, sensory-motor, and locomotory disturbances that occur after return to Earth. Adaptation to weightlessness involves not just adaptation of vestibularly mediated reflexive and orientation effects, but also accommodation of the entire postural and muscular control system of the body to a radically different force environment. This means that the control of body-relative movements, body movements in relation to the surroundings, and object manipulation and use involves a remapping of the motor commands to the musculature necessary to bring about the desired actions. Under terrestrial conditions, alterations in the normal motor control patterns necessary to stabilize the head or to achieve particular movement goals can elicit symptoms of motion sickness, postural disturbances, and sensory-motor performance disruptions analogous to those occurring in space flight and after re-entry. The manipulations evoking such impairments involve changing the apparent weight and inertia of the body segments, decreasing them to mimic weightless conditions and increasing them to simulate re-entry disturbances (Lackner & DiZio 1989). Such changes are independent of vestibular function per se and point to the important role of sensory-motor control and calibration of the body to the ongoing force

6 Dec 2004 10:37

132

AR

AR231-PS56-05.tex

LACKNER




AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

DIZIO

environment. Motion sickness elicited in vehicles tends to be exacerbated by active maintenance of posture or by locomotion within the vehicle. Fully passive exposure with the body recumbent or not supporting itself tends to be least provocative (Lackner et al. 1991). Control over the onset and intensity of passive body motion also decreases susceptibility, e.g., the driver of a car rarely experiences carsickness even when passengers have been made quite ill.

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

ARTIFICIAL GRAVITY ENVIRONMENTS For long-duration space missions lasting many months, or even several years such as a manned mission to Mars, artificial gravity produced by rotation may be necessary to reduce or prevent the loss of bone mineral content and muscle deconditioning induced by mechanical unloading of the musculoskeletal system in weightless conditions (cf. Lackner & DiZio 2000, 2004; Young 1999 for reviews).

Coriolis Forces and Disruption of Movement Control “Artificial gravity” is the centripetal force imposed by a rotating vehicle that keeps objects within it moving in a curvilinear path. Centripetal force is proportional to the radius of rotation (r) times the velocity of rotation (in radians) squared. Accordingly, approximately 1.1 g of artificial gravity would be produced by a vehicle 1000 m in radius rotating at 1 rpm and by one 10 m in diameter rotating at 10 rpm. The latter situation is much more difficult to adapt to because when a person moves in a rotating environment, inertial Coriolis forces are generated on his or her body. These forces are larger the higher the velocity of rotation. For example, when a forward-reaching movement is made during exposure to counterclockwise rotation, a rightward Coriolis force will be generated on the arm proportional to the mass of the arm (m), the linear velocity of the arm relative to the rotating vehicle (v), and the angular velocity of rotation (ω): Fcor = −2m (ω × v) (see Figure 5). This means that for a given movement of the body or a part of the body with velocity, v, the Coriolis force is linearly dependent on the angular velocity of vehicle rotation, rotation at 10 rpm will generate a tenfold greater Coriolis force than rotation at 1 rpm. Vestibular function is one of the key factors affecting the ability of humans to adapt to rotating artificial gravity environments. For example, when a pitch head movement is made out of the plane of rotation, an unusual pattern of stimulation of the semicircular canals will result. One pair of semicircular canals will be brought into the plane of rotation and signal head rotation in the direction of vehicle rotation; simultaneously, another pair will be brought out of the plane of rotation losing angular momentum and consequently signaling head rotation in the opposite direction to that of the vehicle’s motion. The final pair will accurately signal the forward pitch of the head. The otolith organs will also be transiently stimulated by the Coriolis force generated by the head movement (see Figure 6). This bizarre pattern of vestibular activity leads to complex orientation illusions, destabilizes

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

133

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

SPATIAL ORIENTATION MECHANISMS

Figure 5 (A) Illustration of a subject in the center of a counterclockwise rotating room (ω) reaching forward to a target. A rightward Coriolis force, FCor, is generated while the arm is moving forward at velocity v. (B) A top view of the fingertip path shows that reaches are straight to the target before rotation starts, deviate to the right during rotation, return to baseline after additional reaches, and show mirror-image aftereffects when rotations stops. (C) Plots of the endpoints and curvatures of individual movements show that they return to baseline within 10 to 15 trials after rotation starts and stops.

6 Dec 2004 10:37

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

134

AR

AR231-PS56-05.tex

LACKNER




AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

DIZIO

Figure 6 (A) Schematic showing the coordinate system of a set of angular rate sensors attached to the head for a movement in which the subject nods the head forward and returns immediately to the upright position (dotted arrows). (B) Traces from the angular rate sensors when the head movement is made in a normal stationary environment. (C) Traces recorded for a comparable pitch head movement made during 10 rpm counterclockwise rotation. Prior to the movement, the yaw axis of the head is rotating at the speed of the room; pitching the head forward reduces the yaw rate, and returning it to the upright restores it. The roll axis of the head comes into the room rotation plane during pitch forward and out of the plane during pitch up. The semicircular canals that are fixed to the head pick up these cross-coupled angular accelerations, and the subject will experience a tumbling sensation involving illusory yaw and roll motion in addition to the veridical pitch of the head.

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

SPATIAL ORIENTATION MECHANISMS

135

eye movement control, and if repeated can rapidly lead to motion sickness (except in individuals without labyrinthine function). Studies of artificial gravity in slow rotation rooms in the 1960s suggested that 5 rpm was the highest rate of rotation at which people could fully adapt to the vestibular stimulation encountered (Clark & Graybiel 1961; Fregly & Kennedy 1965; Graybiel et al. 1960, 1965; Guedry et al. 1962, 1964; Kennedy & Graybiel 1962). Subsequently, this value became the “accepted” upper velocity limit for space flight artificial gravity environments. In these early studies, posture, locomotion, and arm movement control were also disrupted. On cessation of rotation, normal subjects, but not vestibular-loss subjects, experienced a recurrence of symptoms of motion sickness. Both the normal and labyrinthine-defective subjects experienced postrotation disturbances of posture, gait, and arm movement control. These aftereffects reflect the fact that motor control as well as vestibular function is affected in rotating environments.

Adaptive Accommodations and Aftereffects Recent studies of adaptation to artificial gravity environments, however, have demonstrated that head and arm and leg movement control as well as locomotion can be adapted quite rapidly to rotational velocities of 10–25 rpm if the same movement is attempted repeatedly (Bouyer et al. 2003; DiZio & Lackner 1995, 1997, 2003; Lackner & DiZio 1994, 1998). In this circumstance, movement paths soon become straight again and movement endpoints accurate. On cessation of rotation, mirror-image errors in movement control occur, indicating that the nervous system has constructed a model of the Coriolis forces expected to be forthcoming and planned appropriate compensatory innervations to null their effect on movement paths. The persistence of this compensation even when no longer appropriate leads to the appearance of mirror-image aftereffects on cessation of vehicle rotation. When adaptation to a rotating environment is complete, the Coriolis forces associated with movements no longer are consciously perceived. Although still present during movements, these forces become perceptually transparent. On cessation of rotation when movements are made again, a new “force” is perceived that is felt to act in the direction opposite to the original forces felt during rotation. The internally generated compensations for Coriolis forces that are no longer appropriate and that deviate movement paths and endpoints postrotation thus are perceived as external forces deviating the arm. On Earth, the forces generated by support of the body against gravity and of body movements relative to gravity are also largely transparent. They are not felt or perceptually registered even though they are significantly above thresholds for detection. For example, to raise one’s arm it is necessary to counter the torques on the arm owing to the force of gravity. Nevertheless, the effort associated with the movement feels virtually the same even though the force generated to bring about the movement is varying greatly. Similarly, if balance is shifted from two feet to one while standing, hardly any force change on the sole of the remaining stance

6 Dec 2004 10:37

136

AR

AR231-PS56-05.tex

LACKNER




AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

DIZIO

foot will be felt even though the contact force has doubled (Lackner & Graybiel 1982, 1984). For a 180 lb individual it will increase by 90 lb to 180 lb, a huge change. Such observations mean that the body is dynamically tuned to its force environment, and movements within it feel virtually effortless.

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Self-Generated Coriolis Forces in Everyday Behavior The ability to adapt very rapidly and almost effortlessly to Coriolis force perturbations of head, arm, and leg movement trajectories in rotating environments was initially surprising given the restricted adaptation seen in earlier studies. However, natural locomotion generates significant linear and angular accelerations that are compensated for by automatic adjustments of the eyes and head (Imai et al. 2001). Recently, it has been shown that Coriolis forces are commonly encountered in our everyday activities (Pigeon et al. 2003a). When a natural turn-and-reach movement is made to an external object, the trunk reaches peak velocities much higher (e.g., 150–200◦ /s) than those ever contemplated for use in artificial gravity environments (e.g., 60–90◦ /s). Because turn-and-reach movements are not segmented into turn and then reach, the peak velocities of the trunk movement and of the forward velocity of the arm occur close together in time. As a consequence, very large Coriolis forces are generated on the reaching arm. Nevertheless, reaching accuracy is preserved and is independent of the peak velocity of the trunk and of whether visual feedback is available. This preservation of reaching accuracy means that the nervous system anticipates the Coriolis forces that will be generated, and it institutes appropriate anticipatory compensatory forces to eliminate their deflecting influence on movement paths. Head and leg movements out of the plane of rotation are also common during voluntary trunk rotation and are also made accurately. Such accuracy indicates that compensation for self-generated Coriolis forces is a typical aspect of everyday motor control. It is not surprising, therefore, that individuals are able to adapt to the Coriolis forces associated with movements in passively rotating artificial gravity environments because these forces are much smaller than those generated during their everyday activities. In other words, the body when turning constitutes its own artificial gravity environment from the perspective of eye, head, arm, leg, and torso movement control. An examination of the relative straightness of arm movements and of their velocity profiles during natural turn-and-reach movements supports the view that they are planned in relation to external spatial coordinates rather than joint or intrinsic coordinates. Arm trajectories are straight and their velocity profiles are single-peaked in relation to external space, but are curved with multiple velocity peaks when plotted in relation to joint or torso relative coordinates (Pigeon et al. 2003b).

Coriolis Forces and Motion Sickness Making pitch head movement during passive rotation is highly nauseogenic and elicits a complex, illusory tumbling sensation because it generates Coriolis

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

SPATIAL ORIENTATION MECHANISMS

137

cross-coupled stimulation (CCS) of the semicircular canals (Johnson et al. 1951), as shown in Figure 6. The puzzling finding from space flight experiments is that CCS is not nauseogenic in orbital flight (Graybiel et al. 1977). These investigators were motivated by the desire to ascertain whether the limits of tolerable rotation that had been found on Earth, where CCS occurs in a 1 g force background, predict what would occur in space, where the background, artificial gravity level (centripetal force) will likely be less than 1 g. This question was and still is highly relevant for developing a feasible artificial gravity countermeasure for prolonged space flight because the severity of the side effects of rotation (motion sickness, disorientation, and disruption of movement control) as a function of background force level is a major factor that must be considered in choosing the radius and rotation rate. Many attempts to understand the variable provocativeness of CCS have focused on putative sensory conflicts (e.g., Guedry & Benson 1978). In such theories, conflict occurs when CCS elicits semicircular canal signals encoding an off-vertical tumbling axis that is in contrast with a nonchanging otolith signal from the physically nontumbling head. Such a conflict is not possible in space flight where there is no background GIA whose direction can be encoded by the otolith afferent signals. Parabolic flight experiments have provided an alternate perspective on this problem. The duration and intensity of perceived tumbling are augmented in 1.8 g and almost eliminated in 0 g (as is motion sickness, confirming the space flight results) following the same CSS stimulation of the semicircular canals in both force backgrounds (DiZio et al. 1987). This pattern suggests that semicircular canal afferent signals encoding head velocity may be centrally integrated to produce a representation of body displacement, and that this integration process is GIA-dependent. This possibility has been confirmed recently in parabolic flight experiments. Blindfolded subjects continuously indicated their subjective self-displacement relative to external space by pointing to a fixed location with a joystick while they were exposed to seminatural, brief, suprathreshold tilts about their yaw axis. In 1 g and 1.8 g, pointing responses were nearly constant relative to the aircraft, but in zero gravity, the stick was always kept in a body-fixed orientation. Subjects verbally confirmed that they were aware of being moved in all force backgrounds but that the movement was not associated with any perceived displacement in 0 g. This means that the lack of motion sickness elicited by CCS in zero gravity may be due to a lack of internally represented body displacement. These recent findings also add to the debate about how three-dimensional spatial orientation is represented and remembered, which is surveyed in the following section.

PATH INTEGRATION Place Cells and Head Direction Cells The classical observations of Beritoff (1965) suggested that vestibular afferent signals during passive transport of the body could be integrated to generate a representation of the body’s path through space and of its new position in relation

6 Dec 2004 10:37

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

138

AR

AR231-PS56-05.tex

LACKNER




AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

DIZIO

to the start position. These results have led to renewed interest because of the finding of “place cells” in the hippocampus that are active when an animal is in a specific location within a familiar environment (O’Keefe & Dostrovsky 1971) and of “head direction cells” that seem to code the direction in which the head is oriented with respect to extrinsic spatial coordinates (Taube et al. 1990). Numerous experiments have demonstrated that during passive transport, spatial memory and knowledge of the trajectory and of the spatial location within the environment are maintained (Berthoz et al. 1987; Bloomberg et al. 1991; Israel et al. 1993, 1997; Mittelstaedt 1980). This is manifested in the ability to orient the head and body toward the starting position or to memorized targets, or to recapitulate the path the body has traveled. Vestibular inputs may be sufficient for such path integration because neuronal recordings show that hippocampal cells maintain their spatial tuning even during passive transport in darkness (Gavrilov et al. 1998). However, vestibular inputs are not necessary for updating spatial orientation because humans without functioning labyrinths show path integration during active walking (Glassauer et al. 1994). Path integration is also exhibited by insects lacking a vestibular system (see Wehner 2003 for a review). These creatures must rely, in the absence of vision, on monitoring motor outputs and sensory feedback to generate spatial “representations.”

COMPLEXITY OF SPATIAL REPRESENTATIONS: IMPLICATIONS FOR IMMERSIVE VIRTUAL ENVIRONMENTS One of the most promising applications of virtual environment technology is training users to build cognitive maps of environments (Durlach et al. 2000). Through visual displays and clever “locomotion interfaces,” a user can “navigate” a large virtual space without moving around at all or while moving in very limited fashion. An example of the former would be using a joystick to navigate a virtual suite of rooms shown on a desktop display; an example of the latter would be a pilot trainee viewing a visual model of terrain he or she is “overflying” in a simulator that reproduces a portion of the inertial cues that would be present in a real aircraft. Realistic experiences can be generated in these situations, and subjects seem to acquire survey knowledge (Ruddle et al. 1997) of the virtual environment without normal vestibular inputs. An individual is considered to have survey knowledge when he or she remembers enough of the environmental layout to draw a map that indicates distances between landmarks and to give directions to destinations along novel routes (Siegel & White 1975). Acquiring survey knowledge in virtual environments is slower than in the real world (Witmer et al. 1996). It is not clear yet whether this decrement is due to lack of fidelity of the visually rendered virtual world or to the lack of normal inertial cues, including vestibular cues. However, one study has shown that spatial learning is not hindered when inertial cues are reduced or distorted (Waller et al. 2003).

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

SPATIAL ORIENTATION MECHANISMS

139

Rotating room studies have demonstrated that the spatial representation and recognition of environments is dependent on the means of normally accessing them and that the subenvironment is embedded in a broader spatial representation of the spatial context (Lackner & DiZio 1998). As a consequence, entry into a particular site, e.g., a room from a direction never before experienced, can lead to a sense of unfamiliarity and strangeness in an environment that when entered from the familiar entrance seems familiar and natural. Such observations mean that the cognitive representation of spatial environments is quite complex and is influenced by the modes of access. This has especial significance for recent developments in virtual environment technology. For example, it has been proposed that immersive virtual environments can be used to train knowledge of a real environment. This would make it possible, for example, for firefighters not to have to train in a real environment (e.g., an aircraft carrier) to know its spatial layout intimately. Training from the virtual to the real environment, however, may not transfer unless the “access” into the virtual environment is from modes of entry that are possible in the real environment. ACKNOWLEDGMENT This work has been partially supported by Air Force Office of Scientific Research grant F49620110171 and NASA grants NAG9-1483, NAG9-1263, and NAG91466.

GLOSSARY Afferent discharge rate – Primary afferent neurons of the mammalian vestibular system relay information from hair cells to the central nervous system. They encode head angular velocity and linear acceleration via a rate code, in terms of the number of action potentials per second. Ampulla – An enlargement in the toroidal ring of each semicircular canal, within which resides the sensory apparatus, the cupula/hair cell complex. Cupula – A gelatinous membrane inside the ampulla of each semicircular canal. The inertial tendency of the endolymph to lag head rotary acceleration displaces the cupula and deflects the ensheathed cilia of hair cells affixed to the ampulla base. The elasticity of the cupula causes recoil to the resting position following angular acceleration of the head. Doll’s eye reflex – Reflexive elevation or depression of the eyes relative to the head in the direction opposite to the pitch angle of the body. Eccentric contraction – Active production of tension in a muscle while it is being lengthened. For example, on Earth lowering the body during a deep knee bend involves eccentric contraction of the quadriceps muscle group. Efferent copy – A duplicate of a command signal innervating muscles. For example, efferent copies of signals from brainstem nuclei to eye muscles are

6 Dec 2004 10:37

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

140

AR

AR231-PS56-05.tex

LACKNER




AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

DIZIO

thought to be involved in parcellation of visual signals into components due to self-motion and object motion. Endolymph – The fluid that fills the semicircular canals and bathes the otolith organs. Its mass and viscosity are major factors governing the encoding of head angular velocity by the afferent signal from the semicircular canals. Gain – The ratio of the output of a controlled system to its input. For example, the ratio of eye roll (output) to body tilt (input) in ocular counterrolling is about −0.1. Gravitational acceleration – Acceleration due to the force of gravity. An object at rest on the Earth’s surface is being accelerated by gravity toward the center of the earth and is simultaneously accelerated in the opposite direction by the ground. Gravitoinertial force – The vector resultant of gravitational and inertial force. The otoliths encode acceleration due to gravitoinertial forces. Einstein’s equivalence principle states that no simple physical sensor can detect whether its acceleration is due to inertial or gravitational force. A very active research topic concerns how multiple sensory inputs are processed to yield separate senses of tilt relative to gravity and inertial motion. Hair cell – Goblet-shaped sensory neuron with a bundle of cilia of graded length protruding from the cell body. Maximum depolarization or hyperpolarization is produced when the bundle is bent toward or away, respectively, from its apex, and graded responses are evoked by intermediate directions of deflection. Deflection of large arrays of hair cells is the basis of the afferent response of the semicircular canals and otolith organs (as well as other sensory systems). Haptic perception – A type of perception derived from active touch, in contrast to the passive impressions of external stimuli impressed on the skin. Inertial acceleration – Acceleration relative to a fixed- or constant-velocity frame of reference (such as the earth) by an external force to the object. The semicircular canals are stimulated by angular inertial acceleration and do not respond to gravity. Inversion illusion – The experience of oneself and/or the entire seen and felt environment being upside down relative to the subjective vertical. It is often elicited upon exposure to weightless conditions. Muscle spindle – A specialized receptor in muscle tissue producing afferent signals modulated by muscle length and rate of change in length. Its responses are modulated by the gamma motor system. Nystagmus – Named for a sawtooth pattern involving quick jumps in one direction alternating with slower drifts in the other. Nystagmoid eye movements involve a “field fixing” pattern in which slow eye drifts compensate for selfmotion in one direction alternate with rapid recentering movements. Ocular counterrolling – Reflexive rotation of the eyes about the naso-occipital

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

SPATIAL ORIENTATION MECHANISMS

141

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

axis in the direction opposite lateral body tilt about a parallel axis. Otolith signals related to body tilt are a major influence on this reflex. Optokinetic stimulation – Unidirectional motion of all or part of the external visual array. Otolith organ – Nonauditory sensory organ in the vertebrate inner ear responsible for transduction of linear acceleration and gravity. The otoliths include masses of dense crystals (otoconia) embedded in a gelatinous membrane anchored to the cilia of a skull-fixed array of hair cells. The otoconia are displaced opposite to linear accelerations of the head, thereby deflecting the hair cells and modulating their membrane potential. Different classes of spatially and temporally tuned hair cells are hypothesized to be crucial for distinguishing the inertial and gravitational parts of the net linear acceleration. Parabolic flight – Aircraft flown in a roller-coaster path to produce alternating periods of high gravitoinertial force during the climbs, and rapid transitions to weightlessness, zero gravity, during the pushover at the top of each parabola. Path integration – Knowledge of one’s current position and heading relative to a position with which one is no longer in sensory contact. It is derived from temporal accumulation of sensory and motor signals about movement path through space. Proprioception – The sense of position and motion of one’s body segments, derived from central processing of efferent signals as well as afferent signals from muscles, tendons, joints, and skin. Some investigators include vision and audition as contributing potential proprioceptive information. Push-pull – A traditional control theory name for the manner in which angular head acceleration in a given plane is encoded by the discharge rate of primary vestibular afferent neurons. The afferent neurons have a high resting discharge rate (about 100 imp/sec), which is increased during head angular acceleration in one direction and decreased by acceleration in the opposite direction. Receptor potential – Graded changes in voltage across the cell body membrane of vestibular hair cells, proportional to deflection of the ciliary bundle. Saccules – Bilateral pair of otolith organs anatomically arranged to respond to vertical-plane linear accelerations. Semicircular canals – Nonauditory sensory organs in the vertebrate inner ear that are stimulated by angular acceleration of the head and encode angular velocity of natural movements. The name describes the structure: three approximately orthogonal, fluid-filled toroidal tubes on either side of the head. The mechanism of the sensory response is the deflection of hair cells inside the canal by the inertial lag of the fluid during head acceleration. Spatial neglect – Decreased probability of detection of stimuli in a circumscribed region of extracorporeal space. It can be partial or complete, and can involve multiple sensory modalities.

6 Dec 2004 10:37

142

AR

AR231-PS56-05.tex

LACKNER




AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

DIZIO

Utricles – Bilateral pairs of otolith organs anatomically arranged to respond to horizontal plane linear accelerations when the head is upright. The Annual Review of Psychology is online at http://psych.annualreviews.org

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

LITERATURE CITED Beritoff J. 1965. Neural Mechanisms of Higher Vertebrate Behavior. Boston, MA: Little, Brown Berthoz A, Israel I, Vieville T, Zee D. 1987. Linear head displacement measured by the otoliths can be reproduced through the saccadic system. Neurosci. Lett. 82(3):285–90 Blake AJ, Morgan MJ, Dallosso H, Ebrahim SBJ, Arie THD, et al. 1988. Falls by elderly people at home: prevalence and associated factors. Age Ageing 17:365–72 Bloomberg J, Melvill Jones G, Segal B. 1991. Adaptive modification of vestibularly perceived rotation. Exp. Brain Res. 84:47–56 Bortolami SB, DiZio P, Rabin E, Lackner JR. 2003. Analysis of human postural responses to recoverable falls. Exp. Brain Res. 151:387–404 Bortolami SB, Rocca S, DiZio P, Lackner JR. 2004. Novel concepts of quasi-static spatial orientation mechanics in humans. J. Neurophysiol. In press Bottini G, Karnath HO, Vallar G, Sterzi R, Frith CD, et al. 2001. Cerebral representations for egocentric space—functional-anatomical evidence from caloric vestibular stimulation and neck vibration. Brain 124:1182–96 Bouyer L, DiZio P, Lackner JR. 2003. Adaptive modification of human locomotion by Coriolis force. Program No. 494.13.2003. Abstract viewer/itinerary planner. Washington, DC: Soc. Neurosci. Bronstein AM, Hood JD. 1986. The cervicoocular reflex in normal subjects and patients with absent vestibular function. Brain Res. 373(1–2):399–408 Brown LA, Shumway-Cook A, Wollacott MH. 1999. Attentional demands and postural recovery: the effects of aging. J. Gerontol. Med. Sci. 54A:M165–71

Campbell AJ, Borrie MJ, Spears GF, Jackson SL, Brown JS, Fitzgerald JL. 1990. Circumstances and consequences of fall experienced by a community population 70 years and over during a prospective study. Age Ageing 19:136–41 Chen HC, Schultz AB, Ashton-Miller JA, Giordani B, Alexander NB, et al. 1996. Stepping over obstacles: dividing attention impairs performance of old more than young adults. J. Gerontol. Med. Sci. 51A:M116–22 Clark B, Graybiel A. 1949a. The effect of angular acceleration on sound localization: the audiogyral illusion. J. Psychol. 28:235–44 Clark B, Graybiel A. 1949b. Linear acceleration and deceleration as factors influencing nonvisual orientation during flight. J. Aviat. Med. 20:92–101 Clark B, Graybiel A. 1949c. Apparent rotation of a fixed target associated with linear acceleration in flight. Am. J. Ophthalmol. 32:549– 57 Clark B, Graybiel A. 1961. Human performance during adaptation to stress in the Pensacola Slow Rotation Room. Aerosp. Med. 32:93–106 Cohen B, Wearne S, Dai M, Raphan T. 1999. Spatial orientation of the angular vestibuloocular reflex. J. Vestib. Res. 9(3):163–72 Collins JJ, Priplata AA, Gravelle DC, Niemi J, Harry J, et al. 2003. Noise-enhanced human sensorimotor function. IEEE Eng. Med. Biol. 22(2):76–83 DiZio P, Lackner JR. 1995. Motor adaptation to Coriolis force perturbations of reaching movements: endpoint but not trajectory adaptation transfers to the non-exposed arm. J. Neurophysiol. 74(4):1787–92 DiZio P, Lackner JR. 1997. Circumventing side effects of immersive virtual environments. In

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

SPATIAL ORIENTATION MECHANISMS Advances in Human Factors/Ergonomics, ed. MJ Smith, G Salvendy, RJ Koubek, 21:893– 97. Amsterdam: Elsevier DiZio P, Lackner JR. 2003. Sensory motor adaptation at high rotation speeds. J. Vestib. Res. 12:291–99 DiZio P, Lackner JR, Evanoff JN 1987. The influence of gravitoinertial force level on oculomotor and perceptual responses to Coriolis, cross-coupling stimulation. Aviat. Space Environ. Med. 58:A218–23 DiZio P, Lackner JR, Held RM, ShinnCunningham B, Durlach NI. 2001. Gravitoinertial force magnitude and direction influence head-centric auditory localization. J. Neurophysiol. 85:2455–60 Draganich LF, Zacny J, Klfta J, Karrison T. 2001. The effects of antidepressants on obstructed and unobstructed gait in healthy elderly people. J. Gerontol. Med. Sci. 56A(1):M36–41 Dulac S, Raymond JL, Sejnowski TJ, Lisberger SG. 1995. Learning and memory in the vestibuloocular reflex. Annu. Rev. Neurosci. 18:409–41 Durlach N, Allen G, Darken R, Garnett RL, Loomis J, Templeman J. 2000. Virtual environments and the enhancement of spatial behavior: toward a comprehensive research agenda. Presence 9:593–614 Ebenholtz SM, Shebilske W. 1975. The doll reflex: ocular counterrolling with headbody tilt in the median plane. Vision Res. 15(6):713–17 Fernandez C, Goldberg JM, Abend WK. 1972. Response to static tilts of peripheral neurons innervating otolith organs of the squirrel monkey. J. Neurobiol. 35:978–97 Fregly AR, Kennedy RS. 1965. Comparative effects of prolonged rotation at 10 rpm on postural equilibrium in vestibular normal and vestibular defective human subjects. Aerospace Med. 36(12):1160–67 Gavrilov VV, Wiener SI, Berthoz A. 1998. Discharge correlates of hippocampal complex spike neurons in behaving rats passively displaced on a mobile robot. Hippocampus 8:475–90

143

Gdowski GT, McCrea RA. 1999. Integration of vestibular and head movement signals in the vestibular nuclei during whole-body rotation. J. Neurophysiol. 81:436–49 Glasauer S, Amorim MA, Vitte E, Berthoz A. 1994. Goal-directed linear locomotion in normal and labyrinthine-defective subjects. Exp. Brain Res. 98(2):323–35 Goodwin GM, McCloskey DI, Matthews PBC. 1972. Proprioceptive illusions induced by muscle vibration: contribution by muscle spindles to perception? Science 175:1382– 84 Grant W, Best W. 1987. Otolith-organ mechanics: lumped parameter model and dynamic response. Aviat. Space Environ. Med. 58:970–76 Graybiel A. 1952. The oculogravic illusion. AMA Arch. Ophthalmol. 48:605–15 Graybiel A. 1969. Structural elements in the concept of motion sickness. Aerosp. Med. 40(4):351 Graybiel A, Clark B, Zarriello JJ. 1960. Observations on human subjects living in a “slow rotation room” for periods of two days. Arch. Neurol. 3:55–73 Graybiel A, Hupp DI. 1946. The oculogyral illusion: a form of apparent motion which may be observed following stimulation of the semicircular canals. Aerosp. Med. 3:1– 12 Graybiel A, Kennedy RS, Knoblock EC, Guedry FE, Mertz W, et al. 1965. Effects of exposure to a rotating environment (10 rpm) on four aviators for a period of twelve days. Aerosp. Med. 36(8):733–54 Graybiel A, Lackner JR. 1987. Treatment of severe motion sickness with antimotion sickness drug injections. Aviat. Space Environ. Med. 58:773–76 Graybiel A, Miller EF, Homick JL. 1977. Experiment M131. Human vestibular function. In Biomedical Results from Skylab, ed. RS Johnston, LF Dietlein, pp. 74–103. Washington, DC: Sci. Tech. Inform. Off., NASA Graybiel A, Miller EF, Newsom BD, Kennedy RS. 1968. The effect of water immersion on perception of the oculogravic illusion in

6 Dec 2004 10:37

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

144

AR

AR231-PS56-05.tex

LACKNER




AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

DIZIO

normal and labyrinthine-defective subjects. Acta Otolaryngol. 65:599–610 Graybiel A, Niven JI. 1951. The effect of a change in direction of resultant force on sound localization: the audiogravic illusion. J. Exp. Psychol. 42:227–30 Grigg P, Finerman GA, Riley LH. 1973. Position sense after total hip replacement. J. Bone Joint Surg. A55(5):1016–25 Guedry FE, Benson AJ. 1978. Coriolis crosscoupling effects: disorienting and nauseogenic or not. Aviat. Space Environ. Med. 49(1):29–35 Guedry FE, Collins WE, Graybiel A. 1964. Vestibular habituation during repetitive complex stimulation: a study of transfer effects. BuMed Proj. MR005.13–6001 Substask 1, Rep. No.93, NASA Order No. R-93. Pensacola, FL: Naval Sch. Aviat. Med. Guedry FE, Kennedy RS, Harris S, Graybiel A. 1962. Human performance during two weeks in a room rotating at three rpm. BuMed Proj. MR005.13–6001 Subtask 1, Rep. No.74, NASA Order R-47. Pensacola, FL: U.S. Naval Sch. Aviat. Inst. Guldin WO, Grusser O-J. 1998. Is there a vestibular cortex? Trends Neurosci. 21:254– 59 Hagbarth KE, Eklund G. 1966. Motor effects of vibratory stimuli in man. In Muscular Afferents and Motor Control, ed. R Granit, p. 177–86. Sweden: Almqvist & Wiksell Holden M, Ventura J, Lackner JR. 1994. Stabilization of posture by precision contact of the index finger. J. Vest. Res. 4(4):285–301 Howard IP, Templeton WB. 1966. Human Spatial Orientation. London: Wiley Imai T, Moore ST, Raphan T, Cohen B. 2001. Interaction of the body, head, and eyes during walking and turning. Exp. Brain Res. 136:1– 18 Israel I, Chapuis N, Glasauer S, Charade O, Berthoz A. 1993. Estimation of passive horizontal linear whole-body displacement in humans. J. Neurophysiol. 70(3):1270–73 Israel I, Grasso R, Georges-Francois P, Tsuzuku T, Berthoz A. 1997. Spatial memory and path integration studied by self-driven passive lin-

ear displacement. I. Basic properties. J. Neurophysiol. 77(6):3180–92 Jeka JJ, Lackner JR. 1994. Fingertip contact influences human postural control. Exp. Brain Res. 100:495–502 Jeka JJ, Schoner G, Dijkstra T, Ribeiro P, Lackner JR. 1997. Coupling of fingertip somatosensory information to head and body sway. Exp. Brain Res. 113:475–83 Jennings RT. 1997. Managing space motion sickness. J. Vestib. Res . 8:67–70 Johansson RS, Westling G. 1984. Roles of glabrous skin receptors and sensorimotor memory in automatic control of precision grip when lifting rougher or more slippery objects. Exp. Brain Res. 56:550–64 Johnson WH, Meek JC, Graybiel A. 1962. Effects of labyrinthectomy on canal sickness in squirrel monkey. Ann. Otol. Rhinol. Laryngol. 71:289–98 Johnson WH, Stubbs RA, Kelk GF, Franks WR. 1951. Stimulus required to produce motion sickness. 1. Preliminary report dealing with importance of head movements. J. Aviat. Med . 22:365–74 Jongkees LBW, Philipszoon AJ. 1962. Nystagmus provoked by linear accelerations. Acta Physiol. Pharmacol. Neerl. 10:239–47 Judge JO, King MB, Whipple R, Clive J, Wolfson LI. 1995. Dynamic balance in older persons: effects of reduced visual and proprioceptive input. J. Gerontol. A Biol. Sci. Med. Sci. 50(5):M263–70 Kahane P, Hoffmann D, Minotti L, Berthoz A. 2003. Reappraisal of the human vestibular cortex by cortical electrical stimulation study. Ann. Neurol. 54(5):615–24 Karnath HO, Sievering D, Fetter M. 1994. The interactive contribution of neck muscle proprioception and vestibular stimulation to subjective straight ahead orientation in man. Exp. Brain Res. 101(1):140–46 Kavounoudias A, Roll R, Roll JP. 1998. The plantar sole is a “dynamometric map” for human balance control. NeuroReport 9(14):3247–52 Kennedy RS, Graybiel A. 1962. Symptomatology during prolonged exposure in a

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

SPATIAL ORIENTATION MECHANISMS constantly rotating environment at a velocity of one revolution per minute. Aerosp. Med. 33:817–25 Kentala E, Vivas J, Wall C. 2003. Reduction of postural sway by use of a vibrotactile balance prosthesis prototype in subjects with vestibular deficits. Ann. Otol. Rhinol. Laryngol. 112(5):404–9 Klein BEK, Klein R, Lee KE, Cruickshanks KJ. 1998. Performance-based and self-assessed measures of visual function as related to history of falls, hip fractures, and measured gait time. Opthalmology 105:160–64 Lackner JR. 1988. Some proprioceptive influences on the perceptual representation of body shape and orientation. Brain 111:281– 97 Lackner JR. 1992. Spatial orientation in weightless environments. Perception 21:803–12 Lackner JR, DiZio P. 1989. Altered sensorimotor control of the head as an etiological factor in space motion sickness. Percept. Mot. Skills 68:784–86 Lackner JR, DiZio P. 1993. Multisensory, cognitive, and motor influences on human spatial orientation in weightlessness. J. Vestib. Res. 3(3):361–72 Lackner JR, DiZio P. 1994. Rapid adaptation to Coriolis force perturbations of arm trajectory. J. Neurophysiol. 72(1):299–313 Lackner JR, DiZio P. 1998. Spatial orientation as a component of presence: insights gained from non-terrestrial environments. Presence 7(2):1–12 Lackner JR, DiZio P. 2000. Human orientation and movement control in weightlessness and artificial gravity environments. Exp. Brain Res. 130:2–26 Lackner JR, DiZio P. 2004. Adaptation to rotating artificial gravity environments. J. Vestib. Res. 13(4–6):321–30 Lackner JR, DiZio P, Jeka JJ, Horak F, Krebs D, et al. 1999. Precision contact of the fingertip reduces postural sway of individuals with bilateral vestibular loss. Exp. Brain Res. 126:459–66 Lackner JR, Graybiel A, DiZio P. 1991. Altered sensorimotor control of the body as an etio-

145

logical factor in space motion sickness. Aviat. Space Environ. Med. 62:765–71 Lackner JR, Graybiel A. 1974. Elicitation of vestibular side effects by regional vibration of the head. Aerosp. Med. 45:1267–72 Lackner JR, Graybiel A. 1981. Illusions of postural, visual, and aircraft motion elicited by deep knee bends in the increased gravitoinertial force phase of parabolic flight. Exp. Brain Res. 44:312–16 Lackner JR, Graybiel A. 1982. Rapid perceptual adaptation to high gravitoinertial force levels: evidence for context-specific adaptation. Aviat. Space Environ. Med. 53:766– 69 Lackner JR, Graybiel A. 1984. Elicitation of motion sickness by head movements in the microgravity phase of parabolic flight maneuvers. Aviat. Space Environ. Med. 55:513– 20 Lackner JR, Graybiel A. 1986. Head movements made in non-terrestrial force environments elicit symptoms of motion sickness: implications for the etiology of space motion sickness. Aviat. Space Environ. Med. 57:443– 48 Lackner JR, Graybiel A. 1987. Head movements in low and high gravitoinertial force environments elicit motion sickness: implications for space motion sickness. Aviat. Space Environ. Med. 58:A212–17 Lackner JR, Graybiel A. 1994. Use of promethazine to hasten adaptation to provocative motion. J. Clin. Pharmacol . 34:644– 48 Lackner JR, Levine MS. 1978. Visual direction depends on the operation of spatial constancy mechanisms: the oculobrachial illusion. Neurosci. Lett. 7:207–12 Lackner JR, Levine MS. 1979. Changes in apparent body orientation and sensory localization induced by vibration of postural muscles: vibratory myesthetic illusions. Aviat. Space Environ. Med. 50(4):346–54 Lackner JR, Rabin E, DiZio P. 2000. Fingertip contact suppresses the destabilizing influence of leg muscle vibration. J. Neurophysiol. 84(5):2217–24

6 Dec 2004 10:37

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

146

AR

AR231-PS56-05.tex

LACKNER




AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

DIZIO

Lackner JR, Rabin E, DiZio P. 2004. Haptic stabilization of posture: time course. Exp. Brain Res. Submitted Lajoie Y, Teasdale N, Bard C, Fleury M. 1993. Attentional demands for static and dynamic equilibrium. Exp. Brain Res. 97:139–44 Leipzig RM, Cumming RG, Tinetti ME. 1999. Drugs and falls in older people: a systematic review and meta-analysis: I. psychotropic drugs. J. Am. Geriatr. Soc. 47:30–39 Lewald J, Karnath HO. 2000. Vestibular influence on human auditory space perception. J. Neurophysiol. 84(2):1107–11 Luchies CW, Alexander NB, Schultz AB, Ashton-Miller J. 1994. Stepping responses of young and old adults to postural disturbances: kinematics. J. Am. Geriatr. Soc. 42: 506–12 Mach E. 1875. Grundlinien der Lehre von den Bewegungsempfindungen. Leipzig: Engelmann Maki BE, Holliday PJ, Fernie GR. 1987. A posture control model and balance test for the prediction of relative postural stability. IEEE Trans. Biomed. Eng. MBE-334(10):797–810 Matthews PBC. 1972. Mammalian Muscle Receptors and Their Central Actions. London: Arnold Mayne R. 1974. A system concept of the vestibular organ. In Handbook of Sensory Physiology, Vol. VI, ed. HH Horhuber, pp. 493–580. New York: Springer-Verlag Mbourou GA, Lajoie Y, Teasdale N. 2003. Step length variability at gait initiation in elderly fallers and non-fallers, and young adults. Gerontologist 49(1):21–26 McIllroy WE, Maki B. 1996. Age-related changes in compensatory stepping in response to unpredictable perturbations. J. Gerontol. Med. Sci. 51A:M289–96 Merfeld DM, Zupan LH, Peterka RJ. 1999. Humans use internal models to estimate gravity and linear acceleration. Nature 398:615–18 Miller AD, Wilson VJ. 1983. “Vomiting center” reanalyzed in an electrical stimulation study. Brain Res . 270:154–61 Miller EF II, Graybiel A. 1962. Counterrolling of the human eyes produced by head tilt with

respect to gravity. Acta Otolaryngol. 54:479– 501 Miller EF II, Graybiel A. 1966. Magnitude of gravitoinertial force, an independent variable in egocentric visual localization of the horizontal. J. Exp. Psychol. 71(3):452–60 Mittelstaedt H. 1980. Homing by path integration in mammals. Naturwissenschaften 67:566 Niven JI, Graybiel A. 1953. Postural equilibrium in man following unilateral labyrinthectomy. NSAM-435. Pensacola, FL: Naval Sch. Aviat. Med. O’Keefe J, Dostrovsky J. 1971. The hippocampus as a spatial map: preliminary evidence from unit activity in the freely-moving rat. Brain Res. 34:171–75 Overstall PW, Exton-Smith AN, Imms FJ, Johson AL. 1977. Falls in the elderly related to postural imbalance. Br. Med. J. 1:261–64 Pai YC, Patton J. 1997. Center of mass velocityposition predictions for balance control. J. Biomech. 30(4):347–54 Paige GD, Tomko DL. 1991. Eye-movement responses to linear head motion in the squirrelmonkey. 1. Basic characteristics. J. Neurophysiol. 65(5):1170–82 Persad CC, Giordani B, Chen HC, AshtonMiller JA, Alexander NB, et al. 1995. Neuropsychological predictors of complex obstacle avoidance in healthy older adults. J. Gerontol. B Psychol. Sci. Soc. Sci. 50(5): P272–77 Pigeon P, Bortolami SB, DiZio P, Lackner JR. 2003a. Coordinated turn and reach movements. I. Anticipatory compensation for selfgenerated Coriolis and interaction torques. J. Neurophysiol. 89:276–89 Pigeon P, Bortolami SB, DiZio P, Lackner JR. 2003b. Coordinated turn and reach movements. II. Planning in an external frame of reference. J. Neurophysiol. 89:290–303 Rabin E, DiZio P, Lackner JR. 2004. Time course of visual stabilization of posture. Exp. Brain Res. In press Reason JT. 1970. Motion sickness: a special case of sensory rearrangement. Adv. Sci. 26(130):386–93

6 Dec 2004 10:37

AR

AR231-PS56-05.tex

AR231-PS56-05.sgm

LaTeX2e(2002/01/18)

P1: IKH

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

SPATIAL ORIENTATION MECHANISMS Roberts TDM. 1978. Neurophysiology of Postural Mechanisms. London: Butterworth. 2nd ed. Roy JE, Cullen KE. 1998. A neural correlate for vestibuloocular reflex suppression during voluntary eye-head gaze shifts. Nat. Neurosci. 1:404–10 Roy JE, Cullen KE. 2001. Vestibuloocular reflex signal modulation during voluntary and passive head movements. J. Neurosci. 87:2337–57 Rubenstein LZ, Robbins AS, Schulman BL, Rosado J, Osterweil D, et al. 1988. Falls and instability in the elderly. J. Am. Geriatr. Soc. 36:266–78 Ruddle RA, Payne SJ, Jones DM. 1997. Navigating buildings in desk-top virtual environments: experimental investigations using extended navigational experience . J. Exp. Psychol. Appl. 3:143–59 Rupert AH. 2000. Tactile situation awareness system: proprioceptive prostheses for sensory deficiencies. Aviat. Space Environ. Med. 71(9 Suppl.):A92–99 Schultheis LW, Robinson DA. 1981. Directional plasticity of the vestibuloocular reflex in the cat. Ann. NY Acad. Sci. 374:504–12 Shumway-Cook A, Woollacott MH, Kerns KA, Baldwin M. 1997. The effects of two types of cognitive tasks on postural stability in older adults with and without a history of falling. J. Gerontol. Med. Sci. 52A:M232–40 Siegel AW, White SH. 1975. The development of spatial representations of large-scale environments. In Advances in Child Development and Behavior, ed. H Reeses, 10:10–55. New York: Academic Taube JS, Muller RU, Ranck JB. 1990. Head direction cells recorded from the postsubiculum in freely moving rats. I. Description and quantitative analysis. J. Neurosci. 10:420– 35 Telford L, Seidman SH, Paige GD. 1997. Dynamics of squirrel monkey linear vestibuloocular reflex and interactions with fixation distance. J. Neurophysiol. 78(4):1775–90 Thelen DG, Schultz AB, Alexander NB,

147

Ashton-Miller JA. 1996. Effects of age on rapid ankle torque development. J. Gerontol. 51:M226–32 Thelen DG, Wojcik LA, Schultz AB, AshtonMiller JA, Alexander NB. 1997. Age differences in using a rapid step to regain balance during a forward fall. J. Gerontol. A Biol. Sci. Med. Sci. 52:M8–13 Vallar G, Sterzi R, Bottini G, Cappa S, Rusconi ML. 1990. Temporary remission of left hemianesthesia after vestibular stimulation—a sensory neglect phenomenon. Cortex 26 (1):123–31 Waller D, Loomis JM, Steck SD. 2003. Inertial cues do not enhance knowledge of environmental layout. Psychon. Bull. Rev. 10(4):987–93 Wehner R. 2003. Desert ant navigation: how miniature brains solve complex tasks. J. Comp. Physiol. A Neuroethol. Sens. Neural Behav. Physiol. 189(8):579–88 Wenzel R, Bartensteing P, Dieterich M, Danek A, Weindl A, et al. 1996. Deactivation of human visual cortex by involuntary ocular oscillations—a PET activation study. Brain 119:101–10 Whiteside TCD. 1961. Hand-eye coordination in weightlessness. Aerosp. Med. 32:719–25 Whiteside TCD, Graybiel A, Niven JI. 1965. Visual illusions of movement. Brain 88:193– 219 Wilson VJ, Boyle R, Fukushima K, Rose PK, Shinoda Y, et al. 1995. The vestibulocollic reflex. J. Vestib. Res. 5(3):147–70 Wilson VJ, Melvill Jones G. 1979. Mammalian Vestibular Physiology. New York: Plenum Wilson VJ, Peterson BW. 1978. Peripheral and central substrates of vestibulospinal reflexes. Physiol. Rev. 58(1):80–105 Witmer BG, Bailey JH, Knerr BW, Parsons KC. 1996. Virtual spaces and real world places: transfer of route knowledge. Int. J. Hum. Comp. Stud. 45:413–28 Young LR. 1999. Artificial gravity considerations for a Mars exploration mission. Otolith function in spatial orientation and movement. Ann. NY Acad. Sci. 871:367–78

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

Annual Review of Psychology Volume 56, 2005

CONTENTS

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Frontispiece—Richard F. Thompson

xviii

PREFATORY In Search of Memory Traces, Richard F. Thompson

1

DECISION MAKING Indeterminacy in Brain and Behavior, Paul W. Glimcher

25

BRAIN IMAGING/COGNITIVE NEUROSCIENCE Models of Brain Function in Neuroimaging, Karl J. Friston

57

MUSIC PERCEPTION Brain Organization for Music Processing, Isabelle Peretz and Robert J. Zatorre

89

SOMESTHETIC AND VESTIBULAR SENSES Vestibular, Proprioceptive, and Haptic Contributions to Spatial Orientation, James R. Lackner and Paul DiZio

115

CONCEPTS AND CATEGORIES Human Category Learning, F. Gregory Ashby and W. Todd Maddox

149

ANIMAL LEARNING AND BEHAVIOR: CLASSICAL Pavlovian Conditioning: A Functional Perspective, Michael Domjan

179

NEUROSCIENCE OF LEARNING The Neuroscience of Mammalian Associative Learning, Michael S. Fanselow and Andrew M. Poulos

207

HUMAN DEVELOPMENT: EMOTIONAL, SOCIAL, AND PERSONALITY Behavioral Inhibition: Linking Biology and Behavior Within a Developmental Framework, Nathan A. Fox, Heather A. Henderson, Peter J. Marshall, Kate E. Nichols, and Melissa A. Ghera

235

BIOLOGICAL AND GENETIC PROCESSES IN DEVELOPMENT Human Development: Biological and Genetic Processes, Irving I. Gottesman and Daniel R. Hanson

263 vii

P1: JRX

December 8, 2004

viii

12:13

Annual Reviews

AR231-FM

CONTENTS

SPECIAL TOPICS IN PSYCHOPATHOLOGY The Psychology and Neurobiology of Suicidal Behavior, Thomas E. Joiner Jr., Jessica S. Brown, and LaRicka R. Wingate

287

DISORDERS OF CHILDHOOD Autism in Infancy and Early Childhood, Fred Volkmar, Kasia Chawarska, and Ami Klin

315

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

CHILD/FAMILY THERAPY Youth Psychotherapy Outcome Research: A Review and Critique of the Evidence Base, John R. Weisz, Amanda Jensen Doss, and Kristin M. Hawley

337

ALTRUISM AND AGGRESSION Prosocial Behavior: Multilevel Perspectives, Louis A. Penner, John F. Dovidio, Jane A. Piliavin, and David A. Schroeder

365

INTERGROUP RELATIONS, STIGMA, STEREOTYPING, PREJUDICE, DISCRIMINATION The Social Psychology of Stigma, Brenda Major and Laurie T. O’Brien

393

PERSONALITY PROCESSES Personality Architecture: Within-Person Structures and Processes, Daniel Cervone

423

PERSONALITY DEVELOPMENT: STABILITY AND CHANGE Personality Development: Stability and Change, Avshalom Caspi, Brent W. Roberts, and Rebecca L. Shiner

453

WORK MOTIVATION Work Motivation Theory and Research at the Dawn of the Twenty-First Century, Gary P. Latham and Craig C. Pinder

485

GROUPS AND TEAMS Teams in Organizations: From Input-Process-Output Models to IMOI Models, Daniel R. Ilgen, John R. Hollenbeck, Michael Johnson, and Dustin Jundt

517

LEADERSHIP Presidential Leadership, George R. Goethals

545

PERSONNEL EVALUATION AND COMPENSATION Personnel Psychology: Performance Evaluation and Pay for Performance, Sara L. Rynes, Barry Gerhart, and Laura Parks

571

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

CONTENTS

ix

PSYCHOPHYSIOLOGICAL DISORDERS AND PSYCHOLOGICAL EFFECTS ON MEDICAL DISORDERS Psychological Approaches to Understanding and Treating Disease-Related Pain, Francis J. Keefe, Amy P. Abernethy, and Lisa C. Campbell

601

TIMELY TOPIC

Annu. Rev. Psychol. 2005.56:115-147. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Psychological Evidence at the Dawn of the Law’s Scientific Age, David L. Faigman and John Monahan

631

INDEXES Subject Index Cumulative Index of Contributing Authors, Volumes 46–56 Cumulative Index of Chapter Titles, Volumes 46–56

ERRATA An online log of corrections to Annual Review of Psychology chapters may be found at http://psych.annualreviews.org/errata.shtml

661 695 700

23 Dec 2004 19:45

AR

AR231-PS56-06.tex

AR231-PS56-06.sgm

P1: IKH LaTeX2e(2002/01/18) 10.1146/annurev.psych.56.091103.070217

Annu. Rev. Psychol. 2005. 56:149–78 doi: 10.1146/annurev.psych.56.091103.070217 c 2005 by Annual Reviews. All rights reserved Copyright
First published online as a Review in Advance on September 2, 2004

HUMAN CATEGORY LEARNING F. Gregory Ashby

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Department of Psychology, University of California, Santa Barbara, California 93106; email: [email protected]

W. Todd Maddox Department of Psychology, University of Texas, Austin, Texas 78712; email: [email protected]

Key Words exemplar, prototype, decision bound, multiple systems, striatum ■ Abstract Much recent evidence suggests some dramatic differences in the way people learn perceptual categories, depending on exactly how the categories were constructed. Four different kinds of category-learning tasks are currently popular—rulebased tasks, information-integration tasks, prototype distortion tasks, and the weather prediction task. The cognitive, neuropsychological, and neuroimaging results obtained using these four tasks are qualitatively different. Success in rule-based (explicit reasoning) tasks depends on frontal-striatal circuits and requires working memory and executive attention. Success in information-integration tasks requires a form of procedural learning and is sensitive to the nature and timing of feedback. Prototype distortion tasks induce perceptual (visual cortical) learning. A variety of different strategies can lead to success in the weather prediction task. Collectively, results from these four tasks provide strong evidence that human category learning is mediated by multiple, qualitatively distinct systems.

CONTENTS INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . EARLY CATEGORY-LEARNING THEORIES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CATEGORY-LEARNING TASKS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . RULE-BASED TASKS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Neuropsychological Patient Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Neuroimaging Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Theories of Rule-Based Category Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . INFORMATION-INTEGRATION TASKS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Category-Learning Limits in Information-Integration Tasks . . . . . . . . . . . . . . . . . . Neuropsychological Patient Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Neuroimaging Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Theories of Information-Integration Category Learning . . . . . . . . . . . . . . . . . . . . . . 0066-4308/05/0203-0149$14.00

150 152 152 155 155 156 156 157 158 159 160 160 161

149

23 Dec 2004 19:45

150

AR

ASHBY

AR231-PS56-06.tex




AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

P1: IKH

MADDOX

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

DISSOCIATIONS BETWEEN RULE-BASED AND INFORMATION-INTEGRATION CATEGORY LEARNING . . . . . . . . . . . . . . . . . PROTOTYPE DISTORTION TASKS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Neuropsychological Patient Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Neuroimaging Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Theories of Prototype Distortion Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . WEATHER PREDICTION TASK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Task and Individual Differences Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Neuropsychological Patient and Neuroimaging Studies . . . . . . . . . . . . . . . . . . . . . . CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

163 165 166 166 167 169 170 171 172

INTRODUCTION Is the plant edible or poisonous? Is the person friend or foe? Was the sound made by a predator or by the wind? All organisms assign objects and events in the environment to separate classes or categories. This allows them to respond differently, for example, to nutrients and poisons, and to predators and prey. Any species lacking this ability would quickly become extinct. Given the important role that categorization plays in our day-to-day lives, it is not surprising that there is a huge and old literature on the perceptual, cognitive, and neurobiological processes that mediate this vital skill. This article surveys that literature, with an emphasis on discoveries and developments during the past 10 years. The past decade has seen exciting and profound changes in categorization research. Two important new themes have fundamentally changed the field. First, there has been huge attention, both theoretical and empirical, devoted to the question of whether human category learning is mediated by a single system, or by multiple, qualitatively distinct learning systems. Second, the categorization field has whole-heartedly embraced the cognitive neuroscience revolution. Not only has the past 10 years seen a massive increase in neuropsychological and neuroimaging research on category learning, but this new knowledge also has permeated into, and significantly sharpened, the core theories in the field. These twin themes— multiple systems and cognitive neuroscience—organize and motivate much of the present review. The categorization literature is enormous, and no single article could survey it all. Thus, we focus on a restricted subset of the entire literature. In particular, this article focuses on how humans learn perceptual categories. By restricting ourselves in this way, we must necessarily ignore a number of large and interesting subliteratures. First, we do not discuss category learning in nonhuman animals (Smith et al. 2004). The animal literature is especially relevant to understanding the neural basis of human category learning. Unfortunately, however, this literature is fractionated, largely because, within the behavioral neuroscience literature at least, relatively few papers address animal categorization per se. Rather, the relevant results come from a wide variety of tasks and phenomena (e.g., discrimination learning, memory systems, long-term potentiation). For this reason, perhaps, we know of no

23 Dec 2004 19:45

AR

AR231-PS56-06.tex

AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN CATEGORY LEARNING

P1: IKH

151

recent comprehensive review (although, for reviews of the behavioral literature, see Vauclair 2002 or the 2002 special issue of the Journal of the Experimental Analysis of Behavior). Second, our focus on learning prevents us from considering the categorization behavior of highly experienced experts. This distinction is important because there is good evidence that the neural mechanisms and pathways that mediate the learning of new categories are different from the neural structures that mediate the representation of highly learned categories. For example, many neuropsychological groups that are impaired in category learning (e.g., frontal patients and Parkinson’s disease patients) do not lose old, familiar categories (e.g., fruits and tools). Similarly, there is no evidence that people who lose a familiar category (i.e., who develop a category-specific agnosia) develop any general category-learning deficits. Readers interested in the representation of highly learned categories are referred to any of several excellent reviews of the category representation literature (e.g., Cree & McRae 2003, Humphreys & Forde 2001, Joseph 2001). Third, we discuss how people learn new categories, but not how they use this new information in other cognitive tasks. For example, people use categorical information to make inferences about unobserved features of a stimulus, to facilitate decision making, and to problem solve (e.g., Lewandowsky et al. 2002, Markman & Ross 2003). Finally, our focus is on perceptual categories rather than concepts. By “perceptual category,” we mean a collection of similar objects belonging to the same group. Although the term “concept” often is used interchangeably with category, we refer to a concept as a group of related ideas. For example, the set of all X rays displaying a tumor forms a perceptual category, whereas the many varieties of religious experience form a concept. Although many of the results reviewed below are relevant to understanding both perceptual categorization and concept formation, the representation of categories and concepts is likely different, and a review of both literatures is beyond the scope of this article. Readers interested in concepts are referred to any of several excellent recent reviews (e.g., Barsalou 2003, Murphy 2002). If the goal is to study category learning rather than category representation, then it is necessary to present subjects with unfamiliar categories and observe their behavior during the period when their ability to assign stimuli to these categories rises from chance to some stable level. In experiments with adults, the prevailing method of ensuring unfamiliarity is for the experimenter to create new, arbitrary categories of objects (so-called “artificial categories”). In the past, little attention was paid to the manner in which these arbitrary categories were created. However, much recent evidence suggests some dramatic differences in the way people learn such categories, depending on exactly how the categories are constructed. In fact, these differences are so profound that we take the unusual step of organizing all the research that we review by the type of task that was used. Toward this end, we focus on four different kinds of category learning tasks—rule-based tasks, information-integration tasks, prototype distortion tasks, and the so-called weather

23 Dec 2004 19:45

152

AR

ASHBY

AR231-PS56-06.tex




AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

P1: IKH

MADDOX

prediction task. The next section briefly reviews some important early categorylearning theories. The third section describes the four basic tasks, and then sections four through eight review results from each of these tasks. Finally, we close with some general conclusions.

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

EARLY CATEGORY-LEARNING THEORIES Many theories of human category learning have been proposed. The early theories virtually all assumed that humans have a single category-learning system that they use to learn all types of categories (for an exception, see Brooks 1978). A few of these theories are still important in the sense that they continue to motivate significant amounts of new research. This section briefly introduces the most important of these theories. Other sources should be consulted for a more complete discussion, and for a more thorough review of category-learning theories (e.g., Ashby & Maddox 1998, Estes 1994, Smith & Medin 1981). Prototype theory assumes that category learning is equivalent to learning the category prototype. When an unfamiliar stimulus is then encountered, it is assigned to the category with the most similar prototype (Homa et al. 1981; Posner & Keele 1968, 1970; Reed 1972; Rosch 1973, 1975; Smith & Minda 1998). Exemplar theory assumes that category learning is a process of learning about the exemplars that belong to the category. When an unfamiliar stimulus is encountered, its similarity is computed to the memory representation of every previously seen exemplar from each potentially relevant category. The stimulus is then assigned to the category for which the sum of these similarities is greatest (Brooks 1978; Estes 1986, 1994; Hintzman 1986; Lamberts 2000; Medin & Schaffer 1978; Nosofsky 1986). Decision bound theory assumes subjects partition the stimulus space into response regions. When presented with an unfamiliar stimulus, the subject determines which region the percept is in, and then emits the associated response. The partition between regions associated with competing responses is called the decision bound. Category learning is the process of either learning the decision bound or else of learning the regions associated with each response (Ashby & Gott 1988, Ashby & Townsend 1986, Maddox & Ashby 1993).

CATEGORY-LEARNING TASKS As mentioned above, this article focuses on four different kinds of category learning tasks. Rule-based tasks are those in which the categories can be learned via some explicit reasoning process. Frequently, the rule that maximizes accuracy (i.e., the optimal strategy) is easy to describe verbally (Ashby et al. 1998). In the most common applications, only one stimulus dimension is relevant, and the subject’s task is to discover this relevant dimension and then to map the different

23 Dec 2004 19:45

AR

AR231-PS56-06.tex

AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN CATEGORY LEARNING

P1: IKH

153

dimensional values to the relevant categories. However, there is no requirement that the rule that maximizes accuracy (i.e., the optimal rule) in rule-based tasks is one-dimensional. For example, a conjunction rule (e.g., respond A if the stimulus is small on dimension x and small on dimension y) is a rule-based task because it is easy to describe verbally. Information-integration category learning tasks are those in which accuracy is maximized only if information from two or more stimulus components (or dimensions) is integrated at some predecisional stage (Ashby & Gott 1988). Perceptual integration could take many forms—from computing a weighted linear combination of the dimensional values to treating the stimulus as a gestalt. In many cases, the optimal strategy in information-integration tasks is difficult or impossible to describe verbally (Ashby et al. 1998). Real-world examples of informationintegration tasks are common. For example, deciding whether an X ray shows a tumor requires years of training, and expert radiologists are only partially successful at describing their categorization strategies. Examples of rule-based and information-integration categories that might be used in experimental research are shown in Figure 1. In both cases, the two contrasting categories are composed of circular sine-wave gratings (i.e., disks in which luminance varies sinusoidally). Examples are shown in Figure 1a. The disks are all of equal diameter, but they differ in spatial frequency (i.e., the frequency of the sine wave) and (sine-wave) orientation. Each symbol in Figures 1b and 1c denotes the spatial frequency and orientation of a sine-wave grating. Category A exemplars are denoted by pluses and category B exemplars are denoted by circles. In each condition, there are two distinct categories that do not overlap, so perfect accuracy is possible. Also shown in Figures 1b and 1c are the decision bounds that maximize categorization accuracy. In the rule-based task (Figure 1b), the optimal bound, denoted by the vertical line in Figure 1b, requires observers to attend to spatial frequency and ignore orientation. This bound has a simple verbal description: “Respond A if the bars are thick and B if they are thin.” In the information-integration task (Figure 1c), which was generated by rotating the rule-based categories by 45◦ , equal attention must be allocated to both stimulus dimensions. In this task, there is no simple verbal description of the optimal decision bound. Note that we use the word “rule” more narrowly than is common in the psychological literature, where it is often used to refer to any strategy from an explicit reasoning process to any algorithm that can be expressed formally. In particular, we define “rule-based strategy” narrowly to refer specifically to an explicit reasoning process. Note that according to this criterion, there is no limit on the complexity of the optimal rule in rule-based tasks. However, as the complexity of the optimal rule increases, its salience decreases and it becomes less likely that observers will learn the associated categories through an explicit reasoning process. Thus, the boundary is fuzzy between rule-based and information-integration tasks. Tasks in which the optimal rule is one-dimensional are unambiguously rule-based (at least with separable stimulus dimensions), and tasks in which the optimal rule is

23 Dec 2004 19:45

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

154

AR

ASHBY

AR231-PS56-06.tex




AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

P1: IKH

MADDOX

significantly more complex than a conjunction rule almost never are rule-based. In between, the classification is not so clear-cut. It is also important to emphasize that the terms “rule-based” and “informationintegration” make no assumptions about how people learn these different category structures in any particular application. For example, there is evidence that pigeons can learn both types of category structures (Herbranson et al. 1999), but no one would claim that they learn rule-based categories via an explicit reasoning process. The question of how people learn rule-based and information-integration categories is strictly empirical. As such, this particular classification of categorization tasks is useful only because there are many interesting empirical dissociations between the two tasks (e.g., Ashby et al. 1999, 2002, 2003; Ashby & Waldron 1999; Maddox et al. 2003). Prototype distortion tasks are a third type of category-learning task in which each category is created by first constructing a category prototype (Posner & Keele 1968, 1970). The other exemplars of the category are then created by randomly distorting the prototype. In the most popular prototype distortion task, each stimulus is a random pattern of (often nine) dots. One pattern is selected as the category prototype and then the other category exemplars are created by randomly perturbing the location of each dot in the prototype. Examples are shown in Figure 2. The final category-learning task we consider is the so-called weather prediction task. Stimuli in this task are tarot cards; each displays a unique geometric pattern. The subject’s task is to decide if the particular constellation of cards that is shown signals “rain” or “sun.” The actual outcome is determined by a probabilistic rule based on the individual cards.

Figure 2

Stimuli that might be used in a prototype distortion category-learning task.

23 Dec 2004 19:45

AR

AR231-PS56-06.tex

AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

HUMAN CATEGORY LEARNING

P1: IKH

155

RULE-BASED TASKS

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Introduction As mentioned above, rule-based tasks are those in which it is easy for subjects to describe the optimal strategy verbally. In general, several conditions must be met before a verbal description is possible. First, a semantic label must correspond to each of the stimulus properties that are relevant to the decision. In the Figure 1b rule-based task, the critical stimulus feature has the semantic label “width.” Second, the subject must be able to attend selectively to each relevant stimulus property. For example, it is possible to verbalize a rule such as “Respond A if the saturation of the color patch is high, respond B if saturation is low.” Even so, people are not good at attending selectively to saturation and ignoring irrelevant variations in hue and brightness, so it is unlikely that people would spontaneously experiment with such rules. In the selective attention literature, a stimulus feature that can be attended to selectively is said to be separable from the other stimulus features, whereas features for which it is difficult or impossible to attend selectively are said to be integral. A large and old literature is devoted to this topic (Ashby & Maddox 1994, Ashby & Townsend 1986, Garner 1974, Lockhead 1966, Shepard 1964). The third critical property necessary for easy verbalization is that the rule for combining information from the relevant stimulus features is itself verbalizable. In general, this requires that separate decisions are first made about the level of each feature, and then these separate decisions are combined using logical operations, such as “and” and “or.” Perhaps the most obvious example is a conjunction rule of the sort: “Respond A if the bars are wide and the orientation is steep; otherwise respond B.” Note that to apply this rule the subject must first decide if the bars are narrow or wide and if the orientation is shallow or steep. Next, the outcomes of these two decisions are combined using the word “and.” Thus, information from the various relevant stimulus dimensions is combined after decisions are first made about each dimension. This is in contrast to information-integration tasks in which the raw perceptual information from the relevant stimulus dimensions is combined before any decisions are made. Other rule-based strategies that require combining decisions from separate dimensions include disjunctive and exclusive-or rules. There is no doubt that healthy adults can learn these rules without much difficulty, at least if they are given regular feedback about their response accuracy (e.g., Salatas & Bourne 1974). However, it is also quite clear that such combination rules are much less salient than simple one-dimensional rules, in the sense that people rarely experiment with such rules unless compelled to do so by the feedback they receive (Alfonso-Reese 1996, Ashby et al. 1998). Virtually all category-learning tasks used in neuropsychological assessment are rule-based, including the widely known Wisconsin Card Sorting Test (WCST; Heaton 1981). Stimuli in the WCST are cards containing geometric patterns that vary in color, shape, and symbol number, and in all cases, the correct categorization

23 Dec 2004 19:45

156

AR

ASHBY

AR231-PS56-06.tex




AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

P1: IKH

MADDOX

rule is one-dimensional (and easy to describe verbally). Perseverative errors on the WCST are a classic symptom of frontal dysfunction (e.g., Kimberg et al. 1997).

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Neuropsychological Patient Data Although categorization has been studied in many different neuropsychological groups, the most extensive data come primarily from studies with three different groups: (a) patients with frontal lobe lesions, (b) patients suffering from a disease of the basal ganglia, typically either Parkinson’s or Huntington’s disease, and (c) patients with amnesia. Within this latter group, the most theoretically interesting are those whose amnesia was caused by damage to the medial temporal lobes. In almost all cases, however, studies with amnesiacs include a wide variety of patients, typically including some with Korsakoff’s syndrome and some with medial temporal lobe amnesia. One characteristic feature of the neuropsychological literature on category learning is its inconsistency. For each major patient group, some studies report deficits and some do not. However, as we will see, when the existing studies are partitioned according to the type of task that was used, the discrepancies largely disappear. As mentioned above, perseverative responding on the WCST is among the most classic of all signs of frontal damage. Not surprisingly then, many studies have shown that frontal patients are impaired at rule-based category learning (see, e.g., Kimberg et al. 1997, Robinson et al. 1980). Another group with well-known deficits in rule-based category learning is Parkinson’s disease patients (e.g., Ashby et al. 2003, Brown & Marsden 1988, Cools et al. 1984, Downes et al. 1989). Although later in the disease Parkinson’s patients have frontal damage (primarily the result of cell death in the ventral tegmental area), the disease mainly targets the basal ganglia. The region most affected appears to be the head of the caudate nucleus (van Domburg & ten Donkelaar 1991), which is reciprocally connected to the prefrontal cortex. Thus, the rule-based category-learning deficits of frontal and Parkinson’s disease patients are consistent with the hypothesis that rule-based category learning is mediated, in part, by frontal-striatal circuits (Ashby et al. 1998). In contrast to frontal and basal ganglia disease patients, several studies have reported that amnesiacs with medial temporal lobe damage are normal in rulebased category learning (Janowsky et al. 1989, Leng & Parkin 1988). An obvious possibility is that many rule-based tasks are simple enough (e.g., the WCST) that working memory is sufficient for subjects to keep track of which alternative rules they have tested and rejected. If so, then a natural prediction is that medial temporal lobe amnesiacs should be impaired in complex rule-based tasks (e.g., when the optimal rule is disjunctive).

Neuroimaging Data A number of neuroimaging studies have used the WCST or a rule-based task similar to the WCST. All of these have reported task-related activation in prefrontal cortex,

23 Dec 2004 19:45

AR

AR231-PS56-06.tex

AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN CATEGORY LEARNING

P1: IKH

157

most have reported activation in the head of the caudate nucleus, and at least one has reported task-related activation in the anterior cingulate (Konishi et al. 1999, Lombardi et al. 1999, Rao et al. 1997, Rogers et al. 2000, Volz et al. 1997). Converging evidence for the hypothesis that these are important structures in rulebased category learning comes from several sources. First are the many studies that have implicated these structures as key components of executive attention (Posner & Petersen 1990) and working memory (Goldman-Rakic 1987, 1995), both of which are likely to be critically important to the explicit processes of rule formation and testing that are assumed to mediate rule-based category learning. Second, a recent neuroimaging study identified the (dorsal) anterior cingulate as the site of hypothesis generation in a rule-based category-learning task (Elliott et al. 1999). Third, lesion studies in rats implicate the dorsal caudate nucleus in rule switching (Winocur & Eskes 1998). Fourth, of course, are the neuropsychological data reviewed above, which show that patient groups with damage to any of these structures are impaired in rule-based tasks.

Theories of Rule-Based Category Learning Theories of rule-based category learning can be classified according to whether they assume that learning in rule-based tasks is not fundamentally different from other category-learning tasks, or whether they assume that rule-based learning is special. Most of the attempts to account for the results of rule-based category learning with a single system model have been by exemplar theorists. According to exemplar theory, rule-based tasks, in general, are no different from any other type of category-learning task. However, one-dimensional rules, like the one depicted in Figure 1b, are unique in that they encourage selective attention to a single dimension, which in turn dramatically affects the stimulus-exemplar similarity computations (Kruschke 1992, Nosofsky 1991, Nosofsky et al. 1989). In particular, increasing attention to a dimension will serve to increase the perceived differences on that dimension. As a result, the perceived separation between the Figure 1b rule-based categories will tend to be greater than the perceived separation between the Figure 1c information-integration categories. It is in this way that exemplar theory is able to account for the difficulty differences between the two tasks. The earliest theory that applies directly to rule-based tasks is the so-called classical theory of categorization (e.g., Bruner et al. 1956, Smith & Medin 1981), which assumes that every category is represented by a set of necessary and sufficient features. When a stimulus is presented for categorization, the subject is assumed to retrieve the feature list of the relevant categories and then test whether the stimulus features match one of these feature lists. This theory accounts for performance in many rule-based tasks. For example, category A in Figure 1b is defined by the necessary and sufficient feature “thick bars.” Similarly, the conjunction rule: “Respond A if the bars are thick and the orientation is shallow” is equivalent to the necessary and sufficient features “thick bars and shallow orientation.” On the

23 Dec 2004 19:45

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

158

AR

ASHBY

AR231-PS56-06.tex




AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

P1: IKH

MADDOX

other hand, in some rule-based tasks the optimal rule cannot be expressed as a set of necessary and sufficient conditions. For example, given appropriate feedback, subjects can learn disjunctive-or rules such as: “Respond A if the bars are thick and the orientation is shallow or if the bars are thin and the orientation is steep” (Salatas & Bourne 1974). Thus, it was recognized long ago that the classical theory is incomplete, even for the restricted set of rule-based tasks (e.g., Ashby & Maddox 1998, Smith & Medin 1981). As originally proposed, classical theory was meant to apply to all categorization tasks. There have been several attempts to modernize the theory. Each of these attempts, however, has assumed that rule-based category learning is only one of several category-learning systems that humans have available (Ashby et al. 1998, Brooks 1978, Erickson & Kruschke 1998, Nosofsky et al. 1994b). The rule-based components of these various models are similar. The model that perhaps is best developed, and the only one with a neuropsychological basis, was proposed by Ashby et al. (1998) as part of the COVIS (COmpetition between Verbal and Implicit Systems) model of category learning (which also includes a procedural learning component). The COVIS explicit system assumes that rule-based category learning is mediated primarily by an explicit, hypothesis-testing system that depends heavily on working memory and executive attention. The idea is that candidate rules are stored in working memory during the time they are being tested. COVIS assumes that subjects will continue to use the active rule until feedback or other evidence disconfirms its validity. At this point, a new rule must be instantiated. COVIS assumes that activating a new rule requires two separate processes. First, a new candidate rule must be identified or selected, and second, attention must be switched from the old rule to the new rule. The probability that any given rule will be instantiated is determined by its reinforcement history (which determines its overall salience), the tendency of the subject to select novel hypotheses (which is assumed to depend on cortical dopamine levels; Ashby et al. 1999), and the tendency of the subject to perseverate (which is assumed to depend on basal ganglia dopamine levels). Ashby and his colleagues proposed, and presented evidence in support of the hypothesis, that the selection operation is mediated cortically, by the anterior cingulate and possibly also by the prefrontal cortex, and that switching is mediated by the head of the caudate nucleus. A review of this evidence is beyond the scope of this article (see Ashby et al. 1998, 1999).

INFORMATION-INTEGRATION TASKS As mentioned above, information-integration category-learning tasks are defined as those in which accuracy is maximized only if information from two or more stimulus components (or dimensions) is integrated at some predecisional stage. Typically, the optimal rule is difficult or impossible to describe verbally (Ashby et al. 1998).

23 Dec 2004 19:45

AR

AR231-PS56-06.tex

AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

HUMAN CATEGORY LEARNING

P1: IKH

159

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Category-Learning Limits in Information-Integration Tasks An important theoretical and practical question is whether there are limits on the complexity of information-integration category structures that can be learned. One of the first research efforts to address this question focused on comparing learning in linearly and nonlinearly separable categories. A pair of categories is linearly separable if optimal performance can be achieved by making category decisions based on the magnitude of a linear combination of dimensional values (or equivalently, if a linear decision bound is optimal). Categories are nonlinearly separable if optimal performance depends on a nonlinear combination of dimensional values (i.e., if the optimal bound is nonlinear). Prototype theory predicts that nonlinearly separable category structures should be impossible to learn, at least if each category contains only a single prototype (Ashby & Gott 1988), whereas exemplar models predict no consistent advantage for linearly or nonlinearly separable categories. Medin & Schwanenflugel (1981) compared linearly and nonlinearly separable category learning using a small number of stimuli constructed from binary-valued dimensions and found no advantage for linearly separable structures. Ashby and colleagues (Ashby & Gott 1988; Ashby & Maddox 1990, 1992) compared linearly and nonlinearly separable category learning using a large number of stimuli constructed from continuous-valued dimensions. They found a consistent advantage for linearly separable categories, with nonlinearly separable category learning often requiring numerous experimental sessions. Despite their greater difficulty, however, the fact that people can learn nonlinearly separable categories effectively falsifies the standard prototype-theory account of information-integration category learning. Even so, there is evidence that early in learning people may abstract prototypes in some information-integration tasks (Minda & Smith 2001; Smith & Minda 1998, 2002). McKinley & Nosofsky (1995) examined category learning when the categories were composed of a large number of unique exemplars sampled from mixtures of bivariate normal distributions. Although some subjects were able to learn the categories fairly well, a large number failed to learn even after seeing nearly 4000 exemplars over a full week of training. These data present a challenge to exemplar theory because after such extensive training, exemplar theory predicts nearly optimal performance, no matter what the category structures (Ashby & AlfonsoReese 1995). In a related study, Ashby et al. (2001) compared two-category learning where the optimal decision bound was quadratic, with four-category learning where the optimal bound separating each pair of categories was quadratic. As in previous studies (e.g., Ashby & Maddox 1992), learning was good in the twocategory case, and the best fitting model assumed subjects used quadratic bounds. In the four-category case, however, learning was worse, and the best fitting model assumed subjects used suboptimal linear bounds to separate each pair of categories. Taken together, these two data sets suggest that there is an upper bound on the complexity of information-integration category structures that can be learned (in a reasonable amount of time) and that this upper bound is greater than a single quadratic curve but less than a set of such curves.

23 Dec 2004 19:45

160

AR

ASHBY

AR231-PS56-06.tex




AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

P1: IKH

MADDOX

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Neuropsychological Patient Data Over the past several years, a number of studies of information-integration category learning have been conducted in brain-damaged populations. The focus has been on patients with medial temporal lobe amnesia or striatal damage (e.g., patients with Parkinson’s disease or Huntington’s disease). Filoteo et al. (2001b) tested the ability of amnesiacs to learn a highly nonlinear information-integration rule when the categories were normally distributed and a large number of unique stimuli were sampled from each category. Over the full 600 trials of the experiment, the performance of amnesiacs and controls was equivalent. One patient and one control returned for a second session on the following day. During the first block of trials on the second day, the amnesiac and control again showed equivalent performance, and in fact, performance during the first block of the second session was slightly better than during the final block of trials from the first session. Some researchers have argued that amnesiacs learn categorization rules using working or short-term memory processes (Nosofsky & Zaki 1998, Palmeri & Flanery 1999). The day 2 results of Filoteo et al. (2001b) argue against this possibility. Instead, these findings indicate that the categorization rule was retained over the one-day delay period and argue strongly against the hypothesis that working memory or explicit declarative memory mediates information-integration category learning. Filoteo et al. (2001a) and Maddox & Filoteo (2001) tested the ability of Huntington’s disease and Parkinson’s disease patients to learn the same category structures used by Filoteo et al. (2001b). Over the full 600 trials of the experiment, both patient groups showed a consistent performance decrement, suggesting an involvement of the striatum in nonlinear information-integration category learning. On the other hand, Ashby et al. (2003) found that Parkinson’s disease patients learned as well as an age-matched control group in an information-integration task with linearly separable categories. More recently, Filoteo et al. (2004) compared the ability of Parkinson’s disease patients to learn a linear and a nonlinear information-integration rule. The linear results replicated the Ashby et al. (2003) results—that is, the Parkinson’s disease patients were not impaired in learning linearly separable categories. On the other hand, the same patients were impaired in the nonlinear condition, but only later in training. Thus, these studies suggest that Parkinson’s disease subjects are impaired in information-integration tasks, but only if the category structures are complex (as, e.g., when the categories are nonlinearly separable).

Neuroimaging Data To date, only one neuroimaging study of information-integration category learning has been conducted (although see the section on the weather prediction task). Seger & Cincotta (2002) reported significant striatal and lateral occipital activation by a group of subjects who had already had extensive training in the task.

23 Dec 2004 19:45

AR

AR231-PS56-06.tex

AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

HUMAN CATEGORY LEARNING

P1: IKH

161

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Theories of Information-Integration Category Learning There are a number of successful theories of information-integration category learning. These can be classified into two types: parametric and nonparametric (see Ashby & Alfonso-Reese 1995 for a detailed discussion). Parametric classifiers assume either that the categories have a specific type of structure (e.g., normal distributions) or that the categorization boundary has a specific functional form (e.g., linear). Nonparametric classifiers make no assumptions about category structure or categorization boundaries. Simple prototype models (Reed 1972, Smith & Medin 1981) are parametric because they assume a linear decision bound (Ashby & Gott 1988). For this reason, as mentioned above, they can be rejected as a general theory of informationintegration category learning because humans can learn nonlinear decision bounds (see, e.g., Ashby & Maddox 1992, Medin & Schwanenflugel 1981). Exemplar models are nonparametric because they assume that every exemplar presented is stored in memory along with the appropriate category label. Because all category information is retained (although perhaps in a degraded form), exemplar models predict that, under very general conditions, subjects should eventually respond almost optimally, no matter how complex the categories (Ashby & Alfonso-Reese 1995). For this reason, despite the enormous success of this class of models, the failure of subjects to respond optimally in complex informationintegration tasks (e.g., Ashby et al. 2001, McKinley & Nosofsky 1995) suggests that exemplar models may be too powerful. In addition, the only current neurobiological hypotheses about the exemplar-memory process attach a critical role to the hippocampus (Pickering 1997). As such, the finding that medial temporal lobe amnesiacs are relatively normal at information-integration category learning is problematic for the hypothesis that exemplar theory is adequate as a general theory of categorization in information-integration tasks. Decision bound models can be parametric or nonparametric depending on whether they assume subjects learn decision bounds (parametric) or assign responses to regions (nonparametric). In this latter case, the bound is simply the partition between regions associated with contrasting responses. Ashby & Waldron (1999) conducted a critical test of whether category learning in information-integration tasks is parametric or nonparametric. Previous research (reviewed above) showed that people can learn either linear or nonlinear (e.g., quadratic) information-integration decision bounds. If humans are parametric classifiers, then some categorical information must signal whether they should use a linear bound or a nonlinear bound. Ashby & Waldron (1999) constructed categories in which all statistical information that could be readily estimated (e.g., means, variances, and covariances) signaled that a parametric classifier should use a linear decision bound, but for which the optimal bound was quadratic. In a second condition, the statistical information signaled that a nonlinear bound should be used, but the optimal bound was linear. All known parametric classifiers predict that subjects will use a decision bound of the wrong type in these two conditions.

23 Dec 2004 19:45

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

162

AR

ASHBY

AR231-PS56-06.tex




AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

P1: IKH

MADDOX

Yet, the data of every subject (who did not use a rule-based strategy) were best fit by a bound of the optimal type. These results provide strong evidence against all known parametric classifiers, including prototype models, and decision bound models that assume people learn decision bounds. At the same time, the Ashby & Waldron (1999) results support nonparametric classifiers, such as exemplar models and decision bound models that assume people learn to assign responses to regions of perceptual space. As a model of the latter type, Ashby & Waldron (1999) proposed a nonparametric decision bound model called the striatal pattern classifier. Other nonparametric decision-bound-type models include Anderson’s (1991) rational model, and Love et al.’s (2004) SUSTAIN (Supervised and Unsupervised Stratified Adaptive Incremental Network) model. Each of these models can be loosely described as multiple prototype (or cluster) models because perceptually similar category exemplars tend to be grouped or clustered together. For the most part, all three models can account for the observed complexity limits on the learning of information-integration category structures, but the rational and SUSTAIN models make no attempt to account for the neuropsychological and neuroimaging data, and neither proposes a neurobiological account of information-integration category learning. The striatal pattern classifier, on the other hand, offers a computational model of the procedural-learning-based system proposed in COVIS, and it proposes a neurobiological interpretation. In brief, the model assumes that information-integration category learning is mediated primarily within the tail of the caudate nucleus (for visual stimuli). Two key neurophysiological features make this system a good candidate for information-integration category learning. First, all of visual cortex (except area V1) projects directly to the tail of the caudate, and these projections are characterized by massive convergence (i.e., of approximately 10,000 to 1; Wilson 1995). This convergence causes the decision space represented in the caudate nucleus to have much lower resolution than the perceptual space represented in visual cortex, and is assumed to account for the complexity limits on information-integration category learning reviewed above. Second, the tail of the caudate receives dopaminergic input from the substantia nigra that is widely thought to serve as a reward-mediated feedback signal (e.g., Schultz 1992, Wickens 1993). The idea is that an unexpected reward causes dopamine to be released from the substantia nigra into the tail of the caudate, and that the presence of this dopamine strengthens recently active synapses. The next section reviews a number of empirical results that are thought to be directly attributable to unique features of this feedback system. The striatal pattern classifier is consistent with much of the neuropsychological and neuroimaging data reviewed above. For example, the model predicts that amnesiacs should show normal information-integration category learning, whereas patients with striatal damage (Parkinson’s and Huntington’s patients) should be impaired. In addition, the model predicts that there should be striatal activation during information-integration category learning. In addition, although speculative at this point, it seems reasonable to suppose that the learning of complex nonlinear

23 Dec 2004 19:45

AR

AR231-PS56-06.tex

AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

HUMAN CATEGORY LEARNING

P1: IKH

163

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

information-integration category structures requires a higher resolution in the tailof-the-caudate decision space than the learning of simpler linear informationintegration category structures. Because the tail of the caudate is dysfunctional in Parkinson’s disease, the model makes the natural prediction that Parkinson’s disease patients should be especially impaired in nonlinear information-integration category learning. Each of these predictions was supported by the data reviewed above.

DISSOCIATIONS BETWEEN RULE-BASED AND INFORMATION-INTEGRATION CATEGORY LEARNING In a seminal study, Shepard et al. (1961; Shepard & Chang 1963) examined category learning in six tasks constructed from different stimulus-category assignments of the same 8 three-dimensional binary-valued stimuli (for replications and extensions, see Nosofsky et al. 1994a, Nosofsky & Palmeri 1996, Smith et al. 2004). These included rule-based, information-integration, and unstructured (memorization) tasks. Results showed that the one-dimensional task was easiest to learn and the unstructured, memorization task was the most difficult, with the other tasks (including the information-integration task) of intermediate difficulty. One weakness of this study is that the structural properties of the categories, such as within-category coherence and between-category discriminability, were not controlled. More recently, a number of studies have compared rule-based and information-integration category learning in a variety of settings where these and other structural properties are controlled. Collectively, these data offer a serious challenge to single-system models. Many of the studies in question were motivated by the COVIS model of category learning (Ashby et al. 1998). As outlined earlier, COVIS assumes that learning in rule-based tasks is dominated by an explicit, hypothesis-testing system that uses working memory and executive attention and is mediated primarily by the anterior cingulate, the prefrontal cortex, and the head of the caudate nucleus. In contrast, learning in information-integration tasks is assumed to be dominated by an implicit procedural-learning-based system, which is mediated largely within the tail of the caudate nucleus (Ashby et al. 1998, Ashby & Ell 2001, Willingham 1998). A series of studies attempted to dissociate the processes involved in rule-based and structurally equivalent information-integration category learning by introducing simple experimental manipulations that are predicted by COVIS to affect processing in the procedural-learning system but not the hypothesis-testing system, or vice versa. These tests focused on predictions that the two putative systems would be affected differently by manipulations of the nature and timing of feedback, by changes in the locations of the response keys, and by adding additional demands on working memory and executive attention. The remainder of this section briefly reviews these studies (for a more detailed review, see Maddox & Ashby 2004).

23 Dec 2004 19:45

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

164

AR

ASHBY

AR231-PS56-06.tex




AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

P1: IKH

MADDOX

Because the COVIS hypothesis-testing system is under conscious control and has full access to working memory and executive attention, the nature and timing of the feedback signal should not be critical for rule-based category learning. In contrast, a procedural-learning system that is mediated within the tail of the caudate nucleus would not be accessible to conscious awareness and is far removed from working memory.1 As a result, it would depend more heavily on the nature and timing of the feedback. Several studies tested these predictions. First, observational training was found to be equally effective to traditional feedback training with rulebased categories, but with information-integration categories, a distinct advantage occurred for feedback training (Ashby et al. 2002). During observational training, subjects are informed of the category membership of each stimulus just before it appears, whereas during feedback training, the stimulus is presented and then the category label is shown immediately after the subject responds. Second, some rulebased categories can be learned without feedback of any kind, whereas there is no evidence that information-integration categories can be learned without feedback (Ashby et al. 1999). Third, delaying the feedback by as little as 2.5 seconds after the response significantly interferes with information-integration category learning, but delays as long as 10 seconds have no effect on rule-based learning (Maddox et al.). A second set of studies tested the prediction that information-integration category learning is mediated largely by a form of procedural learning. The quintessential paradigm for studying procedural learning is the serial reaction time task (Nissen & Bullemer 1987), in which subjects press keys as quickly as possible in response to stimuli that appear in various locations on the screen. A large response time improvement is observed when the stimulus sequence is repeated, even when subjects are unaware that a sequence exists. Willingham et al. (2000) showed that changing the location of the response keys interferes with serial reaction time learning, but that changing the fingers that push the keys does not. Thus, if procedural learning is used in information-integration tasks, then switching the locations of the response keys should interfere with learning, but switching the fingers that depress the keys should not. In fact, Ashby et al. (2003) reported evidence that directly supported this prediction. They also reported that neither manipulation had any effect on rule-based category learning. Maddox et al. (2004) reported a similar sensitivity of information-integration category learning to response location. On half the trials, subjects responded “Yes” or “No” depending on whether the stimulus belonged to category A, and on half the trials they responded “Yes” or “No” depending on whether the stimulus belonged to category B. Thus, there was no consistent mapping of category label to response position. Compared to a standard 1

Crick & Koch (1990, 1995, 1998) offered a cognitive neuroscience theory of conscious awareness that states one can have conscious awareness only of activity in brain areas that project directly to the prefrontal cortex. The caudate nucleus does not project to the prefrontal cortex (it first projects through the globus pallidus and then the thalamus), so the Crick-Koch hypothesis predicts that we are not aware of activity within the caudate nucleus.

23 Dec 2004 19:45

AR

AR231-PS56-06.tex

AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN CATEGORY LEARNING

P1: IKH

165

control condition, learning was impaired with the information-integration categories, but not with the rule-based categories. These results provide the first direct evidence of procedural learning in perceptual categorization and suggest that the hypothesis-testing system learns abstract category labels, whereas the procedurallearning system learns response positions (for other examples of response effects in categorization, see Barsalou et al. 2003). A third set of studies tested the prediction that rule-based category learning requires working memory and executive attention, both to select and apply the correct rule and to interpret and process the feedback signal. First, Waldron & Ashby (2001) showed that rule-based category learning was disrupted more than information-integration category learning by the simultaneous performance of a task that required working memory and executive attention (a numerical Stroop task). In a second related study, Maddox et al. (2004) required subjects to alternate categorization trials with trials of a classic memory-scanning task (Sternberg 1966). In one condition, a short delay followed categorization and a long delay followed memory scanning, whereas these delays were reversed in the other condition. Learning of the information-integration categories was unaffected by the location of the short delay, whereas rule-based category learning was significantly worse when the short delay followed categorization. This result supports the hypothesis that feedback processing requires attention and effort in rule-based categorization, but not in information-integration category learning.

PROTOTYPE DISTORTION TASKS In prototype distortion tasks, the category exemplars are created by randomly distorting a single category prototype. As mentioned above, the most widely known example uses a constellation of dots (often 7 or 9) as the category prototype (see Figure 2 for an example), and the other category members are created by randomly perturbing the spatial location of each dot. These random dot stimuli and categories have been used in dozens of studies (e.g., Homa et al. 1979, 1981; Posner & Keele 1968, 1970; Shin & Nosofsky 1992; Smith & Minda 2002). Two different types of prototype distortion tasks are popular—(A, B) and (A, not A). In an (A, B) task, subjects are presented a series of exemplars that are each from some category A or from a contrasting category B. The task of the subject is to respond with the correct category label on each trial (i.e., “A” or “B”). An important feature of (A, B) tasks is that the stimuli associated with both responses each have a coherent structure—that is, they each have a central prototypical member around which the other category members cluster. In an (A, not A) task, on the other hand, there is a single central category A and subjects are presented with stimuli that are either exemplars from category A or random patterns that do not belong to category A. The subject’s task is to respond “Yes” or “No” depending on whether the presented stimulus was or was not a member of category A. In an (A, not A)

23 Dec 2004 19:45

166

AR

ASHBY

AR231-PS56-06.tex




AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

P1: IKH

MADDOX

task, the category A members have a coherent structure, but the stimuli associated with the “not A” (or “No”) response do not. Historically, prototype distortion tasks have been run both in (A, B) form and in (A, not A) form, although (A, not A) tasks are most common.

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Neuropsychological Patient Data Prototype distortion tasks are particularly important because the neuropsychological patient data are profoundly different from those in rule-based or informationintegration tasks. In particular, a variety of patients groups that are known to have deficits in rule-based and information-integration tasks show apparently normal prototype distortion learning, at least in (A, not A) designs. This includes patients with Parkinson’s disease (Reber & Squire 1999), schizophrenia (Keri et al. 2001), or Alzheimer’s disease (Sinha 1999, although see Keri et al. 1999). Normal (A, not A) performance has also been shown in patients with amnesia (Knowlton & Squire 1993, Kolodny 1994, Squire & Knowlton 1995). These results must be interpreted with caution, however, because several studies have shown that if category A is created from low-level distortions of the category A prototype, then healthy young adults can learn in (A, not A) tasks without any feedback (i.e., training) at all (Homa & Cultice 1984, Palmeri & Flanery 1999). Thus, it is not yet clear that all these patient groups would learn normally in a difficult (A, not A) task (i.e., one that requires feedback for optimal performance). At least two studies have compared (A, not A) and (A, B) prototype distortion learning on the same patients—and both studies report the same striking dissociation. Specifically, Sinha (1999) reported normal (A, not A) performance in Alzheimer’s disease patients, but impaired (A, B) performance, and Zaki et al. (2003) reported this same pattern of results with amnesiacs. Sinha (1999) also reported deficits in (A, B) prototype distortion learning in patients with amnesia.

Neuroimaging Data A handful of neuroimaging studies have used prototype distortion tasks. When interpreting these results, it is vital to consider whether an (A, B) or (A, not A) task was used. As in the purely behavioral studies, the most popular choice has been the (A, not A) task. All of these studies have reported learning-related changes in occipital cortex (Aizenstein et al. 2000; Reber et al. 1998a,b)—in general, reduced occipital activation was found in response to category A exemplars, although Aizenstein et al. (2000) found this reduction only under implicit learning conditions. When subjects were given explicit instructions to learn the A category, increased occipital activation was observed. Studies that have used (A, B) tasks have reported quite different results. Seger et al. (2000) did report categorization-related activation in occipital cortex, but they also found significant learning-related changes in prefrontal and parietal cortices. Vogels et al. (2002) reported results from a hybrid task in which subjects were to respond “A,” “B,” or “Neither.” Thus, stimuli were created from distortions of

23 Dec 2004 19:45

AR

AR231-PS56-06.tex

AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

HUMAN CATEGORY LEARNING

P1: IKH

167

an A prototype or a B prototype, or were just random patterns. Like Seger et al. (2000), Vogels et al. (2002) found prefrontal and parietal activation (although in different foci). However, they also reported task-related activation in orbitofrontal cortex and the neostriatum, and they failed to find any task-related activation in occipital cortex.

Theories of Prototype Distortion Learning The most recent debates between prototype and exemplar theories have focused on prototype distortion tasks. Prototype theory assumes a category is represented as a prototype, and that stimuli are categorized by comparing them to the prototypes of each contrasting category (Homa et al. 1981; Posner & Keele 1968; Reed 1972; Rosch 1973, 1975; Smith & Minda 2001). This theory seems ideally suited to prototype distortion tasks where all category members are simple distortions of a central prototype. Indeed, early results seemed to support this prediction. For example, performance is generally better on the prototype and on exemplars similar to the prototype than on distortions, even when subjects are trained on the distortions but not on the prototype (Homa et al. 1979, 1981; Posner & Keele 1970; Strange et al. 1970). Even so, exemplar theorists showed that exemplar theory was also compatible with these results, and they argued that exemplar theory provides at least as good an account of data from prototype distortion tasks as does prototype theory (Hintzman 1986, Hintzman & Ludlam 1980, Shin & Nosofsky 1992). Smith & Minda (2001) identified a critical test between exemplar and prototype accounts of prototype distortion data. Consider a space with a dimension for each stimulus component. In the case of the dot patterns there would thus be two dimensions for each dot—one to identify the horizontal position of the dot and one to identify the vertical position. In this space, each category exemplar is identified by a single point, and the category prototype is represented by a point in the center of the cloud of points denoting all exemplars of the category. Smith & Minda’s (2001) analogy was to the solar system, with the sun representing the category prototype and the planets the category members that were created by distorting the prototype. Now consider the probability of responding “A” in an (A, not A) task and how this probability changes with the position of the stimulus in the dot pattern space. According to prototype theory, the probability of responding “A” is completely determined by the similarity, or equivalently the distance, between the stimulus and the category prototype (i.e., the sun in the solar system analogy). According to exemplar theory, however, the probability of responding “A” depends on the (sum of the) similarities between the stimulus and all of the category A exemplars, or equivalently on the distances between the stimulus point and all the planets in the category A solar system. When the stimulus is outside of the category A cluster or solar system, then prototype and exemplar theories both predict that the probability of responding “A” will increase sharply as the stimulus moves toward the category A prototype,

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

EXEMPLAR VERSUS PROTOTYPE THEORIES

23 Dec 2004 19:45

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

168

AR

ASHBY

AR231-PS56-06.tex




AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

P1: IKH

MADDOX

because the distance between the stimulus and the sun is decreasing as are the distances between the stimulus and each planet. The critical difference between prototype and exemplar theories emerges when the stimulus first enters the category A solar system. For example, as a meteor passes Pluto and enters our solar system, it is still moving closer to the sun and to the planets nearest to the sun, but it is now moving away from Pluto. The closer it moves toward the sun the greater this effect—that is, it is always moving steadily nearer the sun, but as it approaches the sun it begins moving away from more planets and moving nearer to fewer planets. As a result, prototype theory predicts that the probability of responding “A” continually increases as the similarity between the stimulus and the category A prototype is increased. However, exemplar theory predicts that this probability gradient will begin to flatten as the stimulus moves inside the category A cluster. Smith & Minda (2001, 2002; Smith 2002) examined these probabilityof-responding-“A” profiles for a number of new and previously published studies and showed that they were steeper than predicted by exemplar theory, but were in general agreement with the predictions of prototype theory. The neuroimaging results showing learningrelated changes in visual cortex in (A, not A) prototype distortion tasks motivated several proposals that the perceptual representation memory system contributes to learning under these conditions (Ashby & Casale 2002, Reber & Squire 1999). The idea is that performance may be mediated, at least in part, by perceptual learning within visual cortex. If such visual cortical perceptual learning is important in prototype distortion tasks, then it likely will have different effects in (A, not A) and (A, B) tasks. Consider first an (A, not A) task. The category A prototype will induce a graded pattern of activation throughout visual cortex. One particular cell (or small group of cells) will fire most rapidly to the presentation of this pattern. Call this cell A. A lowlevel distortion of the category A prototype will be visually similar to the prototype and therefore will likely also cause cell A to fire. Thus, cell A will repeatedly fire throughout training on the category A exemplars. Perceptual learning is thought to occur any time repeated presentations of the same stimulus occur during some relatively brief time interval (Dosher & Lu 1999). As a result, perceptual learning will cause the magnitude of the cell A response to increase throughout training. In contrast, the stimuli associated with the “not A” response will be visually dissimilar to the category A prototype and therefore will be unlikely to cause cell A to fire. During the transfer or testing phase of the experiment, the subject can use the increased sensitivity of cell A to respond accurately. In particular, stimuli from category A are likely to lead to an enhanced visual response compared to stimuli that do not belong to category A. Thus, to respond with above chance accuracy, subjects need only respond “A” to any stimulus that elicits an enhanced visual response. Note that one could interpret cell A as encoding the representation of the category prototype, and thus, this type of perceptual learning could be interpreted as a neuropsychological basis of prototype theory.

PERCEPTUAL LEARNING THEORIES

23 Dec 2004 19:45

AR

AR231-PS56-06.tex

AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN CATEGORY LEARNING

P1: IKH

169

Next, consider an (A, B) task. In this case, there will be some cell A maximally tuned to the category A prototype, but there will be some other cell B that is tuned to the category B prototype. During training, every presented stimulus is a distortion of either the category A or category B prototype, so it is likely that either cell A or B will fire on many trials. During the testing phase, all stimuli are again from either category A or B, and so stimuli from both categories will be equally likely to elicit an enhanced visual response. As a result, the mere existence of an enhanced visual response will not help subjects decide whether to respond “A” or “B.” The conclusion, therefore, is that perceptual learning could greatly assist in (A, not A) tasks but, by itself, it would be of little help in (A, B) tasks. This is not to say that learning in (A, B) prototype distortion tasks is impossible; only that other learning systems must be used. For example, in low-distortion (A, B) tasks, the enhanced prototype responses caused by perceptual learning might facilitate an explicit memorization strategy in which subjects memorize the A and B prototype patterns and their associated responses. Much of the cognitive neuroscience data reviewed above supports these predictions. First, neuroimaging results of (A, not A) prototype distortion tasks consistently report learning-related activation in visual cortex (Aizenstein et al. 2000; Reber et al. 1998a,b). Second, neuroimaging results of (A, B) tasks have sometimes failed to find such occipital activation (Vogels et al. 2002), and they have consistently reported task-related activation in prefrontal cortex that is not seen in (A, not A) tasks (Seger et al. 2000, Vogels et al. 2002). Third, a variety of neuropsychological studies show normal (A, not A) prototype distortion learning in patient groups that are impaired in rule-based or information-integration category learning (e.g., schizophrenics, Keri et al. 2001; Parkinson’s disease patients, Reber & Squire 1999; Alzheimer’s disease patients, Sinha 1999), and in amnesiacs (Knowlton & Squire 1993). Fourth, amnesiacs and patients with Alzheimer’s disease are impaired in (A, B) prototype distortion learning (Sinha 1999, Zaki et al. 2003). The prefrontal activation in (A, B) tasks, and the impaired learning of amnesiacs, suggest that learning in (A, B) prototype distortion tasks may be mediated by explicit reasoning strategies and/or by explicit memorization.

WEATHER PREDICTION TASK An important distinction in category-learning experiments is whether category membership is deterministic or probabilistic. In deterministic tasks, each stimulus is unambiguously a member of one category (i.e., optimal performance is perfect), whereas in probabilistic tasks, at least some stimuli are probabilistically associated with the contrasting categories. For example, on one trial in a probabilistic classification task a particular stimulus might belong to category A but on the next trial the same stimulus might belong to category B. Obviously, in such tasks, perfect performance is impossible. Although most category-learning studies have used deterministic tasks, probabilistic classification also has a long history (e.g.,

23 Dec 2004 19:45

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

170

AR

ASHBY

AR231-PS56-06.tex




AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

P1: IKH

MADDOX

Ashby & Gott 1988; Ashby & Maddox 1990, 1992; Estes et al. 1989; Gluck & Bower 1988; Kubovy & Healy 1977). One popular probabilistic classification task, which is used extensively in cognitive neuroscience, is the so-called weather prediction task (Eldridge et al. 2002; Knowlton et al. 1994, 1996a,b; Reber et al. 1996; Reber & Squire 1999). On each trial of this task, subjects are shown one, two, or three of four possible tarot cards and are asked to indicate whether the presented constellation signals rain or sun. Each card is labeled with a unique, and highly discriminable, geometric pattern. Fourteen of the 16 possible card combinations are used (the no cards and four card patterns are excluded) and each combination is probabilistically associated with the two outcomes. In the original version of the task, the highest possible accuracy was 76% (Knowlton et al. 1994). Interestingly though, a single-cue strategy in which the subject gives one response if one card is present and the other response if that same card is absent yields an accuracy of 75% correct. Knowlton et al. (1994) reported that performance of a control group increased from approximately 50% to 65% correct during the first 50 trials and continued to improve to approximately 75% correct after 350 trials.

Task and Individual Differences Analysis In order to relate results from the weather prediction task to the other results reviewed in this article, it is important to determine its relationship to the other tasks we have discussed. Because the optimal strategy requires information-integration across cues and is nonverbalizable, the weather prediction task is technically an information-integration task. On the other hand, a single-cue strategy results in nearly optimal performance (75% for a single cue versus 76% for the optimal strategy), so nearly optimal accuracy does not rule out simple rule-based strategies. In addition, because the task uses only a few highly distinct exemplars, explicit memorization is also a plausible strategy. Because a variety of different strategies are all about equally effective, we might expect more individual differences in results obtained with the weather prediction task than with the other tasks we have considered. This possibility makes it especially important to determine what strategy each subject is using before interpreting his or her data. Gluck et al. (2002) provided a “strategy” analysis of data collected in the weather prediction task to address this issue. In one study, they simply asked subjects what strategy they used after the experiment was over. Although many of the protocols lacked detail, those with sufficient detail generally fell into one of three types: (a) single-cue strategies in which responding was based on the presence or absence of one card (as in a rule-based task), (b) multiple-cue learning (as in an informationintegration task), or (c) singleton learning in which correct responses to the singlecard patterns were memorized, and guessing occurred for the remaining patterns. Based on these self-reports, Gluck et al. (2002) developed a model-based analysis to identify each subject’s strategy in two follow-up studies. Based on a full 200-trial session, 90% (Experiment 1) and 80% (Experiment 2) of the subjects used

23 Dec 2004 19:45

AR

AR231-PS56-06.tex

AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

HUMAN CATEGORY LEARNING

P1: IKH

171

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

a singleton strategy (explicit memorization) to learn the categories. When broken down into 50-trial blocks, a shift from singleton toward multiple-cue strategies was observed (although the proportion of subjects using a multiple-cue strategy was still well below 0.5). Interestingly, there was little correspondence between the strategies that subjects self-reported in their protocols and the strategies identified by the modeling approach. Thus, although originally designed as an informationintegration task, these results suggest that subjects adopt a variety of different strategies in the weather prediction task, and the evidence indicates that the most popular choice may be explicit memorization.

Neuropsychological Patient and Neuroimaging Studies The weather prediction task has been used to study category learning in a number of patient groups. In one of the first such studies, Knowlton et al. (1994) found that amnesiacs performed as well as healthy controls during the first 50 trials of learning, but with extended training, amnesiacs showed a learning deficit relative to healthy controls. Declarative memory is impaired in amnesia, so amnesiacs are impaired in explicit memorization strategies. Healthy controls are not, however, so one interpretation of the late training deficit is that controls begin memorizing and the amnesiacs do not (Knowlton et al. 1994; see also Gluck et al. 1996). Unlike amnesiacs, patients with Parkinson’s or Huntington’s disease show learning deficits in the weather prediction task during the first 50 trials that continue throughout training (Knowlton et al. 1996a,b). The weather prediction task has also been used to examine category learning in patients suffering from Alzheimer’s disease or schizophrenia. Patients in the early stages of Alzheimer’s disease are similar to amnesiacs in the sense that both show anterograde amnesia due to neurodegenerative processes in the medial temporal lobes. As one might predict given this similarity, Alzheimer’s patients show intact performance during early trials of the weather prediction task, similar to that seen in amnesia (Eldridge et al. 2002). Schizophrenics exhibit marked abnormalities in executive function and explicit memory. Even so, their performance on the weather prediction task is within normal ranges (Keri et al. 2000). Neuroimaging studies of the weather prediction task indicate that the medial temporal lobes are active early in learning, and gradually become deactivated as learning progresses (Poldrack et al. 2001). This deactivation is mirrored by a simultaneous activation of the basal ganglia. Specifically, early in learning the basal ganglia are inactive, and gradually become more active as learning progresses. The weather prediction task has provided a useful tool for studying classification learning and has offered some important insights into the neurobiology of category learning. Even so, the fact that nearly optimal performance can be achieved by a variety of different strategies (e.g., information-integration, rule-based, explicit memorization) makes it difficult to draw strong inferences from data collected with this task. Although the strategies approach developed by Gluck et al. (2002) helps identify the type of strategy that subjects are using (however, see Shohamy

23 Dec 2004 19:45

172

AR

ASHBY

AR231-PS56-06.tex




AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

P1: IKH

MADDOX

et al. 2004), a better alternative might be to use one of the other tasks discussed in this article that are not so susceptible to identifiability problems.

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

CONCLUSIONS The results reviewed in this article offer important lessons. First, when interpreting a category-learning result, it is critical to consider carefully the specific task that was used. For example, Parkinson’s disease patients are normal in (A, not A) prototype distortion tasks, they are mildly impaired in information-integration category learning, and they are profoundly impaired in rule-based categorization. Several studies support each of these conclusions, but without specifying the task, these studies would appear to just catalog a confusing set of contradictory results. Second, although the issue is far from resolved, the results presented here make a strong case that human category learning is mediated by multiple qualitatively distinct systems. To a large extent, it is also becoming clear that this issue—of whether there are one or more category-learning systems—is tied to the historically older issue of whether there are one or more memory systems. Learning is, by definition, the process of laying down some sort of memory trace, and there is certainly no reason to suspect that any of the separate memory systems that have been hypothesized are incapable of storing memories about categories. Although research efforts to resolve this debate will continue, other work is already attacking the next set of questions. For example, the next decade will likely see a flurry of research activity directed at determining the conditions under which the various systems contribute to category learning, at determining how the different systems interact, and at fleshing out their underlying neurobiology. ACKNOWLEDGMENTS Preparation of this article was supported by Public Health Service Grants MH3760 and MH59196. We thank Michael Casale and David Smith for their helpful comments. The Annual Review of Psychology is online at http://psych.annualreviews.org

LITERATURE CITED Aizenstein HJ, MacDonald AW, Stenger VA, Nebes RD, Larson JK, et al. 2000. Complementary category learning systems identified using event-related functional MRI. J. Cogn. Neurosci. 12:977–87 Alfonso-Reese L. 1996. Dynamics of category learning. Unpubl. PhD dissert. Univ. Calif., Santa Barbara

Anderson JR. 1991. The adaptive nature of human categorization. Psychol. Rev. 98:409– 29 Ashby FG, Alfonso-Reese LA. 1995. Categorization as probability density estimation. J. Math. Psychol. 39:216–33 Ashby FG, Alfonso-Reese LA, Turken AU, Waldron EM. 1998. A neuropsychological

23 Dec 2004 19:45

AR

AR231-PS56-06.tex

AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN CATEGORY LEARNING theory of multiple systems in category learning. Psychol. Rev. 105:442–81 Ashby FG, Casale MB. 2002. The cognitive neuroscience of implicit category learning. In Attention and Implicit Learning, ed. L. Jim´enez, pp. 109–41. Amsterdam/Erdenheim, PA: Benjamins Ashby FG, Ell SW. 2001. The neurobiology of category learning. Trends Cogn. Sci. 5:204– 10 Ashby FG, Ell SW, Waldron EM. 2003. Procedural learning in perceptual categorization. Mem. Cogn. 31:1114–25 Ashby FG, Gott RE. 1988. Decision rules in the perception and categorization of multidimensional stimuli. J. Exp. Psychol.: Learn. Mem. Cogn. 14:33–53 Ashby FG, Isen AM, Turken AU. 1999. A neuropsychological theory of positive affect and its influence on cognition. Psychol. Rev. 106:529–50 Ashby FG, Maddox WT. 1990. Integrating information from separable psychological dimensions. J. Exp. Psychol.: Hum. Percept. Perform. 16:598–612 Ashby FG, Maddox WT. 1992. Complex decision rules in categorization: contrasting novice and experienced performance. J. Exp. Psychol.: Hum. Percept. Perform. 18:50–71 Ashby FG, Maddox WT. 1994. A response time theory of separability and integrality in speeded classification. J. Math. Psychol. 38:423–66 Ashby FG, Maddox WT. 1998. Stimulus categorization. In Handbook of Perception and Cognition: Measurement, Judgment, and Decision Making, ed. MH Birnbaum, pp. 251– 301. New York: Academic Ashby FG, Maddox WT, Bohil CJ. 2002. Observational versus feedback training in rulebased and information-integration category learning. Mem. Cogn. 30:666–77 Ashby FG, Noble S, Filoteo J, Waldron EM, Ell SW. 2003. Category learning deficits in Parkinson’s disease. Neuropsychology 17:115–24 Ashby FG, Queller S, Berretty PM. 1999. On the dominance of unidimensional rules in

P1: IKH

173

unsupervised categorization. Percept. Psychophys. 61:1178–99 Ashby FG, Townsend JT. 1986. Varieties of perceptual independence. Psychol. Rev. 93:154– 79 Ashby FG, Waldron EM. 1999. On the nature of implicit categorization. Psychon. Bull. Rev. 6:363–78 Ashby FG, Waldron EM, Lee WW, Berkman A. 2001. Suboptimality in human categorization and identification. J. Exp. Psychol.: Gen. 130:77–96 Barsalou LW. 2003. Situated simulation in the human conceptual system. Lang. Cogn. Process. Spec. Iss.: Concept. Rep. 18:513– 62 Barsalou LW, Simmons WK, Barbey AK, Wilson CD. 2003. Grounding conceptual knowledge in modality-specific systems. Trends Cogn. Sci. 7:84–91 Brooks L. 1978. Nonanalytic concept formation and memory for instances. In Cognition and Categorization, ed. E Rosch, BB Lloyd, pp. 169–211. Hillsdale, NJ: Erlbaum Brown RG, Marsden CD. 1988. Internal versus external cues and the control of attention in Parkinson’s disease. Brain 111:323–45 Burner JS, Goodnow J, Austin G. 1956. A Study of Thinking. New York: Wiley Cools AR, van den Bercken JHL, Horstink MWI, van Spaendonck KPM, Berger HJC. 1984. Cognitive and motor shifting aptitude disorder in Parkinson’s disease. J. Neurol. Neurosurg. Psychiatry 47:443–53 Cree GS, McRae K. 2003. Analyzing the factors underlying the structure and computation of the meaning of chipmunk, cherry, chisel, cheese, and cello (and many other such concrete nouns). J. Exp. Psychol.: Gen. 132:163–201 Crick F, Koch C. 1990. Towards a neurobiological theory of consciousness. Semin. Neurosci. 2:263–75 Crick F, Koch C. 1995. Are we aware of neural activity in primary visual cortex? Nature 375:121–23 Crick F, Koch C. 1998. Consciousness and neuroscience. Cereb. Cortex 8:97–107

23 Dec 2004 19:45

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

174

AR

ASHBY

AR231-PS56-06.tex




AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

P1: IKH

MADDOX

Dosher B, Lu ZL. 1999. Mechanisms of perceptual learning. Vision Res. 39:3197– 221 Downes JJ, Roberts AC, Sahakian BJ, Evenden JL, Morris RG, Robbins TW. 1989. Impaired extra-dimensional shift performance in medicated and unmedicated Parkinson’s disease: evidence for a specific attentional dysfunction. Neuropsychologia 27:1329–43 Eldridge LL, Masterman D, Knowlton BJ. 2002. Intact implicit habit learning in Alzheimer’s disease. Behav. Neurosci. 116: 722–26 Elliott R, Rees G, Dolan RJ. 1999. Ventromedial prefrontal cortex mediates guessing. Neuropsychologia 37:403–11 Erickson MA, Kruschke JK. 1998. Rules and exemplars in category learning. J. Exp. Psychol.: Gen. 127:107–40 Estes WK. 1986. Array models for category learning. Cogn. Psychol. 18:500–49 Estes WK. 1994. Classification and Cognition. New York: Oxford Univ. Press Estes WK, Campbell JA, Hatsopoulos N, Hurwitz JB. 1989. Base-rate effects in category learning: a comparison of parallel network and memory storage-retrieval models. J. Exp. Psychol.: Learn. Mem. Cogn. 15556–71 Filoteo JV, Maddox WT, Davis JD. 2001a. A possible role of the striatum in linear and nonlinear categorization rule learning: evidence from patients with Huntington’s disease. Behav. Neurosci. 115:786–98 Filoteo JV, Maddox WT, Davis JD. 2001b. Quantitative modeling of category learning in amnesic patients. J. Int. Neuropsychol. Soc. 7:1–19 Filoteo JV, Maddox WT, Salmon DP, Song DD. 2004. Information-integration category learning in patients with striatal dysfunction. Neuropsychology. In press Garner WR. 1974. The Processing of Information and Structure. New York: Wiley Gluck MA, Bower GH. 1988. From conditioning to category learning: an adaptive network model. J. Exp. Psychol.: Gen. 117:227–47 Gluck MA, Oliver LM, Myers CE. 1996. Latetraining amnesic deficits in probabilistic

category learning: a neurocomputational analysis. Learn. Mem. 3:326–40 Gluck MA, Shohamy D, Myers C. 2002. How do people solve the “weather prediction” task? Individual variability in strategies for probabilistic category learning. Learn. Mem. 9:408–18 Goldman-Rakic PS. 1987. Circuitry of the prefrontal cortex and the regulation of behavior by representational knowledge. In Handbook of Physiology, ed. F Plum, V Mountcastle, pp. 373–417. Bethesda, MD: Am. Physiol. Soc. Goldman-Rakic PS. 1995. Cellular basis of working memory. Neuron 14:477–85 Heaton RK. 1981. A Manual for the Wisconsin Card Sorting Test. Odessa, FL: Psychol. Assess. Resourc. Herbranson WT, Fremouw T, Shimp CP. 1999. The randomization procedure in the study of categorization of multidimensional stimuli by pigeons. J. Exp. Psychol. Anim. Behav. Process. 25:113–34 Hintzman DL. 1986. Schema abstraction in a multiple-trace memory model. Psychol. Rev. 93:411–28 Hintzman DL, Ludlam G. 1980. Differential forgetting of prototypes and old instances: simulation by an exemplar-based classification model. Mem. Cogn. 8:378–82 Homa D, Cultice JC. 1984. Role of feedback, category size, and stimulus distortion on the acquisition and utilization of ill-defined categories. J. Exp. Psychol.: Learn. Mem. Cogn. 10:83–94 Homa D, Rhoads D, Chambliss D. 1979. Evolution of conceptual structure. J. Exp. Psychol.: Learn. Mem. Cogn. 5:11–23 Homa D, Sterling S, Trepel L. 1981. Limitations of exemplar-based generalization and the abstraction of categorical information. J. Exp. Psychol.: Learn. Mem. Cogn. 7:418–39 Humphreys GW, Forde EME. 2001. Hierarchies, similarity and interactivity in object recognition: “category-specific” neuropsychological deficits. Behav. Brain Sci. 24:453–509 Janowsky JS, Shimamura AP, Kritchevsky M,

23 Dec 2004 19:45

AR

AR231-PS56-06.tex

AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN CATEGORY LEARNING Squire LR. 1989. Cognitive impairment following frontal lobe damage and its relevance to human amnesia. Behav. Neurosci. 103:548–60 Joseph JE. 2001. Functional neuroimaging studies of category specificity in object recognition: a critical review and metaanalysis. Cogn. Affect. Behav. Neurosci. 1: 119–36 J. Exp. Anal. Behav. 2002. Vol. 78, No. 3 (entire issue) K´eri S, Kalman J, Rapcsak SZ, Antal A, Benedek G, Janka Z. 1999. Classification learning in Alzheimer’s disease. Brain 122: 1063–68 K´eri S, Kelemen O, Benedek G, Janka Z. 2001. Intact prototype learning in schizophrenia. Schizophr. Res. 52:261–64 Keri S, Kelemen O, Szekeres G, Bagoczky N, Erdelyi R, et al. 2000. Schizophrenics know more than they can tell: probabilistic classification learning in schizophrenia. Psychol. Med. 30:149–55 Kimberg DY, D’Esposito M, Farah MJ. 1997. Frontal lobes: neuropsychological aspects. In Behavioral Neurology and Neuropsychology, ed. TE Feinberg, MJ Farah, pp. 409–18. New York: McGraw-Hill Knowlton BJ, Mangels JA, Squire LR. 1996a. A neostriatal habit learning system in humans. Science 273:1399–402 Knowlton BJ, Squire LR. 1993. The learning of natural categories: parallel memory systems for item memory and category-level knowledge. Science 262:1747–49 Knowlton BJ, Squire LR, Gluck MA. 1994. Probabilistic classification learning in amnesia. Learn. Mem. 1:106–20 Knowlton BJ, Squire LR, Paulsen JS, Swerdlow NR, Swenson M, Butters N. 1996. Dissociations within nondeclarative memory in Huntington’s disease. Neuropsychology 10:538– 48 Kolodny JA. 1994. Memory processes in classification learning: an investigation of amnesic performance in categorization of dot patterns and artistic styles. Psychol. Sci. 5:164– 69

P1: IKH

175

Konishi S, Karwazu M, Uchida I, Kikyo H, Asakura I, Miyashita Y. 1999. Contribution of working memory to transient activation in human inferior prefrontal cortex during performance of the Wisconsin Card Sorting Test. Cereb. Cortex 9:745–73 Kruschke JK. 1992. ALCOVE: an exemplarbased connectionist model of category learning. Psychol. Rev. 99:22–44 Kubovy M, Healy AF. 1977. The decision rule in probabilistic categorization: what it is and how it is learned. J. Exp. Psychol.: Gen. 106:427–46 Lamberts K. 2000. Information-accumulation theory of speeded categorization. Psychol. Rev. 107:227–60 Leng NR, Parkin AJ. 1988. Double dissociation of frontal dysfunction in organic amnesia. Br. J. Clin. Psychol. 27:359–62 Lewandowsky S, Kalish M, Ngang SK. 2002. Simplified learning in complex situations: knowledge partitioning in function learning. J. Exp. Psychol.: Gen. 131:163–93 Lockhead GR. 1966. Effects of dimensional redundancy on visual discrimination. J. Exp. Psychol. 72:95–104 Lombardi WJ, Andreason PJ, Sirocco KY, Rio DE, Gross RE, et al. 1999. Wisconsin Card Sorting Test performance following head injury: Dorsolateral fronto-striatal circuit activity predicts perseveration. J. Clin. Exp. Neuropsychol. 21:2–16 Love BC, Medin DL, Gureckis TM. 2004. SUSTAIN: a network model of category learning. Psychol. Rev. 111:309–32 Maddox WT, Ashby FG. 1993. Comparing decision bound and exemplar models of categorization. Percept. Psychophys. 53:49– 70 Maddox WT, Ashby FG. 2004. Dissociating explicit and procedural-learning based systems of perceptual category learning. Behav. Process. 66:309–32 Maddox WT, Ashby FG, Bohil CJ. 2003. Delayed feedback effects on rule-based and information-integration category learning. J. Exp. Psychol.: Learn. Mem. Cogn. 29:650– 62

23 Dec 2004 19:45

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

176

AR

ASHBY

AR231-PS56-06.tex




AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

P1: IKH

MADDOX

Maddox WT, Ashby FG, Ing AD, Pickering AD. 2004. Disrupting feedback processing interferes with rule-based but not information-integration category learning. Mem. Cogn. 32:582–91 Maddox WT, Bohil CJ, Ing AD. 2004. Evidence for a procedural learning-based system in category learning. Psychon. Bull. Rev. In press Maddox WT, Filoteo JV. 2001. Striatal contribution to category learning: quantitative modeling of simple linear and complex nonlinear rule learning in patients with Parkinson’s disease. J. Int. Neuropsychol. Soc. 7: 710–27 Markman AB, Ross BH. 2003. Category use and category learning. Psychol. Bull. 129:592–613 McKinley SC, Nosofsky RM. 1995. Investigations of exemplar and decision bound models in large, ill-defined category structures. J. Exp. Psychol.: Hum. Percept. Perform. 21:128–48 Medin DL, Schaffer MM. 1978. Context theory of classification learning. Psychol. Rev. 85:207–38 Medin DL, Schwanenflugel PJ. 1981. Linear separability in classification learning. J. Exp. Psychol.: Learn. Mem. Cogn. 7:355–68 Minda JP, Smith JD. 2001. Prototypes in category learning: the effects of category size, category structure, and stimulus complexity. J. Exp. Psychol.: Learn. Mem. Cogn. 27:775– 99 Murphy GL. 2002. The Big Book of Concepts. Cambridge, MA: MIT Press Nissen MJ, Bullemer P. 1987. Attentional requirements of learning: evidence from performance measures. Cogn. Psychol. 19:1–32 Nosofsky RM. 1986. Attention, similarity, and the identification categorization relationship. J. Exp. Psychol.: Gen. 115:39–57 Nosofsky RM. 1991. Typicality in logically defined categories: exemplar-similarity versus rule instantiation. Mem. Cogn. 19:131–50 Nosofsky RM, Clark SE, Shin HJ. 1989. Rules and exemplars in categorization, identification, and recognition. J. Exp. Psychol.: Learn. Mem. Cogn. 15:282–304

Nosofsky RM, Gluck MA, Palmeri TJ, McKinley SC, Glauthier P. 1994a. Comparing models of rule-based classification learning: a replication and extension of Shepard, Hovland, and Jenkins 1961. Mem. Cogn. 22:352– 69 Nosofsky RM, Palmeri TJ. 1996. Learning to classify integral-dimension stimuli. Psychon. Bull. Rev. 3:222–26 Nosofsky RM, Palmeri TJ, McKinley SC. 1994b. Rule-plus-exception model of classification learning. Psychol. Rev. 101:53–79 Nosofsky RM, Zaki SR. 1998. Dissociations between categorization and recognition in amnesic and normal individuals: an exemplar-based interpretation. Psychol. Sci. 9:247–55 Palmeri TJ, Flanery MA. 1999. Learning about categories in the absence of training: profound amnesia and the relationship between perceptual categorization and recognition memory. Psychol. Sci. 10:526–30 Pickering AD. 1997. New approaches to the study of amnesic patients: What can a neurofunctional philosophy and neural network methods offer? Memory 5:255–300 Poldrack RA, Clark J, Pare-Blagoev EJ, Shohamy D, Moyano JC, et al. 2001. Interactive memory systems in the human brain. Nature 414:546–50 Posner MI, Keele SW. 1968. On the genesis of abstract ideas. J. Exp. Psychol. 77:353– 63 Posner MI, Keele SW. 1970. Retention of abstract ideas. J. Exp. Psychol. 83:304–8 Posner MI, Petersen SE. 1990. Attention systems in the human brain. Annu. Rev. Neurosci. 13:25–42 Rao SM, Bobholz JA, Hammeke TA, Tosen AC, Woodley SJ, et al. 1997. Functional MRI evidence for subcortical participation in conceptual reasoning skills. Neuroreport 8:1987– 93 Reber PJ, Knowlton BJ, Squire LR. 1996. Dissociable properties of memory systems: differences in the flexibility of declarative and nondeclarative knowledge. Behavioral Neuroscience 110:861–71

23 Dec 2004 19:45

AR

AR231-PS56-06.tex

AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN CATEGORY LEARNING Reber PJ, Squire LR. 1999. Intact learning of artificial grammars and intact category learning by patients with Parkinson’s disease. Behav. Neurosci. 113:235–42 Reber PJ, Stark CEL, Squire LR. 1998a. Contrasting cortical activity associated with category memory and recognition memory. Learn. Mem. 5:420–28 Reber PJ, Stark CEL, Squire LR. 1998b. Cortical areas supporting category learning identified using functional MRI. Proc. Natl. Acad. Sci. USA 95:747–50 Reed SK. 1972. Pattern recognition and categorization. Cogn. Psychol. 3:382–407 Robinson AL, Heaton RK, Lehman RA, Stilson DW. 1980. The utility of the Wisconsin Card Sorting Test in detecting and localizing frontal lobe lesions. J. Consult. Clin. Psychol. 48:605–64 Rogers RD, Andrews TC, Grasby PM, Brooks DJ, Robbins TW. 2000. Contrasting cortical and subcortical activations produced by attentional-set shifting and reversal learning in humans. J. Cogn. Neurosci. 12:142– 62 Rosch E. 1973. Natural categories. Cogn. Psychol. 4:328–50 Rosch E. 1975. Cognitive reference points. Cogn. Psychol. 7:532–47 Salatas H, Bourne LE. 1974. Learning conceptual rules: III. Processes contributing to rule difficulty. Mem. Cogn. 2:549–53 Schultz W. 1992. Activity of dopamine neurons in the behaving primate. Semin. Neurosci. 4:129–38 Seger CA, Cincotta CM. 2002. Striatal activity in concept learning. Cogn. Affect. Behav. Neurosci. 2:149–61 Seger CA, Poldrack RA, Prabhakaran V, Zhao M, Glover GH, Gabrieli JDE. 2000. Hemispheric asymmetries and individual differences in visual concept learning as measured by functional MRI. Neuropsychologia 38:1316–24 Shepard RN. 1964. Attention and the metric structure of the stimulus space. J. Math. Psychol. 1:54–87 Shepard RN, Chang JJ. 1963. Stimulus general-

P1: IKH

177

ization in learning of classifications. J. Exp. Psychol. 65:94–102 Shepard RN, Hovland CI, Jenkins HM. 1961. Learning and memorization of classifications. Psychol. Monogr. 75(13, Whole No. 517) Shin HJ, Nosofsky RM. 1992. Similarityscaling studies of dot-pattern classification and recognition. J. Exp. Psychol.: Gen. 121: 278–304 Shohamy D, Myers CE, Grossman S, Sage J, Gluck MA, Poldrack RA. 2004. Corticostriatal contributions to feedback-based learning: converging data from neuroimaging and neuropsychology. Brain 127:1–9 Sinha RR. 1999. Neuropsychological substrates of category learning. Dissert. Abstr. Int.: Sect. B: Sci. Engineer. 60(5-B), 2381 (UMI No. AEH9932480) Smith EE, Medin DL. 1981. Categories and Concepts. Cambridge, MA: Harvard Univ. Press Smith JD. 2002. Exemplar theory’s predicted typicality gradient can be tested and disconfirmed. Psychol. Sci. 13:437–42 Smith JD, Minda JP. 1998. Prototypes in the mist: the early epochs of category learning. J. Exp. Psychol.: Learn. Mem. Cogn. 24:1411– 30 Smith JD, Minda JP. 2001. Journey to the center of the category: the dissociation in amnesia between categorization and recognition. J. Exp. Psychol.: Learn. Mem. Cogn. 27:984– 1002 Smith JD, Minda JP. 2002. Distinguishing prototype-based and exemplar-based processes in category learning. J. Exp. Psychol.: Learn. Mem. Cogn. 28:800–11 Smith JD, Minda JP, Washburn DA. 2004. Category learning in Rhesus monkeys: a study of the Shepard, Hovland, and Jenkins tasks. J. Exp. Psychol. Gen. 133:398–414 Squire LR, Knowlton BJ. 1995. Learning about categories in the absence of memory. Proc. Natl. Acad. Sci. USA 92:12470– 74 Sternberg S. 1966. High-speed scanning in human memory. Science 153:652–54

23 Dec 2004 19:45

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

178

AR

ASHBY

AR231-PS56-06.tex




AR231-PS56-06.sgm

LaTeX2e(2002/01/18)

P1: IKH

MADDOX

Strange W, Keeney T, Kessel FS, Jenkins JJ. 1970. Abstraction over time of prototypes from distortions of random dot patterns: a replication. J. Exp. Psychol. 83:508– 10 van Domburg PHMF, ten Donkelaar HJ. 1991. The Human Substantia Nigra and Ventral Tegmental Area. Berlin: Springer-Verlag Vauclair J. 2002. Categorization and conceptional behavior in nonhuman primates. In The Cognitive Animal: Empirical and Theoretical Perspectives on Animal Cognition, ed. M Bekoff, C Allen, GM Burghardt, pp. 239–45. Cambridge, MA: MIT Press Vogels R, Sary G, Dupont P, Orban GA. 2002. Human brain regions involved in visual categorization. Neuroimage 16:401–14 Volz H-P, Gaser C, Haeger F, Rzanny R, Mentzel H-J, et al. 1997. Brain activation during cognitive stimulation with the Wisconsin Card Sorting Test—a functional MRI study on healthy volunteers and schizophrenics. Psychiatry Res.: Neuroimaging 75:45– 157 Waldron EM, Ashby FG. 2001. The effects of concurrent task interference on category learning: evidence for multiple category

learning systems . Psychon. Bull. Rev. 8:168– 76 Wickens J. 1993. A Theory of the Striatum. New York: Pergamon Willingham DB. 1998. A neuropsychological theory of motor skill learning. Psychol. Rev. 105:558–84 Willingham DB, Wells LA, Farrell JM, Stemwedel ME. 2000. Implicit motor sequence learning is represented in response locations. Mem. Cogn. 28:366–75 Wilson CJ. 1995. The contribution of cortical neurons to the firing pattern of striatal spiny neurons. In Models of Information Processing in the Basal Ganglia, ed. JC Houk, JL Davis, DG Beiser, pp. 29–50. Cambridge, MA: Bradford Winocur G, Eskes G. 1998. Prefrontal cortex and caudate nucleus in conditional associative learning: dissociated effects of selective brain lesions in rats. Behav. Neurosci. 112:89–101 Zaki SR, Nosofsky RM, Jessup NM, Unversagt FW. 2003. Categorization and recognition performance of a memory-impaired group: evidence for single-system models. J. Int. Neuropsychol. Soc. 9:394–406

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Ashby.qxd 7/27/04 2:02 PM Page 1

HUMAN CATEGORY LEARNING

See legend on next page 179

Ashby.qxd

180

7/27/04

2:02 PM

ASHBY

■

Page 2

MADDOX

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Figure 1 (a) Three stimuli that might be used in a perceptual categorization experiment. Each stimulus is a circular sine-wave grating that varies across trials in spatial frequency and orientation. (b) A plot of stimuli that might be used in a rule-based category-learning task. Each plus denotes the spatial frequency and orientation of an exemplar from category A and each circle denotes a category B exemplar. The vertical line is the optimal category boundary. (c) A plot of stimuli that might be used in an information-integration category-learning task.

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

Annual Review of Psychology Volume 56, 2005

CONTENTS

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Frontispiece—Richard F. Thompson

xviii

PREFATORY In Search of Memory Traces, Richard F. Thompson

1

DECISION MAKING Indeterminacy in Brain and Behavior, Paul W. Glimcher

25

BRAIN IMAGING/COGNITIVE NEUROSCIENCE Models of Brain Function in Neuroimaging, Karl J. Friston

57

MUSIC PERCEPTION Brain Organization for Music Processing, Isabelle Peretz and Robert J. Zatorre

89

SOMESTHETIC AND VESTIBULAR SENSES Vestibular, Proprioceptive, and Haptic Contributions to Spatial Orientation, James R. Lackner and Paul DiZio

115

CONCEPTS AND CATEGORIES Human Category Learning, F. Gregory Ashby and W. Todd Maddox

149

ANIMAL LEARNING AND BEHAVIOR: CLASSICAL Pavlovian Conditioning: A Functional Perspective, Michael Domjan

179

NEUROSCIENCE OF LEARNING The Neuroscience of Mammalian Associative Learning, Michael S. Fanselow and Andrew M. Poulos

207

HUMAN DEVELOPMENT: EMOTIONAL, SOCIAL, AND PERSONALITY Behavioral Inhibition: Linking Biology and Behavior Within a Developmental Framework, Nathan A. Fox, Heather A. Henderson, Peter J. Marshall, Kate E. Nichols, and Melissa A. Ghera

235

BIOLOGICAL AND GENETIC PROCESSES IN DEVELOPMENT Human Development: Biological and Genetic Processes, Irving I. Gottesman and Daniel R. Hanson

263 vii

P1: JRX

December 8, 2004

viii

12:13

Annual Reviews

AR231-FM

CONTENTS

SPECIAL TOPICS IN PSYCHOPATHOLOGY The Psychology and Neurobiology of Suicidal Behavior, Thomas E. Joiner Jr., Jessica S. Brown, and LaRicka R. Wingate

287

DISORDERS OF CHILDHOOD Autism in Infancy and Early Childhood, Fred Volkmar, Kasia Chawarska, and Ami Klin

315

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

CHILD/FAMILY THERAPY Youth Psychotherapy Outcome Research: A Review and Critique of the Evidence Base, John R. Weisz, Amanda Jensen Doss, and Kristin M. Hawley

337

ALTRUISM AND AGGRESSION Prosocial Behavior: Multilevel Perspectives, Louis A. Penner, John F. Dovidio, Jane A. Piliavin, and David A. Schroeder

365

INTERGROUP RELATIONS, STIGMA, STEREOTYPING, PREJUDICE, DISCRIMINATION The Social Psychology of Stigma, Brenda Major and Laurie T. O’Brien

393

PERSONALITY PROCESSES Personality Architecture: Within-Person Structures and Processes, Daniel Cervone

423

PERSONALITY DEVELOPMENT: STABILITY AND CHANGE Personality Development: Stability and Change, Avshalom Caspi, Brent W. Roberts, and Rebecca L. Shiner

453

WORK MOTIVATION Work Motivation Theory and Research at the Dawn of the Twenty-First Century, Gary P. Latham and Craig C. Pinder

485

GROUPS AND TEAMS Teams in Organizations: From Input-Process-Output Models to IMOI Models, Daniel R. Ilgen, John R. Hollenbeck, Michael Johnson, and Dustin Jundt

517

LEADERSHIP Presidential Leadership, George R. Goethals

545

PERSONNEL EVALUATION AND COMPENSATION Personnel Psychology: Performance Evaluation and Pay for Performance, Sara L. Rynes, Barry Gerhart, and Laura Parks

571

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

CONTENTS

ix

PSYCHOPHYSIOLOGICAL DISORDERS AND PSYCHOLOGICAL EFFECTS ON MEDICAL DISORDERS Psychological Approaches to Understanding and Treating Disease-Related Pain, Francis J. Keefe, Amy P. Abernethy, and Lisa C. Campbell

601

TIMELY TOPIC

Annu. Rev. Psychol. 2005.56:149-178. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Psychological Evidence at the Dawn of the Law’s Scientific Age, David L. Faigman and John Monahan

631

INDEXES Subject Index Cumulative Index of Contributing Authors, Volumes 46–56 Cumulative Index of Chapter Titles, Volumes 46–56

ERRATA An online log of corrections to Annual Review of Psychology chapters may be found at http://psych.annualreviews.org/errata.shtml

661 695 700

18 Nov 2004 11:15

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

P1: IKH LaTeX2e(2002/01/18) 10.1146/annurev.psych.55.090902.141409

Annu. Rev. Psychol. 2005. 56:179–206 doi: 10.1146/annurev.psych.55.090902.141409 c 2005 by Annual Reviews. All rights reserved Copyright


PAVLOVIAN CONDITIONING: A Functional Perspective

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Michael Domjan Department of Psychology, University of Texas at Austin, Austin, Texas 78712; email: [email protected]

Key Words naturalistic conditioned stimuli, ecological learning, object learning, conditioned modifications of unconditioned behavior, adaptive significance of learning ■ Abstract From a functional perspective, Pavlovian conditioning involves learning about conditioned stimuli (CSs) that have a pre-existing relation to an unconditioned stimulus (US) rather than learning about arbitrary or neutral CSs. In addition, the most important product of learning involves changes in how the organism responds to the US, not in how it responds to the CS, because the US is the more biologically relevant stimulus. These concepts are illustrated using examples from a variety of behavioral and physiological situations including caloric intake and digestion, breast feeding, poison-avoidance learning, eyeblink conditioning, sexual conditioning, fear conditioning, aggression, and drug tolerance and sensitization.

CONTENTS INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Natural Learning Paradigms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conditioned Modifications of the Unconditioned Response . . . . . . . . . . . . . . . . . . . LEARNING WITH ECOLOGICALLY RELEVANT CONDITIONED STIMULI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Poison-Avoidance Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Taste-Odor Potentiation and Contrablocking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Caloric Conditioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sexual Conditioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fear Conditioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Maternal Nursing and Infant Suckling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CONDITIONED MODIFICATIONS OF RESPONSES TO THE UNCONDITIONED STIMULUS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conditioned Modification of the Eyeblink Response . . . . . . . . . . . . . . . . . . . . . . . . Sexual Behavior and Reproductive Success . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Aggression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Maternal Nursing and Infant Suckling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0066-4308/05/0203-0179$14.00

180 180 181 182 182 183 183 184 184 185 185 188 188 189 190 190

179

18 Nov 2004 11:15

180

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

P1: IKH

DOMJAN Fear-Potentiated Startle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conditioned Hypoalgesia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Digestion and Feeding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conditioning and Drug Tolerance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Drug Conditioning and Sensitization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . SUMMARY AND CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

191 191 192 193 195 195 197

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INTRODUCTION Pavlovian conditioning is one of the oldest and most extensively studied learning paradigms. The paradigm basically involves two stimuli. The unconditioned stimulus (US) elicits vigorous responding without any special prior training, or unconditionally. Because of that, the US originally was labeled the unconditional stimulus (Gantt 1966). In contrast, the conditioned stimulus (CS) elicits little more than an orienting response at first. The effectiveness of the CS depends, or is conditional, upon its pairings with the unconditioned stimulus. Hence, the CS originally was called the conditional stimulus. Learning is identified by the emergence of new responses to the CS, called conditioned responses or CRs. Because the development of conditioned responding depends on the pairing of the CS and US, the learning is considered to involve the establishment of an association between the CS and the US. This has made Pavlovian conditioning a favorite paradigm for the study of associative learning. Staddon (1983), for example, characterized Pavlovian conditioning as “the prototype for all signal learning” (p. 103). The associative tradition encouraged investigators to use conditioned stimuli that are initially unrelated to, or arbitrary, with respect to the US. In fact, the initial independence of the CS and US has been incorporated into the definition of Pavlovian conditioning. Authors have characterized the CS as “arbitrary” (Bower & Hilgard 1981, p. 49) or “neutral” (Anderson 1995, p. 10; Papini 2002, p. 491; Shettleworth 1998, p. 109; Staddon 1983, p. 102) with respect to the US. The view that Pavlovian conditioning involves learning about neutral or arbitrary cues that come to elicit conditioned behavior has provided a great deal of information about associative mechanisms. Parallel to, but in the shadows of, the associative tradition, a functional perspective on Pavlovian conditioning also has been developed (Domjan et al. 2000; Hollis 1982, 1997). The functional perspective is encouraged by the fact that Pavlovian conditioning has been demonstrated in a wide range of species and response systems (Turkkan 1989). The prevalence of Pavlovian conditioning suggests it is an adaptive trait that readily occurs under natural circumstances and serves to promote reproductive fitness, directly and/or indirectly. These are the key assumptions of a functional perspective (Dukas 1998; Hollis 1982, 1997; Shettleworth 1983, 1994).

Natural Learning Paradigms If Pavlovian conditioning is an adaptive trait, it presumably occurs under natural circumstances. However, outside the laboratory, the hand of an experimenter is not

18 Nov 2004 11:15

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

PAVLOVIAN CONDITIONING

P1: IKH

181

available to make sure that occurrences of a conditioned stimulus are reliably paired with presentations of a US. Rather, the CS-US pairings that are necessary for Pavlovian conditioning have to be features of the natural environment. For that to be the case, the CS has to be naturally related to the US instead of being an arbitrary cue or a neutral stimulus. An arbitrary CS may coincide with a US occasionally under natural circumstances, but such accidental pairings are bound to be rare. In addition, an accidental pairing will be preceded and followed by unpaired CS and US encounters, which will undermine the development of conditioned responding (e.g., Benedict & Ayres 1971, Rescorla 2000). Thus, although CS-US associations reside in the nervous system, they no doubt reflect physical relationships between conditioned and unconditioned stimuli that exist in the natural environment of the organism. A pre-existing relation between the CS and the US can take several different forms. One possibility is that the CS is a stimulus early in the causal chain of events that leads to the US (Dickinson 1980, Staddon 1988). Another possibility is that the CS is a feature of the US that initially does not elicit the unconditioned response but comes to do so after repeated encounters with the US. This was the case in the first documented demonstration of Pavlovian conditioning, which was conducted by S.G. Vul’fson in Pavlov’s laboratory (Boakes 1984, Todes 1997). Vul’fson repeatedly presented various substances to dogs (sand, dry food, wet food, or sour water) and measured the quantity and quality of saliva elicited by each. After one of these substances had been placed in the dog’s mouth several times, Vul’fson noticed that the dog would salivate when it was “teased” by having the substance presented at a distance. In Vul’fson’s demonstration, the US was the sand or dry food in the mouth. The CS was the sight and/or smell of the US at a distance. Notice that the CS was not unrelated to the CS at the outset of training. Rather, the CS and the US were different features of the same object (sand, for example). In the absence of experimental intervention, having a CS and a US that are different features of the same object helps to make sure that the CS will occur with the kind of contingent and temporal relation to the US that will result in the establishment of an association. This type of pre-existing relation between the CS and US is probably a feature of most naturally occurring instances of Pavlovian conditioning and therefore has to be carefully considered in a functional analysis. Therefore, I review instances in which the CS is a natural precursor of the US, to see whether these examples of Pavlovian conditioning have any unique properties.

Conditioned Modifications of the Unconditioned Response In addition to fashioning laboratory experiments that better mimic natural conditions, a functional approach requires focusing on aspects of behavior that are of potential adaptive significance. From a functional perspective, the important task for the organism is to interact effectively with the unconditioned stimuli it encounters. By definition, USs are of great biological significance (Pavlov 1927). In contrast, conditioned stimuli are important only because of their relation to a US. If the US does not occur, a response to the CS is a useless “false start.”

18 Nov 2004 11:15

182

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

P1: IKH

DOMJAN

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Conditioned responses are functionally significant only to the extent that they facilitate the subject’s interactions with the US. The significance of conditioned responses in facilitating the subject’s interactions with a US was brought into focus in several important papers by Hollis (1982, 1990, 1997). However, the implications of this claim have not been fully appreciated. If the primary function of Pavlovian conditioning is to improve how an organism interacts with the US, then the critical behavioral consequence that one should measure in studies of classical conditioning is how responses to the US change as a function of learning. Accordingly, I review various lines of evidence showing how Pavlovian conditioning can modify responses to unconditioned stimuli.

LEARNING WITH ECOLOGICALLY RELEVANT CONDITIONED STIMULI Definitions of classical conditioning that call for selecting a conditioned stimulus that is “neutral” or “arbitrary” imply that the choice of a CS is of little consequence for the outcome of conditioning experiments. That is hardly the case. The choice of the CS can make a big difference. One prominent parameter is CS intensity. Generally, more intense conditioned stimuli produce faster learning and more vigorous conditioned responding (e.g., Kamin 1965). The conditioned stimulus can also determine the nature of the conditioned response (Holland 1984). In this section, I review evidence that conditioning effects also vary as a function of the ecological relevance of the CS.

Poison-Avoidance Learning One of the most dramatic examples of learning influenced by the conditioned stimulus was discovered in the course of research on poison-avoidance learning. The ingestion of a poisonous food involves first tasting the food and then swallowing and digesting it. The poisonous effects become prominent as the food becomes absorbed in the body. This sequence of events results in the learning of an aversion to the taste of the food (Garcia et al. 1974, Rozin & Kalat 1971, Rzoska 1953). Taste-aversion learning has been modeled in the laboratory by permitting rats to drink an innocuous flavored solution (e.g., saccharin) and then injecting them with something that makes them sick (e.g., lithium chloride). By experimentally controlling the exposure to the CS and the US, investigators have been able to examine how the learning is influenced by the nature of the CS and the effects of the CS-US interval. Taste aversion learning can occur in a single trial, even if the US is delayed several hours after the CS (Garcia et al. 1966, Smith & Roll 1967). However, the aversion learning occurs much more readily if the CS is a novel taste than if the CS is an auditory and/or visual cue (Domjan & Wilson 1972, Garcia & Koelling 1966). Furthermore, the long-delay learning of taste aversions is at least in part due to the specificity of the learning to taste cues (Revusky 1977).

18 Nov 2004 11:15

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

PAVLOVIAN CONDITIONING

P1: IKH

183

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

For organisms that hold food in their front paws, the aversive consequence of eating poisonous food is reliably preceded not only by taste cues but also by the olfactory and tactile cues of the poisonous food as it is held in preparation for eating. Aversions also can develop to these tactile and olfactory features of the food (Domjan 1973, Domjan & Hanlon 1982). However, food-related tactile and olfactory cues have to be distinguished from the tactile and olfactory cues that are encountered during the course of locomotion or other noningestive activities. Studies have shown that at least in the case of odor aversion learning, the sensations provided by the motor movements involved in eating and drinking facilitate this discrimination (Domjan 1973).

Taste-Odor Potentiation and Contrablocking The taste and odor features of ingested food are closely related because as one gets close enough to a food to taste it, one invariably encounters the food’s odor. This relationship appears to result in some unusual learning effects. When two conditioned stimuli are presented simultaneously and paired with a US, the typical outcome (called overshadowing) is that the presence of one CS interferes with the conditioning of the other (e.g., Kamin 1969). In contrast to overshadowing, when a taste and an odor are presented together prior to illness, the presence of the taste sometimes facilitates rather than interferes with the conditioning of the odor (Bouton et al. 1986; Rusiniak et al. 1982a,b). This phenomenon is called potentiation. Another important compound cue effect is “blocking.” In the blocking design, subjects are conditioned first to asymptote with one CS. A second CS is then added to the first one, as conditioning trials are continued. The interesting result is that the presence of the initially conditioned CS interferes with (or blocks) the conditioning of the added CS. The blocking effect was demonstrated initially in fear conditioning in rats with light and noise conditioned stimuli (Kamin 1969) and became a keystone phenomenon that inspired numerous major theories of learning. However, an effect opposite blocking (contrablocking) occurs if taste and odor cues are used. When an odor cue is added to a previously conditioned taste stimulus (or a taste cue is added to a previously conditioned odor), aversion conditioning of the added stimulus is facilitated rather than blocked (Batsell et al. 2001, Batson & Batsell 2000).

Caloric Conditioning Nutritious foods provide caloric repletion rather than poisoning as the postingestive consequence. Tastes can become associated with caloric repletion, with the outcome that subjects increase their preference for the associated flavor (Fedorchak 1997, Sclafani 1997). The caloric substances may be mixed with the CS flavor or intubated directly into the stomach. The resultant flavor preferences have been characterized as highly resistant to extinction (Fedorchak 1997, Mehiel 1991), but it is not clear to what the resistance to extinction was compared. Interestingly, a caloric

18 Nov 2004 11:15

184

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

P1: IKH

DOMJAN

conditioned preference is sensitive to changes in hunger, with higher preferences evident when the subject is food deprived as compared with being nondeprived (e.g., Capaldi et al. 1994). This latter finding suggests that stimulus-stimulus, or S-S, learning mechanisms mediate caloric conditioning.

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Sexual Conditioning The effects of environmental regularities have been also examined in sexual conditioning with male Japanese quail (Domjan et al. 2004). Sexual conditioning is conducted by presenting a CS shortly before providing the male with access to a female. If a localized stimulus (e.g., a light) serves as the CS, the male quickly comes to approach the CS as the conditioned response (Domjan et al. 1986). Japanese quail are ground birds that live in grassy areas (Schwartz & Schwartz 1949). When a male initially detects a female, he is likely to see just a part of her body, perhaps her head sticking up through the grass. By approaching these limited visual cues, the male will get close to the female and may get a chance to copulate with her. This sequence of events may be modeled in the laboratory by presenting a CS that includes the taxidermically prepared head of a female (see Cusato & Domjan 1998, Figure 1), and following that with access to a live female with whom the male may copulate. Studies have shown that such a naturalistic CS elicits only modest approach behavior unconditionally. However, if the CS is paired with access to a live female, conditioned approach behavior significantly increases (K¨oksal et al. 1994). In addition, the males also come to grab and attempt copulations with the naturalistic CS (Cusato & Domjan 1998). Such conditioned copulatory responses did not develop in subjects that were conditioned with an arbitrary CS that had the same size and shape but lacked female head cues. Additional comparisons revealed a constellation of learning effects that differentiated the naturalistic CS from the more conventional arbitrary CS. The naturalistic CS was resistant to blocking (K¨oksal et al. 1994) and failed to show extinction (Krause et al. 2003). It was also resistant to increases in the CS-US interval (Akins 2000), resulted in stronger second-order conditioning (Cusato & Domjan 2001), and showed a sensitization rather than a habituation effect with repeated unreinforced exposures (Cusato & Domjan 1998).

Fear Conditioning Fear conditioning typically is investigated using experimental procedures in which an auditory or visual CS is presented to laboratory rats shortly before a brief foot shock. Outside the laboratory, however, fear conditioning is likely to cues that are natural precursors of an aversive event. Such precursors are easy to identify in predator-prey interactions. For example, the sight and sound of a rattlesnake preying on a rabbit is a natural precursor of the biting attack. An extensive series of experiments have examined the conditioning of fear to the sight of a snake in

18 Nov 2004 11:15

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

PAVLOVIAN CONDITIONING

P1: IKH

185

¨ monkeys and people (Ohman & Mineka 2001, 2003). In both species, stronger fear conditioning was found when the CS was provided by visual cues of a snake rather than visual cues of flowers or mushrooms. However, the flower and mushroom stimuli were just as effective as the sight of a snake in appetitive conditioning procedures (EW Cook et al. 1986, M Cook & Mineka 1990). This selective advantage of the snake stimuli in fear conditioning was also evident when the stimuli were presented to human participants without their conscious awareness ¨ (Ohman & Soares 1998). The evolutionary basis of selective conditioning of snake stimuli is further supported by evidence that snake cues are more effective in human fear conditioning than are visual cues provided by a gun (EW Cook et al. 1986). Selective associations in fear conditioning also have been demonstrated in pigeons, where auditory cues have been found to be more effective as a CS for shock, whereas visual cues are more effective as a signal for food (LoLordo 1979). In addition, auditory-shock associations are resistant to the blocking effect (LoLordo et al. 1982). The presentation of a previously conditioned CS does not interfere with the development of a tone-shock association. Thus, as in sexual conditioning (K¨oksal et al. 1994), a naturalistic CS is less susceptible to blocking.

Maternal Nursing and Infant Suckling Another situation in which the natural course of events results in the reliable pairing of conditioned and unconditioned stimuli is provided by the interactions between mother and infant that occur during the course of nursing. The unconditioned stimulus for the milk letdown and milk ejection reflex is suckling stimulation of the breast. Olfactory and other cues from the infant typically precede suckling stimulation, and these cues can become conditioned to elicit the release of oxytocin and prolactin, hormones that stimulate milk letdown and ejection (Fuchs et al. 1987; Grosvenor & Mena 1972, 1974; McNeilly et al. 1983). Correspondingly, various exteroceptive cues provided by the mother before a nursing episode can become conditioned to elicit suckling on the part of the infant (Blass 1990, Blass et al. 1984). Research has clearly demonstrated that such conditioned endocrine and suckling responses can develop. However, in this line of research specific experiments have not been conducted to see if naturalistic conditioned stimuli are more effective than arbitrary cues.

Discussion The evidence reviewed above suggests that learning with ecologically relevant stimuli often proceeds differently from learning with arbitrary cues. In foodaversion learning, sexual conditioning, and fear conditioning, the use of an ecologically relevant CS resulted in acquisition that was more robust and was resistant to the blocking effect. In food aversion learning and sexual conditioning, the learning also occurred over longer CS-US delays. These and the other contrasts with

18 Nov 2004 11:15

186

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

P1: IKH

DOMJAN

conventional conditioning phenomena that were described suggest that efforts to understand how Pavlovian conditioning occurs in the natural environment have to consider the role of pre-existing relations between CSs and USs.

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

CONDITIONED STIMULUS INTENSITY, SALIENCE, AND BIOLOGICAL SIGNIFICANCE

One possible explanation for the more robust learning effects that were observed when an ecologically relevant CS was used in food-aversion learning, sexual conditioning, and fear conditioning is that the naturalistic CSs were more intense or salient than the corresponding arbitrary cues. Consistent with this possibility, several investigators have reported that a target CS is less likely to be blocked by a previously conditioned cue if the target CS is of higher intensity (Feldman 1975, Hall et al. 1997, Miller & Matute 1996). However, stimulus salience or intensity is not likely to be the primary dimension that distinguishes naturalistic CSs from arbitrary cues. In studies of food aversion learning, taste cues were more effective than audiovisual cues only if poisoning was used as the US. When shock served as the US, the relative effectiveness of taste versus audiovisual cues was reversed (Domjan & Wilson 1972, Garcia & Koelling 1966). Similarly, in studies of fear conditioning in pigeons, auditory cues were more effective than visual cues only if the US was shock—not if the US was food (LoLordo 1979). In studies with monkeys and people, the flower and mushroom stimuli that were not effective in fear conditioning worked well when appetitive conditioning was conducted (EW Cook et al. 1986, M Cook & Mineka 1990). Another way to characterize the difference between a naturalistic CS and an arbitrary cue is that the former is of greater biological significance. Gunther et al. (1997) defined biological significance by the vigor of responding (conditioned or unconditioned) that is elicited by a stimulus at the outset of a training procedure. Consistent with this interpretation, the naturalistic CS used in the sexual conditioning experiments elicited more responding even in the absence of conditioning than did the arbitrary CS (Cusato & Domjan 1998, K¨oksal et al. 1994), and this may have been responsible for the resistance to blocking, extinction, and increases in the CS-US interval that was observed. Stimuli of greater biological significance also have been shown to be more resistant to blocking, overshadowing, relative validity, and degraded contingency manipulations in fear conditioning (Miller & Matute 1996, Oberling et al. 2000). In many of the examples described above, the pre-existing relation between CS and US was provided by the fact that the stimuli were different features of the same object. Food-aversion learning and caloric conditioning involve learning about different features of ingested food. The CS feature is the taste of the food, and the US feature is its aversive or caloric postingestional consequence. One may also conceptualize the sexual conditioning with a naturalistic CS as an instance of object learning. The naturalistic CS included the taxidermically prepared head of a female quail, whereas the US was copulatory access to a live female. The partial female visual cues that were provided by the CS were just one

OBJECT LEARNING

18 Nov 2004 11:15

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

PAVLOVIAN CONDITIONING

P1: IKH

187

feature of the US, which included additional visual, auditory, and tactile stimuli, as well as movement cues produced by the female’s behavior. In fear conditioning, visual cues provided by a snake were paired with an aver¨ sive US. In the laboratory experiments (Ohman & Mineka 2001, 2003), the aversive US was not a snakebite but a shock or the sight of a demonstrator monkey behaving fearfully. However, the laboratory procedures presumably activated processes that evolved to deal with encounters with dangerous snakes. In the conditioning of oxytocin and prolactin release, the CS and US were both provided by the infant. Suckling on the part of the infant provided the US, and cues provided by the infant before nipple attachment served as the CS. Correspondingly, different features of the mother provided the CS and the US in conditioning infant suckling. In this case, orosensory stimuli of the nipple served as the US and maternal cues preceding access to the nipple served as the CS. Object learning or part-whole associations also have been examined in the context of drug conditioning (Cepeda-Benito & Short 1997, Greeley et al. 1984, Kim et al. 1999) In these experiments, a small dose of a drug was given as a CS for a larger subsequent dose. For control subjects, the two drug doses were administered in an unpaired fashion. Test trials revealed that the small drug dose came to activate a conditioned compensatory response (see Conditioning and Drug Tolerance section below). However, further research is required to determine whether such drug CSs have properties different from more arbitrary CSs in these situations. Learning that involves associating different features of an object is no doubt widespread. Organisms have to learn about all sorts of objects to successfully navigate their environment. We can recognize something as being a chair even if we only see a small part of it because we have learned a constellation of associations involving different features of chairs. The prevalence of object learning makes object learning a useful heuristic for identifying CS-US pairings in the natural environment. If a US object has multiple features, only some of which elicit behavior unconditionally, the initially ineffective features of the object may come to elicit responding as well through Pavlovian conditioning. Whether the concept of object learning helps to explain some of the special properties of learning with ecologically relevant stimuli remains to be seen. The more rapid and robust learning that was observed with ecologically relevant stimuli in food-aversion learning, sexual conditioning, and fear conditioning is yet to be ¨ satisfactorily explained. Ohman & Mineka (2001, 2003) favor an evolutionary account of the special efficacy of snake cues in fear conditioning. An evolutionary explanation also was offered for the special efficacy of taste cues in food aversion learning (Garcia et al. 1974). Evolutionary explanations leave open the question ¨ of proximate mechanism. Both Ohman & Mineka (2001) and Garcia et al. (1974) offered neurophysiological proximate hypotheses. Behavioral mechanisms may also promote object learning. When the CS and the US are different features of the same object, the two events are likely to covary more closely and may have more stimulus elements in common. Stricter covariation and greater similarity can both promote CS-US associations (e.g., Rescorla & Furrow 1977, Testa 1975).

18 Nov 2004 11:15

188

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

P1: IKH

DOMJAN

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

CONDITIONED MODIFICATIONS OF RESPONSES TO THE UNCONDITIONED STIMULUS An associationist approach to classical conditioning emphasizes the learning of an association between the CS and the US. Once this association has been learned, presentation of the CS activates a representation of the US (Rescorla 1988) or a representation of what significant biological event is about to occur. However, it was clear even from the research in Pavlov’s laboratory that subjects learn not only what will occur but also when it will happen. Pavlov (1927) noted that when an extended CS-US interval is used, the CR becomes delayed as training progresses, until it occurs close to the actual time of US delivery. (This phenomenon was termed “inhibition of delay.”) Subsequent research has confirmed that subjects encode not only CS-US associations but also precise information about the temporal relation between the CS and the US (Blaisdell et al. 1998, Denniston et al. 1998, Savastano et al. 1998), and that conditioned responses can be timed beautifully to occur just when the US is about to be presented (Kehoe et al. 1989). If subjects timed their CRs perfectly, the CR would occur exactly when the US was delivered, and no behavior would be evident during the CS-US interval. This has encouraged the use of test trials to measure conditioned behavior. However, such test trials miss the critical function of Pavlovian conditioning, which is to permit the subject to respond to the US more effectively. From a functional perspective, the most important consequence of learning is how the subject’s interactions with the US change as a function of having that US preceded or signaled by a conditioned stimulus. Such conditioned modifications of the UR have been documented in several situations.

Conditioned Modification of the Eyeblink Response Irritation of the eye elicits a blink unconditionally. The eyeblink reflex has been a popular response system for the study of classical conditioning (Gormezano et al. 1983). Furthermore, one of the first demonstrations of conditioned modification of the UR was in eyeblink conditioning. Testing human participants, Kimble & Ost (1961) found that a CS that had been paired with an airpuff US not only elicits a blink as a conditioned response, but the magnitude of the blink response to the airpuff is attenuated by prior presentation of the CS. This phenomenon was called “conditioned diminution of the UR.” Subsequent investigators did not always replicate the conditioned diminution effect and also reported enhanced responding to the US following exposure to a CS on occasion. A more systematic examination of the parameters of the conditioned diminution effect suggested that a critical variable is the intensity of the US. Conditioned facilitation of the UR is more likely with low US intensities, and conditioned diminution of the UR is likely with higher US intensities (Donegan & Wagner 1987). Thus, a CS may attenuate or enhance unconditioned responding under different parametric conditions.

18 Nov 2004 11:15

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

PAVLOVIAN CONDITIONING

P1: IKH

189

Contrasting conditioned modifications of the UR also have been observed in conditioning with pharmacological unconditioned stimuli (see below).

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Sexual Behavior and Reproductive Success As is the case in other domains of Pavlovian conditioning, the emphasis in studies of sexual conditioning has been on conditioned responses that develop to cues that are predictive of sexual reinforcement (Domjan & Holloway 1998). However, a growing body of evidence indicates that a sexually conditioned CS can also facilitate responding to the US in a sexual situation. Exposure to a sexually conditioned stimulus decreases the latency of rats to ejaculate during copulations with a female (Zamble et al. 1985), and decreases the latency of male quail to initiate copulation with a female (Domjan et al. 1986). Furthermore, this decrease in copulatory latency helps to determine the outcome of sexual competition. When two male quail receive access to a single female, the male that is able to predict the sexual encounter because of a CS is able to copulate with the female first (Guti´errez & Domjan 1996). In studies with the blue gourami fish (Trichogaster trichopterus), males that encountered a female after exposure to a sexually conditioned light stimulus showed reduced levels of aggression and more frequent courtship appeasement action patterns in response to the female (Hollis et al. 1989). In other studies, male quail were found to be more responsive to minimal female cues if these cues were presented in a context that was previously paired with access to a female (Hilliard et al. 1997). Other evidence of conditioned modifications of unconditioned behavior is evident in female quail presented with a conditioned stimulus that signals the impending introduction of a male (Guti´errez & Domjan 1997). The females show no response to a sexually conditioned stimulus but the CS makes them more receptive once they encounter a male. The ultimate standard for evidence of functional significance is reproductive success. If a behavioral trait is of adaptive significance, individuals with that behavioral trait should produce greater numbers of offspring. Sexual behavior is one of the few systems in which the contributions of learning to reproductive outcomes can be measured directly. In the first study involving direct measurement of reproductive outcome, Hollis et al. (1997) permitted male blue gourami to copulate and tend eggs after exposure to a sexually conditioned stimulus. Males that encountered a female after exposure to a sexually conditioned stimulus showed less aggression toward the female, more nest-building behavior, more clasping behavior, and shorter latencies to spawn. Most importantly, sexual encounters that were preceded by a conditioned stimulus yielded more than 10 times as many offspring as unsignaled encounters. Sexual conditioning has been also found to increase reproductive parameters in quail. Male quail release greater numbers of sperm following exposure to a sexually conditioned stimulus as compared with a control condition (Domjan et al. 1998). More recently, Adkins-Regan & MacKillolp (2003) found that the sexual conditioning of either the male or the female increases the number of fertilized eggs

18 Nov 2004 11:15

190

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

P1: IKH

DOMJAN

that are produced following a copulatory interaction. Increased numbers of fertilized eggs also were found in a subsequent study (Mahometa & Domjan 2004), but only when both the male and the female were able to anticipate a sexual encounter. In addition, Mahometa & Domjan (2004) showed that as with the blue gourami, the greater fertilization success is correlated with changes in how the males and females react to each other. Exposure to a sexually conditioned stimulus increases the female’s receptivity and the efficiency of the male’s copulatory behavior.

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Aggression Agonistic behavior contributes to reproductive success through the defense of important resources, such as food, territory, nesting sites, or potential mates (Poole 1985). The unconditioned stimulus for aggression is the presence of a rival or intruder male. In many studies of the conditioning of agonistic behavior, the emphasis has been on the development of aggressive responses to a conditioned stimulus that was paired with a territorial intruder (e.g., Jenkins & Rowland 1996, Thompson & Sturm 1965). However, Hollis (1984, 1990) found that such Pavlovian conditioning also increases the effectiveness with which a territorial male fights with an intruder. Male blue gourami for which an aggressive encounter was preceded by a previously conditioned light CS made significantly more bite and tail-beating responses than did subjects in a control group that previously had the CS unpaired with the US. Interestingly, the success that paired subjects experienced in their test encounter with the intruder also made these males more effective combatants in an unsignaled fight conducted two weeks later (Hollis et al. 1995). Thus, the increase in US effectiveness that was initially stimulated by the presentation of the CS was long-lasting and persisted when the US subsequently was presented in the absence of the CS.

Maternal Nursing and Infant Suckling As described above, another social situation that readily lends itself to conditioning effects involves the interactions between mother and infant that occur during the course of nursing. Although the emphasis in this area of research has been on the development of conditioned endocrine and suckling responses to conditioned stimuli, these conditioned stimuli may also enhance the effectiveness of an unconditioned stimulus. Typically, the unconditioned stimulus for oxytocin secretion and milk letdown is suckling stimulation provided by the infant. However, other aspects of the infant can serve as CSs and elicit these as conditioned responses (Fuchs et al. 1987; Grosvenor & Mena 1972, 1974; McNeilly et al. 1983). Furthermore, the CS cues provided by the infant may also enhance responses to the suckling US. In a recent study involving dairy cows, for example, Tancin et al. (2001) found that the presence of a mother’s calf can increase oxytocin release and milk yield in response to standard unconditioned stimulation of the teats by a milking machine. These effects were more prominent in multiparous cows that presumably had a more extensive conditioning history involving the pairing of a calf with tactile stimulation of the teats.

18 Nov 2004 11:15

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

PAVLOVIAN CONDITIONING

P1: IKH

191

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Fear-Potentiated Startle One paradigm in which the focus has always been how the conditioned stimulus modifies responses to an unconditioned stimulus is knows as the fear-potentiated startle. The startle response is a defensive response that can be readily elicited in the laboratory by a brief loud noise (Hoffman & Ison 1980). In this system, the brief loud noise is the unconditioned stimulus and the startle response is the unconditioned response. Brown et al. (1951) demonstrated that conditioned fear elicited by an audiovisual stimulus enhances the startle response elicited by a brief loud noise in rats. The parameters of this fear-potentiated startle procedure have since been examined in detail (e.g., Walker & Davis 2002) and the paradigm has become a highly productive technique for investigating the neural and pharmacological mechanisms of fear and defensive behavior (Davis 1997, Fendt & Fanselow 1999, Hijzen et al. 1995). Most of the research on fearpotentiated startle has been conducted with laboratory rats. However, the procedure has been also extended to rhesus monkeys (Winslow et al. 2002). Fearpotentiated startle has been also studied with human participants, using the eyeblink response as an index of startle, and shock or the threat of shock paired with a CS to elicit fear (Ameli et al. 2001, Grillon & Davis 1997, Riba et al. 2001).

Conditioned Hypoalgesia Aversive conditioning can also lead to changes in pain elicited by an aversive unconditioned stimulus. This phenomenon was initially labeled “conditioned analgesia,” but because one cannot be certain that pain is eliminated entirely by exposure to a conditioned stimulus, a more conservative term for the effect is “conditioned hypoalgesia.” In one study (Fanselow & Baackes 1982), for example, rats received shock in a distinctive experimental chamber and were then tested for conditioned fear and pain sensitivity either in the same context or in a different context that had not been paired with shock. Conditioned fear was measured in terms of time spent freezing, and pain sensitivity was measured by recording recuperative responses to having an irritant (a small dose of formalin) injected into one of the hind paws. Subjects tested in the shock chamber showed extensive freezing but little reactivity to the painful formalin injection. In contrast, subjects tested in the alternative context showed very little freezing and substantial levels of recuperative behavior induced by the formalin injections. Similar context-elicited hypoalgesia has been obtained with cold-water swimming and exposure to carbon dioxide as the aversive US (Blustein et al. 1997, Mongeluzi et al. 1996). Conditioned hypoalgesia also has been observed with discrete conditioned stimuli (Illich & Grau 1991, Matzel & Miller 1987). Conditioned hypoalgesia exhibits many of the properties of other conditioning effects, including extinction (Fanselow 1984, Matzel et al. 1988), blocking (Ross 1985), latent inhibition (Maier & Watkins 1991), conditioned inhibition (Wiertelak et al. 1992), and second-order conditioning (Ross 1986).

18 Nov 2004 11:15

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

192

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

P1: IKH

DOMJAN

The hypoalgesia elicited by the shock-associated context in the study by Fanselow & Baackes (1982) was reversed by treatment with the opiate antagonist naltrexone, suggesting that the reaction was mediated by the release of endogenous opiates (Matzel & Miller 1987). The conditioned hypoalgesia and conditioned freezing are also attenuated by benzodiazepines (Fanselow & Helstetter 1988). Conditioned hypoalgesia has important implications for the mechanisms of aversive conditioning. As Fanselow (1991) noted, “analgesia may act as a negative feedback loop that regulates conditioning” (p. 79). During the course of acquisition, the development of conditioned hypoalgesia will reduce the effectiveness of the US and thereby limit the development of conditioned fear. Conditioned hypoalgesia may also be responsible for the blocking effect (Fanselow 1998). Consistent with these predictions, treatment with naloxone increases fear conditioning (Fanselow 1981, Westbrook et al. 1991), reduces the blocking effect (Schull 1979), and also attenuates the US pre-exposure effect, which is a form of blocking in which the context serves as the previously conditioned stimulus (Matzel et al. 1988).

Digestion and Feeding The first unconditioned stimulus used in studies of Pavlovian conditioning was food. If an important function of Pavlovian conditioning is to modify how the organism interacts with the unconditioned stimulus, evidence of this should be available with food USs. It was recognized early on that the conditioned salivation that occurred in anticipation of dry meat powder can be helpful in digesting the food, and salivation in anticipation of an irritant in the mouth (e.g., a weak acidic solution) can serve to dilute the aversive stimulus. More recently, Woods and his colleagues have argued that such anticipatory responses are critical for the efficient digestion of large meals (Woods 1991, Woods & Ramsay 2000, Woods & Seeley 2002, Woods & Strubbe 1994). Food intake triggers major physiological adjustments involved in the digestion, absorption, and storage of the energy source. Woods has compared eating a large meal to suffering a major physiological assault (Woods 1991). In addition to stimulating the secretion of digestive hormones and enzymes, eating causes the release of a cascade of stress hormones, including adrenocorticotropic hormone, epinephrine, and norepinephrine. Pavlovian conditioning serves to mitigate the disruptive effects of eating by mobilizing the secretion of digestive hormones and enzymes before the food reaches the gut. As Woods noted, “by successfully anticipating the ingestion of food, animals can make appropriate compensatory responses and hence lessen the impact of eating upon the body” (Woods 1991, p. 492). The importance of Pavlovian processes in digestion is clearly illustrated by how insulin secretion is regulated. Insulin is required for the transfer of nutrients from the blood to target tissues and is released by the pancreas in response to elevated serum levels of carbohydrates, fats, and proteins. However, insulin is often released before ingested nutrients are absorbed into the circulatory system.

18 Nov 2004 11:15

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

PAVLOVIAN CONDITIONING

P1: IKH

193

The taste or smell of food can trigger the release of insulin long before the food is digested and enters the bloodstream. This has been labeled “cephalic insulin” because it is mediated by brain mechanisms rather than signals originating in the gut. Cephalic insulin secretions also occur in anticipation of predictable large meals and in response to conditioned stimuli that precede meals (Woods et al. 1977). Anticipatory insulin secretion also can be conditioned with injections of insulin rather than meals serving as the US (Woods & Kulkosky 1976). Cephalic insulin and conditioned insulin secretions are responses to cues that precede the gut stimuli that ordinarily serve as the US for insulin secretion. Hence, these are conditioned responses rather than modified unconditioned responses. However, as Woods & Strubbe (1994) have pointed out, “the increment in cephalic insulin coming when it does circumvents the need for a much greater postprandial insulin response” (p. 149). Measurements of the insulin response consequent to the ingestion of a meal would provide decisive evidence on this point. One interesting implication of these mechanisms is that subjects who are able to anticipate a meal should be able to tolerate the stresses of eating more effectively. Consistent with this prediction, rats eat more following a signal for feeding (Zamble 1973). In the study by Zamble, visual and auditory stimuli served as the conditioned stimulus predictive of a meal. Temporal cues can also serve as conditioned stimuli if meals are provided at fixed intervals (e.g., once a day). Interestingly, rats fed at the same time each day eat less if their usual feeding is delayed so that it does not occur in conjunction with the usual temporal CSs (Bousfield & Elliott 1934). This is a remarkable finding because delaying a feeding increases food deprivation. As Woods and his colleagues have pointed out, the Pavlovian approach to the analysis of eating provides a perspective that contrasts with more traditional negative feedback models. Both approaches start with the axiom that ingestion serves to provide needed nutrients and is part of a homeostatic regulatory system. According to classic negative feedback models, organisms monitor an aspect of energy balance (levels of glucose or lipids, for example), and ingestion is initiated when a physiological index of energy balance indicates a deficit. Food intake then rectifies this deficit. The Pavlovian approach suggests that physiological antecedents of meals such as a drop in blood sugar or a decrease in metabolic rate do not trigger eating but help to process the impending meal more effectively (Woods & Seeley 2002, Woods & Strubbe 1994).

Conditioning and Drug Tolerance The administration of a psychoactive drug also may be viewed as creating a major physiological disturbance, whose anticipation may permit the recruitment of processes to deal more effectively with the drug insult. Thus, the types of conditioning mechanisms that facilitate the digestion of food are potentially also relevant to coping with drug experiences (Woods & Ramsay 2000). That drugs may act as unconditioned stimuli was evident to Pavlov, whose associates observed salivation

18 Nov 2004 11:15

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

194

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

P1: IKH

DOMJAN

and other conditioned responses in dogs that were exposed to cues that reliably preceded the administration of morphine and apomorphine (Pavlov 1927, pp. 35– 37). However, as in his studies with food, Pavlov emphasized the development of responses to conditioned stimuli that did not have much impact before being paired with the drug administrations. How the presentation of a drug-conditioned stimulus may alter the organism’s reactions to the drug itself was not considered until later in the twentieth century, when investigators became interested in the role of Pavlovian conditioning in drug tolerance and sensitization (Eikelboom & Stewart 1982; Siegel 1989, 1991; Siegel et al. 2000; Woods & Ramsay 2000; Young & Goudie 1994). Drug tolerance refers to a decrease in a measured drug effect that is frequently observed with repeated administrations of the drug. Learning has been implicated in drug tolerance in part because drug tolerance can last as long as a year (Cochin & Kornetsky 1964). However, the most significant feature of drug tolerance that has encouraged a learning interpretation is its situation specificity. Drug tolerance is most readily observed when the circumstances of drug administration during a test trial are the same as the circumstances that existed during prior drug treatments. If the context or cues in the presence of which the drug was previously administered are altered, drug tolerance is reduced or abolished. The situation specificity of drug tolerance has been demonstrated with a variety of drugs, including opiates (Siegel 1991), ethanol (Lˆe et al. 1979), nicotine (Cepeda-Benito et al. 2000), pentobarbital (Cappell et al. 1981), haloperidol (Poulos & Hinson 1982), and benzodiazepines (King et al. 1987). Other features of learning evident in drug tolerance include extinction (Siegel et al. 1980); external inhibition (Poulos et al. 1988); latent inhibition (Goodison & Siegel 1995); inhibitory conditioning (Fanselow & German 1982); stimulus generalization and loss of stimulus control over time (Feinberg & Riccio 1990); and sensory preconditioning, blocking, and overshadowing (Dafters & Bach 1985, Dafters et al. 1983). Pavlovian analyses of drug tolerance have emphasized how physiological and behavioral responses to a drug are attenuated by the presence of drug-predictive cues that become conditioned by repeated drug administrations. Efforts to understand why a drug-conditioned stimulus contributes to tolerance have involved examining conditioned responses elicited by the CS in the absence of the drug itself. For drugs that show the development of tolerance, the CS generally elicits physiological and behavioral changes that are opposite to or compensate for the drug effects. Such drug-compensatory CRs have been observed with a variety of drugs, including opiates (Grisel et al. 1994, Krank et al. 1981), ethanol (Larson & Siegel 1998), and caffeine (Andrews et al. 1998, Rozin et al. 1984). The Pavlovian analysis of drug tolerance is based on homeostatic regulatory concepts. It assumes that the administration of a drug creates physiological disturbances that in turn activate compensatory changes that serve to attenuate those perturbations. Initially, the compensatory adjustments occur only as delayed unconditioned responses to the drug. However, as conditioning proceeds, the unconditioned compensatory responses also come to be activated by the drug-predictive CSs and

18 Nov 2004 11:15

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

PAVLOVIAN CONDITIONING

P1: IKH

195

thereby contribute to the attenuated drug effects that are observed (Dworkin 1993, Eikelboom & Stewart 1982). The Pavlovian analysis of drug tolerance is well supported by a large body of evidence and has been extended to analyses of drug abuse and treatment (Siegel et al. 2000, Siegel & Ramos 2002). When originally formulated, the model focused on exteroceptive drug-predictive cues. More recently, the model has been extended to also incorporate internal cues involved in the initiation of an episode of drug self-administration (Donny et al. 1995) as well as internal cues characteristic of the onset of a drug experience (Kim et al. 1999). These extensions help explain why drug tolerance is more evident if the drug is self-administered and why some cases of long-term tolerance are not context specific (Siegel et al. 2000, Siegel & Ramos 2002).

Drug Conditioning and Sensitization Although the Pavlovian model of drug tolerance has enjoyed wide success, it is limited to physiological systems that involve some form of homeostatic regulation. Systems in which the physiological changes induced by a drug do not activate compensatory unconditioned adjustments are not predicted to show conditioned tolerance. Such systems may in fact show the opposite outcome, namely sensitization. Sensitization is an increase in the impact of a drug that occurs with repeated drug administrations. Although research on the contributions of Pavlovian conditioning to drug sensitization is not as extensive as research on conditioned drug tolerance, Pavlovian processes have been implicated in drug sensitization as well (Stewart 1992). Furthermore, Pavlovian sensitization is assumed to play an important role in models of drug abuse. In particular, the development of drug craving has been attributed to a context-specific drug sensitization process (Robinson & Berridge 1993, 2000). Anagnostaras & Robinson (1996), for example, demonstrated sensitization of locomotor behavior in rats elicited by amphetamine. The rats were first given 10 injections of a fairly high dose of amphetamine (3 mg/kg) before being tested with one of several amphetamine doses. Sensitization developed in the context in which the drug was administered but was not observed if the subjects were tested with amphetamine in a different context. Furthermore, extinction of the contextual cues attenuated the drug sensitization effect (see also Drew & Glick 1988, Hinson & Poulos 1981, Post et al. 1981, Terelli & Terry 1999).

Discussion The emphasis on research concerned with CS-induced modifications of responding to the unconditioned stimulus has been on documenting such effects and building a case that such effects obey conventional laws of associative learning. The mechanisms of such effects have garnered much less attention. Perhaps the simplest way to explain such effects is in terms of the summation of responses to the CS and the US. According to the summation model, certain conditioned responses

18 Nov 2004 11:15

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

196

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

P1: IKH

DOMJAN

come to be elicited by the CS. If the US is presented right after the CS, the responses observed during the US period represent the summation of delayed CRs and responses directly elicited by the US. The summation model has been used to explain conditioned drug tolerance (e.g., Siegel 1989). In this explanation, the CS is presumed to elicit a drug-compensatory conditioned response, which serves to attenuate the effects of the drug that is administered in the CS context. Consistent with this interpretation, in many instances of conditioned tolerance, the response elicited by the drug-conditioned CS is opposite the initial physiological disturbance caused by the drug itself. The summation model has also been used to explain conditioned modifications of eyeblink responding to a US (Donegan & Wagner 1987), and the model readily lends itself to explanations of UR modification in nursing, suckling, and feeding, although it has not been precisely tested in those situations. Although the summation model may explain some instances of conditioned modifications of responding to the US, it cannot serve as a general account of such effects. In particular, the summation model is not readily applicable to instances in which responses to the CS are qualitatively different from the responses that are elicited by the US. The fear-potentiated startle paradigm is a case in point. Here the conditioned response is a conditioned suppression or freezing response to a relatively long CS. In contrast, the startle US is a much shorter (e.g., 500 msec) stimulus that elicits a vigorous startle reaction. The summation model is also difficult to reconcile with instances of enhanced sexual and aggressive behavior. Many of the sexual and aggressive responses that occur in the presence of a conspecific sexual partner or intruder are not observed when a conditioned stimulus is presented because the CS typically does not provide supportive stimulation needed for various copulatory and combative action patterns. A simple summative model has also been brought into question by evidence that conditioned modifications of unconditioned behavior do not always correspond to conditioned responses elicited by a CS. In their studies of conditioned amphetamine sensitization, for example, Anagnostaras & Robinson (1996) observed increased locomotion as a CR when they administered saline in the amphetamine-paired context. However, the size and timing of this CR was not adequate to fully account for the sensitized amphetamine response that was observed in the same context. As an alternative to the summation model, Anagnostaras & Robinson (1996) proposed that conditioned stimuli modify unconditioned responding through an occasion-setting mechanism whereby the CS sets the occasion for the US (Holland 1992, Rescorla 1985, Schmajuk & Holland 1998). However, it is unclear how occasion setting might account for CS-induced modifications of responding to the US, since this mechanism deals with interactions between target and modulating CSs rather than interactions between conditioned and unconditioned stimuli. A more promising alternative to the summation model rests on the common observation that the fundamental outcome of Pavlovian conditioning is that the CS comes to activate a representation of the US (Rescorla 1988). This activation of the

18 Nov 2004 11:15

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

PAVLOVIAN CONDITIONING

P1: IKH

197

US representation prior to the occurrence of the US presumably allows for more effective interactions with the US once the US arrives. Hollis (1982) referred to this kind of process as “prefiguring.” Prefiguring or anticipatory activation of the US representation may alter responses to the US in different ways in different response systems. Possible mechanisms include reductions in the threshold for eliciting unconditioned behavior and/or changes in the perception of the US. Additional research is required to document how these mechanisms might operate.

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

SUMMARY AND CONCLUSION Pavlovian conditioning is typically described as a form of learning in which a neutral or arbitrary CS becomes associated with US, and as a consequence the CS comes to elicit a conditioned response. This description emphasizes that the CS is unrelated to the US at the outset of conditioning and that learning is best measured by the development of new responses to the CS. These features accurately characterize many laboratory studies of Pavlovian conditioning (especially those conducted in the associationist tradition) but fail to capture the critical features of Pavlovian conditioning from a functional perspective. A functional perspective assumes that Pavlovian conditioning is an adaptive trait that occurs under natural circumstances. This simple claim has two important implications, one relevant to the conditioned stimulus, and the other relevant to the conditioned response. For Pavlovian conditioning to occur in the ecological niche of an organism, CS-US pairings must be a feature of that environment. For that to be the case, the CS cannot be unrelated or arbitrary with respect to that US. Rather, there must be a pre-existing ecological relation between the CS and the US. Thus, a functional approach to Pavlovian conditioning rejects the common characterization that Pavlovian conditioning involves learning to associate a neutral or arbitrary CS with a US. In addition to focusing on how learning might occur in the natural environment, a functional approach directs us to focus on behavioral consequences of Pavlovian conditioning that are of adaptive significance. The common characterization of Pavlovian conditioning emphasizes how organisms learn new responses to the CS. However, starting with Pavlov himself, researchers have widely acknowledged that conditioned stimuli are not particularly important in their own right. Therefore, the adaptive significance of Pavlovian conditioning probably does not rest with how an organism’s interactions with the CS are improved by learning. From a functional perspective, the critical task for the organism is to cope with the unconditioned stimulus, which is of much greater biological import. Therefore, a functional perspective directs our attention to how an organism’s responses to the US are changed by Pavlovian conditioning. Evidence of learning with conditioned stimuli that are natural precursors of the US is available from studies of poison avoidance, food intake, sexual behavior,

18 Nov 2004 11:15

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

198

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

P1: IKH

DOMJAN

fear and defensive behavior, and maternal nursing and infant suckling. Although the details have not been documented in all of these learning situations, the studies have provided provocative evidence that learning with naturalistic stimuli proceeds differently from learning with arbitrary cues. In particular, learning with naturalistic stimuli is more rapid, more resistant to increases in the CS-US interval, and more resistant to the blocking effect. These findings suggest that the phenomena of Pavlovian conditioning may differ for naturalistic as contrasted with arbitrary conditioned stimuli. The mechanisms mediating these learning effects may not be distinctive, but considerable additional research is required to understand how conventional learning mechanisms may produce some of the special learning effects that have been documented with naturalistic CSs. Although conventional descriptions of Pavlovian conditioning emphasize the development of conditioned responses to the CS, numerous studies have shown that conditioning also alters how organisms react to, and interact with, the unconditioned stimulus. Evidence of conditioned modifications of responding to the US is available from studies of eyeblink conditioning; sexual, aggressive, and maternal behavior; fear conditioning; feeding and digestion; and drug conditioning. These conditioned modifications of responding to the US improve the efficacy of the organism’s interactions with the US and reduce the disruptive effects of encountering the US. If these changes in responding to the US are of adaptive significance, they should also be correlated with increased reproductive fitness. In most of the behavior systems examined, adaptive significance is inferred from parameters that are presumed to be related to reproductive fitness. Adaptive significance can be demonstrated more explicitly in the sexual behavior system, which permits direct measurement of reproductive fitness. Consistent with a functional perspective, recent studies have shown that Pavlovian conditioning enhances sperm output and various aspects of sexual behavior, and increases the number of offspring that result from a sexual interaction. Breland & Breland (1961) warned more than 40 years ago that common laboratory paradigms for the study of learning might not accurately reflect learning in natural ecosystems. Although this warning was voiced in relation to operant and instrumental conditioning, it was soon generalized to include Pavlovian and other forms of learning (Hinde & Stevenson-Hinde 1973, Seligman & Hager 1972). Initially the development of a vibrant functional approach to the study of Pavlovian conditioning was hampered by inadequate and difficult methods of discovery (Domjan & Galef 1983). The behavior systems approach developed by Timberlake and his associates (e.g., Timberlake 2001) represents a promising solution to these difficulties. The issues reviewed in the present chapter are theoretically agnostic and complementary to the behavior systems approach. They emphasize two important factors that are central to a functional approach to Pavlovian conditioning: (a) the use of conditioned stimuli that are natural precursors of a US, and (b) the measurement of changes in behavior directed toward the US rather than the CS as the primary behavioral manifestation of Pavlovian conditioning.

18 Nov 2004 11:15

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

PAVLOVIAN CONDITIONING

P1: IKH

199

ACKNOWLEDGMENT Preparation of this review was supported by grant MH39940 from the National Institute of Mental Health. The Annual Review of Psychology is online at http://psych.annualreviews.org

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

LITERATURE CITED Adkins-Regan E, MacKillop EA. 2003. Japanese quail (Coturnix japonica) inseminations are more likely to fertilize eggs in a context predicting mating opportunities. Proc. R. Soc. Lond. B Biol. Sci. 270:1685–89 Akins CK. 2000. Effects of species-specific cues and the CS-US interval on the topography of the sexually conditioned response. Learn. Motiv. 31:211–35 Ameli R, Ip C, Grillon C. 2001. Contextual fearpotentiated startle conditioning in humans: replication and extension. Psychophysiology 38:383–90 Anagnostaras SG, Robinson TE. 1996. Sensitization to the psychomotor stimulant effects of amphetamine: modulation by associative learning. Behav. Neurosci. 110:1397– 414 Anderson JR. 1995. Learning and Memory. New York: Wiley Andrews SE, Blumenthal TD, Flaten MA. 1998. Effects of caffeine and caffeineassociated stimuli on the human startle eyeblink reflex. Pharmacol. Biochem. Behav. 59: 39–44 Batsell WR Jr, Paschall GY, Gleason DI, Batson JD. 2001. Taste preconditioning augments odor-aversion learning. J. Exp. Psychol: Anim. Behav. Proc. 27:30–47 Batson JD, Batsell WR Jr. 2000. Augmentation, not blocking, in an A+/AX+ flavorconditioning procedure. Psychon. Bull. Rev. 7:466–71 Benedict JO, Ayres JJB. 1971. Factors affecting conditioning in the truly random control procedure in the rat. J. Comp. Physiol. Psych. 78:323–30 Blaisdell AP, Denniston JC, Miller RR. 1998. Temporal encoding as a determinant of over-

shadowing. J. Exp. Psychol.: Anim. Behav. Process. 24:72–83 Blass EM. 1990. Suckling: determinants, changes, mechanisms, and lasting impressions. Dev. Psychol. 26:520–33 Blass EM, Ganchrow JR, Steiner JE. 1984. Classical conditioning in newborn humans 2–48 hours of age. Infant Behav. Dev. 7:223– 35 Blustein JE, Ciccolone L, Bersh PJ. 1997. Evidence that adaptation to cold water swiminduced analgesia is a learned response. Physiol. Behav. 63:147–50 Boakes RA. 1984. From Darwin to Behaviorism. London: Cambridge Univ. Press Bousfield WA, Elliott MH. 1934. The effect of fasting on the eating behavior of rats. J. Genet. Psychol. 45:227–37 Bouton ME, Jones DL, McPhillips SA, Swartzentruber D. 1986. Potentiation and overshadowing in odor-aversion learning: role of method of odor presentation, the distalproximal distinction, and the conditionability of odor. Learn. Motiv. 17:115–38 Bower GH, Hilgard ER. 1981. Theories of Learning. Englewood Cliffs, NJ: Prentice Hall. 5th ed. Breland K, Breland M. 1961. The misbehavior of organisms. Am. Psychol. 16:681– 84 Brown JS, Kalish HI, Farber IE. 1951. Conditioned fear as revealed by magnitude of startle response to an auditory stimulus. J. Exp. Psychol. 41:317–28 Capaldi ED, Owens J, Palmer KA. 1994. Effects of food deprivation on learning and expression of flavor preferences conditioned by saccharin or sucrose. Anim. Learn. Behav. 22:173–80

18 Nov 2004 11:15

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

200

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

P1: IKH

DOMJAN

Cappell H, Roach C, Poulos CX. 1981. Pavlovian control of cross-tolerance between pentobarbital and ethanol. Psychopharmacology 74:54–57 Cepeda-Benito A, Short P. 1997. Morphine’s interoceptive stimuli as cues for the development of associative morphine tolerance in the rat. Psychobiology 25:236–40 Cepeda-Benito A, Reynoso J, Erath SA. 2000. Dose-response analyses of associative tolerance to nicotine analgesia in the rat: tail-flick and hot-plate tests. Exp. Clin. Psychopharmacol. 8:112–16 Cochin J, Kornetsky C. 1964. Development and loss of tolerance to morphine in the rat after single and multiple injections. J. Pharmacol. Exp. Ther. 145:1–10 Cook EW, Hodes RL, Lang PJ. 1986. Preparedness and phobia: effects of stimulus content on human visceral conditioning. J. Abnorm. Psychol. 95:195–207 Cook M, Mineka S. 1990. Selective associations in the observational conditioning of fear in rhesus monkeys. J. Exp. Psychol.: Anim. Behav. 16:372–89 Cusato B, Domjan M. 1998. Special efficacy of sexual conditioned stimuli that include species typical cues: tests with a CS preexposure design. Learn. Motiv. 29:152–67 Cusato B, Domjan M. 2001. Second-order sexual conditioning in male domesticated quail. Paper presented at meet. Southw. Psychol. Assoc., Houston, TX Dafters R, Bach L. 1985. Absence of environment-specificity in morphine tolerance acquired in nondistinctive environments: habituation or stimulus overshadowing? Psychopharmacology 87:101–6 Dafters R, Hetherington M, McCartney H. 1983. Blocking and sensory preconditioning effects in morphine analgesic tolerance: support for a Pavlovian conditioning model of drug tolerance. Q. J. Exp. Psychol. 35B:1– 11 Davis M. 1997. The neurophysiological basis of acoustic startle modulation: research on fear motivation and sensory gating. In Attention and Orienting: Sensory and Motiva-

tional Processes, ed. PJ Lang, RF Simons, MT Balaban, pp. 69–96. Mahwah, NJ: Erlbaum Denniston JC, Blaisdell AP, Miller RR. 1998. Temporal coding affects transfer of serial and simultaneous inhibitors. Anim. Learn. Behav. 26:336–50 Dickinson A. 1980. Contemporary Animal Learning Theory. London: Cambridge Univ. Press Domjan M. 1973. Role of ingestion in odortoxicosis learning in the rat. J. Comp. Physiol. Psychol. 84:507–21 Domjan M, Blesbois E, Williams J. 1998. The adaptive significance of sexual conditioning: Pavlovian control of sperm release. Psychol. Sci. 9:411–15 Domjan M, Cusato B, Krause M. 2004. Learning with arbitrary vs. ecological conditioned stimuli: evidence from sexual conditioning. Psychon. Bull. Rev. 11:232–46 Domjan M, Cusato B, Villarreal R. 2000. Pavlovian feed-forward mechanisms in the control of social behavior. Behav. Brain Sci. 23:235– 49 Domjan M, Galef BG Jr. 1983. Biological constraints on instrumental conditioning and classical conditioning: retrospect and prospect. Anim. Learn. Behav. 11:151–61 Domjan M, Hanlon MJ. 1982. Poison-avoidance learning to food-related tactile stimuli: avoidance of texture cues by rats. Anim. Learn. Behav. 10:293–300 Domjan M, Holloway KS. 1998. Sexual learning. In Comparative Psychology: A Handbook, ed. G Greenberg, MM Haraway, pp. 602–13. New York: Garland Domjan M, Lyons R, North NC, Bruell J. 1986. Sexual Pavlovian conditioned approach behavior in male Japanese quail (Coturnix coturnix japonica). J. Comp. Psychol. 100:413– 21 Domjan M, Wilson NE. 1972. Specificity of cue to consequence in aversion learning in the rat. Psychon. Sci. 26:143–45 Donegan NH, Wagner AR. 1987. Conditioned diminution and facilitation of the UR: a sometimes-opponent-process interpretation.

18 Nov 2004 11:15

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

PAVLOVIAN CONDITIONING In Classical Conditioning, ed. I Gormezano, WF Prokasy, RF Thompson, pp. 339–69. Hillsdale, NJ: Erlbaum. 3rd ed. Donny EC, Caggiula AR, Knopf S, Brown C. 1995. Nicotine self-administration in rats. Psychopharmacology 122:390–94 Drew KL, Glick SD. 1988. Characterization of the associative nature of sensitization to amphetamine-induced circling behavior and the environment dependent placebo-like response. Psychopharmacology 95:482–87 Dukas R, ed. 1998. Cognitive Ecology. Chicago: Univ. Chicago Press Dworkin BR. 1993. Learning and Physiological Regulation. Chicago: Univ. Chicago Press Eikelboom R, Stewart J. 1982. Conditioning of drug-induced physiological responses. Psychol. Rev. 89:507–28 Fanselow MS. 1998. Pavlovian conditioning, negative feedback, and blocking: mechanisms that regulate association formation. Neuron 20:625–27 Fanselow MS. 1981. Naloxone and Pavlovian fear conditioning. Learn. Motiv. 12:398–419 Fanselow MS. 1984. Shock-induced analgesia on the formalin test: effects of shock severity, naloxone, hypophysectomy, and associated variables. Behav. Neurosci. 98:79–95 Fanselow MS. 1991. Analgesia as a response to aversive Pavlovian conditioned stimuli: cognitive and emotional mediators. In Fear, Avoidance, and Phobias, ed. MR Denny, pp. 61–86. Hillsdale, NJ: Erbaum Fanselow MS, Baackes MP. 1982. Conditioned fear-induced opiate analgesia on the formalin test: evidence for two aversive motivational systems. Learn. Motiv. 13:200–21 Fanselow MS, German C. 1982. Explicitly unpaired delivery of morphine and the test situation: extinction and retardation of tolerance to the suppressing effects of morphine on locomotor activity. Behav. Neural Biol. 35:231–41 Fanselow MS, Helmstetter FJ. 1988. Conditional analgesia, defensive freezing, and benzodiazepines. Behav. Neurosci. 102:233–43 Fedorchak PM. 1997. The nature and strength

P1: IKH

201

of caloric conditioning. In Learning, Motivation, and Cognition, ed. ME Bouton, MS Fanselow, pp. 255–69. Washington, DC: Am. Psychol. Assoc. Feinberg G, Riccio DC. 1990. Changes in memory for stimulus attributes: implications for tests of morphine tolerance. Psychol. Sci. 1:265–67 Feldman JM. 1975. Blocking as a function of added cue intensity. Anim. Learn. Behav. 3: 98–102 Fendt M, Fanselow MS. 1999. The neuroanatomical and neurochemical basis of conditioned fear. Neurosci. Biobehav. Rev. 23:743–60 Fuchs A-R, Ayromlooi J, Rasmussen AB. 1987. Oxytocin response to conditioned and nonconditioned stimuli in lactating ewes. Biol. Reprod. 37:301–5 Gantt WH. 1966. Conditional or conditioned, reflex or response? Cond. Reflex 1:69– 74 Garcia J, Ervin ER, Koelling RA. 1966. Learning with prolonged delay of reinforcement. Psychon. Sci. 5:121–22 Garcia J, Hankins WG, Rusiniak KW. 1974. Behavioral regulation of the milieu interne in man and rat. Science 185:824–31 Garcia J, Koelling RA. 1966. Relation of cue to consequence in aversion learning. Psychon. Sci. 4:123–24 Goodison T, Siegel S. 1995. Learning and tolerance to the intake-suppressive effect of cholecystokinin in rats. Behav. Neurosci. 109:62– 70 Gormezano I, Kehoe EJ, Marshall BS. 1983. Twenty years of classical conditioning research with the rabbit. In Progress in Psychobiology and Physiological Psychology, ed. JM Prague, AN Epstein, Vol. 10, pp. 197– 275. New York: Academic Greeley J, Lˆe DA, Poulos CX, Cappell H. 1984. Alcohol is an effective cue in the conditional control of tolerance to alcohol. Psychopharmacology 83:159–62 Grillon C, Davis M. 1997. Fear-potentiated startle conditioning in humans: explicit and contextual cue conditioning following

18 Nov 2004 11:15

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

202

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

P1: IKH

DOMJAN

paired vs. unpaired training. Psychophysiology 34:451–58 Grisel JE, Wiertelak EP, Watkins LR, Meier SF. 1994. Route of morphine administration modulates conditioned analgesic tolerance and hyperalgesia. Pharmacol. Biochem. Behav. 49:1029–35 Grosvenor CE, Mena F. 1972. Evidence that the litter exerts an inhibitory influence on milk secretion in the rat during late lactation. Fed. Proc. (Fed. Am. Stud. Exp. Biol.) 31:275 Grosvenor CE, Mena F. 1974. Neural and hormonal control of milk secretion and milk ejection. In Lactation: A Comprehensive Treatise, Vol. 1. The Mammary Gland: Development and Maintenance, ed. BL Larson, VR Smith, pp. 227–76. New York: Academic Gunther LM, Miller RR, Matute H. 1997. CSs and USs: What’s the difference? J. Exp. Psychol.: Anim. Behav. Process. 23:15–30 Guti´errez G, Domjan M. 1996. Learning and male-male sexual competition in Japanese quail (Coturnix japonica). J. Comp. Psychol. 110:170–75 Guti´errez G, Domjan M. 1997. Differences in the sexual conditioned behavior of male and female Japanese quail (Coturnix japonica). J. Comp. Psychol. 111:135–42 Hall G, Mackintosh NJ, Goodall G, Dal Martello M. 1977. Loss of control by a less valid or a less salient stimulus compounded with a better predictor of reinforcement. Learn. Motiv. 8:145–58 Hijzen TH, Houtzager SWJ, Joordens RJE, Olivier B, Slangen JL. 1995. Predictive validity of the potentiated startle response as a behavioral model for anxiolytic drugs. Psychopharmacology 118:150–54 Hilliard S, Nguyen M, Domjan M. 1997. Onetrial appetitive conditioning in the sexual behavior system. Psychon. Bull. Rev. 4:237–41 Hinde RA, Stevenson-Hinde J, eds . 1973. Constraints on Learning. New York: Academic Hinson RE, Poulos CX. 1981. Sensitization to the behavioral effects of cocaine: modification by Pavlovian conditioning. Pharmacol. Biochem. Behav. 15:559–62 Hoffman HS, Ison JR. 1980. Reflex modifica-

tion in the domain of startle: I. Some empirical findings and their implications for how the nervous system processes sensory input. Psychol. Rev. 87:175–89 Holland PC. 1984. Origins of behavior in Pavlovian conditioning. In The Psychology of Learning and Motivation, ed. DL Medin, Vol. 18, pp. 129–74. San Diego: Academic Holland PC. 1992. Occasion setting in Pavlovian conditioning. In The Psychology of Learning and Motivation, ed. GH Bower, Vol. 28, pp. 69–125. Orlando, FL: Academic Hollis KL. 1982. Pavlovian conditioning of signal-centered action patterns and autonomic behavior: a biological analysis of function. Adv. Stud. Behav. 12:1–64 Hollis KL. 1984. The biological function of Pavlovian conditioning: the best defense is a good offense. J. Exp. Psychol.: Anim. Behav. Process. 10:413–25 Hollis KL. 1990. The role of Pavlovian conditioning in territorial aggression and reproduction. In Contemporary Issues in Comparative Psychology, ed. DA Dewsbury, pp. 197–219. Sunderland, MA: Sinauer Hollis KL. 1997. Contemporary research on Pavlovian conditioning: a “new” functional analysis. Am. Psychol. 52:956–65 Hollis KL, Cadieux EL, Colbert MM. 1989. The biological function of Pavlovian conditioning: a mechanism for mating success in the blue gourami (Trichogaster trichopterus). J. Comp. Psychol. 103:115–21 Hollis KL, Dumas MJ, Singh P, Fackelman P. 1995. Pavlovian conditioning of aggressive behavior in blue gourami fish (Trichogaster trichopterus): Winners become winners and losers stay losers. J. Comp. Psychol. 109:123–33 Hollis KL, Pharr VL, Dumas MJ, Britton GB, Field J. 1997. Classical conditioning provides paternity advantage for territorial male blue gouramis (Trichogaster trichopterus). J. Comp. Psychol. 111:219–25 Illich PA, Grau JW. 1991. Conditioned changes in pain reactivity: I. A discrete CS elicits hypoalgesia, not hyperalgesia, on the tail-flick test. Learn. Motiv. 22:421–38

18 Nov 2004 11:15

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

PAVLOVIAN CONDITIONING Jenkins JR, Rowland WJ. 1996. Pavlovian conditioning of agonistic behavior in male threespine stickleback (Gasterosteus aculeatus). J. Comp. Psychol. 110:396–401 Kamin LJ. 1965. Temporal and intensity characteristics of the conditioned stimulus. In Classical Conditioning, ed. WF Prokasy, pp. 118–47. New York: Appleton-CenturyCrofts Kamin LJ. 1969. Predictability, surprise, attention, and conditioning. In Punishment and Aversive Behavior, ed. BA Campbell, RM Church, pp. 279–96. New York: AppletonCentury-Crofts Kehoe EJ, Graham-Clarke P, Schreurs BG. 1989. Temporal patterns of the rabbit’s nictitating membrane response to compound and component stimuli under mixed CSUS intervals. Behav. Neurosci. 103:283– 95 Kim JA, Siegel S, Patenall VRA. 1999. Drugonset cues as signals: intra-administration associations and tolerance. J. Exp. Psychol.: Anim. Behav. Process. 25:491–504 Kimble GA, Ost JWP. 1961. A conditioned inhibitory process in eyelid conditioning. J. Exp. Psychol. 61:150–56 King DA, Bouton ME, Musty RE. 1987. Associative control of tolerance to the sedative effect of a short-acting benzodiazepine. Behav. Neurosci. 101:104–14 K¨oksal F, Domjan M, Weisman G. 1994. Blocking of the sexual conditioning of differentially effective conditioned stimulus objects. Anim. Learn. Behav. 22:103–11 Krank MD, Hinson RE, Siegel S. 1981. Conditional hyperalgesia is elicited by environmental signals of morphine. Behav. Neural Biol. 32:148–57 Krause MA, Cusato B, Domjan M. 2003. Extinction of conditioned sexual responses in male Japanese quail (Coturnix japonica): role of species typical cues. J. Comp. Psychol. 117:76–86 Larson SJ, Siegel S. 1998. Learning and tolerance to the ataxic effect of ethanol. Pharmacol. Biochem. Behav. 61:131–42 Lˆe AD, Poulos CX, Cappell H. 1979. Condi-

P1: IKH

203

tioned tolerance to the hypothermia effect of ethyl alcohol. Science 206:1109–10 LoLordo VM. 1979. Selective associations. In Mechanisms of Learning and Motivation, ed. A Dickinson, RA Boakes, 367–98. Hillsdale, NJ: Erlbaum LoLordo VM, Jacobs WJ, Foree DD. 1982. Failure to block control by a relevant stimulus. Anim. Learn. Behav. 10:183–92 Mahometa MJ, Domjan M. 2004. Classical conditioning increases reproductive success in Japanese quail (Coturnix japonica). Anim. Behav. In press Maier SF, Watkins LR. 1991. Conditioned and unconditioned stress-induced analgesia: stimulus preexposure and stimulus change. Anim. Learn. Behav. 19:295–304 Matzel LD, Hallam SC, Miller RR. 1988. Contribution of conditioned opioid analgesia to the shock-induced associative USpreexposure deficit. Anim. Learn. Behav. 16: 486–92 Matzel LD, Miller RR. 1987. Recruitment time of conditioned opioid analgesia. Physiol. Behav. 39:135–40 McNeilly AS, Robinson ICAF, Houston MJ, Howie PW. 1983. Release of oxytocin and prolactin in response to suckling. Br. Med. J. 286:257–59 Mehiel R. 1991. Hedonic-shift conditioning with calories. In The Hedonics of Taste, ed. RC Bolles, pp. 107–26. Hillsdale, NJ: Erlbaum Miller RR, Matute H. 1996. Biological significance in forward and backward blocking: resolution of a discrepancy between animal conditioning and human causal judgment. J. Exp. Psychol.: Gen. 25:370–86 Mongeluzi DL, Rosellini RA, Caldarone BJ, Stock HS, et al. 1996. Pavlovian aversive context conditioning using carbon dioxide as the unconditional stimulus. J. Exp. Psychol.: Anim. Behav. Process. 22:244–57 Oberling P, Bristol AS, Matute H, Miller RR. 2000. Biological significance attenuates overshadowing, relative validity, and degraded contingency effects. Anim. Learn. Behav. 28:172–86

18 Nov 2004 11:15

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

204

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

P1: IKH

DOMJAN

¨ Ohman A, Mineka S. 2001. Fear, phobias and preparedness: toward an evolved module of fear and fear learning. Psychol. Rev. 108:483–22 ¨ Ohman A, Mineka S. 2003. The malicious serpent: snakes as a prototypical stimulus for an evolved module of fear. Curr. Dir. Psychol. Sci. 12:5–9 ¨ Ohman A, Soares JJF. 1998. Emotional conditioning to masked stimuli: expectancies for aversive outcomes following nonrecognized fear-irrelevant stimuli. J. Exp. Psychol.: Gen. 127:69–82 Papini MR. 2002. Comparative Psychology. Upper Saddle River, NJ: Prentice Hall Pavlov IP. 1927. Conditioned Reflexes. Transl. GV Anrep. London: Oxford Univ. Press. Reprinted 1960, Dover Publ., New York Poole TB. 1985. The Social Behavior of Mammals. London: Chapman & Hall Post RM, Lockfeld A, Squillace KM, Contel NR. 1981. Drug-environment interaction: context dependency of cocaine-induced behavioral sensitization. Life Sci. 28:755–60 Poulos CX, Hinson RE. 1982. Pavlovian conditional tolerance to haloperidol catalepsy: evidence of dynamic adaptations in the dopaminergic system. Science 218:491–92 Poulos CX, Hunt T, Cappell H. 1988. Tolerance to morphine analgesia is reduced by the novel addition or omission of an alcohol cue. Psychopharmacology 94:412–16 Rescorla RA. 1985. Conditioned inhibition and facilitation. In Information Processing in Animals: Conditioned Inhibition, ed. RR Miller, NE Spear, pp. 299–326. Hillsdale, NJ: Erlbaum Rescorla RA. 1988. Pavlovian conditioning: It’s not what you think it is. Am. Psychol. 43:151– 60 Rescorla RA. 2000. Associative changes with a random CS-US relationship. Q. J. Exp. Psychol. 53B:325–40 Rescorla RA, Furrow DR. 1977. Stimulus similarity as a determinant of Pavlovian conditioning. J. Exp. Psychol.: Anim. Behav. Process. 3:203–15 Revusky SH. 1977. The concurrent interfer-

ence approach to delayed learning. In Learning Mechanisms in Food Selection, ed. LM Barker, MR Best, M Domjan, pp. 319–66. Waco, TX: Baylor Univ. Press Riba J, Rodriguez-Fornells A, Urbano G, Morte A, Antonijoan R, Barbanoj MJ. 2001. Differential effects of alprazolam on the baseline and fear-potentiated startle reflex in humans: a dose-response study. Psychopharmacology 157:358–67 Robinson TE, Berridge KC. 1993. The neural basis of drug craving: an incentivesensitization theory of addiction. Brain Res. Rev. 18:247–91 Robinson TE, Berridge KC. 2000. The psychology and neurobiology of addiction: an incentive-sensitization view. Addiction 95(Suppl. 2):S91–117 Ross RT. 1985. Blocking and unblocking of conditioned analgesia. Learn. Motiv. 16: 173–89 Ross RT. 1986. Pavlovian second-order conditioned analgesia. J. Exp. Psychol.: Anim. Behav. Process. 12:32–39 Rozin P, Kalat JW. 1971. Specific hungers and poison avoidance as adaptive specializations of learning. Psychol. Rev. 78:459– 86 Rozin P, Reff D, Mark M, Shull J. 1984. Conditioned responses in human tolerance to caffeine. Bull. Psychon. Soc. 22:117–20 Rusiniak KW, Palmerino CC, Garcia J. 1982a. Potentiation of odor by taste in rats: tests of some nonassociative factors. J. Comp. Physiol. Psychol. 96:775–80 Rusiniak KW, Palmerino CC, Rice AG, Forthman DL, Garcia J. 1982b. Flavor-illness aversions: potentiation of odor by taste with toxin but not shock in rats . J. Comp. Physiol. Psychol. 96:527–39 Rzoska J. 1953. Bait shyness, a study in rat behaviour. Br. J. Anim. Behav. 1:128–35 Savastano HI, Hua Y, Barnet RC, Miller RR. 1998. Temporal coding in Pavlovian conditioning: Hall-Pearce negative transfer. Q. J. Exp. Psychol. 51:139–53 Schmajuk NA, Holland PC, eds. 1998. Occasion Setting: Associative Learning and

18 Nov 2004 11:15

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

PAVLOVIAN CONDITIONING Cognition in Animals. Washington, DC: Am. Psychol. Assoc. Schull J. 1979. A conditioned opponent theory of Pavlovian conditioning and habituation. In The Psychology of Learning and Motivation, ed. GH Bower, 13:57–90. New York: Academic Schwartz CW, Schwartz ER. 1949. A Reconnaissance of the Game Birds in Hawaii, 90– 99. Hilo: Hawaii Bd. Commiss. Agric. Forest. Sclafani A. 1997. Learned controls of ingestive behaviour. Appetite 29:153–58 Seligman MEP, Hager JL, eds. 1972. Biological Boundaries of Learning. New York: Appleton-Century-Crofts Shettleworth SJ. 1983. Function and mechanism in learning. In Advances in Analysis of Behaviour, ed. MD Zeiler, P Harzem, 3:1– 39. Chichester, UK: Wiley-Intersci. Shettleworth SJ. 1994. Biological approaches to the study of learning. In Animal Learning and Cognition, ed. NJ Mackintosh, pp. 185–219. San Diego: Academic Shettleworth SJ. 1998. Cognition, Evolution, and Behavior. New York: Oxford Univ. Press Siegel S. 1989. Pharmacological conditioning and drug effects. In Psychoactive Drugs: Tolerance and Sensitization, ed. AJ Goudie, MW Emmett-Oglesby, pp. 115–80. Clifton, NJ: Humana Siegel S. 1991. Feedforward processes in drug tolerance. In Perspectives in Cognitive Neuroscience, ed. RG Lister, HJ Weingartner, pp. 405–16. New York: Oxford Univ. Press Siegel S, Baptista MAS, Kim JA, McDonald RV, Weise-Kelly L. 2000. Pavlovian psychopharmacology: the associative basis of tolerance. Exp. Clin. Psychopharmacol. 8:276–93 Siegel S, Ramos BMC. 2002. Applying laboratory research: drug anticipation and the treatment of drug addiction. Exp. Clin. Psychopharmacol. 10:162–83 Siegel S, Sherman JE, Mitchell D. 1980. Extinction of morphine analgesic tolerance. Learn. Motiv. 11:289–301 Smith JC, Roll DL. 1967. Trace conditioning

P1: IKH

205

with X-rays as an aversive stimulus. Psychon. Sci. 9:11–12 Staddon JER. 1983. Adaptive Behavior and Learning. London: Cambridge Univ. Press Staddon JER. 1988. Learning as inference. In Evolution and Learning, ed. RC Bolles, MD Beecher, pp. 59–77. Hillsdale, NJ: Erlbaum Stewart J. 1992. Conditioned stimulus control of the expression of sensitization of the behavioral activating effects of opiate and stimulant drugs. In Learning and Memory: The Behavioral and Biological Substrates, ed. I Gormezano, EA Wasserman, pp. 129–51. Hillsdale, NJ: Erlbaum Tancin V, Kraetzl W-D, Schams D, Bruckmaier RM. 2001. The effect of conditioning to suckling, milking and of calf presence on the release of oxytocin in dairy cows. Appl. Anim. Behav. Sci. 72:235–46 Terelli E, Terry P. 1999. Amphetamine-induced conditioned activity and sensitization: the role of habituation to the test context and the involvement of Pavlovian processes. Behav. Pharmacol. 9:409–19 Testa TJ. 1975. Effects of similarity of location and temporal intensity pattern of conditioned and unconditioned stimuli on the acquisition of conditioned suppression in rats. J. Exp. Psychol.: Anim. Behav. Process. 1:114–21 Thompson T, Sturm T. 1965. Classical conditioning of aggressive display in Siamese fighting fish. J. Exp. Anal. Behav. 8:397–403 Timberlake W. 2001. Motivational modes in behavior systems. In Handbook of Contemporary Learning Theories, ed. RR Mowrer, SB Klein, pp. 155–209. Mahwah, NJ: Erlbaum Toades DP. 1997. From the machine to the ghost within: Pavlov’s transition from digestive physiology to conditioned reflexes. Am. Psychol. 52:947–55 Turkkan JS. 1989. Classical conditioning: the new hegemony. Behav. Brain Sci. 12:121– 79 Walker DL, Davis M. 2002. Quantifying fear potentiated startle using absolute versus proportional increase scoring methods: implications for neurocircuitry of fear and anxiety. Psychopharmacology 164:318–28

18 Nov 2004 11:15

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

206

AR

AR231-PS56-07.tex

AR231-PS56-07.sgm

LaTeX2e(2002/01/18)

P1: IKH

DOMJAN

Westbrook RF, Greeley JD, Nabke CP, Swinbourne AL. 1991. Aversive conditioning in the rat: effects of benzodiazepine and of an opioid agonist and antagonist on conditioned hypoalgesia and fear. J. Exp. Psychol.: Anim. Behav. Proc. 17:219–30 Wiertelak EP, Watkins LR, Maier SF. 1992. Conditioned inhibition of analgesia. Anim. Learn. Behav. 20:339–49 Winslow JT, Parr LA, Davis M. 2002. Acoustic startle, prepulse inhibition, and fearpotentiated startle measured in Rhesus monkeys. Biol. Psychiatry 51:859–66 Woods SC. 1991. The eating paradox: how we tolerate food. Psychol. Rev. 98:488–505 Woods SC, Kulkosky PJ. 1976. Classically conditioned changes of blood glucose level. Psychosom. Med. 38:201–19 Woods SC, Ramsay DS. 2000. Pavlovian influences over food and drug intake. Behav. Brain Res. 110:175–82 Woods SC, Seeley RJ. 2002. Hunger and energy homeostasis. In Stevens’ Handbook of

Experimental Psychology, ed. H Pashler, R Gallistel, 3:633–68. New York: Wiley Woods SC, Strubbe JH. 1994. The psychobiology of meals. Psychon. Bull. Rev. 1:141– 55 Woods SC, Vasselli JR, Kaestner E, Szakmary GA, Milburn P, Vitiello MV. 1977. Conditioned insulin secretion and meal-feeding in rats. J. Comp. Physiol. Psychol. 91:128–33 Young AM, Goudie AJ. 1994. Adaptive processes regulating tolerance to behavioral effects of drugs. In Psychopharmacology: The Fourth Generation of Progress, ed. FE Bloom, DJ Kupfer, pp. 657–811. New York: Raven Zamble E. 1973. Augmentation of eating following a signal for feeding in rats. Learn. Motiv. 4:138–47 Zamble E, Hadad GM, Mitchell JB, Cutmore TR. 1985. Pavlovian conditioning of sexual arousal: first- and second-order effects. J. Exp. Psychol.: Anim. Behav. Process. 11:598–610

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

Annual Review of Psychology Volume 56, 2005

CONTENTS

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Frontispiece—Richard F. Thompson

xviii

PREFATORY In Search of Memory Traces, Richard F. Thompson

1

DECISION MAKING Indeterminacy in Brain and Behavior, Paul W. Glimcher

25

BRAIN IMAGING/COGNITIVE NEUROSCIENCE Models of Brain Function in Neuroimaging, Karl J. Friston

57

MUSIC PERCEPTION Brain Organization for Music Processing, Isabelle Peretz and Robert J. Zatorre

89

SOMESTHETIC AND VESTIBULAR SENSES Vestibular, Proprioceptive, and Haptic Contributions to Spatial Orientation, James R. Lackner and Paul DiZio

115

CONCEPTS AND CATEGORIES Human Category Learning, F. Gregory Ashby and W. Todd Maddox

149

ANIMAL LEARNING AND BEHAVIOR: CLASSICAL Pavlovian Conditioning: A Functional Perspective, Michael Domjan

179

NEUROSCIENCE OF LEARNING The Neuroscience of Mammalian Associative Learning, Michael S. Fanselow and Andrew M. Poulos

207

HUMAN DEVELOPMENT: EMOTIONAL, SOCIAL, AND PERSONALITY Behavioral Inhibition: Linking Biology and Behavior Within a Developmental Framework, Nathan A. Fox, Heather A. Henderson, Peter J. Marshall, Kate E. Nichols, and Melissa A. Ghera

235

BIOLOGICAL AND GENETIC PROCESSES IN DEVELOPMENT Human Development: Biological and Genetic Processes, Irving I. Gottesman and Daniel R. Hanson

263 vii

P1: JRX

December 8, 2004

viii

12:13

Annual Reviews

AR231-FM

CONTENTS

SPECIAL TOPICS IN PSYCHOPATHOLOGY The Psychology and Neurobiology of Suicidal Behavior, Thomas E. Joiner Jr., Jessica S. Brown, and LaRicka R. Wingate

287

DISORDERS OF CHILDHOOD Autism in Infancy and Early Childhood, Fred Volkmar, Kasia Chawarska, and Ami Klin

315

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

CHILD/FAMILY THERAPY Youth Psychotherapy Outcome Research: A Review and Critique of the Evidence Base, John R. Weisz, Amanda Jensen Doss, and Kristin M. Hawley

337

ALTRUISM AND AGGRESSION Prosocial Behavior: Multilevel Perspectives, Louis A. Penner, John F. Dovidio, Jane A. Piliavin, and David A. Schroeder

365

INTERGROUP RELATIONS, STIGMA, STEREOTYPING, PREJUDICE, DISCRIMINATION The Social Psychology of Stigma, Brenda Major and Laurie T. O’Brien

393

PERSONALITY PROCESSES Personality Architecture: Within-Person Structures and Processes, Daniel Cervone

423

PERSONALITY DEVELOPMENT: STABILITY AND CHANGE Personality Development: Stability and Change, Avshalom Caspi, Brent W. Roberts, and Rebecca L. Shiner

453

WORK MOTIVATION Work Motivation Theory and Research at the Dawn of the Twenty-First Century, Gary P. Latham and Craig C. Pinder

485

GROUPS AND TEAMS Teams in Organizations: From Input-Process-Output Models to IMOI Models, Daniel R. Ilgen, John R. Hollenbeck, Michael Johnson, and Dustin Jundt

517

LEADERSHIP Presidential Leadership, George R. Goethals

545

PERSONNEL EVALUATION AND COMPENSATION Personnel Psychology: Performance Evaluation and Pay for Performance, Sara L. Rynes, Barry Gerhart, and Laura Parks

571

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

CONTENTS

ix

PSYCHOPHYSIOLOGICAL DISORDERS AND PSYCHOLOGICAL EFFECTS ON MEDICAL DISORDERS Psychological Approaches to Understanding and Treating Disease-Related Pain, Francis J. Keefe, Amy P. Abernethy, and Lisa C. Campbell

601

TIMELY TOPIC

Annu. Rev. Psychol. 2005.56:179-206. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Psychological Evidence at the Dawn of the Law’s Scientific Age, David L. Faigman and John Monahan

631

INDEXES Subject Index Cumulative Index of Contributing Authors, Volumes 46–56 Cumulative Index of Chapter Titles, Volumes 46–56

ERRATA An online log of corrections to Annual Review of Psychology chapters may be found at http://psych.annualreviews.org/errata.shtml

661 695 700

23 Nov 2004 18:57

AR

AR231-PS56-08.tex

AR231-PS56-08.sgm

P1: IKH LaTeX2e(2002/01/18) 10.1146/annurev.psych.56.091103.070213

Annu. Rev. Psychol. 2005. 56:207–34 doi: 10.1146/annurev.psych.56.091103.070213 c 2005 by Annual Reviews. All rights reserved Copyright
First published online as a Review in Advance on August 30, 2004

THE NEUROSCIENCE OF MAMMALIAN ASSOCIATIVE LEARNING Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Michael S. Fanselow Department of Psychology and Brain Research Institute, University of California, Los Angeles, Los Angeles, California 90095-1563; email: [email protected]

Andrew M. Poulos Neuroscience Program, University of Southern California, Los Angeles, California 90089-2520; email: [email protected]

Key Words Pavlovian conditioning, amygdala, hippocampus, cerebellum, fear, eyeblink ■ Abstract Mammalian associative learning is organized into separate anatomically defined functional systems. We illustrate the organization of two of these systems, Pavlovian fear conditioning and Pavlovian eyeblink conditioning, by describing studies using mutant mice, brain stimulation and recording, brain lesions and direct pharmacological manipulations of specific brain regions. The amygdala serves as the neuroanatomical hub of the former, whereas the cerebellum is the hub of the latter. Pathways that carry information about signals for biologically important events arrive at these hubs by circuitry that depends on stimulus modality and complexity. Within the amygdala and cerebellum, neural plasticity occurs because of convergence of these stimuli and the biologically important information they predict. This neural plasticity is the physical basis of associative memory formation, and although the intracellular mechanisms of plasticity within these structures share some similarities, they differ significantly. The last Annual Review of Psychology article to specifically tackle the question of mammalian associative learning (Lavond et al. 1993) persuasively argued that identifiable “essential” circuits encode memories formed during associative learning. The next dozen years saw breathtaking progress not only in detailing those essential circuits but also in identifying the essential processes occurring at the synapses (e.g., Bi & Poo 2001, Martinez & Derrick 1996) and within the neurons (e.g., Malinow & Malenka 2002, Murthy & De Camilli 2003) that make up those circuits. In this chapter, we describe the orientation that the neuroscience of learning has taken and review some of the progress made within that orientation.

CONTENTS CORE QUESTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208 TECHNIQUES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209 MODEL SYSTEMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209 0066-4308/05/0203-0207$14.00

207

23 Nov 2004 18:57

208

AR

AR231-PS56-08.tex

FANSELOW




AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

P1: IKH

POULOS

Fear Conditioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209 Eyeblink Conditioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218 CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

CORE QUESTIONS Three core questions dominate the neuroscience of associative learning. At the systems level we ask, what are the brain circuits that mediate between environmental stimuli and acquired behavior? Then, within those circuits, what synapses must undergo modification for learning to occur? The third question is decidedly more molecular: What intracellular events occur at those sites of critical plasticity to confer the necessary changes in the synaptic efficacy that underlies the formation of a new memory? The preponderance of research on these issues has used Pavlovian conditioning preparations, where dependent relationships between two stimuli result in altered reactions to those stimuli. The behavioral change of interest is to a stimulus that provokes new behavior because receiving a biologically significant stimulus is conditional upon its presence. The stimulus that provokes new behavior is the conditional stimulus (CS) because the reaction to it is conditional upon experiencing a relationship with the biologically significant stimulus. Because behavior is evoked by the biologically relevant stimulus independent of a conditional relationship with another stimulus, it is called an unconditional stimulus (US). A fundamental truth, recognized by neuroscientists of learning for some time, is that the US and the type of reaction it causes determine what neural circuits and what sites of plasticity mediate particular changes in behavior. One implication of this is that there is not a single mechanism for associative learning. Rather, there are many separate systems of associative learning, and although these systems may share some characteristics, they are just as likely to be different from one another. It also means that for analytic purposes it is critical to focus on specific model systems to understand the neural basis of learning. We have chosen two model systems, Pavlovian conditioning of the eyeblink response and Pavlovian conditional fear, as the model systems for the subject of this review. In eyeblink conditioning, a relatively neutral stimulus such as a tone becomes a CS because it is paired with a US such as an air puff or mild shock delivered near the eye to produce an eyeblink. In fear conditioning, a tone CS also may be used, but it is paired with a more threatening US such as foot shock. These are the appropriate models to focus on for several reasons. For both of these systems, staggering progress has been made on each of the three questions raised above. Additionally, the two employ almost completely nonoverlapping circuits. Finally, the behavioral phenomenology of these systems is very different. Fear conditioning is rapidly learned and fear responses are relatively diffuse in timing and topography. Eyeblink is slowly learned, but is characterized by precise timing of a very exact behavioral topography. Additionally, in eyeblink conditioning, the conditional response (CR) to the CS is a near-replica of the unconditional response (UR) to the US; both CR and UR are an eyeblink. In fear conditioning, the CR has

23 Nov 2004 18:57

AR

AR231-PS56-08.tex

AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

NEUROSCIENCE OF ASSOCIATIVE LEARNING

P1: IKH

209

little topographical relationship to the UR, except that it is defensive. For example, a rat or a mouse will show a vigorous burst of activity in response to the shock, but the CR used in most neuroscientific studies of fear is a complete suppression of activity called freezing (Fanselow 1980).

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

TECHNIQUES The use of genetically engineered mice has taken on an increased role in the arsenal of techniques used to understand the neural basis of learning. These molecular genetic techniques are unique in their ability to target specific proteins and eliminate, overexpress, or alter their composition. These tools have become increasingly sophisticated, such that tissue selectivity and temporal control of the mutations are becoming available. However, traditional techniques such as lesions, pharmacological manipulations, and electrophysiological stimulation and recording remain important contributors to our understanding. It needs to be recognized that there is no one superior technique; rather, the greatest analytical power derives from combining various approaches. We try to represent that in this review.

MODEL SYSTEMS Fear Conditioning The neural hub for fear conditioning is the amygdala, which is composed of several nuclei in the anterior portion of the medial temporal lobe. Sensory stimuli corresponding to potential CSs arrive at what Swanson & Petrovich (1998) refer to as the frontotemporal amygdala (FTA) because it interconnects frontal and temporal cortices. This region corresponds to the almond-shaped region for which the amygdala originally derived its name, and often is referred to as the basolateral complex because it contains the lateral and basal nuclei. CS information from the thalamus, hippocampus, and several cortical regions reaches the FTA via glutamatergic projections (i.e., neurons releasing the excitatory neurotransmitter glutamate; see Figure 1). The most straightforward and best-characterized CS pathway is a direct projection from the medial geniculate nucleus of the thalamus to the dorsal portion of the lateral nucleus of the amygdala (LeDoux et al. 1991). This pathway carries auditory information to the amygdala and is sufficient for delay conditioning to simple auditory CSs, where tone onset occurs briefly before US onset and CS termination is coincident with either US onset or termination (Romanski & LeDoux 1992). When the CS requires greater processing, polysynaptic routes to the FTA become necessary and the amygdala receives CS information from cortex. For example, the apparatus or context cues present at the time of shock reach the basolateral amygdala via the ventral angular bundle after processing by the hippocampus and entorhinal cortex (Anagnostaras et al. 1999, Maren & Fanselow 1995) and also reach the lateral amygdala from the perirhinal and postrhinal cortex (Amaral et al. 1992). Although damage to the

NEURAL CIRCUITRY OF FEAR CONDITIONING

210

AR AR231-PS56-08.tex

FANSELOW


AR231-PS56-08.sgm LaTeX2e(2002/01/18)

Figure 1 The basic circuit for contextual fear conditioning organized according to various types of conditional stimuli at the top and conditional fear responses at the bottom. LA refers to the lateral nucleus of the amygdala and BLA refers to the basolateral nucleus of the amygdala. The BLA and the LA are parts of the frontotemporal amygdala and comprise the neuroanatomical hub for learned fear.

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

23 Nov 2004 18:57 P1: IKH

POULOS

23 Nov 2004 18:57

AR

AR231-PS56-08.tex

AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

NEUROSCIENCE OF ASSOCIATIVE LEARNING

P1: IKH

211

ventral angular bundle, perirhinal cortex, and postrhinal cortex produces deficits in context conditioning, it has little effect on delay conditioning to a simple auditory cue (Bucci et al. 2000, Burwell et al. 2004, Maren & Fanselow 1995). Interestingly, when the auditory stimulus is similar to the ultrasonic distress call of a rat, perirhinal lesions will attenuate conditioning (Lindquist et al. 2004), again indicating that stimulus complexity determines whether the monosynaptic projections from the thalamus will be sufficient to inform the FTA about the presence of the CS. One can increase the complexity of the fear conditioning procedure to an auditory cue simply by having CS termination occur briefly before US onset. Pavlov called this procedure trace conditioning because he believed a trace of the CS must remain in the nervous systems to bridge the gap between the CS and US. Like context conditioning, trace fear conditioning requires the hippocampus (McEchron et al. 1998, Quinn et al. 2002). However, lesions of the anterior cingulate cortex reduce trace conditioning but do not affect context conditioning (Han et al. 2003). Additionally, using gene expression as a marker for cellular activity, trace conditioning caused greater activation of the anterior cingulate cortex than delay conditioning with the same stimuli. Thus, the anterior cingulate may play the bridging role to which Pavlov referred. Figure 1 groups various types of CS information with the regions that contribute to processing this information. Each of these regions contains monosynaptic projections to the FTA (Amaral et al. 1992). Information about the reinforcing US arrives at the amygdala from a number of routes, all of which appear to be sufficient, but none appear to be necessary. Pain information arrives at the FTA directly from the posterior thalamus as well as via the insular cortex (Brunzell & Kim 2001, Jasmin et al. 2004, Lanuza et al. 2004, Shi & Davis 1999). However, although combined lesions of both of these forebrain routes attenuate conditioning to a tone, context conditioning proceeds normally (Brunzell & Kim 2001). Pain information from subcortical structures such as the parabrachial nucleus, nucleus of the solitary tract, and even the dorsal horn of the spine reach the central nucleus of the amygdala (Benarroch 2001, Burstein & Potrebic 1993, Gauriau & Bernard 2002). However, there do not seem to be any major monosynaptic projections from the central nucleus to the FTA, at least in the rat (DeOlmos et al. 1985). The primary output of the FTA for the generation of conditional fear responses is the central nucleus of the amygdala, which projects to a wide range of regions responsible for emotional responses (see Figure 1). Autonomic reactions are triggered by direct projections to regions of the brainstem and hypothalamus responsible for controlling autonomic function (Kapp et al. 1979, LeDoux et al. 1988). Central nucleus projections to the brainstem modulate several simple reflexes, including the auditory startle response (Canli & Brown 1996, Rosen & Davis 1990, Rosen et al. 1991). Projections to the hypothalamus lead to an elevation of activity in the pituitary-adrenal axis (Feldman & Weidenfeld 1998). Defensive responses are triggered by projections to the periaqueductal gray, which also trigger an endogenous opioid-mediated analgesic state (Fanselow 1991). This analgesia is of particular importance because it provides a negative feedback regulation of the

23 Nov 2004 18:57

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

212

AR

AR231-PS56-08.tex

FANSELOW




AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

P1: IKH

POULOS

ascending painful information that supports fear conditioning (Fanselow 1998). This regulation serves to keep the level of fear appropriate to the level of threat and directs fear toward the best predictors of aversive outcomes. Not all conditional fear–relevant behaviors are controlled via the central nucleus because termination of a fear CS still serves as a reinforcer in rats with central nucleus lesions (Amorapanth et al. 2000, Killcross et al. 1997). It is possible that fear modulates instrumental behavior via projections from the basolateral nucleus to the ventral striatum (French & Totterdell 2003). Some anxiety-related responses are mediated by projections from the FTA to the bed nucleus of the stria terminalis, but it is not clear if these projections play a role in normal fear conditioning (Walker et al. 2003). Fear not only produces behavior, it affects cognitive activity as well. The FTA has extensive projections to the frontal and temporal lobes; for example, all the cortical-FTA projections in Figure 1 are reciprocal (Amaral et al. 1992). The influence of the central nucleus is certainly not entirely descending, either. It can generate cortical arousal through its control of the ascending cholinergic projections from the basal forebrain (Jolkkonen et al. 2002). Together, this array of ascending and descending projections generates the diffuse set of behavioral manifestations referred to as a fear state. An assumption of neuroscientific theories of associative learning is that convergence of CS and US information onto particular cells leads to changes in synaptic strength at the synapses mediating the CS input to those commonly activated cells. This synaptic plasticity is the mechanism that underlies association formation. Learning is the induction of these synaptic changes and the presence of memory depends on the stability of these changes. Long-term potentiation (LTP) induced by activation of the N-methyl-D-aspartate (NMDA) type of glutamate receptor is the most heavily emphasized form of plasticity for fear conditioning. This long-lasting form of synaptic strengthening was first discovered in the hippocampus and can be induced when glutamate activity at initially “weak” synapses is paired with stimulation that causes the cell to spike (Bliss & Gardner-Medwin 1973, Bliss & Lomo 1973). Such a dual pattern is necessary for NMDA receptor activation and the ensuing calcium influx at the ion channel of this receptor leads to strengthening at those NMDA receptor-containing synapses (Collingridge et al. 1983, Malinow & Miller 1986). This is an attractive model for Pavlovian conditioning because a CS-generated glutamatergic input that at first weakly activates a synapse will be potentiated if the US causes the cell to fire within a temporally limited window. Thus, the cells that participate in this plasticity must receive both CS and US inputs. It is clear that individual cells within the lateral nucleus respond to tones that can serve as an auditory CS and shocks that can serve as a US (Romanski et al. 1993). Furthermore, electrical stimulation of auditory input to the FTA (medial geniculate to lateral nucleus) supports long-term plasticity (Clugnet & LeDoux 1990). Indeed, LTP induction in this pathway produced by electrical stimulation

SITES OF CRITICAL PLASTICITY

23 Nov 2004 18:57

AR

AR231-PS56-08.tex

AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

NEUROSCIENCE OF ASSOCIATIVE LEARNING

P1: IKH

213

increases the FTA’s response to a tone (Rogan et al. 1997). Following fear conditioning, cells within the amygdala show increased firing to the CS, suggesting that the CS input has been potentiated following conditioning (Quirk et al. 1997). Finally, McKernan & Shinnick-Galagher (1997) compared brain slices containing the auditory pathway from the auditory thalamus to the lateral nucleus taken from fear-conditioned and control animals. Stimulation of this projection caused greater activation of the amygdala in trained animals, suggesting that fear conditioning induces potentiation of this pathway. Thus, there is a very convincing case that auditory fear conditioning normally supports long-term potentiation of the pathway carrying simple tone information from the medial geniculate nucleus to the lateral amygdala. There is also good evidence that the pathways described above carrying information from the cortex to the FTA and from the hippocampus to the FTA are capable of long-term potentiation (Chapman et al. 1990, Maren & Fanselow 1995, Yaniv et al. 2000). Because administration of NMDA antagonists to the amygdala prevents the acquisition of fear conditioning, it is highly likely that LTP in the FTA supports changes necessary for fear conditioning (Fanselow & Kim 1994, Miserendino et al. 1990). If LTP is a mechanism of memory formation and LTP in the FTA is normally critical for fear conditioning, then the FTA is presumably encoding some important aspect of the fear-conditioning experience. This must be a general aspect of the experience because the FTA is equally important for fear of simple and complex conditional stimuli (Gale et al. 2004). Because the level of fear is highly dependent on the emotional significance of the US, the memory for emotional significance is one possible candidate for the information that the amygdala has encoded. One method that has proven successful in probing the content of associative memory, particularly for the memory of a US, is the reinforcer revaluation technique. For example, Rescorla (1974) gave rats pairings of tone and weak shock such that the tone alone would provoke a low level of fear. In one group of rats, Rescorla changed how the rats valued the memory of shock by giving them a series of strong shocks without the tone. These rats then behaved as if the tone had been paired with a strong shock. Since the tone-shock association should not be strengthened because of receiving shock by itself, the changes in behavior must be related to changes in the memory for shock. Variants of this procedure have been used to probe the content of associations in a range of situations (Dayan & Balleine 2002). If the FTA is encoding the emotional significance of the US, amygdala activity should be important during memory revaluation, and this appears to be the case. If the amygdala is inactivated only during the memory-inflating strong shocks, rats respond to the tone at levels appropriate to the weak shock with which it was paired while amygdala activity was normal and not at the inflated level (Fanselow & Gale 2003). That is, inactivation of the amygdala during memory revaluation prevents revaluation. FTA lesions made after training eliminate conditional responding even if the memory is allowed to consolidate for a period that approaches the adult life span of the rat (17 months), which suggests that the amygdala permanently encodes the emotional significance of the shock reinforcer (Gale et al. 2004).

23 Nov 2004 18:57

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

214

AR

AR231-PS56-08.tex

FANSELOW




AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

P1: IKH

POULOS

Of course, for conditioning you must have not only a representation of emotional significance but also a representation of the stimuli that are associated with that emotional significance. During the course of conditioning to a tone, changes in the medial geniculate cells code the auditory signal (Edeline & Weinberger 1992). Individual cells respond differentially to tones of different frequencies, and during conditioning, many of these cells retune so that the frequency specificity shifts in the direction of the tone paired with shock. These changes in the representation of the tone appear to depend on the amygdala because inactivation of the amygdala prevents the plasticity in the medial geniculate (Maren et al. 2001, Poremba & Gabriel 2001). Individual cells in the auditory cortex also show shifts in their preferred tuning toward the frequency of reinforced CS (Edeline et al. 1993). NMDA-dependent activity plays a role in the learning about the contextual CS associated with shock as NMDA antagonists injected into the hippocampus or genetic deletion of NMDA receptors from the CA1 region of the hippocampus interfere with contextual fear conditioning (Shimizu et al. 2000, Young et al. 1994). Indeed, genetic manipulations that enhance NMDA receptor function can enhance contextual fear learning (Tang et al. 1999). This learning seems to be specifically about contexts because blocking NMDA antagonists only during a time where the animal is exploring a context eliminates the benefits of that exploration (Stote & Fanselow 2004). Inactivating the amygdala at the time of exploration also prevents the benefits of exploration (Huff & Rudy 2004), similar to the effect of amygdala inactivation on retuning of auditory-responsive neurons in the thalamus. This is somewhat surprising, given that there is no emotion-provoking shock during the exploration period; however, this may reflect emotional arousal due to novelty or handling by the experimenter. During fear conditioning, theta rhythm activity generated by a tone paired with shock synchronizes in hippocampus and amygdala (Seidenbecher et al. 2003). Additionally, hippocampal neurons will differentially respond to tones that were paired, as opposed to not paired, with shock, but only in their place fields (Moita et al. 2003). Thus, it is clear that fear conditioning represents a rich interplay between the structures that encode the emotional, signaling, and contextual aspects of the learning. Questions about plasticity for Pavlovian conditioning typically focus on those related to sensory inputs as opposed to response outputs. This is no doubt because Pavlovian conditioning is thought of as learning about stimulus relationships, which results in preprogrammed responses to new stimuli. However, there is no a priori reason to assume that synaptic strengthening cannot play a role at the other relays in the circuit mediating between environment and behavior. A suggestion that there is also significant plasticity within the response-generation pathways comes from analyses of development of fear systems. The ability to condition fear to different CS appears at different ages, as does the ability to express fear in different responses (Hunt & Campbell 1997). Fear-induced freezing occurs at an earlier age than fear-induced potentiation of the startle reflex (Hunt 1999). Fan & Richardson (2002) used odor-shock pairings to train rats at an age when they were old enough to express fear in terms of odor avoidance, but too young to express fear

23 Nov 2004 18:57

AR

AR231-PS56-08.tex

AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

NEUROSCIENCE OF ASSOCIATIVE LEARNING

P1: IKH

215

in terms of potentiation of the startle response. The intriguing result is that if the rats were not tested until they were old enough to express potentiated startle, they showed avoidance but not enhanced startle. The finding that animals only express behaviors available at the time of training suggests that some component of the plasticity occurs within the response-generation machinery. Thus, it seems that in fear conditioning, important plasticity occurs in the pathways corresponding to CS, US, and CR. All of this plasticity must be rapidly induced because significant fear conditioning occurs with a single pairing of CS and US. If CS and US convergence in the FTA causes potentiation of the glutamatergic synapses activated by the CS, changes must be occurring within the presynaptic neuron, the postsynaptic neuron, or both. The majority of what we know about the intracellular events that support synaptic strengthening comes from studies of long-term potentiation in rodent hippocampal neurons or from studies of gill and siphon withdrawal reflexes of the marine mollusk aplysia (Pittenger & Kandel 2003). There has been considerable debate within each of these two domains as to whether presynaptic or postsynaptic changes are of primary importance, although the hippocampal work has tended to emphasize postsynaptic processes and the aplysia work has emphasized presynaptic processes (Antonov et al. 2003, Lisman 2003, Roberts & Glanzman 2003). Presynaptic changes could take the form of greater neurotransmitter release per action potential arriving at the relevant synaptic terminals. Postsynaptic changes typically take the form of changes that make the postsynaptic cell more responsive to a fixed amount of neurotransmitter release. This could happen through restructuring the postsynaptic region to be more responsive to glutamate, such as the insertion of more of the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)type glutamate receptors that mediate the majority of excitatory glutamatergic transmission (Isaac et al. 1995, Liao et al. 1995). Obviously, such changes are not logically exclusive. Rather, coordinated changes in both pre- and postsynaptic regions would be synergistic, and there is some evidence that postsynaptic activity may trigger retrograde messengers that affect their presynaptic contacts (Boehring & Snyder 2003). Finally, some forms of plasticity result in increased synaptic contacts through the growth of new dendritic spines (Muller et al. 2002, Trachtenberg et al. 2002). Thus, we have a rich field of candidates for learning-related cellular changes in the FTA. Determination of which changes contribute to learning requires indirect methods because it is not feasible at this time to identify, in vivo, whether or not neurotransmitter release or receptor density has changed at the individual synapses responsible for learning. The first suggestion that increased transmitter release may be critical to fear-related plasticity in the FTA was made by Maren & Fanselow (1995). They used an in vivo LTP preparation that stimulated the ventral angular bundle, which carries information from the hippocampus and entorhinal cortex to the basolateral nucleus. High-frequency stimulation of this pathway induced a form of long-term potentiation that reduced paired-pulse facilitation. Paired-pulse

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

INTRACELLULAR MEDIATORS OF PLASTICITY

23 Nov 2004 18:57

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

216

AR

AR231-PS56-08.tex

FANSELOW




AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

P1: IKH

POULOS

facilitation is an enhanced postsynaptic response to the second of two closely spaced electrical stimulations of the presynaptic neuron. This observation of an interaction between paired-pulse facilitation and FTA-LTP is critical because pairedpulse facilitation results from increased neurotransmitter release by the presynaptic terminal. The idea is that if neurotransmitter release is already optimized by LTP, then additional manipulations designed to enhance neurotransmitter release, such as paired-pulse facilitation, should produce little effect. Like acquisition of fear conditioning, induction of the LTP was prevented by application of an NMDA antagonist (Maren & Fanselow 1995). These researchers went on to confirm that either severing the ventral angular bundle with electrolytic lesions or killing the cells in the region of the FTA that received these projections with a neurotoxin prevented contextual fear conditioning, providing evidence that this pathway participates in contextual fear conditioning. The pattern of results suggests that the acquisition of fear depends on postsynaptic activation of NMDA receptors, which in turn sets in motion processes that lead to increased neurotransmitter release by the presynaptic neuron. Subsequently, Huang & Kandel (1998) replicated the interaction of FTA LTP and paired-pulse facilitation using an in vitro slice preparation where stimulation of the external capsule caused LTP in the lateral nucleus. The external capsule contains projections from the auditory cortex to the lateral amygdala, which suggests that in this pathway, too, expression of potentiation involves increased neurotransmitter release by the presynaptic neuron. Again consistent with the earlier report, NMDA receptors were critical for induction of LTP. Additionally, because injecting a calcium chelator into the postsynaptic neuron blocked LTP, calcium influx through the NMDA receptor appears to be critical for the postsynaptic contribution to the LTP. Paired-pulse facilitation and LTP were attenuated by blocking protein kinase A (PKA) activity in both the presynaptic and postsynaptic cells. However, blocking PKA in the postsynaptic cell alone did not affect LTP. Thus, the increased transmitter release caused by LTP depends on presynaptic activation of PKA (Huang & Kandel 1998). Brain slices containing the auditory pathway from either the thalamus (McKernan & Shinnick-Gallagher 1997) or cortex (Tsvetkov et al. 2002) to the lateral amygdala taken from fear-conditioned animals show enhanced neurotransmission accompanied by an increased probability of neurotransmitter release, which suggests that presynaptic changes are generally involved in the formation of fear memories. The finding that infusion of PKA inhibitors into the amygdala blocks acquisition of long- but not short-term memory for conditional fear is consistent with the hypothesis that PKA-regulated presynaptic changes mediate fear learning (Goosens et al. 2000, Schafe & LeDoux 2000). Induction of the presynaptic changes mediating learning requires NMDAreceptor activation of the postsynaptic neuron. At this point, there is reasonable detail about the postsynaptic cascade within the FTA that contributes to the formation of memory, and these events are roughly parallel to the postsynaptic events that contribute to plasticity in the CA1 region of the hippocampus (see, e.g., Lynch 2004, Pittenger & Kandel 2003). The influx of calcium ions through the NMDA

23 Nov 2004 18:57

AR

AR231-PS56-08.tex

AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

NEUROSCIENCE OF ASSOCIATIVE LEARNING

P1: IKH

217

receptor results in persistent activation of calcium-calmodulin-dependent protein kinases (CaMK). Genetically induced deletions of one type of CaMK (CaMKII) impair several forms of learning in mice, although much of the work has focused on hippocampus-dependent learning (e.g., Silva 2003). Mayford et al. (1996) targeted a mutation to a specific amino acid within the CaMKII molecule that interfered with its normal activation. Mice bearing this mutation in a manner restricted to amygdala and striatal neurons had a deficit in fear conditioning. Similarly, genetic deletion of CaMKIV appears to produce selective deficits in fear learning and FTA-LTP (Wei et al. 2002). Both auditory fear conditioning and LTP-inducing electrical stimulation of the medial geniculate-amygdala pathway cause activation and movement of CaMKII into dendritic spines of lateral nucleus neurons (Rodrigues et al. 2004). Electron and light microscopy confirmed that this activated CaMKII is associated with postsynaptic NMDA receptors located in spines that synapse with the terminals of neurons that originate in the medial geniculate. Rodrigues et al. also found that infusions of pharmacological inhibitors of CaMKII into the FTA blocked both fear conditioning and LTP. Activation of CamK is known to activate cyclic-AMP response element-binding protein (CREB), which is a regulator of gene expression and memory (Impey et al. 1998, Kida et al. 2002, Tully 1997). Along with a loss of fear conditioning, mice with a genetic deletion of CaMKIV show reduced CREB activation in the FTA (Wei et al. 2002). Increasing CREB expression in the FTA by introducing additional copies of the CREB gene carried by a virus enhances long-term memory for fear (Josselyn et al. 2001). Gene expression is further implicated because mRNA for the immediate early gene Zif268 is expressed specifically in the lateral amygdala following fear conditioning (Rosen et al. 1998). Blockade of the transcription of DNA to RNA within the amygdala prevents the acquisition of fear memories (Bailey et al. 1999), as does translation of RNA into protein (Schafe & LeDoux 2000). Some important issues have not been resolved. It is not known what these new proteins are or how they function to produce enduring changes in synaptic efficacy in the FTA. Protein synthesis in the postsynaptic cell seems necessary for the induction of the essential plasticity, but the critical changes in the expression of this plasticity appear to be presynaptic. Some combination of postsynaptic calcium influx, activation and movement of CamK into postsynaptic spines, and/or gene expression-mediated protein synthesis in the nucleus of the postsynaptic neuron initiates PKA-dependent presynaptic changes confined to the synapses that participated in the learning. Of course, this is such a fast-evolving field that these questions may well be answered by the time this review appears in print. As pointed out in the section above, fear conditioning involves plasticity at a number of sites. Here we have chosen to focus on the FTA because it is the site most relevant to the formation of new associations between the CS and US, and the presently available data make the strongest link between intracellular events and specific learned behaviors in vertebrate animals. In addition, with the exception of the hippocampus, little is known about the intracellular events mediating

23 Nov 2004 18:57

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

218

AR

AR231-PS56-08.tex

FANSELOW




AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

P1: IKH

POULOS

plasticity in these other regions. Certainly, the knowledge of the intracellular events mediating plasticity for the CA1 field of the hippocampus is more detailed than it is for the amygdala. Although much of the postsynaptic cascade is similar in the two, a critical mediator of hippocampal plasticity is the insertion of new AMPAtype glutamate receptors into the postsynaptic region (Isaac et al. 1995, Liao et al. 1995, Pickard et al. 2001). However, the current state of knowledge does not preclude such postsynaptic changes in the FTA, and there is certainly evidence for a presynaptic influence on hippocampal LTP (Lisman 2003). Indeed, one can be reasonably confident that the plasticity underlying learning in these structures will turn out to be a result of an active dialogue between the pre- and postsynaptic neurons. Generalization to humans Although it is impossible to perform equivalently detailed analysis in humans, to the extent that such work has been done the specifics are remarkably consistent. Bechara et al. (1995) examined a patient with UrbachWiethe disease, a condition that results in bilateral calcification and atrophy of the amygdala with little damage to any other medial temporal lobe structures. The patient had pronounced loss in delay fear conditioning assessed by autonomic arousal despite the fact that her memory for the details (declarative memory) of the conditioning session was completely intact. Consistent with the lesion data, functional magnetic resonance imaging (fMRI) shows enhanced blood flow to the amygdala during delay fear conditioning (LaBar et al. 1998), and this amygdala activation is highly correlated with the conditional fear response (Cheng et al. 2003). However, hippocampal activity is correlated more with the cognitive appreciation of the situation than with the emotional response (Cheng et al. 2003). Similarly, humans with hippocampal damage and an intact amygdala show normal fear conditioning with a loss of declarative memory for the conditioning situation (Bechara et al. 1995). Contextual fear conditioning has not been studied in humans. Knight et al. (2004) found that trace and delay CS produced similar activation of the human anterior cingulate; this contrasts with the differential activation found in mice. It is not clear if this is because the animal studies examined cellular markers of activity that were more precise (Han et al. 2003) or because the human studies trained trace, delay, and unpaired stimuli in the same subjects (Knight et al. 2004).

Eyeblink Conditioning The core neural component for eyeblink conditioning is the cerebellum, a brain structure located just caudal to the cerebral hemispheres and overlying the dorsal surface of the brainstem. The neuronal cell bodies of the cerebellum form a thick cortical layer that covers the underlying white matter (axons) and deep cerebellar nuclei. These nuclei, organized in medial-lateral orientation, consist of the fastigial, interpositus, and dentate nuclei. The relays of cerebellum are organized in highly regular manner, with well-defined sensory input and motor output pathways.

NEURAL CIRCUITRY OF EYEBLINK CONDITIONING

23 Nov 2004 18:57

AR

AR231-PS56-08.tex

AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

NEUROSCIENCE OF ASSOCIATIVE LEARNING

P1: IKH

219

Sensory information corresponding to potential tone or light CSs is relayed to the cerebellum by auditory and visual pathways primarily via pontine mossy fibers (see Figure 2). Mossy fibers form monosynaptic connections to neurons in the interpositus nucleus (IPN) and disynaptic connections to the Purkinje cells of the cerebellar cortex by way of granule cell parallel fibers. Damage to appropriate portions of the pontine nuclei can abolish CRs established to a tone CS while leaving conditioning to a visual CS, such as light, intact (Steinmetz et al. 1987). At the level of the IPN, single neurons receive CS information from multiple sensory modalities. For example, single cell recordings reveal IPN neurons that respond singly and in combination to both tone and light CSs and air puffs (Tracy et al. 2001), and appropriately directed lesions of the IPN prevent conditioning to all CS modalities tested (Ivkovich et al. 1993, Lavond et al. 1985, Steinmetz et al.

Figure 2 A simplified schematic of the putative neural circuitry essential for delay eyeblink conditioning. Arrows and fork terminals denote excitatory synapses; bar terminals denote inhibitory synapses. Solid ovals denote fibers (axons) and dotted ovals denote proposed sites of plasticity underlying eyeblink conditioning. A/V, auditory and visual pathways; CC, cerebellar cortex; cf, climbing fiber; CS, conditioned stimulus; CR/UR, conditioned response/unconditioned response; GC, granule cell; IO, inferior olive; IPN, interpositus nucleus; mf, mossy fibers; MN, motor nuclei; PC, Purkinje cell; pf, parallel fiber; PN, pontine nucleus; RN, red nucleus; RF, reticular formation; TN, trigeminal nucleus; US, unconditioned stimulus. (Modified from Christian & Thompson 2003.)

23 Nov 2004 18:57

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

220

AR

AR231-PS56-08.tex

FANSELOW




AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

P1: IKH

POULOS

1992, Thompson & Krupa 1994, Yeo et al. 1985). Substitution of peripheral CSs with direct electrical stimulation of either the pontine nucleus or IPN serves as a very powerful CS, producing conditioning that is more rapidly acquired than to light or tone (Poulos & Thompson 2004, Steinmetz et al. 1986, Tracy et al. 1998). Somatosensory information corresponding to an air-puff or mild eye-shock US arrives at the cerebellum primarily by climbing fibers from the inferior olive (see Figure 2). This essential US pathway includes the trigeminal nucleus, which in turn innervates the inferior olive, which sends axonal processes to the cerebellum via climbing fibers. The climbing fibers and their collaterals synapse on Purkinje cells and IPN neurons of the cerebellum and utilize glutamate and CRH as primary neurotransmitters (Ito 2002). Prior to training, lesions of the inferior olive prevent eyeblink conditioning, whereas similar lesions following conditioning result in extinction with continued paired training (McCormick et al. 1985). Conversely, electrical stimulation of the inferior olive that elicits behavioral responses when used as a US produces conditioning at a rate, magnitude, and topography similar to peripheral USs (Mauk et al. 1986). The IPN is not only a primary site of CS-US convergence, but also a primary cerebellar motor output responsible for the generation of the conditional eyeblink response. The IPN, like the motor cortex, contains a somatotopic representation of the entire body (Schultz et al. 1979). For example, in the rabbit, direct electrical stimulation of the medial aspects of IPN elicits movements in the lower trunk and hind limb areas, whereas stimulation of lateral portions of IPN evokes movements in the head area that include eyelid closure (McCormick et al. 1983). This eyelid region of the IPN sends heavy projections to the contralateral magnocellular red nucleus. From the red nucleus, projections to a set of motor nuclei trigger the expression of the conditional eyeblink response. The same motor nuclei also receive direct and indirect projections from the trigeminal nucleus, responsible for producing the unconditional eyeblink reflex. Under simple delay conditioning procedures, the cerebellum and its associated brainstem regions are necessary and sufficient for the acquisition and expression of conditional eyeblink responses (Figure 2). However, damage to the medial septum, a primary source of cholinergic projections to the hippocampus, retards the rate of delay conditioning (Berry & Thompson 1979). Conversely, intraseptal injections of scopolamine, an acetylcholine receptor antagonist that suppresses hippocampal functioning, also slow eyeblink conditioning (Salvatierra & Berry 1989, Solomon & Gottfried 1981). However, if the hippocampus is lesioned first, scopolamine no longer impairs learning (Solomon et al. 1983). Thus, for delay eyeblink conditioning, a functionally compromised hippocampus is more detrimental to learning than is the absence of the hippocampus. As similarly described in fear conditioning, the introduction of a stimulus-free period between the CS and US requires the hippocampus. Lesions of the hippocampus, which produce no discernable effects on delay conditioning, markedly impair trace eyeblink conditioning (Beylin

23 Nov 2004 18:57

AR

AR231-PS56-08.tex

AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

NEUROSCIENCE OF ASSOCIATIVE LEARNING

P1: IKH

221

et al. 2001, Solomon et al. 1986). In trace conditioning, lesions made one month following training do not impair performance of trace CRs (Kim et al. 1995) in a manner that parallels hippocampus-dependent consolidation of context fear (Kim & Fanselow 1992). Anatomical evidence suggests that interactions between the cerebellum and hippocampus occur indirectly, with the hippocampus modulating CS and US input by interacting with the pontine nucleus and inferior olive (Lee & Kim 2004). In eyeblink conditioning, it is quite clear that both neurons of IPN and Purkinje cells of the cerebellar cortex receive CS and US inputs. Therefore, it is conceivable that plasticity at one or both sites mediates the formation and storage of eyeblink memory traces. Furthermore, anatomical evidence reveals that both cerebellar regions send and receive reciprocal connections among each other (see Figure 2). For this reason, determining the relative contributions of IPN and cerebellar cortex has been difficult. Initial findings by McCormick et al. (1981) demonstrated that electrolytic lesions of the dentate-IPN region and large aspirations of cerebellar cortex completely abolished all expression of conditional eyeblink responses without affecting the UR. Later studies revealed that lesions of the anterior lateral IPN as small as 1 mm3 are sufficient to completely abolish CRs (Lavond et al. 1984). Further, temporary inactivation methods demonstrate that inactivation of the IPN during training completely prevents learning, as evidenced by the lack of CRs following inactivation (Clark et al. 1992; Krupa & Thompson 1995, 1997). In contrast, learning is not prevented by inactivation of an efferent pathway or its target, the red nucleus, which completely abolishes CR expression (Clark & Lavond 1993, Krupa & Thompson 1995, Krupa et al. 1993). These results suggest that the essential plasticity cannot occur efferent to the IPN and that essential memory trace (or traces) is established and maintained in the cerebellum. Single- and multiple-unit recordings in the IPN show learning-related unit activity, also found in the cerebellar cortex, that precedes and predicts occurrence of the behavioral CR (Berthier & Moore 1986, 1990; Gould & Steinmetz 1994; King et al. 2001; McCormick & Thompson 1984; Rogers et al. 2001). Moreover, lesions or temporary inactivation of the IPN (Thompson & Krupa 1994) abolish neuronal recording of conditioning-related unit activity in many regions of the essential eyeblink circuitry (see Figure 2). Removal of cerebellar cortical tissue has been reported to result in varying effects in eyeblink conditioning. Although it has been reported that lesions of cerebellar cortex prevent the acquisition of conditional eyeblink responses (Garcia et al. 1999), all other studies to date report either no effect or impairments in acquisition, retention, and/or CR timing (Lavond & Steinmetz 1989, Logan et al. 1994, McCormick & Thompson 1983, Perrett et al. 1993). Perhaps the clearest interpretation of these lesion studies is provided by Purkinje cell degeneration (PCD) mice. Several weeks after birth of these mice, Purkinje cells, the sole output of the cerebellar cortex, completely degenerate. PCD mice acquire eyeblink

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

SITES OF CRITICAL PLASTICITY

23 Nov 2004 18:57

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

222

AR

AR231-PS56-08.tex

FANSELOW




AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

P1: IKH

POULOS

conditioning at a slower rate and lower level than wild-type mice, but they do learn (Chen et al. 1996). Further, CRs expressed tended to have shorter peak latencies. Recent work by Bao et al. (2002) suggests that CR expression and timing may be completely dissociable and that memory traces for eyeblink CRs may be encoded in a functionally distinct manner in both the cerebellar cortex and IPN. In their study, rabbits displaying well-timed CRs were given sequential IPN application of the gamma-aminobutyric acid (GABAA) agonist muscimol and the GABAA antagonist picrotoxin, which resulted in the expression of reduced onset and peak CR latencies poorly timed to the US. Such results are consistent with studies by Mauk and associates using high concentrations of picrotoxin (Garcia & Mauk 1998, Medina et al. 2001). This disruption of CR timing following blockade of cortical outputs suggests that the memory traces for learned timing and for basic associative eyeblink memory may be expressed within two distinct sites of plasticity, the cerebellar cortex and IPN, respectively. Early cerebellar learning theories proposed plasticity within the cerebellar cortex, specifically at the parallel fiber-Purkinje cell (PF-PC) synapses, as a general mechanism of motor learning (Albus 1971, Marr 1969). Albus (1971) further hypothesized that the high tonic-firing rate of Purkinje cells resulting in sustained GABAergic inhibition of the IPN could be released by a decrease in PF-PC synaptic strength. Empirical evidence for this theory came with the discovery that long-term depression (LTD) resulting from simultaneous low-frequency stimulation of parallel fibers and climbing fibers could result in persistent decreases in PF-PC synaptic efficacy (Ito & Kano 1982). However, under such parameters of simultaneous CS and US presentation, eyeblink conditioning does not develop. Further work by Chen & Thompson (1995) demonstrated that under specific in vitro conditions that did not block GABA receptors, LTD could be induced by repeated parallel fiber and climbing fiber stimulation separated by 250 msec, matching the optimal eyeblink conditioning interstimulus intervals. In contrast, simultaneous stimulation of parallel and climbing fibers can elicit LTD in the presence of the GABA antagonist biccuculline. Indeed, in the intact animal, direct electrical stimulation of parallel fibers as a CS and climbing fibers as a US produces appropriately timed eyeblink conditioning (Shinkman et al. 1996). Recent investigations both in vivo and in vitro have begun to examine plasticity specific to the IPN and its synapses. One attractive mechanism of memory formation in eyeblink conditioning is long-term plasticity of the mossy fiber-IPN synapse. Racine (1986) previously reported that tetanic electrical stimulation of mossy fibers induces LTP of the mossy fiber-IPN synapse. Perhaps the most compelling evidence of conditioning-specific synaptic plasticity comes from recent findings by Kleim and colleagues (2002) using unbiased stereological synapse quantification methods. In this study, rats trained in delay eyeblink conditioning exhibited an increased number of excitatory IPN synapses compared to unpaired and untrained controls. Moreover, because the expression of CRs is driven solely by the CS, increases in excitatory IPN synapses are likely to occur along the CS

23 Nov 2004 18:57

AR

AR231-PS56-08.tex

AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

NEUROSCIENCE OF ASSOCIATIVE LEARNING

P1: IKH

223

pathway, specifically at mossy fiber-IPN synapses. There was no increase in inhibitory synapses in the IPN (presumably from Purkinje axons). A form of nonsynaptic plasticity also has been implicated in cerebellar learning. Aizenman & Linden (2000) showed that direct high-frequency stimulation of deep nuclear neurons of the cerebellum under reduced levels of inhibition resulted in persistent increases in maximum firing rate. Interestingly, IPN electrical stimulation to eyeblink thresholds are markedly reduced following eyeblink conditioning, suggesting possible conditioning-related increases in IPN excitability (Poulos et al. 2002). Conversely, infusion of NMDA antagonist APV, in the IPN, which in vitro blocks the induction of intrinsic excitability (Aizenman & Linden 2000), selectively impairs conditioning but not expression of eyeblink CRs (Chen & Steinmetz 2000a). In eyeblink conditioning the combination of evidence from several levels of analysis strongly suggests that the development and maintenance of the adaptive (well-timed) associative memory (CRs) are mediated by plasticity occurring in at least two cerebellar regions, the cortex and IPN. Moreover, the forms of plasticity that contribute to learning have yet to be directly identified in vivo; therefore, it is imperative to employ molecular and genetic methods to isolate key components of plasticity and learning. At the level of the IPN, if an increase in synaptic strength is a mechanism of the primary CSUS memory, then alterations of intracellular mediators of LTP or synaptogenesis should compromise conditioning. Further, if depression of synaptic function in the cortex mediates the appropriate timing of the associative memory, utilizing methods to manipulate expression of key molecular components of cerebellar cortical LTD should affect conditional responding and timing. The demonstration that eyeblink conditioning is associated with an increase in the number of IPN excitatory synapses prompts important questions as to possible mechanisms of synapse formation in the adult central nervous system that may mediate memory formation. Almost certainly any physical or chemical changes in neurons involve alterations in gene expression and/or protein structure. Gomi et al. (1999) demonstrated that inhibition of RNA synthesis in the IPN profoundly impairs learning but not the expression of eyeblink CRs. Conversely, inhibition of protein synthesis similarly reduces the rate of conditioning (Bracha et al. 1998). Gomi et al. (1999) further identified a kinase whose expression increased in the IPN with eyeblink conditioning. Isolation of cDNA and sequence analysis revealed that the expressed gene was KKIAMRE kinase, a member of the CDC2-related and mitogen-activated protein (MAP) family. Both inhibitors of protein kinases and specific MAP kinase p38 markedly impair learning but not CR expression (Chen & Steinmetz 2000b, Zhen et al. 2001). Evidence of synaptogenesis, or gene transcription and translation, does not preclude a role for mossy fiber-IPN LTP as a mechanism of learning, such that LTP may be an antecedent or act in concert with synaptogenesis to promote memory formation. Like amygdaladependent fear conditioning and hippocampal-dependent maze learning, application of an NMDA receptor antagonist, which prevents learning and amygdala and

INTRACELLULAR MEDIATORS OF PLASTICITY

23 Nov 2004 18:57

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

224

AR

AR231-PS56-08.tex

FANSELOW




AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

P1: IKH

POULOS

hippocampal LTP, markedly impairs eyeblink conditioning (Chen & Steinmetz 2000a). A reduction in PF-PC synaptic strength at multiple and spatially distinct synapses is an attractive model that could promote the appropriate temporal release of IPN inhibition and hence expression of appropriately timed conditional eyeblink responses. Indeed, single-unit recordings of Purkinje cells reveal decreases and increases in firing rates that immediately precede the expression of well-timed CRs (King et al. 2001). The requirements for eyeblink conditioning, like those of PF-PC LTD, are typically associative. Both can be induced by parallel fiber and Purkinje cell stimulation. To date, molecular mechanisms of cerebellar LTD are more clearly understood. The induction of cerebellar LTD is initially triggered by activation of mGluR1 metabotropic and AMPA receptors. Mutant mice deficient in mGluR1 and impaired in the induction of LTD are also impaired in eyeblink conditioning (Aiba et al. 1994). Further, blockade of cerebellar cortical AMPA receptors, which block the induction of LTD (Wang & Linden 2000), also has been reported to impair the acquisition and expression of CRs (Attwell et al. 1999, 2001). Following activation of glutamate receptors, transient activation of protein kinase C (PKC) is essential. Transgenic mice with Purkinje cell–specific inhibition of PKC, which prevents the induction of LTD, show impairments in eyeblink conditioning (Goossens et al. 2001, Koekkoek et al. 2003). Conversely, eyeblink conditioning results in an increase in membrane-bound PKC specific to cerebellar cortical tissue (Freeman et al. 1998). Further, inhibition of nitric oxide, which blocks the induction of LTD (Shibuki & Okada 1991), results in attenuated eyeblink acquisition and learning-related neural activity in the IPN (Allen & Steinmetz 1996). Expression of PF-PC LTD is expressed postsynaptically, possibly as a reduction in the number of functional AMPA receptors produced by endocytosis. Interestingly, quantitative autoradiography reveals eyeblink conditioning– related decreases in (3H) AMPA binding to synaptic subpopulations of cerebellar cortical AMPA receptors (Hauge et al. 1998). Additional support for the LTD hypothesis comes from mice deficient in glial fibrillary acidic protein; these mice show normal excitatory synaptic transmission and have no observable motor deficits, but are markedly deficient in PF-PC LTD and have impaired eyeblink conditioning (Shibuki et al. 1996). It is likely that the eyeblink conditioning results in plasticity at a number of sites. Here we have chosen to focus primarily on two critical sites of plasticity, the IPN and cerebellar cortex. It should be noted that within each of these regions, synaptic plasticity at sites other than PF-PC and mossy fiber-IPN synapses has been identified and discussion of such sites and mediators of plasticity and their potential contributions to eyeblink conditioning has been reviewed by Hansel et al. (2001). However, these potential mechanisms have yet to be tested in eyeblink conditioning. As pointed out above, much of the research in cerebellar synaptic plasticity has focused on PF-PC LTD. However, because most of the evidence suggests that eyeblink conditioning can occur independent of the cerebellar cortex, it seems likely that elucidating the molecular mechanisms of IPN plasticity (LTP

23 Nov 2004 18:57

AR

AR231-PS56-08.tex

AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

NEUROSCIENCE OF ASSOCIATIVE LEARNING

P1: IKH

225

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

and/or synaptogenesis) may yield for the first time a cellular substrate of cerebellar memory. Generalization to humans The results obtained in animal studies of eyeblink conditioning correspond closely to the evidence available in human eyeblink conditioning. The combination of lesion, neuroimaging, and observation in amnesic patients confirms that the cerebellum plays a critical role in eyeblink conditioning in humans as well. Positron emission tomography in humans reveals changes in glucose metabolism in the cerebellum correlated with eyeblink conditioning (Blaxton et al. 1996, Logan & Grafton 1995, Molchan et al. 1994, Schreurs et al. 1997). Moreover, fMRI shows increases in cerebellar cortex and deep nuclear activity over the course of eyeblink conditioning (Lemieux & Woodruff-Pak 2000). In addition, both positron emission tomography and fMRI reveal activation of the medial temporal lobe, which includes the hippocampus (Lemieux & WoodruffPak 2000, McIntosh & Schreurs 2000). In amnesic patients suffering from bilateral medial temporal lobe damage, trace conditioning under a long trace interval is markedly impaired, whereas delay eyeblink conditioning is normal (Clark & Squire 1998, McGlinchey-Berroth et al. 1997, McGlinchey-Berroth 2000). However, patients with bilateral as well as with unilateral cerebellar lesions ipsilateral to the trained eye are severely impaired in delay and trace eyeblink conditioning (Daum et al. 1993, Lye et al. 1988, Solomon et al. 1989, Topka et al. 1993). In sum, across a wide spectrum of mammals including humans the cerebellum is critically for the acquisition, retention, and expression of conditional eyeblink responses.

CONCLUSIONS Different types of learning and memory must be analyzed as separate systems that evolved to solve specific functions. Emotional memory, examined through Pavlovian fear conditioning, and specific motor responses, examined through Pavlovian eyeblink conditioning, arise from largely different neural circuits. Even the specific mechanisms that produce memory at the cellular level are different. We also need to recognize that there are a few important similarities. At the circuit level, both systems require input from similar cortical structures when conditioning becomes more complex. The hippocampal contribution to trace conditioning is the clearest example. Additionally, a negative feedback circuit regulates learning in both systems, although the specific circuit is very different (Fanselow 1998). In terms of plasticity, there is considerable overlap in the specific molecules that regulate synaptic plasticity, although these molecules may play different roles in the different forms of learning. Two emerging findings are critical. The first finding is that the mechanisms of learning involve a coordinated interplay between pre- and postsynaptic regions, and the second is that learning and gene expression are so intimately related that the old nature-nurture distinction is of little value.

23 Nov 2004 18:57

226

AR

AR231-PS56-08.tex

FANSELOW




AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

P1: IKH

POULOS

The Annual Review of Psychology is online at http://psych.annualreviews.org

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

LITERATURE CITED Aiba A, Kano M, Chen C, Stanton ME, Fox GD, et al. 1994. Deficient cerebellar longterm depression and impaired motor learning in mGluR1 mutant mice. Cell 79:377–88 Aizenman CD, Linden DJ. 2000. Rapid, synaptically driven increases in the intrinsic excitability of cerebellar deep nuclear neurons. Nat. Neurosci. 3:109–11 Albus JS. 1971. A theory of cerebellar function. Math. Biosci. 10:25–61 Allen MT, Steinmetz JE. 1996. A nitric oxide synthase inhibitor delays the formation of learning-related neural activity in the cerebellar interpositus nucleus during rabbit eyelid conditioning. Pharmacol. Biochem. Behav. 53:147–53 Amaral DG, Price JL, Pitkanen A, Carmichael ST. 1992. Anatomical organization of the primate amygdaloid complex. In The Amygdala: Neurological Aspects of Emotion, Memory and Mental Dysfunction, ed. JP Aggleton, pp. 1–66. New York: Wiley-Liss Amorapanth P, LeDoux JE, Nader K. 2000. Different lateral amygdala outputs mediate reactions and actions elicited by a fear-arousing stimulus. Nat. Neurosci. 3(1):74–79 Anagnostaras SG, Maren S, Fanselow MS. 1999. Temporally graded retrograde amnesia of contextual fear after hippocampal damage in rats: within-subjects examination. J. Neurosci. 19:1106–14 Antonov I, Antonova I, Kandel ER, Hawkins RD. 2003. Activity-dependent presynaptic facilitation and Hebbian LTP are both required and interact during classical conditioning in aplysia. Neuron 37(1):135–47 Attwell PJ, Rahman S, Ivarsson M, Yeo CH. 1999. Cerebellar cortical AMPA-kainate receptor blockade prevents performance of classically conditioned nictitating membrane responses. J. Neurosci. 19:RC45 Attwell PJ, Rahman S, Yeo CH. 2001. Acquisition of eyeblink conditioning is critically dependent on normal function in cerebellar

cortical lobule HVI. J. Neurosci. 21:5715– 22 Bailey DJ, Kim JJ, Sun W, Thompson RF, Helmstetter FJ. 1999. Acquisition of fear conditioning in rats requires the synthesis of mRNA in the amygdala. Behav. Neurosci. 113(2):276–78 Bao S, Chen L, Kim JJ, Thompson RF. 2002. Cerebellar cortical inhibition and classical eyeblink conditioning. Proc. Natl. Acad. Sci. USA 99:1592–97 Bechara A, Tranel D, Damasio H, Adolphs R, Rockland C, Damasio AR. 1995. Double dissociation of conditioning and declarative knowledge relative to the amygdala and hippocampus in humans. Science 269:1115– 18 Benarroch EE. 2001. Pain-autonomic interactions: a selective review. Clin. Auton. Res. 11(6):343–49 Berry SD, Thompson RF. 1979. Medial septal lesions retard classical conditioning of the nicitating membrane response in rabbits. Science 205:209–11 Berthier NE, Moore JW. 1986. Cerebellar Purkinje cell activity related to the classically conditioned nictitating membrane response. Exp. Brain Res. 63:341–50 Berthier NE, Moore JW. 1990. Activity of deep cerebellar nuclear cells during classical conditioning of nictitating membrane extension in rabbits. Exp. Brain Res. 83:44–54 Beylin AV, Gandhi CC, Wood GE, Talk AC, Matzel LD, Shors TJ. 2001. The role of the hippocampus in trace conditioning: temporal discontinuity or task difficulty? Neurobiol. Learn. Mem. 76:447–61 Bi G, Poo M. 2001. Synaptic modification by correlated activity: Hebb’s postulate revisited. Annu. Rev. Neurosci. 24:139–66 Blaxton TA, Zeffiro TA, Gabrieli JD, Bookheimer SY, Carrillo MC, et al. 1996. Functional mapping of human learning: a positron emission tomography activation study of

23 Nov 2004 18:57

AR

AR231-PS56-08.tex

AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

NEUROSCIENCE OF ASSOCIATIVE LEARNING eyeblink conditioning. J. Neurosci. 16:4032– 34 Bliss TV, Gardner-Medwin AR. 1973. Longlasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path. J. Physiol. 232(2):357–74 Bliss TV, Lomo T. 1973. Long-lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path. J. Physiol. 232:331–56 Boehning D, Snyder SH. 2003. Novel neural modulators. Annu. Rev. Neurosci. 26:105–31 Bracha V, Irwin KB, Webster ML, Wunderlich DA, Stachowiak MK, Bloedel JR. 1998. Microinjections of anisomycin into the intermediate cerebellum during learning affects the acquisition of classically conditioned responses in the rabbit. Brain Res. 788:169–78 Brunzell DH, Kim JJ. 2001. Fear conditioning to tone, but not to context, is attenuated by lesions of the insular cortex and posterior extension of the intralaminar complex in rats. Behav. Neurosci. 115(2):365–75 Bucci DJ, Phillips RG, Burwell RD. 2000. Contributions of postrhinal and perirhinal cortex to contextual information processing. Behav. Neurosci. 114(5):882–94 Burstein R, Potrebic SJ. 1993. Retrograde labeling of neurons in the spinal cord that project directly to the amygdala or the orbital cortex in the rat. Comp. Neurol. 335(4):469–85 Burwell RD, Saddoris MP, Bucci DJ, Wiig KA. 2004. Corticohippocampal contributions to spatial and contextual learning. J. Neurosci. 24(15):3826–36 Canli T, Brown TH. 1996. Amygdala stimulation enhances the rat eyeblink reflex through a short-latency mechanism. Behav. Neurosci. 110(1):51–59 Chapman PF, Kairiss EW, Keenan CL, Brown TH. 1990. Long-term synaptic potentiation in the amygdala. Synapse 6(3):271–78 Chen C, Thompson RF. 1995. Temporal specificity of long-term depression in parallel fiber–Purkinje synapses in rat cerebellar slice. Learn. Mem. 2:185–98

P1: IKH

227

Chen G, Steinmetz JE. 2000a. Intra-cerebellar infusion of NMDA receptor antagonist AP5 disrupts classical eyeblink conditioning in rabbits. Brain Res. 887:144–56 Chen G, Steinmetz JE. 2000b. Microinfusion of protein kinase inhibitor H7 into the cerebellum impairs the acquisition but not the retention of classical eyeblink conditioning in rabbits. Brain Res. 856:193–201 Cheng DT, Knight DC, Smith CN, Stein EA, Helmstetter FJ. 2003. Functional MRI of human amygdala activity during Pavlovian fear conditioning: stimulus processing versus response expression. Behav. Neurosci. 117:3– 10 Christian KM, Thompson RF. 2003. Neural substrates of eyeblink conditioning: acquisition and retention. Learn. Mem. 10:427–55 Clark RE, Lavond DG. 1993. Reversible lesions of the red nucleus during acquisition and retention of a classically conditioned behavior in rabbits. Behav. Neurosci. 107:264–70 Clark RE, Squire LR. 1998. Classical conditioning and brain systems: the role of awareness. Science 280:77–81 Clark RE, Zhang AA, Lavond DG. 1992. Reversible lesions of the cerebellar interpositus nucleus during acquisition and retention of a classically conditioning behavior. Behav. Neurosci. 106:879–88 Clugnet MC, LeDoux JE. 1990. Synaptic plasticity in fear conditioning circuits: induction of LTP in the lateral nucleus of the amygdala by stimulation of the medial geniculate body. J. Neurosci. 10(8):2818–24 Collingridge GL, Kehl SJ, McLennan H. 1983. Excitatory amino acids in synaptic transmission in the Schaffer collateral-commissural pathway of the rat hippocampus. J. Physiol. 334:33–46 Daum I, Schugens MM, Ackermann H, Lutzenberger W, Dichgans J, Birbaumer N. 1993. Classical conditioning after cerebellar lesions in humans. Behav. Neurosci. 107:748– 56 Dayan P, Balleine BW. 2002. Reward, motivation, and reinforcement learning. Neuron 36(2):285–98

23 Nov 2004 18:57

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

228

AR

AR231-PS56-08.tex

FANSELOW




AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

P1: IKH

POULOS

DeOlmos JS, Alheid GF, Beltramino CA. 1985. Amygdala. In The Rat Nervous System, ed. G Paxinos, pp 223–334. Orlando, FL: Academic Edeline JM, Pham P, Weinberger NM. 1993. Rapid development of learning-induced receptive field plasticity in the auditory cortex. Behav. Neurosci. 107(4):539–51 Edeline JM, Weinberger NM. 1992. Associative retuning in the thalamic source of input to the amygdala and auditory cortex: receptive field plasticity in the medial division of the medial geniculate body. Behav. Neurosci. 106(1):81–105 Fanselow MS. 1980. Conditional and unconditional components of post-shock freezing in rats. Pavlov. J. Biol. Sci. 15:177–82 Fanselow MS. 1991. The midbrain periaqueductal gray as a coordinator of action in response to fear and anxiety. In The Midbrain Periaqueductal Gray Matter: Functional, Anatomical and Immunohistochemical Organization, ed. A Depaulis, R Bandler, pp 151–73. New York: Plenum Fanselow MS. 1998. Pavlovian conditioning, negative feedback, and blocking: mechanisms that regulate association formation. Neuron 20:625–27 Fanselow MS, Gale GD. 2003. The amygdala, fear and memory. Ann. NY Acad. Sci. 985:125–34 Fanselow MS, Kim JJ. 1994. Acquisition of contextual Pavlovian fear conditioning is blocked by application of an NMDA receptor antagonist D,L-2-amino-5-phosphonovaleric acid to the basolateral amygdala. Behav. Neurosci. 108:210–12 Feldman S, Weidenfeld J. 1998. The excitatory effects of the amygdala on hypothalamopituitary-adrenocortical responses are mediated by hypothalamic norepinephrine, serotonin, and CRF-41. Brain Res. Bull. 45(4):389–93 Freeman JH Jr., Scharenberg AM, Olds JL, Schreurs BG. 1998. Classical conditioning increases membrane-bound protein kinase C in rabbit cerebellum. Neuroreport 9:2669–73 French SJ, Totterdell S. 2003. Individual

nucleus accumbens-projection neurons receive both basolateral amygdala and ventral subicular afferents in rats. Neuroscience 119(1):19–31 Gale GD, Anagnostaras SG, Godsil BP, Mitchell S, Nozawa T, et al. 2004. Role of the basolateral amygdala in the storage of fear memories across the adult lifetime of rats. J. Neurosci. 24:3810–15 Garcia KS, Mauk MD. 1998. Pharmacological analysis of cerebellar contributions to the timing and expression of conditioned eyelid responses. Neuropharmacology 37:471–80 Garcia KS, Steele PM, Mauk MD. 1999. Cerebellar cortex lesions prevent acquisition of conditioned eyelid responses. J. Neurosci. 19:10940–47 Gauriau C, Bernard JF. 2002. Pain pathways and parabrachial circuits in the rat. Exp. Physiol. 87(2):251–58 Gomi H, Sun W, Finch CE, Itohara S, Yoshimi K, Thompson RF. 1999. Learning induces a CDC2-related protein kinase, KKIAMRE. J. Neurosci. 19: 9530–37 Goosens KA, Holt W, Maren S. 2000. A role for amygdaloid PKA and PKC in the acquisition of long-term conditional fear memories in rats. Behav. Brain Res. 114(1–2):145–52 Goossens J, Daniel H, Rancillac A, van der Steen J, Oberdick J, et al. 2001. Expression of protein kinase C inhibitor blocks cerebellar long-term depression without affecting Purkinje cell excitability in alert mice. J. Neurosci. 21:5813–23 Gould TJ, Steinmetz JE. 1994. Multiple-unit activity from rabbit cerebellar cortex and interpositus nucleus during classical discrimination/reversal eyelid conditioning. Brain Res. 652:98–106 Han CJ, O’Tuathaigh CM, van Trigt L, Quinn JJ, Fanselow MS, et al. 2003. Trace but not delay fear conditioning requires attention and the anterior cingulate cortex. Proc. Natl. Acad. Sci. USA 100:13087–92 Hansel C, Linden DJ, D’Angelo E. 2001. Beyond parallel fiber LTD: the diversity of synaptic and non-synaptic plasticity in the cerebellum. Nat. Neurosci. 4:467–75

23 Nov 2004 18:57

AR

AR231-PS56-08.tex

AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

NEUROSCIENCE OF ASSOCIATIVE LEARNING Hauge SA, Tracy JA, Baudry M, Thompson RF. 1998. Selective changes in AMPA receptors in rabbit cerebellum following classical conditioning of the eyelid-nictitating membrane response. Brain Res. 803:9–18 Huang YY, Kandel ER. 1998. Postsynaptic induction and PKA-dependent expression of LTP in the lateral amygdala. Neuron 21(1): 169–78 Huff NC, Rudy JW. 2004. The amygdala modulates hippocampus-dependent context memory formation and stores cue-shock associations. Behav. Neurosci. 118(1):53–62 Hunt PS. 1999. A further investigation of the developmental emergence of fear-potentiated startle in rats. Dev. Psychobiol. 34(4):281– 91 Hunt PS, Campbell BA. 1997. Developmental dissociation of the components of conditioned fear. In Learning, Motivation, and Cognition: The Functional Behaviorism of Robert C. Bolles, ed. ME Bouton, MS Fanselow, pp. 53–74. Washington, DC: Am. Psychol. Assoc. Impey S, Smith DM, Obrietan K, Donahue R, Wade C, Storm DR. 1998. Stimulation of cAMP response element (CRE)-mediated transcription during contextual learning. Nat. Neurosci. 1(7):595–601 Isaac JT, Nicoll RA, Malenka RC. 1995. Evidence for silent synapses: implications for the expression of LTP. Neuron 15(2):427–34 Ito M. 2002. The molecular organization of cerebellar long-term depression. Nat. Neurosci. Rev. 3:896–902 Ito M, Kano M. 1982. Long-lasting depression of parallel fiber–Purkinje cell transmission induced by conjunctive stimulation of parallel fibers and climbing fibers in the cerebellar cortex. Neurosci. Lett. 33:253–58 Ivkovich D, Lockard JM, Thompson RF. 1993. Interpositus lesion abolition of the eyeblink conditioned response is not due to effects on performance. Behav. Neurosci. 107:530–32 Jasmin L, Granato A, Ohara PT. 2004. Rostral agranular insular cortex and pain areas of the central nervous system: a tract-tracing study in the rat. J. Comp. Neurol. 468(3):425–40

P1: IKH

229

Jolkkonen E, Miettinen R, Pikkarainen M, Pitkanen A. 2002. Projections from the amygdaloid complex to the magnocellular cholinergic basal forebrain in rat. Neuroscience 111(1):133–49 Josselyn SA, Shi C, Carlezon WA Jr., Neve RL, Nestler EJ, Davis M. 2001. Long-term memory is facilitated by cAMP response elementbinding protein overexpression in the amygdala. J. Neurosci. 21(7):2404–12 Kapp BS, Frysinger RC, Gallagher M, Haselton JR. 1979. Amygdala central nucleus lesions: effect on heart rate conditioning in the rabbit. Physiol. Behav. 23:1109–17 Kida S, Josselyn SA, de Ortiz SP, Kogan JH, Chevere I, et al. 2002. CREB required for the stability of new and reactivated fear memories. Nat. Neurosci. 5(4):348–55 Killcross S, Robbins TW, Everitt BJ. 1997. Different types of fear-conditioned behaviour mediated by separate nuclei within amygdala. Nature 388(6640):377–80 Kim JJ, Clark RE, Thompson RF. 1995. Hippocampectomy impairs the memory of recently, but not remotely, acquired trace eyeblink conditioned responses. Behav. Neurosci. 109:195–203 Kim JJ, Fanselow MS. 1992. Modality specific retrograde amnesia of fear following hippocampal lesions. Science 256:675–77 King DA, Krupa DJ, Foy MR, Thompson RF. 2001. Mechanisms of neuronal conditioning. Int. Rev. Neurobiol. 45:313–37 Kleim JA, Freeman JH Jr., Bruneau R, Nolan BC, Cooper NR, et al. 2002. Synapse formation is associated with memory storage in the cerebellum. Proc. Natl. Acad. Sci. USA 99:13228–31 Knight DC, Cheng DT, Smith CN, Stein EA, Helmstetter FJ. 2004. Neural substrates mediating human delay and trace fear conditioning. J. Neurosci. 24:218–28 Koekkoek SK, Hulscher HC, Dortland BR, Hensbroek RA, Elgersma Y, et al. 2003. Cerebellar LTD and learning-dependent timing of conditioned eyelid responses. Science 301:1736–39 Krupa DJ, Thompson JK, Thompson RF. 1993.

23 Nov 2004 18:57

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

230

AR

AR231-PS56-08.tex

FANSELOW




AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

P1: IKH

POULOS

Localization of a memory trace in the mammalian brain. Science 260:989–91 Krupa DJ, Thompson RF. 1995. Inactivation of the superior cerebellar peduncle blocks expression but not acquisition of the rabbit’s classically conditioned eye-blink response. Proc. Natl. Acad. Sci. USA 92:5097– 101 Krupa DJ, Thompson RF. 1997. Reversible inactivation of the cerebellar interpositus nucleus completely prevents acquisition of the classically conditioned eye-blink response. Learn. Mem. 3:545–56 LaBar KS, Gatenby JC, Gore JC, LeDoux JE, Phelps EA. 1998. Human amygdala activation during conditioned fear acquisition and extinction: a mixed-trial fMRI study. Neuron 20:937–45 Lanuza E, Nader K, LeDoux JE. 2004. Unconditioned stimulus pathways to the amygdala: effects of posterior thalamic and cortical lesions on fear conditioning. Neuroscience 125(2):305–15 Lavond DG, Hembree TL, Thompson RF. 1985. Effect of kainic acid lesions of the cerebellar interpositus nucleus on eyelid conditioning in the rabbit. Brain Res. 326:179–82 Lavond DG, Kim JJ, Thompson RF. 1993. Mammalian brain substrates of aversive classical conditioning. Annu. Rev. Psychol. 44:317–42 Lavond DG, Lincoln JS, McCormick DA, Thompson RF. 1984. Effect of bilateral lesions of the dentate and interpositus cerebellar nuclei on conditioning of heart-rate and nictitating membrane/eyelid responses in the rabbit. Brain Res. 305:323–30 Lavond DG, Steinmetz JE. 1989. Acquisition of classical conditioning without cerebellar cortex. Behav. Brain Res. 33:113–64 LeDoux JE, Farb CR, Romanski LM. 1991. Overlapping projections to the amygdala and striatum from auditory processing areas of the thalamus and cortex. Neurosci. Lett. 134(1):139–44 LeDoux JE, Iwata J, Cicchetti P, Reis DJ. 1988. Different projections of the central amygdaloid nucleus mediate autonomic and be-

havioral correlates of conditioned fear. J. Neurosci. 8(7):2517–29 Lee T, Kim JJ. 2004. Differential effects of cerebellar, amygdalar, and hippocampal lesions on classical eyeblink conditioning in rats. J. Neurosci. 24:3242–50 Lemieux SK, Woodruff-Pak DF. 2000. Functional MRI studies of eyeblink classical conditioning. See Woodruff-Pak & Steinmetz 2000, Vol. 1, pp. 71–93 Liao D, Hessler NA, Malinow R. 1995. Activation of postsynaptically silent synapses during pairing-induced LTP in CA1 region of hippocampal slice. Nature 375(6530):400– 4 Lindquist DH, Jarrard LE, Brown TH. 2004. Perirhinal cortex supports delay fear conditioning to rat ultrasonic social signals. J. Neurosci. 24(14):3610–17 Lisman J. 2003. Long-term potentiation: outstanding questions and attempted synthesis. Philos. Trans. R. Soc. Lond. B Biol. Sci. 358(1432):829–42 Logan CG, Grafton ST. 1995. Functional anatomy of human eyeblink conditioning determined with regional cerebral glucose metabolism and positron-emission tomography. Proc. Natl. Acad. Sci. USA 92:7500–4 Logan CG, Lavond DG, Wong JT, Thompson RF. 1994. Acquisition of classically conditioned eyeblink response following bilateral lesions of flocculus and paraflocculus. Behav. Neural. Biol. 61:102–6 Lye RH, O’Boyle DJ, Ramsden RT, Schady W. 1988. Effects of unilateral cerebellar lesions on the acquisition of eye-blink conditioning in man. J. Physiol. 403:58P Lynch MA. 2004. Long-term potentiation and memory. Physiol. Rev. 84(1):87–136 Malinow R, Malenka RC. 2002. AMPA receptor trafficking and synaptic plasticity. Annu. Rev. Neurosci. 25:103–26 Malinow R, Miller JP. 1986. Postsynaptic hyperpolarization during conditioning reversibly blocks induction of long-term potentiation. Nature 320(6062):529–30 Maren S, Fanselow MS. 1995. Synaptic plasticity in the basolateral amygdala induced by

23 Nov 2004 18:57

AR

AR231-PS56-08.tex

AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

NEUROSCIENCE OF ASSOCIATIVE LEARNING hippocampal formation stimulation in vivo. J. Neurosci. 15:7548–64 Maren S, Yap SA, Goosens KA. 2001. The amygdala is essential for the development of neuronal plasticity in the medial geniculate nucleus during auditory fear conditioning in rats. J. Neurosci. 21(6):RC135 Marr D. 1969. A theory of cerebellar cortex. J. Physiol. 202:437–70 Martinez JL Jr., Derrick BE. 1996. Long-term potentiation and learning. Annu. Rev. Psychol. 47:173–203 Mauk MD, Steinmetz JE, Thompson RF. 1986. Classical conditioning using stimulation of the inferior olive as the unconditioned stimulus. Proc. Natl. Acad. Sci. USA 83:5349– 53 Mayford M, Bach ME, Huang YY, Wang L, Hawkins RD, Kandel ER. 1996. Control of memory formation through regulated expression of a CaMKII transgene. Science 274(5293):1678–83 McCormick DA, Lavond DG, Clark GA, Kettner RE, Thompson RF. 1981. The engram found? Role of the cerebellum in classical conditioning of the nictitating membrane and eyelid responses. Bull. Psychon. Soc. 28:769–75 McCormick DA, Lavond DG, Thompson RF. 1983. Neuronal responses of the rabbit brainstem during performance of the classically conditioned nictitating membrane (NM)/eyelid response. Brain Res. 271:73–88 McCormick DA, Steinmetz JE, Thompson RF. 1985. Lesions of the inferior olivary complex cause extinction of the classically conditioned eyeblink response. Brain Res. 359:120–30 McGlinchey-Berroth R. 2000. Eyeblink classical conditioning in amnesia. See WoodruffPak & Steinmetz 2000, Vol. 1, pp. 205–27 McGlinchey-Berroth R, Carrillo MC, Gabrieli JD, Brawn CM, Disterhoft JF. 1997. Impaired trace eyeblink conditioning in bilateral, medial-temporal lobe amnesia. Behav. Neurosci. 111:873–82 McIntosh AR, Schreurs BG. 2000. Functional networks underlying human eyeblink con-

P1: IKH

231

ditioning. See Woodruff-Pak & Steinmetz 2000, Vol. 1, pp. 51–77 McKernan MG, Shinnick-Gallagher P. 1997. Fear conditioning induces a lasting potentiation of synaptic currents in vitro. Nature 390(6660):607–11 Medina JF, Garcia KS, Mauk MD. 2001. A mechanism for savings in the cerebellum. J. Neurosci. 21:4081–89 Miserendino MJ, Sananes CB, Melia KR, Davis M. 1990. Blocking of acquisition but not expression of conditioned fear-potentiated startle by NMDA antagonists in the amygdala. Nature 345(6277):716–18 Moita MA, Rosis S, Zhou Y, LeDoux JE, Blair HT. 2003. Hippocampal place cells acquire location-specific responses to the conditioned stimulus during auditory fear conditioning. Neuron 37(3):485–97 Molchan SE, Sunderland T, McIntosh AR, Herscovitch P, Schreurs BG. 1994. A functional anatomical study of associative learning in humans. Proc. Natl. Acad. Sci. USA 91:8122–26 Muller D, Nikonenko I, Jourdain P, Alberi S. 2002. LTP, memory and structural plasticity. Curr. Mol. Med. 2(7):605–11 Murthy VN, De Camilli 2003. Cell biology of the presynaptic terminal. Annu. Rev. Neurosci. 26:701–28 Perrett SP, Ruiz BP, Mauk MD. 1993. Cerebellar cortex lesions disrupt learning-dependent timing of conditioned eyelid responses. J. Neurosci. 13:1708–18 Pickard L, Noel J, Duckworth JK, Fitzjohn SM, Henley JM, et al. 2001. Transient synaptic activation of NMDA receptors leads to the insertion of native AMPA receptors at hippocampal neuronal plasma membranes. Neuropharmacology 41(6):700–13 Pittenger C, Kandel ER. 2003. In search of general mechanisms for long-lasting plasticity: aplysia and the hippocampus. Philos. Trans. R. Soc. Lond. B. Biol. Sci. 358(1432):757–63 Poremba A, Gabriel M. 2001. Amygdalar efferents initiate auditory thalamic discriminative training-induced neuronal activity. J. Neurosci. 21(1):270–78

23 Nov 2004 18:57

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

232

AR

AR231-PS56-08.tex

FANSELOW




AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

P1: IKH

POULOS

Poulos AM, Robleto K, Thompson RF. 2002. Effects of sensory stimulation and learning on excitability of the interpositus nucleus. Abstr. Soc. Neurosci. 28:79.10 Poulos AM, Thompson RF. 2004. Timing of conditioned responses utilizing electrical stimulation in the region of the interpositus nucleus as a CS. Integ. Physiol. Behav. Sci. In press Quinn JJ, Oommen SS, Morrison GE, Fanselow MS. 2002. Post-training excitotoxic lesions of the dorsal hippocampus attenuate forward trace, backward trace, and delay fear conditioning in a temporally specific manner. Hippocampus 12:495–504 Quirk GJ, Armony JL, LeDoux JE. 1997. Fear conditioning enhances different temporal components of tone-evoked spike trains in auditory cortex and lateral amygdala. Neuron 19(3):613–24 Racine RJ, Wilson DA, Gingell R, Sunderland D. 1986. Long-term potentiation in the interpositus and vestibular nuclei in the rat. Exp. Brain Res. 63:158–62 Rescorla RA. 1974. Effect of inflation of the unconditioned stimulus value following conditioning. J. Comp. Physiol. Psychol. 86:101–6 Richardson R, Fan M. 2002. Behavioral expression of learned fear in rats is appropriate to their age at training, not their age at testing. Anim. Learn. Behav. 30(4):394–404 Roberts AC, Glanzman DL. 2003. Learning in aplysia: looking at synaptic plasticity from both sides. Trends Neurosci. 26(12):662– 70 Rodrigues SM, Farb CR, Bauer EP, LeDoux JE, Schafe GE. 2004. Pavlovian fear conditioning regulates Thr286 autophosphorylation of Ca2+/calmodulin-dependent protein kinase II at lateral amygdala synapses. J. Neurosci. 24(13):3281–88 Rogan MT, Staubli UV, LeDoux JE. 1997. Fear conditioning induces associative longterm potentiation in the amygdala. Nature 390(6660):604–7 Rogers RF, Britton GB, Steinmetz JE. 2001. Learning-related interpositus activity is conserved across species as studied during eye-

blink conditioning in the rat. Brain Res. 905:171–77 Romanski LM, Clugnet MC, Bordi F, LeDoux JE. 1993. Somatosensory and auditory convergence in the lateral nucleus of the amygdala. Behav. Neurosci. 107(3):444–50 Romanski LM, LeDoux JE. 1992. Equipotentiality of thalamo-amygdala and thalamocortico-amygdala circuits in auditory fear conditioning. J. Neurosci. 12(11):4501–9 Rosen JB, Davis M. 1990. Enhancement of electrically elicited startle by amygdaloid stimulation. Physiol. Behav. 48(2):343–49 Rosen JB, Fanselow MS, Young SL, Sitcoske M, Maren S. 1998. Immediate-early gene expression in the amygdala following footshock stress and contextual fear conditioning. Brain Res. 796:132–42 Rosen JB, Hitchcock JM, Sananes CB, Miserendino MJ, Davis M. 1991. A direct projection from the central nucleus of the amygdala to the acoustic startle pathway: anterograde and retrograde tracing studies. Behav. Neurosci. 105(6):817–25 Salvatierra AT, Berry SD. 1989. Scopolamine disruption of septo-hippocampal activity and classical conditioning. Behav. Neurosci. 103:715–21 Schafe GE, LeDoux JE. 2000. Memory consolidation of auditory Pavlovian fear conditioning requires protein synthesis and protein kinase A in the amygdala. J. Neurosci. 20(18):RC96 Schreurs BG, McIntosh AR, Bahro M, Herscovitch P, Sunderland T, Molchan SE. 1997. Lateralization and behavioral correlation of changes in regional cerebral blood flow with classical conditioning of the human eyeblink response. J. Neurophysiol. 77:2153–63 Schultz W, Montgomery EB Jr., Marini R. 1979. Proximal limb movements in response to microstimulation of primate dentate and interpositus nuclei mediated by brain-stem structures. Brain 102:127–46 Seidenbecher T, Laxmi TR, Stork O, Pape HC. 2003. Amygdalar and hippocampal theta rhythm synchronization during fear memory retrieval. Science 301(5634):846–50

23 Nov 2004 18:57

AR

AR231-PS56-08.tex

AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

NEUROSCIENCE OF ASSOCIATIVE LEARNING Shi C, Davis M. 1999. Pain pathways involved in fear conditioning measured with fearpotentiated startle: lesion studies. J. Neurosci. 19(1):420–30 Shibuki K, Gomi H, Chen L, Bao S, Kim JJ, et al. 1996. Deficient cerebellar long-term depression, impaired eyeblink conditioning, and normal motor coordination in GFAP mutant mice. Neuron 16:587–99 Shibuki K, Okada D. 1991. Endogenous nitric oxide release required for long-term synaptic depression in the cerebellum. Nature 349:326–28 Shimizu E, Tang YP, Rampon C, Tsien JZ. 2000. NMDA receptor-dependent synaptic reinforcement as a crucial process for memory consolidation. Science 290(5494):1170– 74 Shinkman PG, Swain RA, Thompson RF. 1996. Classical conditioning with electrical stimulation of cerebellum as both conditioned and unconditioned stimulus. Behav. Neurosci. 110:914–21 Silva AJ. 2003. Molecular and cellular cognitive studies of the role of synaptic plasticity in memory. J. Neurobiol. 54(1):224–37 Solomon PR, Gottfried KE. 1981. The septohippocampal cholinergic system and classical conditioning of the rabbit’s nictitating membrane response. J. Comp. Physiol. Psychol. 95:322–30 Solomon PR, Pomerleau D, Bennett L, James J, Morse DL. 1989. Acquisition of the classically conditioned eyeblink response in humans over the life span. Psychol. Aging 4:34– 41 Solomon PR, Solomon SD, Schaaf EV, Perry HE. 1983. Altered activity in the hippocampus is more detrimental to classical conditioning than removing the structure. Science 220:329–31 Solomon PR, Vander Schaaf ER, Thompson RF, Weisz DJ. 1986. Hippocampus and trace conditioning of the rabbit’s classically conditioned nictitating membrane response. Behav. Neurosci. 100:729–44 Steinmetz JE, Lavond DG, Ivkovich D, Logan CG, Thompson RF. 1992. Disruption of clas-

P1: IKH

233

sical eyelid conditioning after cerebellar lesions: damage to a memory trace system or a simple performance deficit? J. Neurosci. 12:4403–26 Steinmetz JE, Logan CG, Rosen DJ, Thompson JK, Lavond DG, Thompson RF. 1987. Initial localization of the acoustic conditioned stimulus projection system to the cerebellum essential for classical eyelid conditioning. Proc. Natl. Acad. Sci. USA 84:3531– 35 Steinmetz JE, Rosen DJ, Chapman PF, Lavond DG, Thompson RF. 1986. Classical conditioning of the rabbit eyelid response with a mossy-fiber stimulation CS: I. Pontine nuclei and middle cerebellar peduncle stimulation. Behav. Neurosci. 100:878–87 Stote DL, Fanselow MS. 2004. NMDA receptor modulation of incidental learning in Pavlovian context conditioning. Behav. Neurosci. 118:253–57 Swanson LW, Petrovich GD. 1998. What is the amygdala? Trends Neurosci. 21(8):323–31 Tang YP, Shimizu E, Dube GR, Rampon C, Kerchner GA, et al. 1999. Genetic enhancement of learning and memory in mice. Nature 401(6748):63–69 Thompson RF, Krupa DJ. 1994. Organization of memory traces in the mammalian brain. Annu. Rev. Neurosci. 17:519–49 Topka H, Valls-Sole J, Massaquoi SG, Hallett M. 1993. Deficit in classical conditioning in patients with cerebellar degenerations. Brain 116:961–69 Trachtenberg JT, Chen BE, Knott GW, Feng G, Sanes JR, et al. 2002. Long-term in vivo imaging of experience-dependent synaptic plasticity in adult cortex. Nature 420(6917): 788–94 Tracy JA, Britton GB, Steinmetz JE. 2001. Comparison of single unit responses to tone, light, and compound conditioned stimuli during rabbit classical eyeblink conditioning. Neurobiol. Learn. Mem. 76:253–67 Tracy JA, Thompson JK, Krupa DJ, Thompson RF. 1998. Evidence of plasticity in the pontocerebellar conditioned stimulus pathway during classical conditioning of the

23 Nov 2004 18:57

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

234

AR

AR231-PS56-08.tex

FANSELOW




AR231-PS56-08.sgm

LaTeX2e(2002/01/18)

P1: IKH

POULOS

eyeblink response in the rabbit. Behav. Neurosci. 112:267–85 Tsvetkov E, Carlezon WA, Benes FM, Kandel ER, Bolshakov VY. 2002. Fear conditioning occludes LTP-induced presynaptic enhancement of synaptic transmission in the cortical pathway to the lateral amygdala. Neuron 34(2):289–300 Tully T. 1997. Regulation of gene expression and its role in long-term memory and synaptic plasticity. Proc. Natl. Acad. Sci. USA 94(9):4239–41 Walker DL, Toufexis DJ, Davis M. 2003. Role of the bed nucleus of the stria terminalis versus the amygdala in fear, stress, and anxiety. Eur. J. Pharmacol. 463(1–3):199–216 Wang YT, Linden DJ. 2000. Expression of cerebellar long-term depression requires postsynaptic clathrin-mediated endocytosis. Neuron 25:635–47 Wei F, Qiu CS, Liauw J, Robinson DA, Ho N, et al. 2002. Calcium calmodulin-dependent protein kinase IV is required for fear memory. Nat. Neurosci. 5(6):573–79

Woodruff-Pak DS, Steinmetz JE, eds. 2000. Eyeblink Classical Conditioning: Applications in Humans. Norwell, MA: Kluwer Acad. Yaniv D, Schafe GE, LeDoux JE, Richter-Levin G. 2000. Perirhinal cortex and thalamic stimulation induces LTP in different areas of the amygdala. Ann. NY Acad. Sci. 911:474–76 Yeo CH, Hardiman MJ, Glickstein M. 1985. Classical conditioning of the nictitating membrane response of the rabbit. I. Lesions of the cerebellar nuclei. Exp. Brain Res. 60:87–98 Young SL, Bohenek DL, Fanselow MS. 1994. NMDA processes mediate anterograde amnesia of contextual fear conditioning induced by hippocampal damage: immunization against amnesia by contextual preexposure. Behav. Neurosci. 108:19–29 Zhen X, Du W, Romano AG, Friedman E, Harvey JA. 2001. The p38 mitogen-activated protein kinase is involved in associative learning in rabbits. J. Neurosci. 21:5513– 19

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

Annual Review of Psychology Volume 56, 2005

CONTENTS

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Frontispiece—Richard F. Thompson

xviii

PREFATORY In Search of Memory Traces, Richard F. Thompson

1

DECISION MAKING Indeterminacy in Brain and Behavior, Paul W. Glimcher

25

BRAIN IMAGING/COGNITIVE NEUROSCIENCE Models of Brain Function in Neuroimaging, Karl J. Friston

57

MUSIC PERCEPTION Brain Organization for Music Processing, Isabelle Peretz and Robert J. Zatorre

89

SOMESTHETIC AND VESTIBULAR SENSES Vestibular, Proprioceptive, and Haptic Contributions to Spatial Orientation, James R. Lackner and Paul DiZio

115

CONCEPTS AND CATEGORIES Human Category Learning, F. Gregory Ashby and W. Todd Maddox

149

ANIMAL LEARNING AND BEHAVIOR: CLASSICAL Pavlovian Conditioning: A Functional Perspective, Michael Domjan

179

NEUROSCIENCE OF LEARNING The Neuroscience of Mammalian Associative Learning, Michael S. Fanselow and Andrew M. Poulos

207

HUMAN DEVELOPMENT: EMOTIONAL, SOCIAL, AND PERSONALITY Behavioral Inhibition: Linking Biology and Behavior Within a Developmental Framework, Nathan A. Fox, Heather A. Henderson, Peter J. Marshall, Kate E. Nichols, and Melissa A. Ghera

235

BIOLOGICAL AND GENETIC PROCESSES IN DEVELOPMENT Human Development: Biological and Genetic Processes, Irving I. Gottesman and Daniel R. Hanson

263 vii

P1: JRX

December 8, 2004

viii

12:13

Annual Reviews

AR231-FM

CONTENTS

SPECIAL TOPICS IN PSYCHOPATHOLOGY The Psychology and Neurobiology of Suicidal Behavior, Thomas E. Joiner Jr., Jessica S. Brown, and LaRicka R. Wingate

287

DISORDERS OF CHILDHOOD Autism in Infancy and Early Childhood, Fred Volkmar, Kasia Chawarska, and Ami Klin

315

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

CHILD/FAMILY THERAPY Youth Psychotherapy Outcome Research: A Review and Critique of the Evidence Base, John R. Weisz, Amanda Jensen Doss, and Kristin M. Hawley

337

ALTRUISM AND AGGRESSION Prosocial Behavior: Multilevel Perspectives, Louis A. Penner, John F. Dovidio, Jane A. Piliavin, and David A. Schroeder

365

INTERGROUP RELATIONS, STIGMA, STEREOTYPING, PREJUDICE, DISCRIMINATION The Social Psychology of Stigma, Brenda Major and Laurie T. O’Brien

393

PERSONALITY PROCESSES Personality Architecture: Within-Person Structures and Processes, Daniel Cervone

423

PERSONALITY DEVELOPMENT: STABILITY AND CHANGE Personality Development: Stability and Change, Avshalom Caspi, Brent W. Roberts, and Rebecca L. Shiner

453

WORK MOTIVATION Work Motivation Theory and Research at the Dawn of the Twenty-First Century, Gary P. Latham and Craig C. Pinder

485

GROUPS AND TEAMS Teams in Organizations: From Input-Process-Output Models to IMOI Models, Daniel R. Ilgen, John R. Hollenbeck, Michael Johnson, and Dustin Jundt

517

LEADERSHIP Presidential Leadership, George R. Goethals

545

PERSONNEL EVALUATION AND COMPENSATION Personnel Psychology: Performance Evaluation and Pay for Performance, Sara L. Rynes, Barry Gerhart, and Laura Parks

571

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

CONTENTS

ix

PSYCHOPHYSIOLOGICAL DISORDERS AND PSYCHOLOGICAL EFFECTS ON MEDICAL DISORDERS Psychological Approaches to Understanding and Treating Disease-Related Pain, Francis J. Keefe, Amy P. Abernethy, and Lisa C. Campbell

601

TIMELY TOPIC

Annu. Rev. Psychol. 2005.56:207-234. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Psychological Evidence at the Dawn of the Law’s Scientific Age, David L. Faigman and John Monahan

631

INDEXES Subject Index Cumulative Index of Contributing Authors, Volumes 46–56 Cumulative Index of Chapter Titles, Volumes 46–56

ERRATA An online log of corrections to Annual Review of Psychology chapters may be found at http://psych.annualreviews.org/errata.shtml

661 695 700

18 Nov 2004 11:19

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

P1: IKH LaTeX2e(2002/01/18) 10.1146/annurev.psych.55.090902.141532

Annu. Rev. Psychol. 2005. 56:235–62 doi: 10.1146/annurev.psych.55.090902.141532 c 2005 by Annual Reviews. All rights reserved Copyright
First published online as a Review in Advance on August 18, 2004

BEHAVIORAL INHIBITION: Linking Biology and

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Behavior within a Developmental Framework Nathan A. Fox,1 Heather A. Henderson,2 Peter J. Marshall,3 Kate E. Nichols,1 and Melissa M. Ghera1 1

Department of Human Development, University of Maryland, College Park, Maryland 20742; 2Department of Psychology, University of Miami, Coral Gables, Florida 33149; 3Department of Psychology, Temple University, Philadelphia, Pennsylvania 19122; email: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]

Key Words amygdala, fear, anxiety, attention, temperament, social anxiety ■ Abstract Behavioral inhibition refers to a temperament or style of reacting that some infants and young children exhibit when confronted with novel situations or unfamiliar adults or peers. Research on behavioral inhibition has examined the link between this set of behaviors to the neural systems involved in the experience and expression of fear. There are strong parallels between the physiology of behaviorally inhibited children and the activation of physiological systems associated with conditioned and unconditioned fear. Research has examined which caregiving behaviors support the frequency of behavioral inhibition across development, and work on the interface of cognitive processes and behavioral inhibition reveal both how certain cognitive processes moderate behavioral inhibition and how this temperament affects the development of cognition. This research has taken place within a context of the possibility that stable behavioral inhibition may be a risk factor for psychopathology, particularly anxiety disorders in older children. The current chapter reviews these areas of research and provides an integrative account of the broad impact of behavioral inhibition research.

CONTENTS INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . THE BIOLOGY OF BEHAVIORAL INHIBITION . . . . . . . . . . . . . . . . . . . . . . . . . . . Heart Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cortisol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Electroencephalogram, Event-Related Potentials, and Functional Magnetic Resonance Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CONTINUITY AND DISCONTINUITY IN BEHAVIORAL INHIBITION OVER DEVELOPMENT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Longitudinal Studies of Behavioral Inhibition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Role of Caregiving Behaviors and Environments . . . . . . . . . . . . . . . . . . . . . . . 0066-4308/05/0203-0235$14.00

236 238 239 239 240 243 243 245

235

18 Nov 2004 11:19

236

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

P1: IKH

FOX ET AL.

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

COGNITION-EMOTION INTERACTIONS IN BEHAVIORAL INHIBITION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Effects of Emotion on Attentional Orienting . . . . . . . . . . . . . . . . . . . . . . . . . . . Voluntary Attentional Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Attention and the Development of Social Cognition in Behaviorally Inhibited Children . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . BEHAVIORAL INHIBITION AND PSYCHOPATHOLOGY . . . . . . . . . . . . . . . . . . FUTURE DIRECTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . EPILOGUE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

248 248 250 251 252 254 256

INTRODUCTION Behavioral inhibition to the unfamiliar refers to “the child’s initial behavioral reactions to unfamiliar people, objects, and contexts, or challenging situations” (Kagan et al. 1985, p. 53). The initial research reports on behavioral inhibition (Garcia-Coll et al. 1984, Kagan et al. 1988) described a group of toddlers who, by both parent report and observation in the laboratory, avoided unfamiliar events and people. When confronted with such challenges, these children ceased their play behavior and withdrew to the proximity of their caregivers. They remained vigilant of their surroundings during these situations and rarely approached novel objects or unfamiliar people. Our goal in this chapter is to provide a broad overview of the work on behavioral inhibition. After a brief introduction, we begin with a review of the research identifying the biological underpinnings of behavioral inhibition. We next review the longitudinal studies of behavioral inhibition outlining the findings on continuity and discontinuity and identifying factors both within the child and in the environment that may affect these different developmental trajectories. In the third section, we focus on the role of attentional processes as an example of a within-child factor that contributes to patterns of reactivity and regulation among behaviorally inhibited children. Our fourth section details the findings from the developmental psychopathology literature on relations between behavioral inhibition and the heightened risk for general and specific anxiety disorders and more global problems in adjustment. In the final section of this review, we suggest future directions for the study of behavioral inhibition. Research on behavioral inhibition has in many ways provided a model of interdisciplinary integration for other areas of developmental psychology that are now at the forefront of psychological science. Among these areas are the study of links between basic neuroscience and emotional development, examinations of the mutual influences of affect and cognition on behavior, and the identification of precursors to psychopathology in early childhood. There are a number of reasons why the work on behavioral inhibition has been successful in creating these links. First, unlike most previous research on temperament, this work has relied less on questionnaire data (e.g., parent report of temperament) and more on behavioral

18 Nov 2004 11:19

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BEHAVIORAL INHIBITION

P1: IKH

237

description. The focus on behavioral observation and clear descriptions allowed scientists to identify certain responses (e.g., freezing, avoidance) that are similar to those described in animal models of fear or anxiety. The ability to relate behavioral descriptions of humans to descriptions of other animals provided an important initial link between temperament and the neurosciences. A second reason why the work on behavioral inhibition has had such a broad impact in psychology is the emphasis placed on the categorical nature of extreme temperamental behavioral inhibition, particularly by Kagan and colleagues (e.g., Kagan et al. 1984). Kagan appealed to notions in biology and medicine where categories serve an important function in identifying a species or a specific disease state (Kagan 1994). Correct or not, the idea of a categorical trait with its own unique biology and behavioral constellation forged a link with biologists and neuroscientists. Kagan’s reading of the then current work in behavioral neuroscience, including the studies of LeDoux and Davis (e.g., Davis 1986, LeDoux et al. 1988) enhanced his interest in describing the underlying biology of behavioral inhibition. In separate research programs, LeDoux and Davis focused on the amygdala as the brain structure responsible for the enhancement of fear conditioning and the potentiation of fear behaviors (Davis 1992, LeDoux et al. 1988). Building on the psychophysiological data that he and his colleagues had collected, Kagan suggested that individual differences in behavioral inhibition were the result, in part, of an overactive amygdala, creating an enhanced fear response to novelty and unfamiliarity. This attempt to bridge the behavior-neuroscience gap came at a time when the work of Davis and LeDoux was receiving widespread attention, and it facilitated a broader discussion of the ways in which the interplay of biology and behavior could be understood over time and within the context of human development. The initial research on behavioral inhibition was also inherently developmental in approach and theory. Two key observations were highlighted in longitudinal studies of behavioral inhibition. First, across development, children developed a greater repertoire of behaviors in response to novel social situations. Second, while the behavioral manifestation changed somewhat over development, there was significant preservation of individual differences in inhibition. That is, behaviorally inhibited children displayed marked continuity in their distinctive pattern of responding to unfamiliar social and nonsocial stimuli. At the same time, variations in the developmental trajectories of behaviorally inhibited children necessitated broadening the model to include both endogenous and exogenous factors that might influence these different developmental paths, and merited examination of the contextual factors and cognitive processes that may mediate the expression of behavioral inhibition as children get older. A primary source of hypotheses regarding factors influencing different developmental trajectories for behaviorally inhibited children is the model of temperament postulated by Rothbart and colleagues. Rothbart’s model proposed two components: reactivity and self-regulation (Rothbart & Derryberry 1981). Reactivity reflects the infant’s or child’s physiological and behavioral responses to sensory stimuli, and is assessed in terms of the latency and intensity of responding. In the context

18 Nov 2004 11:19

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

238

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

P1: IKH

FOX ET AL.

of behavioral inhibition, individual differences in reactivity are reflected by differences in the strength of the disposition to express particular discrete emotions to novel or unfamiliar stimuli or challenging situations. Regarding regulation, Rothbart’s argument is twofold. First, certain cognitive processes such as voluntary attentional control and response inhibition that develop over infancy and early childhood serve to modulate reactive responses. Rothbart refers to these processes under the broad label of “effortful control.” Second, the child’s style of reactivity may influence the manner in which these processes emerge. Thus, temperament reflects individual differences not only in reactivity, but also in the manner in which effortful control processes are shaped by that reactivity (Derryberry & Rothbart 1997). There are many implications of Rothbart’s model of temperament for the study of behavioral inhibition. Individual differences in reactions to novelty may be identified early in the first year of life and may be described with respect to the disposition to express negative affect (fear or distress) when presented with novel, unfamiliar, or challenging events. Across early childhood, self-regulatory processes (voluntary attention, inhibitory control) will be influenced by this pattern of negative reactivity to novelty and may become biased to focus on negative affect or prevent disengagement from novel, unfamiliar, and often threatening stimuli. The manner in which the development of attention and other executive-function processes underlying effortful control either heighten or dampen the expression of behavioral inhibition in older children is thus of great interest. A further area of study to which research on behavioral inhibition has contributed is the role of early temperament as a risk factor for the development of psychopathology. A number of studies have suggested that by middle to late childhood, socially withdrawn children are rejected and victimized by their peers (Boivin et al. 1995, Hanish & Guerra 2000). As well, by middle childhood, these children are more likely to report feelings of loneliness, low self-esteem, and even depression or anxiety (Hymel et al. 1993, Rubin 1993). Behavioral inhibition may be an antecedent of social withdrawal, which may lead to peer rejection, which in turn may exacerbate inhibited or isolated behavior. Recent evidence has found an increased prevalence of diagnosed anxiety disorders among children identified as behaviorally inhibited (Rosenbaum et al. 1993, Schwartz et al. 1999). Thus, the research on behavioral inhibition has been of great interest to child psychiatrists and clinical child psychologists as they identify early risk factors associated with the onset of anxiety and mood disorders in children.

THE BIOLOGY OF BEHAVIORAL INHIBITION The contrast in behavioral reactions to novelty of inhibited and uninhibited children has been proposed to arise from variation in the excitability of neural circuits of the limbic system (Kagan & Snidman 1991). In particular, this model focuses on the amygdala, which has been implicated in the generation of fear (Davis 1992, 1998). Increased activity of the amygdala (especially the central nucleus)

18 Nov 2004 11:19

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

BEHAVIORAL INHIBITION

P1: IKH

239

would be expected to result in increased activity across response systems that have extensive connections with the central nucleus (see Marshall & Stevenson-Hinde 2001). The work examining this supposition is described below for the cardiac and neuroendocrine response systems, as well as for certain aspects of cortical processing.

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Heart Rate One prediction of the above model is that inhibited children should show consistently lower heart period (HP), corresponding to higher heart rate, and larger decreases in HP (i.e., heart rate acceleration) in response to unfamiliarity, compared with uninhibited children (Kagan 1994). Indeed, HP during quiet or active tasks was significantly negatively correlated with behavioral inhibition at assessments in toddlerhood up to 7.5 years of age (Kagan et al. 1984, Kagan et al. 1988, Reznick et al. 1986). In addition, inhibited children tended to show larger decreases in HP to stressors compared with uninhibited children (Kagan et al. 1988). Low baseline HP was associated with increased behavioral inhibition in an unselected sample of more than one thousand 3-year-old Mauritanian children (Scarpa et al. 1997). Other studies have found mixed results: Marshall & Stevenson-Hinde (1998) found no significant relation between behavioral inhibition and HP in a group of 4.5-year-olds who had been selected for high or low levels of inhibition. However, HP at 4.5 years predicted which of the children would remain inhibited when assessed 2.5 years later at 7 years of age: HP at 4.5 years was significantly lower for highly inhibited children who remained highly inhibited at 7 years, compared with those children who became less inhibited. In an unselected sample of 2-year-olds, Calkins & Fox (1992) found that behavioral inhibition was unrelated to baseline levels of HP, which suggests that large sample sizes or a focus on extremes via the use of selected samples are more likely to yield associations between behavioral inhibition and HP.

Cortisol A perceived threat may activate the hypothalamic-pituitary-adrenal system, with the secretion of the stress hormone cortisol as one of the products of this activation. Salivary cortisol levels have been studied in relation to various aspects of child temperament, including behavioral inhibition (Stansbury & Gunnar 1994). Evidence for an association between cortisol levels and inhibited behavior has been mixed: Some researchers have found that high baseline cortisol levels are associated with behavioral inhibition (e.g., Kagan et al. 1987, Schmidt et al. 1997), but other studies examining the relations between behavioral inhibition and changes in adrenocortical activity in response to stress have been more equivocal. For instance, De Haan and colleagues (De Haan et al. 1998) found home cortisol levels to be associated with more anxious, internalizing behavior in 2-year-olds, but found that an increased cortisol response to starting preschool was associated with more assertive, angry, and aggressive behavior rather than with socially inhibited or anxious

18 Nov 2004 11:19

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

240

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

P1: IKH

FOX ET AL.

behavior. Gunnar (1994) suggested that one reason inhibited children might not show elevated cortisol reactivity during such transitions is that unlike less fearful children, inhibited children tend to avoid the kinds of social and physical activities that would elicit elevations in cortisol. In addition, adrenocortical activity may not necessarily map onto fear-related constructs, but instead cortisol levels may be related to the maintenance or failure of coping strategies. Nachmias et al. (1996) examined cortisol responses of inhibited and uninhibited 18-month-olds to the Ainsworth Strange Situation as well as to a challenging coping episode. Infants who were highly inhibited and insecurely attached showed larger cortisol responses to the Strange Situation and the challenging coping episode compared with children who were highly inhibited but securely attached. The cortisol increase for inhibited-insecure infants was greater than that for the uninhibited infants, whether securely or insecurely attached. In this sense, mothers in secure dyads who have inhibited children may support their children’s strategies for coping with an unfamiliar and/or stressful situation.

Electroencephalogram, Event-Related Potentials, and Functional Magnetic Resonance Imaging A number of studies have examined the relation of behavioral inhibition to electrophysiological measures derived from the electroencephalogram (EEG). The majority of work in this domain has focused on hemispheric asymmetries in EEG activation, although event-related potentials (ERPs) more recently have been used to probe electrophysiological responses to stimuli in inhibited infants and children. In addition, a recent study has incorporated the use of functional magnetic resonance imaging (fMRI) into the study of inhibited temperament. Several studies have related individual differences in approach or withdrawal behaviors in infancy and childhood to patterns of asymmetrical activation in EEG signals recorded over the frontal region of the brain. Fox (1991, 1994) and Davidson (1992) argue that the functional significance of frontal EEG asymmetry may be conceptualized in terms of motivational systems of approach and withdrawal. In this perspective, the left frontal region promotes appetitive, approach-directed emotional responses, while the right frontal region promotes withdrawal-directed responses to perceived aversive stimuli. Hemispheric asymmetries in EEG alpha band activity have been used to probe individual differences in the relative activation of these motivational systems. The use of EEG alpha power in this respect is driven by the fact that when sensory cortex receives incoming stimuli, EEG alpha power over the same part of the cortex is decreased in amplitude (desynchronized). In this sense, alpha power has been used as a proxy for cortical activation, being inversely related to cortical activation (see Marshall et al. 2002). A “right frontal” pattern of EEG asymmetry refers to a pattern of decreased alpha power in electrodes over the right frontal region, relative to the homologous electrodes in the left hemisphere. This right frontal pattern has been taken to indicate activation of the motivational system associated with withdrawal. Indeed, infants who

18 Nov 2004 11:19

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BEHAVIORAL INHIBITION

P1: IKH

241

displayed a pattern of stable right frontal EEG asymmetry across the first two years of life tended to be more inhibited at both 14 and 24 months of age compared with infants who exhibited a pattern of stable left frontal EEG (Fox et al. 1994). Infants who went on to be consistently inhibited up to 4 years of age exhibited greater right frontal EEG asymmetry at 9 and 14 months of age than infants who were to become less inhibited (Fox et al. 2001). In addition, Calkins et al. (1996) found that infants who were selected at 4 months of age for high frequencies of motor behavior and negative affect in response to novel visual and auditory stimuli tended to show right frontal EEG asymmetry at 9 months of age. They also were more behaviorally inhibited at 14 months of age compared with infants who showed either high positive affect or low levels of either positive or negative reactivity at 4 months of age. The best predictor of the tendency to be socially reticent with unfamiliar peers in 4-year-old children was the combination of both temperamental negative affect and right frontal EEG asymmetry (Henderson et al. 2001). Another recent study examined EEG asymmetry in a sample of 10- to 12-yearolds who had been followed since 4 months of age. Children who had exhibited high levels of behavioral inhibition to the unfamiliar in laboratory assessments at ages 14 and 21 months and who had shown high levels of emotional and motor reactivity at 4 months of age were more likely to show right frontal asymmetry in the late-childhood assessment (McManis et al. 2002). In addition to relations with baseline EEG asymmetry, behavioral inhibition is associated with changes in EEG asymmetry during tasks designed to elicit an anxious state. For example, Schmidt et al. (1999) examined changes in the EEG in inhibited and uninhibited children over a task in which the participants had to mentally prepare for giving a speech about their most embarrassing moments. In the end, they were not required to give such a speech, but the EEG patterns during anticipation of this event differed significantly between inhibited and uninhibited children. Relative to the uninhibited children, the inhibited children showed an increase in activation over the right frontal region over the course of the anticipation period. Interestingly, the EEG changes also were coupled with an increase in heart rate over the same period in the inhibited children. We recently examined EEG asymmetries in relation to two different forms of nonsocial behavior in preschoolers during a play session with unfamiliar peers: social reticence and solitary-passive behavior (Henderson et al. 2004). In contrast to solitary-passive children, who occupy themselves with exploratory and constructive activities such as drawing and working on puzzles while in the company of unfamiliar peers, reticent children remain visually focused and oriented toward other children, yet do not join them in their activities. In this sense, reticence is related to behavioral inhibition: Indeed there is evidence for continuity between behavioral inhibition in toddlerhood and reticence in the preschool years (e.g., Rubin et al. 2002). Both reticent and solitary-passive children showed a pattern of resting right frontal EEG asymmetry, which suggests that these different forms of solitude may share a common withdrawal motivation. However, other physiological and

18 Nov 2004 11:19

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

242

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

P1: IKH

FOX ET AL.

behavioral evidence from this study suggested that reticence is associated with a particularly aroused, vigilant physiological profile. For instance, reticent children showed a pattern of increased generalized EEG activation (decreased alpha-band power) across the scalp, which is consistent with Eysenck’s model of increased generalized arousal in introverted individuals (Eysenck & Eysenck 1985), due to tonic differences in the ascending reticular activating system. It is also consistent with animal studies demonstrating that generalized cortical activation can reflect activity in the central nucleus of the amygdala (e.g., Kapp et al. 1994). One promising line of research, which also uses electrophysiological techniques, is the examination of central nervous system responses to stimulation as indexed by ERPs. ERP techniques have the benefit of superlative temporal precision and can give insight into the nature and timing of mental events such as novelty detection and orienting. One relatively recent development in studies of behavioral inhibition is the use of auditory ERPs to examine individual differences in stimulus processing in inhibited and uninhibited children. One question of interest in this respect is the ability of descending projections from limbic structures such as the amygdala to influence stimulus processing at early stages of transmission and stimulus processing. Woodward and colleagues (Woodward et al. 2001) examined the brainstem auditory evoked response in 10- to 12-year-old children who had previously been selected on the basis of high or low affective reactivity in infancy. The high-reactive infants, who had gone on to become more inhibited young children, had a higher amplitude of a certain component of the brainstem auditory evoked response thought to originate in the inferior colliculus. Woodward et al. (2001) interpreted this finding as indicating that projections from the amygdala to the inferior colliculus are more excitable in children with an inhibited, fearful temperament. However, the precise origin of these effects has yet to be elucidated. In a similar vein, we recently examined the mismatch negativity (MMN) in the auditory ERP of both socially withdrawn and more outgoing control children aged 7 to 12 years (Bar-Haim et al. 2003). The MMN indexes a change-detection mechanism in primary auditory cortex, and it is elicited using auditory “oddball” paradigms without specific task demands (Picton et al. 2000). The MMN usually is derived from a comparison of the ERPs to infrequent “deviant” and frequent “standard” auditory stimuli, although its characteristics in terms of morphology and latency vary over infancy and childhood (Cheour et al. 2001). We found that socially withdrawn children have reduced MMN amplitude compared with more outgoing children. Such individual differences in sensory processing either could be a consequence of “top-down” influences by higher affective centers such as the amygdala, or may reflect “bottom-up” differences in early processing that may feed forward to affect the later processing and evaluation of sensory information. Other recent work on auditory ERP responses to novelty in infancy suggests that temperamentally different infants may show different degrees of electrophysiological reactivity to novelty. Infants who have a temperamental tendency to respond to stimulation with high levels of positive affect also show an enhanced ERP response to complex novel stimuli that are interspersed in a train of repetitive tonal

18 Nov 2004 11:19

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BEHAVIORAL INHIBITION

P1: IKH

243

“standard” stimuli, but do not show elements of orienting to stimuli that are only mildly deviant from the standards. In contrast, temperamentally high-negative infants respond with an indication of an orienting response to the mildly deviant stimuli, but with a reduced response to the complex novel stimuli (PJ Marshall, MG Hardin, & NA Fox, in preparation). One explanation for these group differences is that the high-negative and high-positive infants may have differing optimal levels of novelty at which engagement with a novel stimulus is promoted (Berlyne 1960). Although the use of functional neuroimaging techniques such as positron emission tomography (PET) and fMRI in infants and young children presents specific ethical and practical challenges, there is increasing interest in using these techniques, especially fMRI, to probe brain responses to sensory stimulation in older children. One recent study used fMRI to directly assess amygdala activity in response to novel face stimuli in young adults who had been classified as inhibited or uninhibited in the second year of life in Kagan’s longitudinal studies (Schwartz et al. 2003). Participants were familiarized with a set of faces, and then were exposed to a stimulus set that interspersed novel, previously unseen faces among the previously seen, familiar faces. Analyses of the fMRI signal revealed that the adults who as toddlers had been more inhibited showed increased bilateral activation of the amygdala in response to the novel faces compared to adults who initially had been categorized as uninhibited. Although the contemporaneous behavioral profiles of the adults were not assessed, this study provides intriguing evidence of continuity in the reactivity of the physiological systems proposed to underlie an inhibited temperament.

CONTINUITY AND DISCONTINUITY IN BEHAVIORAL INHIBITION OVER DEVELOPMENT Behavioral inhibition and shyness are among the most stable individual differences in the personality development literature, with continuities found throughout early childhood, middle childhood, and adulthood. Despite this relatively high degree of continuity across samples, examination of individual differences in patterns of behavior over time reveals that many children show markedly changed patterns of behavior across the course of childhood. Understanding these patterns of continuity and change has become a vital focus of behavioral inhibition research.

Longitudinal Studies of Behavioral Inhibition More than four decades ago, Kagan focused on the temperamental quality of fearfulness, noting that it showed a remarkably high degree of continuity from toddlerhood through adulthood (Kagan & Moss 1962). Using data from the Fels longitudinal study, Kagan & Moss (1962) reported that the behavioral tendency to express fear and avoidant behaviors to novelty and challenge in the first three years

18 Nov 2004 11:19

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

244

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

P1: IKH

FOX ET AL.

of life showed the highest degree of continuity through adolescence compared to all other behavioral tendencies. Into young adulthood, a remarkable number of the children displaying fearful behaviors at younger ages were described as introverted. In Kagan’s later studies, two cohorts of toddlers were selected based on their extreme reactions to novelty in order to try to tap into these individual differences (e.g., Kagan et al. 1984). These children were followed up later in childhood, when they were observed in play sessions with an unfamiliar peer, and in different unfamiliar situations including a “risk room” in which children’s willingness to engage in mild risk-taking behaviors was measured. The major finding from these longitudinal studies was that the expression of behavioral inhibition, although elicited in different contexts, showed a moderate degree of continuity. Inhibited toddlers who were quiet and restrained tended to be quiet, cautious, and socially reticent children at 7.5 years of age. Conversely, uninhibited toddlers who were social remained talkative and interacted easily with unfamiliar adults and children at 7.5 years (Kagan et al. 1988). The behavioral inhibition index at 7 years correlated significantly with indices of behavioral inhibition at 21 months (Pearson r = 0.67), 4 years (r = 0.54), and 5.5 years (r = 0.57) (Kagan et al. 1988). Other researchers have noted that continuity in behavioral inhibition is lower in samples that are unselected for early patterns of reactivity compared to the selected samples reported on in this original work by Kagan and colleagues. For instance, Stevenson-Hinde & Shouldice (1995) found lower consistency in inhibition toward an unfamiliar adult in an unselected sample from 4.5 to 7 years of age (Pearson r = 0.24) than in a selected sample over the same age range (r = 0.46; Marshall & Stevenson-Hinde 1998). Fox and colleagues (Fox et al. 2001) found similar results in a study of an extreme group of infants who displayed high negative affect and motor reactivity in response to unfamiliar visual and auditory stimuli at 4 months and who were followed up from 9 to 48 months. Approximately half of the infants in this high negative reactivity group continued to show high levels of behavioral inhibition through 24 months of age, and approximately one third of these infants continued to show extreme social reticence during interactions with unfamiliar peers at 48 months (Fox et al. 2001). In the Australian Temperament Project, Sanson and colleagues studied nearly 500 randomly selected infants who were assessed initially at 4 to 8 months of age and followed until they were 5 to 6 years old (Sanson et al. 1996). Maternal reports of shyness showed moderate continuity from infancy to childhood, with continuity being higher at older ages. In addition, similar to Kagan et al. (1988), they found that children at the behavioral extremes showed the greatest degree of continuity in behavior over time. These results have also been replicated in both European and Asian samples. Broberg et al. (1990) used maternal reports and observational measures of temperament in a study of firstborn children in Sweden. An inhibition composite was created that aggregated maternal reports of fear, observer-rated peer noninvolvement in play, and the reversed observer-rated sociability with a stranger score.

18 Nov 2004 11:19

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BEHAVIORAL INHIBITION

P1: IKH

245

Toddlers who were inhibited at 16 months were more likely to be inhibited at 28 and 40 months of age than were uninhibited toddlers. In another unselected Swedish sample, inhibition ratings showed continuity from 21 months to 6 years only in children at the behavioral extremes (Kerr et al. 1994). Reports of continuity have also been found in Chinese studies of temperament (Zhengyan et al. 2003) and in Asendorpf’s (1994) longitudinal study of behavioral inhibition in German children from 4 to 10 years. In a study of Mauritanian children, Scarpa et al. (1995) found that inhibited children at age 3 and age 8 had higher inhibition scores at age 8 and age 11, using questionnaire measures and brief observational ratings. The set of studies finding moderate continuity of inhibition across early and middle childhood has led researchers to assess whether such continuity follows into adulthood. Caspi and colleagues (Caspi et al. 1989) identified shy and reserved children using archival data from the Berkeley Guidance Study (Eichorn 1981). Although there was evidence for continuity of shyness into adulthood, the psychosocial outcomes associated with shyness differed by sex. Shy girls were more likely to follow conventional patterns of marriage, homemaking, and motherhood whereas shy boys were more likely to delay marriage, parenthood, and stable careers, and attain less achievement in their careers. Adults who were previously shy children described themselves as nonassertive and overcontrolled; they reported experiencing few positive emotions and had little desire to influence others. An informant who knew them well described them as less affiliative and less interested in engaging in their surroundings (Caspi et al. 2003). A similar profile of characteristics was reported in young adults who had been behaviorally inhibited as 8- to 12-year-olds (Gest 1997). These young adults reported a less active social life and were less likely to move away from family; men reported experiencing greater emotional distress and negative emotionality. These studies and others have established that behavioral inhibition shows at least moderate continuity across childhood and to a certain extent is also associated with aspects of adult personality. However, given the multitude of influences on development, detailed models are needed that allow for the investigation of specific factors affecting continuity and discontinuity in temperamental tendencies over time. As such, there has been much theoretical discussion regarding the moderating effects of environmental factors on the associations between temperament and outcomes (e.g., Bates 2001, Rothbart & Bates 1998). To date, however, there have been few empirical investigations of these moderating effects. In the following sections, we discuss the possible moderating effects of parenting behaviors and nonparental care environments on patterns of continuity and change in behavioral inhibition.

The Role of Caregiving Behaviors and Environments In early childhood, the most salient environmental influence on the child is the caregiving environment. However, only a few studies have investigated the specific parenting behaviors or styles associated with continuity and discontinuity

18 Nov 2004 11:19

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

246

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

P1: IKH

FOX ET AL.

in behavioral inhibition. The parenting characteristics that have been examined in relation to continuity and discontinuity are acceptance, warmth, sensitivity, responsiveness, control, and overprotection (Park et al. 1997, Wood et al. 2003). Some evidence suggests that parental sensitivity can reduce emotional negativity and perhaps behavioral inhibition by enhancing feelings of self-worth. For example, infants who became less negative from 3 to 9 months of age had interaction patterns with their mothers at 3 months that were more complementary and harmonious (Belsky et al. 1991). Similar findings were observed in a study involving preterm infants (Washington et al. 1986). In contrast to these findings suggesting that sensitive parenting functions to reduce negative reactivity in temperamentally prone infants, Kagan (1994) hypothesized that sensitive parenting behaviors may increase negative reactivity in temperamentally distress-prone infants. He suggested that a firm parenting style including limit setting might help the child cope and lead to reduced inhibition. Consistent with this hypothesis, Park and colleagues (Park et al. 1997) found that more intrusive parenting by mothers across the second and third year led to less inhibited behavior at age 3, after controlling for early emotionality. Less inhibited behavior was also more likely when fathers were less sensitive and affectionate in the second year, and more intrusive and expressive of negative affect in the third year. Park et al. (1997) suggest that sensitive parenting sends the message to the child that it is fine to be who you are, while negative and intrusive controlling behavior sends the message to change. Because Park et al. used global ratings of parenting, it is not possible to discern the specific circumstances or contexts surrounding the parenting behavior. Under certain circumstances, pushing a child to control his anxieties may actually reflect a sensitive awareness instead of a controlling intrusive parenting style (Park et al. 1997). Rubin and colleagues (Rubin et al. 2002) investigated whether the interaction of parenting behaviors and behavioral inhibition at age 2 years explained child characteristics at 4 years of age, either directly or through the moderation of earlier inhibition. A maternal parenting style that consists of overly warm, intrusive, unresponsive, and derisive behavior moderated the concurrent association between shyness and behavioral inhibition at 2 years (Rubin et al. 1997). These associations remained two years later when children were reassessed at 4 years of age (Rubin et al. 2002). Inhibition at 2 years only predicted reticence with unfamiliar peers at 4 years when mothers behaved in a psychologically controlling or derisive manner. Wood and colleagues (Wood et al. 2003) suggest two heuristics for guiding research focused on refining and extending models of parenting and childhood anxiety, including clarifying issues of timing and direction of effects. One heuristic, based on the work of Rubin and colleagues, suggests that parents who are solicitous and overresponsive (in situations in which the child does not need help) may reinforce child anxiety or shyness by rewarding their child’s initial signs of anxiety or distress with parental warmth, and by preventing the child from using and developing self-regulatory skills. In contrast, parents who encourage children to engage in social activities may help prevent their child from developing greater

18 Nov 2004 11:19

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BEHAVIORAL INHIBITION

P1: IKH

247

distress related to social interactions. The other heuristic suggested involves ideas of control and mastery: Children whose parents provide them with opportunities to excel and master their environment are less likely to feel dependent on their parents. The ideas set forth in these heuristics are not inconsistent with the idea that sensitive parenting leads to reduced negative affect. Sensitive and responsive parents are aware that sensitivity and responsiveness needs to change with development and they will encourage their children to explore their world because they are sensitive and responsive to their child’s burgeoning independence. The studies by Park and colleagues, Rubin and colleagues, and others are a good first step to understanding what types of parenting behaviors are associated with maintaining inhibition. Nonparental caregiving may also influence patterns of continuity and discontinuity in behavioral inhibition. Fox and colleagues (Fox et al. 2001) found that infants who showed high negative emotionality at 4 months were more likely to change their behavior and become less inhibited over toddlerhood when they were placed in nonparental caregiving environments for at least 10 hours per week. They suggested several ways that the caregiving environment could lead to different patterns of continuity and discontinuity. For example, children in nonparental care may receive less responsive and sensitive caregiving and/or they may gain more experience interacting with unfamiliar peers at an earlier age compared to their peers who remained inhibited over time. Ahnert & Lamb (2003) have noted that mothers may be less sensitive than care providers in quality day care settings when interacting one-on-one with a child. Even though children may be competing for caregiver attention and be limited in the amount of time they have to interact with the caregiver, this does not necessarily affect the quality of the relationship formed or the quality of the care. The personality of parents who choose to keep their infant at home may explain the higher degree of continuity because a parent’s own anxious or fearful personality may lead to both the decision to keep a child at home and to an overprotective parenting style that contributes to the continuity of inhibition. Alternatively, a working mother may have stress at work above and beyond the stress related to responsibilities in her home, and this stress may affect her parenting style and interactions with her child. Some evidence suggests that mothers of children in childcare took longer to respond to their toddlers’ signals of distress compared to mothers who stay at home. Other observations suggest that toddlers in daycare display more negative behaviors when interacting with their parents outside of daycare (Nelson & Garduque 1991). Taken together, these may explain some of the observed differences in continuity and discontinuity of behavioral inhibition between young children who are in the exclusive care of their parents versus those who spend time in nonparental care. Children who stay at home may be more likely to receive parenting that is more overcontrolling and oversolicitous, whereas children who go to daycare may be more likely to receive parenting that fosters independence. Yet another reason experiences in out-of-home care could promote discontinuity is that children in out-of-home care may simply gain experience and practice

18 Nov 2004 11:19

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

248

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

P1: IKH

FOX ET AL.

interacting with unfamiliar people, therefore reducing their fears. Furman et al. (1979) found that socially withdrawn preschoolers who were given the opportunity to interact with other preschoolers showed an increase in the amount of peer interactions they engaged in compared to preschoolers who were not given an opportunity to interact with others. Greco & Morris (2001) reviewed peerbased interventions for childhood shyness and related behaviors and found that peer-mediated and peer-pairing approaches seem to be effective in treating internalizing behaviors that are related to low social status. Experience with other children combined with mother-child conversations about peers may be especially associated with children’s competence. Laird et al. (1994) studied mother-child conversations about the child’s peers and found that children’s competence was associated with the frequencies of conversations, maternal advice giving, and the presence of emotional themes in conversations.

COGNITION-EMOTION INTERACTIONS IN BEHAVIORAL INHIBITION The tendency to display behavioral inhibition is associated with the experience and expression of moderately high levels of negative emotions including fear, anxiety, and distress (Eisenberg & Fabes 1992). Thus, one of the biggest challenges for children with a history of behavioral inhibition is to learn to modulate or regulate these relatively intense emotional reactions. At both behavioral and neural levels of analysis, a gradual transfer of control over regulation has been described in which infants’ behaviors and emotions are initially governed by their reactive tendencies, thus making them relatively dependent on external sources, such as their caregivers, for regulation (Kopp 1982, Rothbart & Derryberry 1981). Based on an anatomical argument, Panksepp (1998) paralleled the views of Rothbart and Kopp that early in development, infant emotional responses are governed primarily by upward controls from the limbic system. However, with age and cortical development, cognitive control capacities (i.e., response inhibition and attentional control) increase, allowing for greater downward or top-down control over initial reaction tendencies (Panksepp 1998, Rueda et al. 2004).

The Effects of Emotion on Attentional Orienting One of the adaptive functions of the behavioral inhibition system is that it functions to increase vigilance and orienting to environmental cues that could be indicative of threat (LeDoux 2000). This is an example of a bottom-up influence of emotional biases on information processing, where emotional tendencies guide the direction and patterns of engagement/disengagement of attention (Ochsner & Gross 2004). Individual differences in sensory orienting are attributed to the functioning of the posterior attention system, which includes the superior parietal cortex, the

18 Nov 2004 11:19

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BEHAVIORAL INHIBITION

P1: IKH

249

temporal parietal junction, frontal eye fields, and superior colliculus (Corbetta & Shulman 2002). The posterior attention system is functional early in life, and relates to individual differences in orienting responses to novelty, early attentional persistence, duration and latency of orienting, and early state control (Derryberry & Rothbart 1997, Eisenberg et al. 2004). Related to the study of fear and behavioral inhibition, numerous studies have evaluated the effect of extraversion and neuroticism, in combination with anxiety, on various aspects of cognitive functioning. Such studies generally adopt a bottom-up approach to studying emotion-cognitive interactions, as it is hypothesized that individual differences in emotional tendencies will influence the quality of functioning or the ability to implement specific cognitive functions, including attention focusing and shifting. It has long been known that anxiety and stressful states promote attentional narrowing (Easterbrook 1959). Specifically, it has been theorized that attention selectivity facilitates early processing of potential threat, influencing subsequent cognitive and emotional processing (Mathews 1990). However, the exact mechanisms and the direction of these emotion-attention relations are still unclear. Beck et al. (1985) noted that emotional dysfunctions co-occur with activation of a cognitive schema biased toward mood-congruent information. This biased information processing, brought about by anxiety-related schema, leads to better encoding of threat-related information. Consistent with this, when anxious individuals perform an emotional Stroop task, they tend to take longer to report the color of threatening words compared to nonthreatening words. This pattern of performance is thought to be due to increased attention to the threatening word itself, thereby leading to distraction from the color-naming task (Williams et al. 1996). Studies of individual differences on a dot-probe paradigm have led to similar conclusions. Specifically, anxious individuals are quick to detect targets in threatening word locations, possibly due to attention being drawn covertly to the location of the threatening word (MacLeod & Mathews 1988, Wells & Matthews 1994). In detailed analyses of the processes of attentional engagement and disengagement in anxious individuals, recent studies have found that compared to nonanxious individuals, anxious individuals may not attend more quickly or intensely to threatening stimuli, but rather have difficulty disengaging their attention from such stimuli (Derryberry & Reed 1994; E. Fox et al. 2001, 2002). Derryberry and Reed hypothesized that previous findings of differences in performance on emotional Stroop and dot-probe tasks can be attributed to the fact that anxious individuals cannot disengage from the threatening stimuli to attend to other aspects of the task (i.e., word or color), thereby perpetuating the attentional bias to negatively biased, threatening cues (see Derryberry & Reed 2002). Thus, children with a bias toward anxious, fearful reaction tendencies face the challenge of learning to intentionally control their attention at several different levels including the shifting of visual attention away from distressing stimuli and situations and cognitively reappraising stimuli and situations that elicit distress.

18 Nov 2004 11:19

250

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

P1: IKH

FOX ET AL.

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Voluntary Attentional Control Whereas individual differences in attentional orienting have been related to the functioning of the posterior attention system, voluntary control of attention is considered a higher-order cognitive process associated with the functioning of a more anterior system of attention including the anterior cingulate cortex and the lateral prefrontal cortex (Posner & DiGirolamo 2000). The emergence of voluntary control over attention, and specifically the ability to flexibly focus and shift attention, contributes to the developing system of behaviors or responses that allow for greater self-regulation of thought, behavior, and emotion (Posner & Rothbart 1998). The relation between the flexible and intentional employment of attention and the regulation of emotion is apparent early in life and characterizes many of the early interactions that take place between caregivers and infants. During adult-infant interactions, adults engage and disengage infants’ attention in order to manage the infants’ levels of arousal. States of engaged attention between infants and their caregivers tend to be associated with play, states of joy, and general states of positive affect (Gottman et al. 1997). In addition to engaging infants’ attention, adults also tend to be sensitive to infants’ needs to disengage their attention in order to dampen or reduce levels of arousal. When attention is engaged, arousal is heightened, and by disengaging attention, adults give infants the opportunity to dampen their levels of arousal. Early on, infants learn to use such strategies to manage their own levels of arousal. When parents respond contingently to their infants’ needs to disengage and re-engage interactions, infants learn about the efficacy of attentional control as a means of self-regulation (Gottman et al. 1997). The relation between attentional control and self-regulation is supported by the fact that these individual differences have been associated with differences in temperamental reactivity and regulation (Johnson et al. 1991, Rothbart et al. 1994). These findings suggest that the development of attentional control over the first several years of life may provide children with an important source of regulation over their reactive temperamental tendencies. For temperamentally fearful children, the shifting of attention to a different aspect of a situation, or distracting oneself, may provide an effective means of regulating emotional distress. Our observations suggest that disengaging attention from aspects of unfamiliar situations may be particularly challenging for behaviorally inhibited children. In unfamiliar social situations, temperamentally fearful children face the competing challenges of managing feelings of anxiety and interacting with peers. These challenges are particularly salient for a subtype of socially withdrawn children referred to as socially reticent, who spend their time primarily watching other children when put in a group of unfamiliar peers (Coplan et al. 1994, Henderson et al. 2004). These children tend to hover on the fringe of social activity, display behavioral signs of anxiety, carefully watch the other children, and remain unengaged in any other activities. These children appear to be unable to disengage their attention from the other children and make few attempts to join group activities. This fixation appears to be not only ineffective in reducing wariness, but it also

18 Nov 2004 11:19

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BEHAVIORAL INHIBITION

P1: IKH

251

may increase wariness over the course of the play period. These observations are consistent with others’ hypotheses that across the life span, individuals who are biased toward the experience of high levels of negative emotion, including fear and anxiety, and who are unable to employ attentional mechanisms in order to regulate their emotions, would be easily overwhelmed by negative emotion (Derryberry & Rothbart 1988; Eisenberg et al. 1998; Rothbart et al. 1992, 2004). Aspects of attentional control that contribute to the regulation of negative emotional states include attention shifting, attention focusing, and cognitively manipulating one’s assessment of a situation (Eisenberg et al. 1998, Ochsner & Gross 2004). Shifting attention away from a fear-inducing stimulus, or cognitively reframing environmental stimuli that might be interpreted as distressing, appear particularly important for the regulation of shyness (Eisenberg et al. 1998). Adults who report high levels of shyness also report being poor at shifting and focusing attention (Eisenberg et al. 1995). Similarly, in children, Eisenberg et al. (1998) found that teachers’ reports of shyness are negatively correlated with attention shifting at school. In adult populations, it has been found that anxious individuals with good attentional control were better able to shift their attention away from threatening stimuli compared with anxious individuals with poor attentional control, thereby showing that skilled control of voluntary attention may provide anxious individuals with an important form of self-regulation (Derryberry & Reed 2002). Similarly, for children, Eisenberg et al. (1998) reported a significant interaction effect between shyness and attention shifting in the prediction of internalizing emotions, such that internalizing emotions were positively related to parent reports of shyness, but only for children who were relatively low in attention shifting. Shyness was not predictive of internalizing emotions for children who were moderate or high in attention-shifting abilities. Therefore, the ability to shift attention may moderate the associations between shyness and the development of social anxiety and more global internalizing difficulties.

Attention and the Development of Social Cognition in Behaviorally Inhibited Children Social-cognitive models of social adjustment emphasize the cognitive processes or mental steps children engage in before enacting a social behavior. These steps are (a) the encoding of situational cues, (b) the representation and interpretation of those cues, (c) a mental search for possible responses to the situation, and (d) the selection of a response. Individual differences in the processing of information at any of these stages are believed to provide a mechanism through which individual differences in social adjustment develop (Dodge 1989, Ladd & Mize 1983, Rubin & Krasnor 1986). Given the effects of temperament or emotional biases on the direction and control of children’s attention, over time these biases may influence several stages of social information processing, including the encoding and interpretation of social and emotional cues. In turn, differences in social

18 Nov 2004 11:19

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

252

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

P1: IKH

FOX ET AL.

information processing may affect patterns of continuity and discontinuity in behavioral inhibition across childhood and into adolescence. Behaviorally inhibited children interact with others in relatively ineffective ways, and others’ reactions to them likely influence their social cognitions over time. Although behaviorally inhibited children tend to use nonassertive strategies during interactions with peers, they are more likely than are their peers to have their requests refused (Rubin & Borwick 1984). Such findings suggest that in early childhood, social withdrawal is associated with the experience of poor peer relationships (e.g., Rubin 1985). Over time, the experience of social failure may influence children’s perceptions of social situations and their attributions with regard to social success and failure. Specifically, children may begin to interpret otherwise ambiguous social situations as threatening and believe that poor social outcomes are a result of internal causes (Goetz & Dweck 1980, Rubin et al. 1998). The relations between emotional biases and social information processing have been demonstrated in studies of clinical populations. Emotional disorders including depression, anxiety, and phobias are associated with cognitive biases, or the selective processing of emotion-relevant information (Mineka & Tomarken 1989). Attentional biases, as reviewed above, appear to affect the nature of the information that initially is attended to among anxious children and adults. However, emotions, and anxiety in particular, appear to influence later stages of information processing as well, including the interpretation of otherwise ambiguous situations. Specifically, anxious children and adults tend to interpret ambiguous stimuli or situations as disproportionately negative or threatening (Vasey & MacLeod 2001). For example, in a sample of 7- to 9-year-old children, self-reported trait anxiety was positively associated with the number of threatening interpretations children gave for a series of homophones with both a threatening and neutral meaning (Hadwin et al. 1997). Beyond the interpretation of single words or images, anxiety is associated with more global interpretive biases in the context of scenarios involving social interactions. Negative or anxious social schemas have been evaluated by presenting children or adults with ambiguous social scenarios and asking for their interpretations. For example, Chorpita et al. (1996) assessed 9- to 13-year-old children’s interpretations and behavioral plans regarding ambiguous situations. Children’s trait anxiety scores significantly predicted the total proportion of both anxious interpretations and anxious plans. Similarly, socially anxious adolescents predict more socially threatening outcomes for hypothetical social scenarios compared to their nonanxious peers (Magnusdottir & Smari 1999).

BEHAVIORAL INHIBITION AND PSYCHOPATHOLOGY The pattern of anxious behaviors, social withdrawal, negative affect, and low selfesteem, all reported in the developmental literature as characteristic of behavioral inhibition, are symptoms also used to diagnose certain anxiety and mood

18 Nov 2004 11:19

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BEHAVIORAL INHIBITION

P1: IKH

253

disorders. A good deal of work on the relation between behavioral inhibition and psychopathology has been done by Biederman, Rosenbaum, and their colleagues. In their first study, Rosenbaum et al. (1988) assessed 56 Caucasian children ages 2 to 7 whose parents were undergoing treatment for one of three possible diagnoses: (a) panic disorder with or without agoraphobia (PD-AG), (b) comorbid PD with major depressive disorder (PD-MDD), or (c) major depressive disorder alone (MDD). These groups were compared to a “control” group of children whose parents did not have PD or MDD. This control group consisted of siblings of children in treatment for attention deficit disorder or children of parents in treatment for other disorders (i.e., tobacco dependence, obesity, or generalized anxiety disorder). Researchers found that children of parents with PD, with or without comorbid MDD, had longer latencies to speak and fewer spontaneous comments than did children from the control group, whereas children of MDD parents did not differ from either group. Results also indicated a higher incidence of behavioral inhibition in children of parents with PD-AG, with or without MDD, compared to children from the control group. The authors suggested that the increased prevalence of behavioral inhibition in children of parents with PD (i.e., anxiety disorders) suggests a familial or biological tie between psychopathology and inhibition. Additional studies by this research group examined the prevalence and familial loadings of psychopathology in children with and without behavioral inhibition in nonclinical samples. Biederman et al. (1990) used a subset of the sample from their 1988 study whose parents were being treated for PD-AG (classified as the Massachusetts General Hospital high-risk sample) and compared these children to inhibited and uninhibited Caucasian children (ages 7 to 8 years) from Kagan’s longitudinal study of behavioral inhibition. Both groups were also compared to healthy controls (ages 4 to 10 years) who were not classified into temperament categories. Results revealed that behaviorally inhibited children in the Massachusetts General Hospital high-risk sample were significantly more likely to have multiple diagnoses (>4), two or more anxiety disorders, overanxious disorder, and oppositional disorder when compared to healthy controls, although comparisons to not-inhibited children failed to reach statistical significance. Additionally, although not statistically significant, rates of major depression and attention deficit disorder were also higher in inhibited children as compared to not-inhibited children or healthy controls. In Kagan’s sample that had been selected for comparison as a nonclinical group, rates of oppositional disorder were found to be significantly lower in inhibited children as compared to uninhibited children, whereas rates of phobic disorder were higher in inhibited versus uninhibited children. Based on these results, the authors concluded that behavioral inhibition was associated with childhood anxiety disorders. Other studies also have found links between inhibition in adolescents and young adults and various disorders. Using behavioral assessments, Schwartz et al. (1999) assessed a group of adolescents who had been previously classified as inhibited or uninhibited at 2 years of age. At 13 years of age, participants were observed in a laboratory battery of physiological, behavioral, and cognitive procedures.

18 Nov 2004 11:19

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

254

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

P1: IKH

FOX ET AL.

Participants were also interviewed directly by a psychologist to assess past and current anxiety symptoms. No differences were found between inhibited and uninhibited teens with respect to specific fears, separation anxiety, or performance anxiety, although the two groups did differ in respect to social anxiety: Sixtyone percent of teens who had been inhibited as toddlers reported social anxiety symptoms, compared to 27% of adolescents who had been uninhibited earlier in life. Additional studies have relied on self-report measures in an attempt to further elucidate the relationship between inhibition and psychopathology. The Australian Temperament Project data revealed associations between shyness and internalizing problems at 5 to 6 years (Sanson et al. 1996) and anxiety at 13 to 14 years of age (Prior et al. 2000). In the latter study, participants who were rated as shy from infancy onward had a higher incidence of later anxiety problems, with persistently shy children being at increased risk compared to children who were never shy or whose shyness rating had changed over time. Forty-two percent of children rated at the higher end on a shyness scale had anxiety problems in adolescence, compared to 11% who never were rated as shy. Muris et al. (1999; see also Muris et al. 2001) conducted a study that examined the relation between self-reported behavioral inhibition and psychopathology symptoms in high school students. Adolescents who rated themselves high on inhibition also reported increased symptoms of anxiety, worry, and depression. In a similar study, Hayward et al. (1998) administered a retrospective self-report of inhibition (in childhood) to high school students. Results indicated social avoidance in childhood to be predictive of the onset of social phobia during high school, but unrelated to depression. Fearfulness in childhood also was found to increase the risk for later diagnoses of social phobia and depression. Research has demonstrated that stable behavioral inhibition is a risk factor in the development of anxiety disorders. Existing research has focused on the link between inhibition and the development of anxiety; however, the relation between inhibition and externalizing behaviors has not been as extensively investigated. It has been hypothesized that behavioral inhibition may actually have a protective effect against externalizing disorders (Hirshfeld-Becker et al. 2003, Kerr et al. 1997). Indeed, not only does anxiety protect children from becoming delinquent adolescents (Tremblay et al. 1994), but behavioral inhibition also has been proposed to be a protective factor against the development of delinquency (Kerr et al. 1997).

FUTURE DIRECTIONS With the approach of 20 years of study of behavioral inhibition, a number of issues present challenges to current researchers in this area. These include enhanced characterization of the early manifestations of the inhibited temperament pattern, an increased understanding of the role of cognitive processes in moderating

18 Nov 2004 11:19

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BEHAVIORAL INHIBITION

P1: IKH

255

behavioral inhibition, and an expanded understanding of the manner in which behavioral inhibition is displayed in older children and adolescents. This latter point is critical for understanding possible mechanisms to the etiology of psychopathology among behaviorally inhibited children. Indeed, the boundary between risk and disorder may be apparent in the behavior of behaviorally inhibited older children. The assessment of behavioral inhibition has differed for infants, young children, and older children. The infant temperament pattern has been identified by presenting novel auditory and visual stimuli to infants and selecting those infants who reflect the top 10% to 15% of the population in motor arousal and negative affect to these stimuli. Behavioral inhibition has been assessed in toddlers by observing their reactions to novel objects (e.g., a moving robot or car) or unfamiliar adults. Aggregate measures of inhibition have included wariness and avoidance of both the nonsocial novel objects and social situations. As children get older, there has been a shift in focus of the assessment, with a greater emphasis on the child’s response to unfamiliar peers. Indeed, much of the work on behavioral inhibition in preschool and older children has been in situations with unfamiliar (and sometimes familiar) peers. Only one study examined the difference in response to novel social and nonsocial stimuli in young children (Rubin et al. 1997). There is a need to characterize the specific stimuli that challenge the young child and elicit behavioral inhibition reliably (Stevenson-Hinde 1989). It may be that there are different subgroups of children: some who show overall avoidance or wariness to any novel event (social or nonsocial) and others who show responses to a distinct class of stimuli. These fine distinctions will be helpful in understanding and characterizing early in life the multiple trajectories that behavioral inhibition may take over the life span. There has been a growing recognition of the importance of the role of specific cognitive processes in the control or modulation of emotion. The interface of cognition and emotion is salient for understanding the different developmental trajectories seen with behaviorally inhibited children. For example, there is great interest in the manner in which both inhibitory control and attentional processes facilitate emotion control. Recent work by Henderson (2004) demonstrates the manner in which inhibitory control interacts with child temperament in a somewhat counterintuitive fashion. She found that behaviorally inhibited children who demonstrated heightened inhibitory control in a delay task were more likely to exhibit heightened social reticence than those inhibited children who did not demonstrate heightened inhibitory control. Because inhibitory control is often viewed as a positive facilitative process for emotion regulation, these results at first appear counter to expectations. Upon further reflection, though, they signal the manner in which certain cognitive processes may actually enhance temperamental dispositions. Future studies of the interaction of temperament and these cognitive processes will further elucidate the importance of certain individual differences in understanding social behavior. Most of the work in behavioral inhibition has focused on young or school-age children. However, children who are avoidant of social interaction may be at risk

18 Nov 2004 11:19

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

256

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

P1: IKH

FOX ET AL.

for social isolation during the critical years of adolescence. The consequences of this social isolation may result in increased feelings of loneliness, low self-worth, and perhaps the emergence of behavior patterns leading to psychopathology. At the moment, there are preciously few data on the role of temperament as a predictor of adolescent psychological development. Such studies would be highly informative for understanding the dramatic increase in anxiety and mood disorders found during this period of development. Finally, the importance of understanding the role that culture plays in behavioral inhibition has become increasingly recognized. The notion that there might be differences in the distribution of certain temperaments as a function of ethnicity, or that certain cultures might hold different perceptions of these temperamental dispositions, only recently has been explored. For example, Rubin (1998) has suggested that behavioral inhibition in North American and Western European countries is different from behavioral inhibition in Chinese culture because the cultural meaning of shyness and the social responses to children with shyness differ in the two areas. Chen et al. (1995, 1999) found that inhibited behavior is related to positive adolescent adjustment, including teacher-assessed competence, academic achievement, and leadership, and that behavioral inhibition is not seen as maladaptive in China. In a related study, in which toddlers from China and Canada were observed, Chen et al. (1998) found that Chinese children were more inhibited than Canadian children. They used observational methods that focused on the child’s latencies to approach novel objects and unfamiliar individuals, and durations of behaviors such as the child’s proximity to his or her mother. Differences in parenting behaviors have been used to explain some of the differences in behavioral inhibition between cultures. Parenting practices and childrearing beliefs are important factors that may mediate cultural influences on child development (Super & Harkness 1986). Chen et al. (1998) found that in the Chinese sample, mothers’ warmth and acceptance was positively associated with inhibition, and maternal rejection and punishment was negatively associated with inhibition. The opposite was found in the Canadian sample. Rubin (1998) suggests that the meaning associated with a social behavior is a function of the context where the behavior is produced. If a behavior is perceived as maladaptive in a culture, then parents will discourage its development, while encouraging the development of adaptive behaviors.

EPILOGUE The study of behavioral inhibition in some ways is like a complex puzzle. Each of the pieces by themselves tells an interesting story, but together they provide a broad view of the phenomenon. Study of behavioral inhibition has ranged from understanding the underlying biology of the fear system, to the effects of this temperament on parenting behavior, and to research on the development of certain forms of psychopathology. Each piece of this puzzle or area related to this

18 Nov 2004 11:19

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

BEHAVIORAL INHIBITION

P1: IKH

257

phenomenon has provided rich data on the psychology of behavioral inhibition. Ultimately, however, the union of these pieces will afford the greatest explanatory power. The Annual Review of Psychology is online at http://psych.annualreviews.org

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

LITERATURE CITED Ahnert L, Lamb ME. 2003. Shared care: establishing a balance between home and child care settings. Child Dev. 74:1044–49 Asendorpf JB. 1994. The malleability of behavior inhibition: a study of individual developmental functions. Dev. Psychol. 30:912– 19 Bar-Haim Y, Marshall PJ, Fox NA, Schorr E, Gordon -Salant S. 2003. Mismatch negativity in socially withdrawn children. Biol. Psychiatry 54:17–24 Bates JE. 2001. Adjustment style in childhood as a product of parenting and temperament. In Temperament in Context, ed. TD Wachs, GA Kohnstamm, pp. 173–200. Mahwah, NJ: Erlbaum Beck AT, Emery G, Greenberg RL. 1985. Anxiety Disorders and Phobias: A Cognitive Perspective. New York: Basic Books Belsky J, Fish M, Isabella R. 1991. Continuity and discontinuity in infant negative and positive emotionality: family antecedents and attachment consequences. Dev. Psychol. 27:421–31 Berlyne DE. 1960. Conflict, Arousal, and Curiosity. New York: McGraw Hill Biederman J, Rosenbaum JF, Hirshfeld DR, Faraone SV, Bolduc EA, et al. 1990. Psychiatric correlates of behavioral inhibition in young children of parents with and without psychiatric disorders. Arch. Gen. Psychiatry 47:21–26 Boivin M, Hymel S, Bukowski W. 1995. The roles of social withdrawal, peer rejection, and victimization by peers in predicting loneliness and depressed mood in childhood. Dev. Psychopathol. 7:765–85 Broberg A, Lamb ME, Hwang P. 1990. Inhibition: its stability and correlates in sixteen-

to forty-month-old children. Child Dev. 61: 1153–63 Calkins SD, Fox NA. 1992. The relations among infant temperament, security of attachment, and behavioral inhibition at twenty-four months. Child Dev. 63:1456–72 Calkins SD, Fox NA, Marshall TR. 1996. Behavioral and physiological antecedents of inhibited and uninhibited behavior. Child Dev. 67:523–40 Caspi A, Bem DJ, Elder GH. 1989. Continuities and consequences of interactional styles across the life course. J. Pers. 57:375–406 Caspi A, Harrington HL, Milne B, Amell JW, Theodore RF, Moffitt TE. 2003. Children’s behavioral styles at age 3 are linked to their adult personality traits at age 26. J. Pers. 71:495–513 Chen X, Hastings PD, Rubin KH. 1998. Childrearing attitudes and behavioral inhibition in Chinese and Canadian toddlers: a crosscultural study. Dev. Psychol. 34:677–86 Chen X, Rubin KH, Li B. 1999. Adolescent outcomes of social functioning in Chinese children. Int. J. Behav. Dev. 23:199–223 Chen X, Rubin KH, Li Z. 1995. Social functioning and adjustment in Chinese children: a longitudinal study. Dev. Psychol. 31:531– 39 Cheour M, Korpilahti P, Martynova O, Lang AH. 2001. Mismatch negativity and late discriminative negativity in investigating speech perception and learning in children and infants. Audiol. Neuro-otol. 6:2–11 Chorpita BF, Albano AM, Barlow DH. 1996. Cognitive processing in children: relation to anxiety and family influences. J. Clin. Child Psychol. 25:170–76 Coplan RJ, Rubin KH, Fox NA, Calkins

18 Nov 2004 11:19

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

258

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

P1: IKH

FOX ET AL.

SD, Stewart SL. 1994. Being alone, playing alone, and acting alone: distinguishing among reticence and passive- and activesolitude in young children. Child Dev. 65: 129–37 Corbetta M, Shulman GL. 2002. Control of goal-directed and stimulus-driven attention in the brain. Nat. Neurosci. Rev. 3:201–15 Davidson RJ. 1992. Anterior cerebral asymmetry and the nature of emotion. Brain Cogn. 20:125–51 Davis M. 1986. Pharmacological and anatomical analysis of fear conditioning using the fear-potentiated startle paradigm. Behav. Neurosci. 100:814–24 Davis M. 1992. The role of the amygdala in fear and anxiety. Annu. Rev. Neurosci. 15:353–75 Davis M. 1998. Are different parts of the extended amygdala involved in fear versus anxiety? Biol. Psychiatry 44:1239–47 de Haan M, Gunnar MR, Tout K, Hart J, Stansbury K. 1998. Familiar and novel contexts yield different associations between cortisol and behavior among 2-year-old children. Dev. Psychobiol. 33:93–101 Derryberry D, Reed MA. 1994. Temperament and attention: orienting toward and away from positive and negative signals. J. Personal. Soc. Psychol. 66:1128–39 Derryberry D, Reed MA. 2002. Anxietyrelated attentional biases and their regulation by attentional control. J. Abnorm. Psychol. 111:225–36 Derryberry D, Rothbart MK. 1988. Arousal, affect, and attention as components of temperament. J. Pers. Soc. Psychol. 55:958–66 Derryberry D, Rothbart MK. 1997. Reactive and effortful processes in the organization of temperament. Dev. Psychopathol. 9:633–52 Dodge KA. 1989. Coordinating responses to aversive stimuli: introduction to a special section on the development of emotion regulation. Dev. Psychol. 25:339–42 Easterbrook JA. 1959. The effect of emotion on cue utilization and the organization of behaviour. Psychol. Rev. 66:183–201 Eichorn D. 1981. Samples and procedures. In Present and Past in Middle Life, ed. D. Ei-

chorn, JA Clausen, N Haan, MP Honzik, P Mussen, pp. 33–51. New York: Academic Eisenberg N, Fabes RA. 1992. Emotion, regulation, and the development of social competence. In Review of Personality and Social Psychology: Vol. 14. Emotion and Social Behavior, ed. MS Clark, pp. 119–50. Newbury Park, CA: Sage Eisenberg N, Fabes RA, Murphy B. 1995. The relations of shyness and low sociability to regulation and emotionality. J. Personal. Soc. Psychol. 68:505–17 Eisenberg N, Shepard SA, Fabes RA, Murphy BC, Guthrie IK. 1998. Shyness and children’s emotionality, regulation, and coping: contemporaneous, longitudinal, and acrosscontext relations. Child Dev. 69:767–90 Eisenberg N, Smith CL, Sadovsky A, Spinrad TL. 2004. Effortful control: relations with emotion regulation, adjustment, and socialization in childhood. In Handbook of Selfregulation: Research, Theory, and Applications, ed. RF Baumeister, KD Vohs, pp. 259– 82. New York: Guilford Eysenck HJ, Eysenck MW. 1985. Personality and Individual Differences: A Natural Science Approach. New York: Plenum Fox E, Russo R, Bowles R, Dutton K. 2001. Do threatening stimuli draw or hold attention in subclinical anxiety? J. Exp. Psychol. Gen. 130:681–700 Fox E, Russo R, Dutton K. 2002. Attentional bias for threat: evidence for delayed disengagement from emotional faces. Cogn. Emot. 16:355–79 Fox NA. 1991. If it’s not left, it’s right. Electroencephalograph asymmetry and the development of emotion. Am. Psychol. 46:863– 72 Fox NA. 1994. Dynamic cerebral processes underlying emotion regulation. In The Development of Emotion Regulation: Biological and Behavioral Considerations, ed. NA Fox. Monogr. Soc. Res. Child Dev. 59(2–3, Ser. No. 240):152–66 Fox NA, Calkins SD, Bell MA. 1994. Neural plasticity and development in the first two years of life: evidence from cognitive and

18 Nov 2004 11:19

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BEHAVIORAL INHIBITION socioemotional domains of research. Dev. Psychopathol. 6:677–96 Fox NA, Henderson HA, Rubin KH, Calkins SD, Schmidt LA. 2001. Continuity and discontinuity of behavioral inhibition and exuberance: psychophysiological and behavioral influences across the first four years of life. Child Dev. 72:1–21 Furman W, Rahe DF, Hartup WW. 1979. Rehabilitation of socially withdrawn preschool children through mixed-age and same-age socialization. Child Dev. 50:915–22 Garcia-Coll C, Kagan J, Reznick JS. 1984. Behavioral inhibition in young children. Child Dev. 55:1005–19 Gest SD. 1997. Behavioral inhibition: stability and associations with adaptation from childhood to early adulthood. J. Personal. Soc. Psychol. 72:467–75 Goetz TE, Dweck CS. 1980. Learned helplessness in social situations. J. Personal. Soc. Psychol. 39:246–55 Gottman JM, Katz LF, Hooven C. 1997. Metaemotion: How Families Communicate Emotionally. Mahwah, NJ: Erlbaum Greco LA, Morris TL. 2001. Treating childhood shyness and related behavior: empirically evaluated approaches to promote positive social interactions. Clin. Child Fam. Psychol. Rev. 4:299–318 Gunnar MR. 1994. Psychoendocrine studies of temperament and stress in early childhood: expanding current models. In Temperament: Individual Differences at the Interface of Biology and Behavior, ed. JE Bates, TD Wachs, pp. 175–98. Washington, DC: Am. Psychol. Assoc. Hadwin J, Frost S, French CC, Richards A. 1997. Cognitive processing and trait anxiety in typically developing children: evidence for an interpretation bias. J. Abnorm. Psychol. 106:486–90 Hanish LD, Guerra NG. 2000. Predictors of peer victimization among urban youth. Soc. Dev. 9:521–43 Hayward C, Killen JD, Kraemer HC, Taylor CB. 1998. Linking self-reported childhood behavioral inhibition to adolescent so-

P1: IKH

259

cial phobia. J. Am. Acad. Child Adolesc. Psychiatry 37:1308–16 Henderson HA, Fox NA, Rubin KH. 2001. Temperamental contributions to social behavior: the moderating roles of frontal EEG asymmetry and gender. J. Am. Acad. Child Adolesc. Psychiatry 40:68–74 Henderson HA, Marshall PJ, Fox NA, Rubin KH. 2004. Psychophysiological and behavioral evidence for varying forms and functions of nonsocial behavior in preschoolers. Child Dev. 75:251–63 Hirshfeld-Becker DR, Biederman J, Calltharp S, Rosenbaum ED, Faraone SV, Rosenbaum JF. 2003. Behavioral inhibition and disinhibition as hypothesized precursors to psychopathology: implications for pediatric bipolar disorder. Biol. Psychiatry 53:985–99 Hymel S, Bowker A, Woody E. 1993. Aggressive versus withdrawn unpopular children: variations in peer and self-perceptions in multiple domains. Child Dev. 64:879–96 Johnson MH, Posner MI, Rothbart MK. 1991. Components of visual orienting in early infancy: contingency learning, anticipatory looking and disengaging. J. Cogn. Neurosci. 3:335–44 Kagan J. 1994. Galen’s Prophecy. New York: Basic Books Kagan J, Moss HA. 1962. Birth to Maturity: A Study in Psychological Development. London: Wiley Kagan J, Reznick JS, Clarke C, Snidman N, Garcia-Coll C. 1984. Behavioral inhibition to the unfamiliar. Child Dev. 55:2212–25 Kagan J, Reznick JS, Snidman N. 1985. Temperamental inhibition in early childhood. In The Study of Temperament: Changes, Continuities, and Challenges, ed. R. Plomin, J Dunn, pp. 53–65. Hillsdale, NJ: Erlbaum Kagan J, Reznick JS, Snidman N. 1987. The physiology and psychology of behavioral inhibition in children. Child Dev. 58:1459– 73 Kagan J, Reznick JS, Snidman N, Gibbons J, Johnson MO. 1988. Childhood derivatives of inhibition and lack of inhibition to the unfamiliar. Child Dev. 59:1580–89

18 Nov 2004 11:19

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

260

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

P1: IKH

FOX ET AL.

Kagan J, Snidman N. 1991. Temperamental factors in human development. Am. Psychol. 46:856–62 Kapp BSS, William F, Whalen PJ. 1994. Effects of electrical stimulation of the amygdaloid central nucleus on neocortical arousal in the rabbit. Behav. Neurosci. 108:81–93 Kerr M, Lambert WW, Stattin H, KlackenbergLarsson I. 1994. Stability of inhibition in a Swedish longitudinal sample. Child Dev. 65:138–46 Kerr M, Tremblay RE, Pagani L, Vitaro F. 1997. Boys’ behavioral inhibition and the risk of later delinquency. Arch. Gen. Psychiatry 54:809–16 Kopp CB. 1982. Antecedents of self-regulation: a developmental perspective. Dev. Psychol. 18:199–214 Ladd GW, Mize J. 1983. A cognitive-social learning model of social-skill training. Psychol. Rev. 90:127–57 Laird RD, Pettit GS, Mize J. 1994. Motherchild conversations about peers: contributions to competence. Fam. Relat.: Interdiscip. J. Appl. Fam. Stud. 43:425–32 LeDoux JE. 2000. Emotion circuits in the brain. Annu. Rev. Neurosci. 23:155–84 LeDoux JE, Iwata J, Cicchetti P, Reis DJ. 1988. Different projections of the central amygdaloid nucleus mediate autonomic and behavioral correlates of conditioned fear. J. Neurosci. 8:2517–29 MacLeod C, Mathews A. 1988. Anxiety and the allocation of attention to threat. Q. J. Exp. Psychol. 40:653–70 Magnusdottir I, Smari J. 1999. Social anxiety in adolescents and appraisal of negative events: specificity or generality of bias? Behav. Cogn. Psychother. 27:223–30 Marshall PJ, Bar-Haim Y, Fox NA. 2002. Development of the EEG from 5 months to 4 years of age. Clin. Neurophysiol. 113:1199– 208 Marshall PJ, Hardin MG, Fox NA. 2005. Electrophysiological responses to auditory novelty in temperamentally different 9-monthold infants. In preparation Marshall PJ, Stevenson-Hinde J. 1998. Behav-

ioral inhibition, heart period, and respiratory sinus arrhythmia in young children. Dev. Psychobiol. 33:283–92 Marshall PJ, Stevenson-Hinde J. 2001. Behavioral inhibition: physiological correlates. In International Handbook of Social Anxiety, ed. WR Crozier, LE Alden, pp. 53–76. Chichester, UK: Wiley Mathews A. 1990. Why worry? The cognitive function of anxiety. Behav. Res. Ther. 28:455–68 McManis MH, Kagan J, Snidman NC, Woodward SA. 2002. EEG asymmetry, power, and temperament in children. Dev. Psychobiol. 41:169–77 Mineka S, Tomarken AJ. 1989. The role of cognitive biases in the origins and maintenance of fear and anxiety disorders. In Aversion, Avoidance, and Anxiety: Perspectives on Aversively Motivated Behavior, ed. T Archer, L-G Nilsson, pp. 195–221. Hillsdale, NJ: Erlbaum Muris P, Merkelbach H, Schmidt H, Gadet B, Bogie N. 2001. Anxiety and depression as correlates of self-reported behavioral inhibition in normal adolescents. Behav. Res. Ther. 39:1051–61 Muris P, Merkelbach H, Wessel I, van de Ven M. 1999. Psychopathological correlates of selfreported behavioral inhibition in normal children. Behav. Res. Ther. 37:575–84 Nachmias M, Gunnar M, Mangelsdorf S, Parritz RH, Buss K. 1996. Behavioral inhibition and stress reactivity: the moderating role of attachment security. Child Dev. 67:508–22 Nelson F, Garduque L. 1991. The experience and perception of continuity between home and day care from the perspectives of child, mother, and caregiver. Early Child Dev. Care 68:99–111 Ochsner KN, Gross JJ. 2004. Thinking makes it so: a social cognitive neuroscience approach to emotion regulation. In Handbook of Selfregulation: Research, Theory, and Applications, ed. RF Baumeister, KD Vohs, pp. 229– 55. New York: Guilford Panksepp J. 1998. Affective Neuroscience. New York: Oxford Univ. Press

18 Nov 2004 11:19

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

BEHAVIORAL INHIBITION Park S, Belsky J, Putnam S, Crnic K. 1997. Infant emotionality, parenting, and 3-year inhibition: exploring stability and lawful discontinuity in a male sample. Dev. Psychol. 33:218–27 Picton TW, Alain C, Otten L, Ritter W, Achim A. 2000. Mismatch negativity: different water in the same river. Audiol. Neuro-otol. 5:111–39 Posner MI, DiGirolamo GJ. 2000. Cognitive neuroscience: origins and promise. Psychol. Bull. 126:873–89 Posner MI, Rothbart MK. 1998. Attention, selfregulation, and consciousness. Trans. Philos. Soc. Lond. B 1915–27 Prior M, Smart D, Sanson A, Oberklaid F. 2000. Does shy-inhibited temperament in childhood lead to anxiety problems in adolescence? J. Am. Acad. Child Adolesc. Psychiatry 39:461–68 Reznick JS, Kagan J, Snidman N, Gersten M, Baak K, Rodenberg A. 1986. Inhibited and uninhibited children: a follow-up study. Child Dev. 57:660–80 Rosenbaum JF, Biederman J, Bolduc-Murphy BA, Faraone SV, Chaloff J, et al. 1993. Behavioral inhibition in childhood: a risk factor for anxiety disorders. Harv. Rev. Psychiatry 1:2–16 Rosenbaum JF, Biederman J, Gersten M, Hirshfeld DR, Meminger SR, et al. 1988. Behavioral inhibition in children of parents with panic disorder and agoraphobia. Arch. Gen. Psychiatry 45:463–70 Rothbart MK, Ahadi SA, Hershey KL. 1994. Temperament and social behavior in childhood. Merrill-Palmer Q. 40:31–39 Rothbart MK, Bates JE. 1998. Temperament. In Handbook of Child Psychology, Volume 3: Social, Emotional and Personality Development, ed. W Damon (ser.), N Eisenberg (vol.), pp. 105–76. New York: Wiley Rothbart MK, Derryberry D. 1981. Development of individual differences in temperament. In Advances in Developmental Psychology, ed. M Lamb, A Brown, vol. 1, pp. 37–86. Hillsdale, NJ: Erlbaum Rothbart MK, Ellis LK, Posner MI. 2004. Tem-

P1: IKH

261

perament and self-regulation. In Handbook of Self-regulation: Research, Theory, and Applications, ed. RF Baumeister, KD Vohs, pp. 357–70. New York: Guilford Rothbart MK, Ziaie H, O’Boyle CG. 1992. Selfregulation and emotion in infancy. New Dir. Child Dev. 55:7–23 Rubin KH. 1985. Socially withdrawn children: an “at risk” population? In Children’s Peer Relations: Issues in Assessment and Intervention, ed. B Schneider, KH Rubin, J Ledingham, pp. 125–39. New York: SpringerVerlag Rubin KH. 1993. The Waterloo Longitudinal Project: correlates and consequences of social withdrawal from childhood to adolescence. In Social Withdrawal, Inhibition and Shyness in Childhood, ed. KH Rubin, J Asendorpf, pp. 291–314. Hillsdale, NJ: Erlbaum Rubin KH. 1998. Social and emotional development from a cultural perspective. Dev. Psychol. 34:611–15 Rubin KH, Borwick D. 1984. The communicative skills of children who vary with regard to sociability. In Communication by Children and Adults: Social Cognitive and Strategic Processes, ed. H Sypher, J Applegate, pp. 152–70. Beverly Hills, CA: Sage Rubin KH, Bukowski W, Parker J. 1998. Peer interactions, relationship, and groups. In Handbook of Child Psychology: Social, Emotional, and Personality Development, ed. N Eisenberg, pp. 619–700. New York: Wiley. 5th ed. Rubin KH, Burgess KB, Hastings PD. 2002. Stability and social-behavioral consequences of toddlers’ inhibited temperament and parenting behaviors. Child Dev. 73:483–95 Rubin KH, Hastings PD, Stewart SL, Henderson HA, Chen X. 1997. The consistency and concomitants of inhibition: some of the children, all of the time. Child Dev. 68:467–83 Rubin KH, Krasnor LR. 1986. Social cognition and social behavioral perspectives on problem-solving. In Minnesota Symposium on Child Psychology, ed. M Perlmutter, vol. 16, pp. 1–68. Hillsdale, NJ: Erlbaum

18 Nov 2004 11:19

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

262

AR

AR231-PS56-09.tex

AR231-PS56-09.sgm

LaTeX2e(2002/01/18)

P1: IKH

FOX ET AL.

Rueda MR, Posner MI, Rothbart MK. 2004. Attentional control and self-regulation. In Handbook of Self-regulation: Research, Theory, and Applications, ed. RF Baumeister, KD Vohs, pp. 283–300. New York: Guilford Sanson A, Pedlow R, Cann W, Prior M, Oberklaid F. 1996. Shyness ratings: stability and correlates in early childhood. Int. J. Behav. Dev. 19:705–24 Scarpa A, Raine A, Venables PH. 1995. The stability of inhibited/uninhibited temperament from ages 3 to 11 years in Mauritian children. J. Abnorm. Child Psychol. 23:607– 18 Scarpa A, Raine A, Venables PH, Mednick SA. 1997. Heart rate and skin conductance in behaviorally inhibited Mauritian children. J. Abnorm. Psychol. 106:182–90 Schmidt LA, Fox NA, Rubin KH, Sternberg E, Gold PW, et al. 1997. Behavioral and neuroendocrine responses in shy children. Dev. Psychobiol. 30:127–40 Schmidt LA, Fox NA, Schulkin J, Gold PW. 1999. Behavioral and psychophysiological correlates of self-presentation in temperamentally shy children. Dev. Psychobiol. 35:119–35 Schwartz CE, Snidman N, Kagan J. 1999. Adolescent social anxiety as an outcome of inhibited temperament in childhood. J. Am. Acad. Child Adolesc. Psychiatry 38:1008–15 Schwartz CE, Wright CI, Shin LM, Kagan J, Rauch SL. 2003. Inhibited and uninhibited infants “grown up”: adult amygdalar response to novelty. Science 300:1952– 53 Stansbury K, Gunnar MR. 1994. Adrenocortical activity and emotion regulation. In The Development of Emotion Regulation: Biological and Behavioral Considerations, ed. NA Fox. Monogr. Soc. Res. Child Dev. 59(2–3, Ser. No. 240):108–34 Stevenson-Hinde J. 1989. Behavioral inhibition: issues of context. In Perspectives on Be-

havioral Inhibition, ed. JS Reznick, pp. 125– 38. Chicago: Univ. Chicago Press Stevenson-Hinde J, Shouldice A. 1995. Maternal interactions and self-reports related to attachment classifications at 4.5 years. Child Dev. 66:583–96 Super CM, Harkness S. 1986. The developmental niche: a conceptualization at the interface of child and culture. Int. J. Behav. Dev. 9:545–69 Tremblay RE, Pihl RO, Vitaro F, Dobkin PL. 1994. Predicting early onset of male antisocial behavior from preschool behavior. Arch. Gen. Psychiatry 51:732–38 Vasey MW, MacLeod C. 2001. Informationprocessing factors in childhood anxiety: a review and developmental perspective. In The Developmental Psychopathology of Anxiety, ed. MW Vasey, MR Dadds, pp. 253–77. London: Oxford Univ. Press Washington J, Minde K, Goldberg S. 1986. Temperament in preterm infants: style and stability. J. Am. Acad. Child Psychiatry 25: 493–502 Wells A, Matthews G. 1994. Attention and Emotion: A Clinical Perspective. Hillsdale, NJ: Erlbaum Williams JMG, Mathews A, MacLeod C. 1996. The emotional Stroop task and psychopathology. Psychol. Bull. 120:3–24 Wood JJ, McLeod BD, Sigman M, Hwang W, Chu B. 2003. Parenting and childhood anxiety: theory, empirical findings, and future directions. J. Child Psychol. Psychiatry 44:134–51 Woodward SA, McManis MH, Kagan J, Deldin P, Snidman NC, et al. 2001. Infant temperament and the brainstem auditory evoked response in later childhood. Dev. Psychol. 37:533–38 Zhengyan W, Huichang C, Xinyin C. 2003. The stability of children’s behavioral inhibition: a longitudinal study from two to four years of age. Acta Psychol. Sinica 35:93–100

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

Annual Review of Psychology Volume 56, 2005

CONTENTS

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Frontispiece—Richard F. Thompson

xviii

PREFATORY In Search of Memory Traces, Richard F. Thompson

1

DECISION MAKING Indeterminacy in Brain and Behavior, Paul W. Glimcher

25

BRAIN IMAGING/COGNITIVE NEUROSCIENCE Models of Brain Function in Neuroimaging, Karl J. Friston

57

MUSIC PERCEPTION Brain Organization for Music Processing, Isabelle Peretz and Robert J. Zatorre

89

SOMESTHETIC AND VESTIBULAR SENSES Vestibular, Proprioceptive, and Haptic Contributions to Spatial Orientation, James R. Lackner and Paul DiZio

115

CONCEPTS AND CATEGORIES Human Category Learning, F. Gregory Ashby and W. Todd Maddox

149

ANIMAL LEARNING AND BEHAVIOR: CLASSICAL Pavlovian Conditioning: A Functional Perspective, Michael Domjan

179

NEUROSCIENCE OF LEARNING The Neuroscience of Mammalian Associative Learning, Michael S. Fanselow and Andrew M. Poulos

207

HUMAN DEVELOPMENT: EMOTIONAL, SOCIAL, AND PERSONALITY Behavioral Inhibition: Linking Biology and Behavior Within a Developmental Framework, Nathan A. Fox, Heather A. Henderson, Peter J. Marshall, Kate E. Nichols, and Melissa A. Ghera

235

BIOLOGICAL AND GENETIC PROCESSES IN DEVELOPMENT Human Development: Biological and Genetic Processes, Irving I. Gottesman and Daniel R. Hanson

263 vii

P1: JRX

December 8, 2004

viii

12:13

Annual Reviews

AR231-FM

CONTENTS

SPECIAL TOPICS IN PSYCHOPATHOLOGY The Psychology and Neurobiology of Suicidal Behavior, Thomas E. Joiner Jr., Jessica S. Brown, and LaRicka R. Wingate

287

DISORDERS OF CHILDHOOD Autism in Infancy and Early Childhood, Fred Volkmar, Kasia Chawarska, and Ami Klin

315

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

CHILD/FAMILY THERAPY Youth Psychotherapy Outcome Research: A Review and Critique of the Evidence Base, John R. Weisz, Amanda Jensen Doss, and Kristin M. Hawley

337

ALTRUISM AND AGGRESSION Prosocial Behavior: Multilevel Perspectives, Louis A. Penner, John F. Dovidio, Jane A. Piliavin, and David A. Schroeder

365

INTERGROUP RELATIONS, STIGMA, STEREOTYPING, PREJUDICE, DISCRIMINATION The Social Psychology of Stigma, Brenda Major and Laurie T. O’Brien

393

PERSONALITY PROCESSES Personality Architecture: Within-Person Structures and Processes, Daniel Cervone

423

PERSONALITY DEVELOPMENT: STABILITY AND CHANGE Personality Development: Stability and Change, Avshalom Caspi, Brent W. Roberts, and Rebecca L. Shiner

453

WORK MOTIVATION Work Motivation Theory and Research at the Dawn of the Twenty-First Century, Gary P. Latham and Craig C. Pinder

485

GROUPS AND TEAMS Teams in Organizations: From Input-Process-Output Models to IMOI Models, Daniel R. Ilgen, John R. Hollenbeck, Michael Johnson, and Dustin Jundt

517

LEADERSHIP Presidential Leadership, George R. Goethals

545

PERSONNEL EVALUATION AND COMPENSATION Personnel Psychology: Performance Evaluation and Pay for Performance, Sara L. Rynes, Barry Gerhart, and Laura Parks

571

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

CONTENTS

ix

PSYCHOPHYSIOLOGICAL DISORDERS AND PSYCHOLOGICAL EFFECTS ON MEDICAL DISORDERS Psychological Approaches to Understanding and Treating Disease-Related Pain, Francis J. Keefe, Amy P. Abernethy, and Lisa C. Campbell

601

TIMELY TOPIC

Annu. Rev. Psychol. 2005.56:235-262. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Psychological Evidence at the Dawn of the Law’s Scientific Age, David L. Faigman and John Monahan

631

INDEXES Subject Index Cumulative Index of Contributing Authors, Volumes 46–56 Cumulative Index of Chapter Titles, Volumes 46–56

ERRATA An online log of corrections to Annual Review of Psychology chapters may be found at http://psych.annualreviews.org/errata.shtml

661 695 700

23 Dec 2004 19:46

AR

AR231-PS56-10.tex

AR231-PS56-10.sgm

P1: IKH LaTeX2e(2002/01/18) 10.1146/annurev.psych.56.091103.070208

Annu. Rev. Psychol. 2005. 56:263–86 doi: 10.1146/annurev.psych.56.091103.070208 c 2005 by Annual Reviews. All rights reserved Copyright
First published online as a Review in Advance on July 21, 2004

HUMAN DEVELOPMENT: Biological and Genetic Processes

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Irving I. Gottesman Department of Psychiatry and Department of Psychology, University of Minnesota, Minneapolis, Minnesota 55454; email: [email protected]

Daniel R. Hanson Departments of Psychiatry and Psychology, University of Minnesota, Minneapolis, and Veterans Administration Hospital, Minneapolis, Minnesota 55417; email: [email protected]

Key Words adaptive systems, endophenotypes, CNS plasticity, schizophrenia, autism ■ Abstract Adaptation is a central organizing principle throughout biology, whether we are studying species, populations, or individuals. Adaptation in biological systems occurs in response to molar and molecular environments. Thus, we would predict that genetic systems and nervous systems would be dynamic (cybernetic) in contrast to previous conceptualizations with genes and brains fixed in form and function. Questions of nature versus nurture are meaningless, and we must turn to epigenetics—the way in which biology and experience work together to enhance adaptation throughout thick and thin. Defining endophenotypes—road markers that bring us closer to the biological origins of the developmental journey—facilitates our understanding of adaptive or maladaptive processes. For human behavioral disorders such as schizophrenia and autism, the inherent plasticity of the nervous system requires a systems approach to incorporate all of the myriad epigenetic factors that can influence such outcomes. CONTENTS INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . COMPLEX ADAPTIVE SYSTEMS IN HUMAN DEVELOPMENT . . . . . . . . . . . . The Meaning of “Gene-Environment Interaction” . . . . . . . . . . . . . . . . . . . . . . . . . . Epigenesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Endophenotypes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Generation and Degeneration as a Continuum of Development . . . . . . . . . . . . . . . . BUILDING THE BRAIN FROM THE GENES UP . . . . . . . . . . . . . . . . . . . . . . . . . . . Cellular Differentiation and Migration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Angiogenesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . PLASTICITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . EXAMPLE: AUTISM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Epidemiology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0066-4308/05/0203-0263$14.00

264 265 265 266 268 270 271 271 271 272 274 274

263

23 Dec 2004 19:46

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

264

AR

AR231-PS56-10.tex

GOTTESMAN




AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

P1: IKH

HANSON

Twin Data Paradoxes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Phenotypes and Endophenotypes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . EXAMPLE: OFFSPRING OF SCHIZOPHRENIC PARENTS, CHILDHOOD ONSET CASES, AND COTWINS AS “MEDIA” FOR GROWING ENDOPHENOTYPES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Systems Biology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Prospective High-Risk Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Candidate Genes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Clues from Childhood Schizophrenia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CONCLUSIONS AND A GLIMPSE AT THE FUTURE . . . . . . . . . . . . . . . . . . . . . . CONFLICT OF INTEREST STATEMENT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

275 276

276 276 277 278 279 279 280

INTRODUCTION Newton’s laws of physics and Einstein’s theories of relativity anchor the physical sciences. Biological sciences have a parallel organizing principle in the concept of adaptation (Holland 1975). Adaptation can be viewed from many perspectives, including adaptation of species, of populations, or of individuals. The time scale for evolutionary adaptation of species occurs over a very long period, measured in multiple generations. During this process, genotypes (DNA) are altered through mutation followed by natural selection to reconfigure the architectures of species. Populations adapt on an intermediate time scale that is often within one, or a few, life spans. Genetic diversity reflected in population individual differences allows for deployment of different partially heritable skill sets for different challenges. For example, human populations have both created and then adapted to the changes from a largely agrarian existence to a complex industrial world within one or two life spans. Individuals adapt over short time frames measured in fractions of a life span. This ability to adapt quickly is an evolutionarily derived trait. We humans, with opposable thumbs, hands freed by upright locomotion, expanded forebrains, and skills with symbolic language, are the most adaptable of all creatures. This chapter is about adaptation and how, in large part, our genetic and biological systems foster adaptation in health and detract from it in illness. By understanding the ramifications of the concept of adaptation in biological systems, most of the details of biological development and behavioral function can be derived (Turkheimer 1998, 2004). In so doing, we quickly confront two old dogmas and hope to teach them new tricks. First, if we accept the idea that our genomes have been “designed” by evolution to maximize adaptability, we could not conclude that genetically mediated traits are fixed and immutable. On the contrary, if our genomes are tuned to maximize adaptability, we would expect genetic factors to play an important role in change even within the individual and over short time frames. The idea that genetic factors influencing behavior are fixed stems mostly from the study of human genetic diseases such as the inborn errors of metabolism and chromosomal anomalies. However, these are examples of broken genes, which cannot be generalized to genes functioning via “rheostatic” control. For example, phenylketonuria (PKU), a recessive disorder producing one kind of mental retardation, is due to a dysfunctional genotype for phenylalanine

23 Dec 2004 19:46

AR

AR231-PS56-10.tex

AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN DEVELOPMENT

P1: IKH

265

hydroxylase. In the absence of normal function, people cannot metabolize ubiquitous dietary phenylalanine, leading to high concentrations that damage the brain. However, the normal or even half-normal genotype allows the individual to adapt to widely varying amounts of phenylalanine, thus preventing toxic levels even with massive ingestion. The expression of any one gene is embedded within a biological system influenced by a multitude of other genetic and environmental influences; concepts of gene regulation (expression) and epigenesis are now essential for understanding development (Carey 2003, Petronis 2004). The classic Mendelian disorders involve genes that are so broken as to overshadow ancillary modifiers. Such dramatic errors of nature divert us from appreciating the subtle, complex, and dynamic nature of biological systems (Dipple & McCabe 2000) that allows them to be adaptive and self-regulating. Second, it has been a long-held belief that the central nervous system is hardwired and cannot be changed easily by the time we reach adulthood. From the perspective of adaptability, this would make no sense because we continue to learn, change, and adapt throughout the life span. The brain does change with experience and the underlying physiology is guided by genetic factors (Grossman et al. 2003, Insel & Fernald 2004, Kennedy et al. 2003, Weaver et al. 2002). Just as observations of broken genes lead to the erroneous conclusion that genetic effects are unmodifiable, the observations that broken brains (e.g., stroke, trauma) do not heal has led to the belief that the central nervous system (CNS) is a fixed structure. However, in the intact brain, synaptic connections and neuronal circuits are shaped continuously to enhance adaptation. Everything that is genetic is biological, but not all things biological are genetic. Having stated that our genomes and CNS are designed to promote adaptation, we must then consider to what are the adaptations responding. The obvious answer is the environment. However, a global concept such as environment is not helpful. In the PKU example, environmental factors could mean the difference between normal IQ and mental retardation. But, by “environment,” we are not referring to air quality, climate, economic class, amount of parental affection, or hours spent watching Sesame Street. All of these factors may have an impact on a child with PKU, but the trait-relevant environment is the amount of phenylalanine in the diet (Meehl 1977). Thus, the outcomes for a person with the PKU mutation (genotype) are determined to a large degree by diet (trait-relevant environment), mediated through the biology of phenylalanine blood levels (endophenotype) that affect the eventual IQ (phenotype).

COMPLEX ADAPTIVE SYSTEMS IN HUMAN DEVELOPMENT The Meaning of “Gene-Environment Interaction” The phrase “gene-environment interaction” is used in a variety of ways and with a variety of meanings (Carey 2003, pp. 291–297; Rutter & Silberg 2002). A few words of clarification are offered as a prelude to the discussion following about

23 Dec 2004 19:46

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

266

AR

AR231-PS56-10.tex

GOTTESMAN




AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

P1: IKH

HANSON

epigenesis. The G × E interaction concept originated from early quantitative genetics (Falconer 1960), especially agricultural genetics, and represents, in a strict sense, an interaction effect in an analysis of variance. As such, G × E interaction means that different genotypes respond differently to different environments. As a quick example, suppose you took some rice seeds and wheat seeds (the different genotypes) and planted some of each in wet conditions and some in dry conditions (the different environments). In the wet conditions, the rice grew vigorously but the wheat drowned and failed. In the dry condition, the wheat thrived but the rice withered. Another way to phrase the gene environment interaction concept is to indicate that both genes and environment make a difference for the development of some trait. Height is a simple example: A person’s height depends on genes promoting height as well as on good nutrition. No matter what your genotype, good nutrition will help individuals grow to their full potential. However, no matter what the genotype, good nutrition always works in the same direction to enhance height. This may be thought of as coaction, but it is not G × E interaction in the strict analysis-of-variance sense. The study of human development does not allow the easy quantification and manipulation of genetic and environmental variables in the same manner as agricultural genetics, thus limiting our abilities to assess G × E interaction in the analysis-of-variance sense. Simple additive models that suggest the phenotype is the sum of environmental and genetic effects do not conform to biological realities (cf. Meaney 2001). The additive model reduces human development to a simple recipe, but there is more to it than adding two parts genes to three parts environment plus a pinch of luck. Turkheimer and colleagues (Turkheimer et al. 2003) demonstrate that socioeconomic status modified the heritability of IQ in a nonlinear fashion such that, in impoverished families, 60% of the variance in IQ was attributed to shared environment, and genetic effects were negligible; in affluent families the reverse was true. Thus, the concept of gene-environment interaction, however defined, is difficult to apply in studies of human development (Caspi et al. 2002, 2003; Gunnar 2003; Kagan 2003). A further limitation of simple G × E interaction models arises from the fact that gene expression is dynamic over time. This is illustrated by studies of the effects of caloric restriction on longevity. When mice were placed on restricted diets, there was a rapid change in the expression of genes associated with longevity, including genes involved in metabolism, signal transduction, stress response, and inflammation (Dhahbi et al. 2004). To introduce a dimension of time into human developmental models and to allow for changes in both the environment and the expressed genotype, the concept of epigenesis is rapidly supplementing the ideas of gene-environment interaction.

Epigenesis The term epigenesis originated with embryological theories suggesting that complex organisms originate from undifferentiated cells, and the term has been broadly

23 Dec 2004 19:46

AR

AR231-PS56-10.tex

AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN DEVELOPMENT

P1: IKH

267

defined to include all the forces that lead to the phenotypic expression of an individual’s genotype (Petronis 2000, 2004; Waddington 1957). Gottesman & Shields (1972) transduced this concept of epigenesis into human behavioral genetics in the early 1970s, with later elaboration (Gottesman et al. 1982). The definition of epigenetic continues to evolve and, to many molecular biologists, the term refers to the mechanisms by which cells change form or function and then transmit that form or function to future cells in that cell line (Jablonka & Lamb 2002, Jaenisch & Bird 2003, Morange 2002). Examples include transformation of an undifferentiated embryo cell into a liver cell or transformation of a normal liver cell into a cancerous cell. Once a cell type acquires a new form through selective gene expression and environmental influences, that cell, through cell division, transmits that acquired characteristic to future cells in the lineage. The previously spurned concept of the inheritance of acquired characteristics is resurfacing at the molecular level (Varmuza 2003). The best-studied mechanisms for the epigenetic regulation of mammalian gene expression involve the addition of a methyl group to cytosine that, along with adenine, thiamine, and guanine, forms the four-letter alphabet of DNA. This methylation of cytosine changes the configuration of the DNA such that the genetic information encoded in that area cannot be read and is nullified (Jaenisch & Bird 2003, Jones & Takai 2001)—the gene essentially is turned off. Conversely, removing DNA methylation allows the gene to be expressed. The variety of factors that influence DNA methylation is huge and includes such things as developmental processes, diet, viral infections, aging, and chance. Failure of methylation systems leads to clinical syndromes, such as Rett syndrome, that involve mental retardation, autistic-like behaviors, and other neurodevelopmental anomalies in girls (Shahbazian & Huda 2002). The impact of prenatal and early postnatal nutrition on the adult development of type 2 diabetes, cardiovascular disease, obesity, and cancer are also thought to be mediated by epigenetic factors mediated by DNA methylation (Waterland & Jirtle 2004). Such epigenetic mechanisms may explain why maternal behavior toward young offspring affects the size of the offspring’s hippocampus in adulthood, depending on the offspring’s genotypes (Weaver et al. 2002). More speculatively, epigenetic theorizing is being applied to the development of schizophrenia (Petronis et al. 2003) and depression (Caspi et al. 2003, Charney & Manji 2004). Although not approaching a biochemical analysis, even traits such as talent are being rethought in epigenetic terms (Simonton 1999). Epigenetic perspectives grapple with the complexities of how multiple genetic factors and multiple environmental factors become integrated over time through dynamic, often nonlinear, sometimes nonreversible, processes to produce behaviorally relevant endophenotypes and phenotypes. How an embryonic cell differentiates into a liver cell while a genetically identical cell in the same embryo develops into a neuron is an epigenetic question. Identical twins discordant for a given trait or disease provide other examples of epigenetic processes (Cannon et al. 2002, Pol et al. 2004, van Erp et al. 2004). Diverse reviews of epigenetic concepts relevant to human development are available (Gottesman & Gould 2003, Nijhout 2003, Petronis 2000, Petronis et al.

23 Dec 2004 19:46

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

268

AR

AR231-PS56-10.tex

GOTTESMAN




AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

P1: IKH

HANSON

2003, Sing et al. 2003). A stellar example of a systems biology approach to studying epigenesis is in the research mapping the developmental sequences in the sea urchin from fertilized egg onward (Davidson et al. 2002). Epigenetic thinking builds on the notion that only a small fraction of our DNA codes for structures (proteins, enzymes, etc.) while, in keeping with the central theme of adaptation, the majority of our DNA codes for regulatory processes. In response to transduced environmental stimuli, genes are turned on or off as the organism proceeds through life. At any time, any one genotype may have a wide array of potential phenotypes. The actual phenotype will depend on the influence of the individual’s other genes and on the specific contexts of environments experienced among a wide range of possible environments. Which environment is experienced may be stochastic (luck) or may be a function of the individual’s past phenotypes. Indeed, an individual’s phenotype (which is partially a result of his or her genotype) may lead the individual to select environments, thereby establishing a correlation between genotype and environment (Carey 2003). The array of possible outcomes could be plotted, in theory, in multidimensional space, as functions of genotypes, environments, and time. The plot would produce an undulating surface that would represent the phenotype for that unique combination of genotype, environment, and time. Such a surface has been referred to as a reaction surface (Gottesman & Gould 2003, Sing et al. 2003) or phenotypic surface (Nijhout 2003); these articles provide informative graphics. Figure 1 provides such an example (Manji et al. 2003) applied to the ontogenesis of schizophrenia with provision for the changing reaction surface and a threshold, suggested endophenotypes, some already connected to candidate genes, and a dimension of environmental inputs (harmful versus protective), all “bathed” in epigenetic influences.

Endophenotypes One of the primary obstacles to progress in connecting the genotypic contributors to many human phenotypes is that the traits submitted to genetic analyses lack biological meaning. It is a long road from genotype through epigenetic pathways to ultimate phenotype, as seen above. When we study the product of this process that may have encompassed decades, we often have too much “pheno” and not enough “geno” to make sense of the trait. What is needed is some kind of intermediate trait that sits closer to the genotype in the developmental scheme. In spite of the best efforts to improve the reliability of psychiatric classification, the diagnoses in the official nomenclatures are still syndromal and lack validating pathophysiological markers. Traits such as IQ and personality have demonstrable genetic effects, but efforts to understand the genetic component suffer from the lack of any intermediary connection between the behavior and the biological underpinnings. The missing links have been referred to as “endophenotypes” (Gottesman & Gould 2003). Alternative concepts with similar but different meanings include “biological markers,” “intermediate phenotypes,” “risk factors,” “vulnerability markers,” and “subclinical traits.” Attempts to develop those that are genetically mediated endophenotypes would require that:

23 Dec 2004 19:46

AR

AR231-PS56-10.tex

AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

HUMAN DEVELOPMENT

P1: IKH

269

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

1. Endophenotypes would be associated with the trait in the population. 2. The endophenotype would be demonstrably heritable. 3. The endophenotype is present whether the trait/disease is or is not present (e.g., vulnerability marker). 4. Within families, endophenotype and trait cosegregate (but not perfectly; see 3 above). 5. The endophenotype found in families with the trait (especially an illness) is found in nonaffected family members at a higher rate than in the general population. An instructive example comes from cardiology and the long QT syndrome. It was known that phenotypes including syncope, ventricular arrhythmias, and sudden death aggregated in families. The common denominator turned out to be QT elongation on electrocardiogram. Using QT elongation as the endophenotype, and by excluding or including family members with this finding, genetic linkage studies were successful in identifying the associated genes (Keating et al. 1991, Keating & Sanguinetti 2001). Putative endophenotypes for schizophrenia include, but are not limited to, those shown in Figure 1. Current strategies for identifying behavioral phenotypes such as psychiatric diagnoses, cognitive abilities, personality traits, or special talents all lack a biological “handle” to submit to genetic analysis. A search for endophenotypes will move us closer to establishing the biological underpinnings of these traits. Returning to the criteria for an endophenotype, we can observe that most implicate features that are closer to the phenotype end of the genotype-to-phenotype pathway. What will we do when we finally traverse the entire phenotype-to-gene pathway and discover the genetic contributors to the trait under study? More than 40 years ago, in a remarkably prescient anticipation of epigenetic and endophenotype thinking, Paul Meehl (1962) emphasized a perspective closer to the genotype end of the pathway. Writing about schizophrenia, he cautioned that knowing “specific (genetic) etiology” does not imply any of the following (p. 828): 1. The etiological factor always, or even usually, produces clinical illness. 2. If illness occurs, the particular form and content of symptoms is derivable by reference to the specific etiology alone. 3. The course of the illness can be influenced materially only by the procedures directed against the specific etiology. 4. All persons who share the specific etiology will have closely similar histories, symptoms, and course. 5. The largest single contributor to symptom variance is the specific etiology. Meehl offered these cautions not to suggest hopeless complexity but to challenge us to rethink our received paradigms (Hanson 2004) and to underscore the importance of epigenetic perspectives.

23 Dec 2004 19:46

270

AR

AR231-PS56-10.tex

GOTTESMAN




AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

P1: IKH

HANSON

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Generation and Degeneration as a Continuum of Development Concepts of epigenesis and endophenotypes help us climb out of our paradigm ruts toward understanding disease states. When things go wrong in the course of our lives, we tend to classify the cause of the problem either as a phenomenon of faulty assembly (developmental models) or of breakdown after normal function (degenerative models). Degenerative models of adaptive failure imply that after a period of normal development, the organism, or one of its parts, takes an unhappy turn in life trajectory and begins to disintegrate. This, of course, describes the eventual outcome for all life forms and is a biological restatement of the second law of thermodynamics. Because degeneration is (eventually) universal, stating that an illness is degenerative is not particularly enlightening. It would be helpful to determine when in the life course the degeneration begins, and how. Answers to the “when” and “how” questions would describe the degenerative process in developmental terms. Developmental models of adaptive dysfunction (Grossman et al. 2003) implicate early brain development as setting the stage for the future (Monk et al. 2001, Webb et al. 2001). The proponents of the developmental models further argue that the perturbations of development are limited to the early times of development and are not continuous. Without this qualifier, developmental models are indistinguishable from degenerative models where the degeneration starts early in the life span (Lewis & Levitt 2002). The early abnormalities are not necessarily the cause of adaptive failure, but instead create a vulnerable risk state for future dysfunction. Consequently, there must be factors later in life that convert the vulnerability to an actuality. These additional factors are presumed to somehow damage development in such a way that dysfunction becomes manifest (cf. Sing et al. 1994). To gain a complete understanding of the syndrome, we must return to the questions of what happens and when. Following this line of reasoning, the distinction between degenerative and developmental models blurs. In fact, a medical-behavioral condition can be both developmental and degenerative, as exemplified by Down syndrome (Head & Lott 2004, Kornberg et al. 1990, Opitz & Gilbert-Barness 1990). Individuals with trisomy 21 exhibit a number of developmental anomalies, including cardiac malformations, abnormal dermatoglyphics, skeletal changes, and muscular hypotonia. As infants with trisomy 21 mature, they exhibit mental retardation. By about age 50, these individuals invariably develop Alzheimer-like CNS degenerative changes. Given the above cited evidence that prenatal and neonatal nutritional deficiencies (developmental) lead to adult diseases such as cancer and heart disease (usually thought of as degenerative), we need to redirect our thinking away from developmental versus degenerative dichotomies just as we have moved away from the nature versus nature mind-sets (Gottesman 2001). In his book Unheard Cry for Meaning, Viktor Frankl (1978) suggested we are not fully developed until we die—we continue to change right up to the last moment of life. Generation and degeneration go hand in hand as we traverse our own epigenetic landscapes.

23 Dec 2004 19:46

AR

AR231-PS56-10.tex

AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

HUMAN DEVELOPMENT

P1: IKH

271

BUILDING THE BRAIN FROM THE GENES UP

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Cellular Differentiation and Migration Our brains can be simplified schematically into three primary components: neurons, glia, and the vascular system. Conventionally, the neuron has been the target of attention in the study of human behavior. This is especially apparent in the various neuronal transmission theories of mental illness where there has been a preponderant focus on dopaminergic function in schizophrenia, and serotonergic and noradrenergic transmission in affective (mood) disorders (Schatzberg & Nemeroff 2004). However, neurons comprise only about 50% of brain volume. Glial cells, which comprise close to the additional 50%, are receiving increasing attention for their possible role in behavioral disorders (Moises et al. 2002). A third component of the CNS, the vascular system, comprises only about 0.1% of total brain volume, but this small component forms the “blood brain barrier” and plays a vital role in regulation of brain metabolism (Schusta & Boado 2002). Development of the brain starts in the primordial inner layer of the neural tube with totipotent stem cells that differentiate into neurons and glial cells (Monk et al. 2001, Webb et al. 2001). Brain development proceeds through proliferation and migration of neurons and glia responding to epigenetic regulation of a large spectrum of genes coding for growth factors. Some neuroblasts move by radial migration, building tissue from the inner depths to the outer layers of cortex (Mehler & Kessler 1999, Nadarajah & Parnavelas 2002). In radial migration, neurons formed in proliferative tissues zones move perpendicular to the brain surface following radially oriented glial fibers. Tangential migration connects brain components across regions when neurons moving parallel to the brain surface following precursor neurons. A multitude of genetic factors guides the migratory processes. Mutations in these genes prevent normal migration and have behavioral consequences (Nadarajah & Parnavelas 2002, Taylor et al. 2004). One of the migration modulators, reelin, has received particular attention in relation to autism and schizophrenia. Glial cells continue to play important roles in regulation and repair in the adult brain (Jessen & Richardson 2001, Levine et al. 2001). Additionally, astroglia cells act as intermediaries between neurons and the vascular system. The cell bodies have two arms; one reaches out and embraces the neuron while the other reaches out to the microcapillaries and is interspersed with cells of the capillary endothelium.

Angiogenesis The brain does not work well if deprived of adequate blood flow. Because the CNS has virtually no reserves of energy and cannot function on anaerobic metabolism, the brain requires constant and precise delivery of glucose and oxygen. Developmentally, the CNS vascular system originates from mesodermic capillary endothelial cells that migrate into developing neuroexoderm under the influence of

23 Dec 2004 19:46

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

272

AR

AR231-PS56-10.tex

GOTTESMAN




AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

P1: IKH

HANSON

neuron-derived trophic factors such as vascular endothelial growth factor (Risau et al. 1998) and erythropoietin (Sasaki 2003), both produced by astroglia. Rather than being a passive conduit, the CNS vascular system is the most precisely managed and complex fluid dynamic system known. Regulation of cerebral blood flow (Kety & Schmidt 1948) is managed primarily by a partnership between astrocytic glial cells (Coyle & Schwarcz 2002, Haydon 2001, Kurosinski & Gotz 2002) and capillary endothelium (Abott 2002, Kety & Schmidt 1948, Medhora et al. 2001, Paulson 2002, Virgintino et al. 2002, Yoder 2002, Zonta et al. 2003). Astrocytes sense local neuronal metabolic activity and adjust blood flow as needed. Cerebral vessels change diameter in response to vasoactive substances released by astrocytes activated by glutamate receptors, serotonin (Cohen et al. 1999), acetylcholine (Elhusseiny et al. 1999), and dopamine (Bacic 1991, Favard 1990). When neuronal activation of discrete areas is sustained over longer periods, vasoactive substances stimulate angiogenesis (blood vessel formation), resulting in capillary density increases (Harder et al. 2002) and thus enhancing local neuronal circuitry. Conversely, a decrease in capillary density is likely to reduce the functional capacity of brain areas so affected (Harder et al. 2002). Consequently, capillary beds in the cortex are not distributed uniformly (Cavaglia et al. 2001). Close relationships exist among local neuronal activity, density of capillary bed, and the distribution of valve-like flow control structures (Harrison et al. 2002). The growing awareness of the dynamic circulation of the CNS vascular flow has important consequences for studying CNS metabolic activity. Imaging studies (e.g., fMRI, positron emission tomography) assume the vascular flow is a constant, so a measured change in cerebral blood flow is attributed to reduction in neural activity. However, we must consider reversals of the causal arrow, with the possibility that a primary vascular disease leads to the deranged cell metabolism, as is being considered in Alzheimer disease (Borroni & Akkawi 2002, Preston & Steart 2003).

PLASTICITY Psychologist D.O. Hebb postulated more than a half century ago that experience modifies cortical connections (Hebb 1947, 1949), yet the adult brain has been primarily viewed as a fixed structure (Grossman et al. 2003, Webb et al. 2001). Recent developments indicate Hebb was correct and that the brain is constantly changing in response to experience. The changes in synaptic connections and recruitment of expanded representational areas devoted to a particular function are referred to as plasticity. Plastic changes are associated with learning/memory, skill acquisition, recovery from injury, and even addiction. A myriad of factors influence brain plasticity, including pre- and postnatal experience, genes, drugs, hormones, maturation/aging, diet, disease, stress, and trauma (Kolb et al. 2003). Details of the molecular mechanisms of CNS plasticity are beyond the scope of this chapter; we limit ourselves to a few examples. Hodge & Boakye (2001) and Johansson (2000) review details about molecular mechanism. Thompson &

23 Dec 2004 19:46

AR

AR231-PS56-10.tex

AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN DEVELOPMENT

P1: IKH

273

Nelson (2001), Shonkoff (2003), and Grossman et al. (2003) summarize many of the social-political ramifications of the growing understanding of brain plasticity. Much of the recent progress in understanding brain plasticity in humans has been made possible by advances such as fMRI (Casey 2002), which has been used to demonstrate and map plasticity, as well as to partial out effects due to maturation, per se, versus experience/practice. Some of the most detailed research on plasticity comes from studies of musicians; their skill development involves unlearning of existing synaptic connections and establishment of new sensory-motor-memory-affective connections (cf. Peretz & Zatorre 2005). The virtuoso performance involves memorizing the music at the motor level, coordinating tactile and auditory sensory inputs, and adding an emotional interpretation that transcends the mechanical reproduction of notes. However, remodeling the brain to achieve such feats involves risks as well. Guitar players, for example, are subject to dystonic movements of their hands that, alternatively, are described as an overuse syndrome. Pascual-Leon (2001) has demonstrated with fMRI that musicians with motor dystonias showed significantly greater activation of contralateral sensorimotor cortex and conspicuous bilateral underactivation of premotor areas. The authors suggest that extensive practice of coordinated hand motions in which fingers function as a single unit might induce changes in sensory-motor field representation of the hand with blurring of the normal separation of each digit. Another illustration of the downside of CNS plasticity comes from addiction research (Ujike et al. 2002). Behavioral sensitization to addicting drugs arises from structural modification of neural networks. Repeated exposure to amphetamines and cocaine alter the cytoarchitecture of the nucleus accumbens and frontal cortex by increasing the length of dendrites and the density of dendritic spines. These changes are regulated by a host of genetic factors regulating synaptogenesis, including genes for neurite sprouting, neuritic elongation, and cell division regulators throughout the brain. The tenacity of addictions is then explained by the difficulty of remodeling these drug-induced structural changes or, possibly, by long-term alteration in gene expression supporting the CNS-mediated behavioral sensitization to the drugs. Responses to stress, including the development of depression and stress syndromes such as posttraumatic stress disorder, are attributed to failures in CNS plasticity, more so in predisposed persons. Chronic stress is implicated in CNS signal transduction cascades that normally allow neuronal plasticity. Chronic stress damages a wide variety of plasticity modulators and, at the biochemical level, causes a reduction in expression of genes associated with synaptic plasticity, resulting in diminished frontal cortical activity (Kuipers et al. 2003). Assembling and maintaining the brain involves a sometimes choreographed and sometimes extemporaneous dance among the partners of neuron, glia, vascular supply, and experience. Damage to the brain interrupts this performance. Repair may be more an issue of probabilities than of potentials. Damage does not mean that the dancers will never dance again. However, the probability of exactly

23 Dec 2004 19:46

274

AR

AR231-PS56-10.tex

GOTTESMAN




AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

P1: IKH

HANSON

repeating the prior performance is extremely low. Efforts focused on CNS repair are increasingly using epigenetic strategies to modulate gene expression in tissues combined with intensive rehabilitative experiences. It is too early to talk about gene transplants.

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

EXAMPLE: AUTISM Some 30 years ago, we (Hanson & Gottesman 1976) reviewed the scant literature on the genetics of autism and early onset schizophrenia and concluded that autism was not connected genetically to schizophrenia, an opinion that has been sustained (Gousse et al. 2003). We also found little in the extant literature providing convincing evidence for genetic factors in autism. Over the ensuing years, this opinion has been overruled so that it is common to find opinions in the literature that autism is strongly genetic (Muhle et al. 2004). Within the Annual Reviews series, there are numerous comprehensive reviews of genetic factors in autism (Cowan et al. 2002, Plomin & McGuffin 2003, Veenstra-VanderWeele et al. 2004) and it would be redundant to repeat them. Instead, we will take an epigenetic perspective to examine the consensus evidence and highlight topics needing more research.

Epidemiology Published prevalence rates for autism and autism spectrum disorders have skyrocketed from early estimates of 1 case in 10,000 to rates as high as 60 in 10,000 (Fombonne 2003). This rise in a few decades is much too fast to be blamed on genomic changes. If the increase is real and anywhere near this dramatic, it would implicate a dramatic rise in the environmental contributors that could involve epigenetic regulation of genetic factors. Continuing public expressions of concerns that agents used to immunize infants could be such a factor cannot be substantiated by careful epidemiology (DeStefano 2002, DeStefano & Thompson 2004). An analysis of pervasive developmental disorders in Israeli residents compared to immigrants to Israel from Ethiopia found much lower rates in the Ethiopians, which could point to environmental factors in the industrialized setting compared to nonindustrial environments; barriers to migration of Ethiopian families with PDD children is an alternate explanation (Kamer et al. 2004). The consensus explanation for the increased rate of autism-like disorders is that the changes are most likely due to broader definitions of the illness and improved case finding (Fombonne 2003, Gernsbacher et al. 2004, Lingam et al. 2003, Wing & Potter 2002). One epidemiological finding that remains undisputed is the higher rate of autism (299.00 in the DSM) in males as compared to females, a factor of about 4:1. The difference may even be greater as broader definitions of the disorder are applied (Veenstra-VanderWeele et al. 2004). Despite common knowledge of this fact, gender is rarely taken into consideration in genetic analyses of autism spectrum disorders. Countless linkage studies implicate loci on autosomes (especially chromosomes 2, 3, and 7). However, autosomes are freely exchanged between the

23 Dec 2004 19:46

AR

AR231-PS56-10.tex

AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN DEVELOPMENT

P1: IKH

275

genders and could not account for the gender differences in prevalence of autism. If autosomal factors were involved, then gender-related epigenetic differences, possibly hormonal, would have to be invoked to explain the gender differences. Alternatively, genes on the X chromosome could explain the gender differences if they were expressed as partial Mendelian recessive traits. Genetic modeling for X-linkage would lead to testable hypotheses. Alternatively, multifactorial threshold models of transmissions would lead to the prediction that the lower frequency group (females in this case) would have higher loadings of the risk factors, thus their siblings would have higher risk compared to the siblings of affected males. Future genetic studies will benefit from capitalizing on the clues provided by the apparent different expressivity in girls versus boys. Simply invoking the notion that males appear more vulnerable to development insults is not a sufficient response because there still must be detailed explanation for the vulnerability.

Twin Data Paradoxes Much of the support for a genetic component in autism came from the study of twins (see reviews cited in the Epidemiology section above). The twin data consistently show high concordance rates in identical twins (monozygotic, MZ) in the range of 60%–90%. By contrast, fraternal twins (dizygotic, DZ) have a low concordance rate, typically close to zero and ranging up to 10% with large standard errors, given the rarity of such samples. Fraternal twins are genetically no more similar than ordinary siblings, and the DZ twin concordance falls in the same ballpark rate of 4%–5% as siblings of autistic people also being affected. Two issues arise. First, for twin studies to be valid, the trait under study cannot be an outcome simply of the twinning process. Twin pregnancies are considered high risk and are associated with increased rates of a wide range of developmental disorders, including mental retardation and cerebral palsy. Two recent reports indicate that autism is more prevalent in twins, suggesting the twin method may not be completely valid for this trait (Betancur et al. 2002, Greenberg et al. 2001). Counterarguments suggest diagnosis inflation and that sampling artifacts could explain the perceived increased rate of twins with autism (Visscher 2002). Other studies fail to find increased twinning among people with autism (Hallmayer et al. 2002). Patience is called for as the debate goes on (Hodge et al. 2002) and we await the definitive answer from population-based international collaborative studies of autism in twins following the precedents applied to the study of cerebral palsy (Petterson et al. 1993, Scher et al. 2002) and the discovery of how environmental factors such as infection, in combination with inherited variations in response to infection, can lead to cerebral palsy (Gibson et al. 2003, Nelson & Willoughby 2000). The very high concordance rates for autism in MZ twins in contrast to very low concordance rates in DZ twins gives rise to very high heritability estimates for autism when these twin rates are plugged into standard formulas for computing heritability from twin data (Carey 2003). At the same time, the low DZ concordance rates (often zero) and low sibling risk rates suggest that autism has low recurrence risks or transmissibility. As a counterpoint, consider a trait like

23 Dec 2004 19:46

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

276

AR

AR231-PS56-10.tex

GOTTESMAN




AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

P1: IKH

HANSON

Huntington disease. This also rare autosomal dominant disorder is little modified by environmental factors (highly genetic), and the disorder conveys risks of 50% to DZ twins or siblings (highly transmissible). For autism, the high heritability contrasted with the low transmissibility creates dissonance in genetic theorizing, especially given the “background noise” of 3%–5% rates of serious developmental disabilities in the general population. A simple explanation is that genes really are not so important after all. Alternatively, epistasis (gene x gene interaction within and across loci) could be invoked—it is more the combination of genes than the specific genes that holds the key to this illness. Another possibility is that an epigenetic or chance change in the germ line leads to autism (Keller & Persico 2003). Once we figure out the heritability versus transmissibility issue, we will then have to understand why it affects males more than females.

Phenotypes and Endophenotypes Autism is certainly heterogeneous, etiologically and clinically. Known genetic conditions with autism-like characteristics include syndromes of tuberous sclerosis, fragile-X, and Prader-Willi, Angelman, Rett, and various chromosomal abnormalities, to name a few (Gillberg & Coleman 2000). Individuals with the syndrome may or may not have evidence of in utero growth retardation, EEG abnormalities or epilepsy, profound (IQ < 50) intellectual impairment, increased head growth in infancy (Courchesne et al. 2003, Lainhart 2003), altered immune systems (Licinio et al. 2002, Lipkin & Hornig 2003), structural brain abnormalities on MRI (Brambilla et al. 2003), and altered neuroarchitecture (Casanova et al. 2002, 2003). Developmental characteristics such as these may point to endophenotypes that are more suitable for genetic analysis than are the phenotypes based on overt behavioral syndromes utilized by our current clinical diagnoses (Baird et al. 2003). While we laud the efforts to further refine the behaviorally based assessment strategies (Constantino et al. 2004), it is our bias to look for classification strategies that incorporate biological as well as behavioral characteristics. The use of eye tracking technology to assess social visual pursuit in autism (Klin et al. 2002) is one such example. Returning to our earlier comments on adaptation, we can view behavior itself as an evolutionary strategy to maximize adaptability. However, the resulting variability of human behavior may mean that strictly behavioral phenotypes are too imprecise to lead us to the neurobiological and genetic underpinnings of disorders such as autism.

EXAMPLE: OFFSPRING OF SCHIZOPHRENIC PARENTS, CHILDHOOD ONSET CASES, AND COTWINS AS “MEDIA” FOR GROWING ENDOPHENOTYPES Systems Biology Our next exemplars maintain a focus on the biological and genetic processes that affect developmental psychopathology by examining recent research on the

23 Dec 2004 19:46

AR

AR231-PS56-10.tex

AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN DEVELOPMENT

P1: IKH

277

precursors to adult schizophrenia, childhood onset schizophrenia, and pervasive developmental disorders (but not autism, already discussed above). We have tried to select research that emphasizes a systems biological strategy (Sing et al. 1994, Zerba et al. 2002)—everything involving brain and behavior is connected to everything else—to integrate across the regions of Figure 1. Excellent reviews complement our selections (Cowan et al. 2002, Erlenmeyer-Kimling 2000, Kennedy et al. 2003, Lewis et al. 2003, Walker et al. 2004). Keeping track of the “gene of the month” in regard to the psychoses is a full-time occupation that can be mentioned only briefly. Contrary to some jaundiced views of such enterprises, replications of candidate genes and new candidates are here to stay; a good two-dozen candidate genes, each of modest impact on total liability to developing schizophrenia in their respective populations, deserve attention and following up (Lewis et al. 2003, McGuffin 2004, Moises et al. 2004, Owen et al. 2004, Sklar 2002).

Prospective High-Risk Studies Prospective longitudinal studies of the offspring born to schizophrenic parents have been a staple in the search for antecedent traits and endophenotypes related to schizophrenia even though they may require 30 or more years of intensive efforts. The high-risk strategy exemplifies a systems approach by the breadth of variables studied including genetics, birth and pregnancy complications, and a host of behavioral variables followed developmentally over decades. The strategy (Pearson & Kley 1957) was suggested with the foreknowledge that 10% or so of such children in such cohorts would go on to develop the illness, thus permitting the observation of illness precursors unconfounded by the effects of the illness per se. Adoption strategies reinforced the continuing efforts by showing that rearing by ill parents was not essential for the dire outcomes (Ingram & Kety 2000, Tienari et al. 2003), and left open the question as to the role of other contributors to liability that were not genetic (cf. Murray et al. 2003). The theme of the endophenotype strategy (Glahn et al. 2004, Gottesman & Gould 2003, Lenox et al. 2002) embedded within a systems biology context is straightforward: deconstruct the disease phenotype into its precursors and correlates not available to the naked eye or ear (neurocognitive tasks, measured personality indicators, fMRI, neuronal growth factors, etc.), and focus on those that are heritable with implied genes and their polymorphisms as distal causes in the complex genes-to-behaviors pathways. Identify the polymorphisms and then explore their expression over the life course, sensitive to the agents of epigenetics. The process may also begin at the genotype level, reversing the plan of research, and starting with those genes now being implicated by whole genome scans that rely on patients and controls without involving relatives (Lewis et al. 2003, Owen et al. 2004). Many of the genes identified do not yet have known disease polymorphisms or CNS-impacting functions, as it is still early in the game (Merikangas & Risch 2003, Varmus 2002). The New York High-Risk Project systematically sampled 358 children ages 7 to 12, starting in 1971, whose mothers or fathers were in one of three groups:

23 Dec 2004 19:46

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

278

AR

AR231-PS56-10.tex

GOTTESMAN




AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

P1: IKH

HANSON

schizophrenics or affectively ill who had been admitted to a state hospital, and community controls without psychiatric diagnoses or treatment. Erlenmeyer-Kimling (2000), after Herculean efforts, followed these offspring through seven rounds of testing until the late 1990s (when the mean age of the offspring was the early thirties), by which time 15%, 7%, and 1% of each of the respective risk groups had developed a schizophrenia-related psychosis (only Sample A, the first half, is reported here). Because their data were gathered prospectively, the researchers could look back to see which neuropsychological indicators (cf. Meehl 1962), if any, would have predicted the outcomes. Briefly, three sets of indicators, when configured, yielded noteworthy results. Using a criterion of failure on all three indicators in early childhood—attentional deviance index, verbal short-term or working memory index, and impairment (subtle) in gross motor skills index—the sensitivity of the battery correctly predicted for 50% of the offspring of schizophrenics, with a false positive rate of 10.4%. The battery was 100% accurate in not predicting any child in the other two groups as a future sufferer from a schizophreniarelated psychosis. Thus, a foundation has been laid for carrying out the remaining steps in the larger strategy, with some of the evidence for success illustrated in Figure 1.

Candidate Genes The COMT gene on the long or q arm of chromosome 22 at band 11 has attracted considerable attention as a candidate gene for schizophrenia given (a) its involvement in coding for a dopamine enzyme and its functional mutations, (b) its overlap with the gene deletion leading to velocardiofacial syndrome or DiGeorge syndrome, wherein excess diagnoses of schizophrenia are seen in some samples (Ivanov et al. 2003), and (c) research programs connecting COMT polymorphisms to measures of frontal lobe function such as working memory in patients (Egan et al. 2001, Malhotra et al. 2002). The genetic association with schizophrenia is equivocal in that a box score shows four positive and four negative results (Owen et al. 2004). Studies of mice with a deletion in the equivalent region of their genome provide interesting and encouraging leads (Maynard et al. 2003). The mix of data requiring integration to this COMT story includes data from studies of unaffected relatives of schizophrenic patients (Goldberg et al. 2003) and of normal children. Diamond et al. (2004) looked at COMT polymorphisms in 39 children under age 12 (mean = 9) who had been tested with a variant of the Stroop test and two control tasks. Their results showed that only some of the supposed prefrontal cortex cognitive functions were sensitive to the dopamine levels inferred from the genotypes, adding to the complexity of what is being tested in the brain–behavior relationship. The story and the validity of genes in the 22 q region, as with all of the other promising candidates already in hand and about to be reported, leave many hurdles left to be traversed successfully; many will fail the tests and others will open windows to neighboring genes that were not even suspected beforehand (Mowry et al. 2004).

23 Dec 2004 19:46

AR

AR231-PS56-10.tex

AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

HUMAN DEVELOPMENT

P1: IKH

279

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Clues from Childhood Schizophrenia Childhood onset schizophrenia (COS) cases, although rare, are considered especially valuable for casting light on the biology and genetic processes of adult onset cases. Criteria are not different for COS, but the criteria are met before the thirteenth birthday. Cases of COS appear to be more severe and more homogeneous, with more familiality for schizophrenia spectrum disorders (Asarnow et al. 2001, Hanson & Gottesman 1976, Nicholson et al. 2003). We would expect COS to have a larger dose of the genes that predispose to schizophrenia as well as to have biology that, if different, cannot be attributed to the wear and tear that confound the phenotypes of adult cases. Lingering problems about the diagnostic overlaps in COS are a challenge to unraveling etiologies. Psychosis not otherwise specified (NOS), pervasive developmental disorder (PDD), and numerous organic diseases of the CNS have symptoms that overlap one another (Sporn et al. 2003, 2004). An important observation made on follow-up of psychotic children with an initial diagnosis of psychosis NOS was that 13 of 27 were diagnosed as bipolar disorder (Addington et al. 2004). In that same study, G72—a gene at 13q33.2—was found to be significantly associated with both COS and early onset bipolar cases; the gene is one of those already with replicated positive results in adult cases of schizophrenia (Owen et al. 2004). Sporn et al. (2003) have observed a likely biological marker in 60 COS cases; a “striking progressive reduction in cortical gray matter volume” was detected with MRI at the follow-up time into adolescence of the COS cases. Altered cortical thickness and surface morphology possibly due to aberrant cellular migration patterns may provide another marker (White et al. 2003). With the growing interest in discovering endophenotypes and the availability of advanced imaging techniques has come a renewed interest in studying identical and fraternal twins with the new equipment. The findings are too new to be integrated into this review, but they can be expected to provide important milestones on the research road map in Figure 1 (Cannon et al. 2002, Pol et al. 2004, Torrey et al. 1994).

CONCLUSIONS AND A GLIMPSE AT THE FUTURE The charge of responsibly overviewing—while being limited to a finite number of words—a “planet” in the solar system of the planet Psychology is both a daunting and enlightening task: daunting because conscientious and highly motivated scientists have produced too many important, and less important, advances since the subject was last reviewed; enlightening because it compels opening one’s mind to the fact that our planet is so enmeshed with others in the system that no one or even three topical Annual Reviews chapters could accomplish the mission of summarizing selected and relevant nuggets of wisdom. These points are made obvious by what is present and even more so by what has been omitted from our cited literature. We hope we will be accused of stretching your mind muscles but not of spraining them if our reach sometimes exceeded our grasp. Consider each citation as a seed for turning on an Internet search engine with a mission of

23 Dec 2004 19:46

280

AR

AR231-PS56-10.tex

GOTTESMAN




AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

P1: IKH

HANSON

incorporating biology and genetics into your own research program—you will be pleasantly surprised by the planets you encounter and what they are doing there that is relevant to your own work. Incorporation followed by integration will take place, we predict, and you will discover the benefits of hybrid vigor for yourselves.

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

ACKNOWLEDGMENTS Preparation of this chapter was supported in part by a grant from the Stanley Foundation to DRH. We dedicate this chapter to the memory of Paul E. Meehl (1920–2003). His far-ranging intellect and his self-described interdisciplinary mind (“I am more of a knowledge-absorber, knowledge-integrator, and knowledge-transmitter”) taught us to search widely for answers to questions posed by science and humanity. Paul Meehl’s autobiography can be found in A History of Psychology in Autobiography, 1989, ed. G. Lindzey, volume 3, pp. 337–389, published by Stanford University Press, Stanford, California.

CONFLICT OF INTEREST STATEMENT IIG and DRH know of no conflicts of interest related to this manuscript. The Annual Review of Psychology is online at http://psych.annualreviews.org

LITERATURE CITED Abott N. 2002. Astrocyte-endothelial interactions and blood-brain permeability. J. Anat. 200:629–38 Addington A, Gornick M, Sporn A, Gogtay N, Greenstein D, et al. 2004. Polymorphisms in the 13q33.2 gene G72/G30 are associated with childhood-onset schizophrenia and psychosis not otherwise specified. Biol. Psychiatry 55(10):989–94 Asarnow R, Nuechterlien K, Fofelson D, Subotnik K, Payne D, et al. 2001. Schizophrenia and schizophrenia-spectrum personality disorders in the first-degree relatives of children with schizophrenia: the UCLA family study. Arch. Gen. Psychiatry 58:581–88 Bacic F, Uematsu S, McCarron RM, Spatz M. 1991. Dopamine receptors linked to adenylate cyclase in human cerebromicrovascular endothelium. J. Neurochem. 57:1774– 80

Baird G, Cass H, Slonims V. 2003. Diagnosis of autism. Br. Med. J. 327:488–93 Betancur C, Leboyer M, Gillberg C. 2002. Increased rate of twins among affected sibling pairs with autism. Am. J. Hum. Genet. 70: 1381–83 Borroni B, Akkawi N. 2002. Microvascular damage and platelet abnormalities in early Alzheimer’s disease. J. Neurol. Sci. 203–204: 189–93 Brambilla P, Hardan A, Ucelli di Nemi S, Perez J, Soares J, Baral F. 2003. Brain anatomy and development in autism: review of structural MRI studies. Brain Res. Bull. 61:557–69 Cannon T, Thompson P, van Erp T, Toga A, Poutanen V, et al. 2002. Cortex mapping reveals regionally specific patterns of genetic and disease-specific gray-matter deficits in twins discordant for schizophrenia. Proc. Natl. Acad. Sci. USA 99:3228–33

23 Dec 2004 19:46

AR

AR231-PS56-10.tex

AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN DEVELOPMENT Carey G. 2003. Human Genetics for the Social Sciences. Thousand Oaks, CA: Sage Casanova M, Buxhoeveden D, Gomez J. 2003. Disruption in the inhibitory architecture of the cell minicolumn: implication for autism. Neuroscientist 9:496–507 Casanova M, Buxhoeveden D, Switala A, Roy E. 2002. Minicolumnar pathology in autism. Neurology 58:428–32 Casey B. 2002. Neuroscience: windows into the human brain. Science 5572:1408–9 Caspi A, McClay J, Moffitt T, Mill J, Martin J, et al. 2002. Role of genotype in the cycle of violence in maltreated children. Science 5582:851–54 Caspi A, Sugden K, Moffitt T, Taylor A, Craig I, et al. 2003. Influence of life stress on depression: moderation by a polymorphism in the 5-HTT gene. Science 301:386–89 Cavaglia M, Dombrowski S, Drazba J, Vasanji A, Bokesch P, Janigro D. 2001. Regional variation in brain capillary density and vascular response to ischemia. Brain Res. 910:81– 93 Charney D, Manji H. 2004. Life stress, genes, and depression: multiple pathways lead to increased risk and new opportunities for depression. Sci. STKE 225:re5 Cohen Z, Bouchelet I, Olivier A, Villemure J, Ball R, et al. 1999. Multiple microvascular and astroglial 5-hydroxytryptamine receptor subtypes in human brain: molecular and pharmacologic characterization. J. Cereb. Blood Flow Metab. 19:908–17 Constantino J, Gruber C, Davis S, Hayes S, Passanante N, Przybeck T. 2004. The factor structure of autistic traits. J. Child Psychol. Psychiatry 45:719–26 Courchesne E, Carper R, Akshoomoff N. 2003. Evidence of brain overgrowth in the first year of life in autism. JAMA 290:337–44 Cowan W, Kopnisky K, Hyman S. 2002. The human genome project and its impact on psychiatry. Annu. Rev. Neurosci. 25:1–50 Coyle J, Schwarcz R. 2000. Mind glue: implications of glial cell biology for psychiatry. Arch. Gen. Psychiatry 57:90–93 Davidson E, Rast J, Oliveri P, Ransick A,

P1: IKH

281

Calestani C, et al. 2002. A genomic regulatory network for development. Science 295:1669–78 DeStefano F. 2002. MMR vaccine and autism: a review of the evidence for a causal association. Mol. Psychiatry 7:S51–S52 DeStefano F, Thompson W. 2004. MMR vaccine and autism: an update of the scientific evidence. Expert Rev. Vaccines 3:19–22 Dhahbi J, Kim H-J, Note P, Beaver R, Spindler S. 2004. Temporal linkage between the phenotypic and genomic responses to caloric restriction. Proc. Natl. Acad. Sci. USA 101: 5524–29 Diamond A, Briand L, Fossella J, Gehlbach L. 2004. Genetic and neurochemical modulation of prefrontal cognitive functions in children. Am. J. Psychiatry 161:125–32 DiLalla LF, ed. 2004. Behavior Genetics Principles: Perspectives in Development, Personality, and Psychopathology. Washington, DC: Am. Psychol. Assoc. Dipple K, McCabe E. 2000. Phenotypes of patients with “simple” Mendelian disorders are complex traits: thresholds, modifiers, and systems dynamics. Am. J. Hum. Genet. 66:1729–35 Egan M, Goldberg T, Kolachana B, Callicott J, Mazzanti C, Straub R. 2001. Effect of COMT Val108/158 Met genotype on the frontal lobe function and risk for schizophrenia. Proc. Natl. Acad. Sci. USA 98:6917–22 Elhusseiny A, Cohen Z, Olivier A, Stanimirovic D, Hamel E. 1999. Functional acetylcholine muscarinic receptor subtypes in human brain microcirculation: identification and cellular localization. J. Cereb. Blood Flow Metab. 19:794–802 Erlenmeyer-Kimling L. 2000. Neurobehavioral deficits in offspring of schizophrenic parents. Am. Med. J. 97:65–71 Falconer D. 1960. Introduction to Quantative Genetics. New York: Ronald Press Favard CSA, Vigny A, Nguyen-Legros J. 1990. Ultrastructural evidence for a close relationship between dopamine cell process and blood capillary walls in Macaca monkey and rat retina. Brain Res. 523:127–33

23 Dec 2004 19:46

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

282

AR

AR231-PS56-10.tex

GOTTESMAN




AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

P1: IKH

HANSON

Fombonne E. 2003. Epidemiological surveys of autism and other pervasive developmental disorders: an update. J. Autism Dev. Disord. 33:365–82 Frankl V. 1978. The Unheard Cry for Meaning. New York: Simon & Schuster Gernsbacher M, Dawson M, Goldsmith H. 2004. Three reasons NOT to believe in an autism epidemic. Curr. Dir. Psychol. Sci. In press Gibson C, MacLennan A, Goldwater P, Dekker G. 2003. Antenatal causes of cerebral palsy: associations between inherited thrombophilias, viral and bacterial infection, and inherited susceptibility to infection. Obstet. Gynecol. Surv. 58:209–20 Gillberg C, Coleman M. 2000. The Biology of the Autistic Syndromes. London: Mac Keith Press/Cambridge Univ. Press. 3rd ed. Glahn D, Niendam T, Escamilla M. 2004. The feasibility of neuropsychological endophenotypes in the search for genes associated with bipolar affective disorder. Bipolar Disord. 6:171–82 Goldberg T, Egan M, Gscheidle T, Coppola R, Weickert T, et al. 2003. Executive subprocesses in working memory: relationship to catechol-o-methyltransferase Val158Met genotype and schizophrenia. Arch. Gen. Psychiatry 60:889–96 Gottesman I. 2001. Nature and nurture. In Encyclopedia of Genetics, ed. S Brenner, J Miller, E Broughton. San Diego: Academic Gottesman I, Gould T. 2003. The endophenotype concept in psychiatry: etymology and strategic intentions. Am. J. Psychiatry 160:1– 10 Gottesman I, Shields J. 1972. Schizophrenia and Genetics: a Twin Study Vantage Point. New York: Academic Gottesman I, Shields J, Hanson D. 1982. Schizophrenia: the Epigenetic Puzzle. London: Cambridge Univ. Press Gousse V, Plumet M-H, Chabane N, MourenSimeoni M-C, Ferradian N, Leboyer M. 2003. Fringe phenotypes in autism: a review of clinical, biochemical and cognitive studies. Eur. Psychiatry 17:120–28

Greenberg D, Hodge S, Sowinski J, Nicoll D. 2001. Excess of twins among affected sibling pairs with autism: implications for the etiology of autism. Am. J. Hum. Genet. 69:1062– 67 Grossman A, Churchill J, McKinney B, Kodish I, Otte S, Greenough W. 2003. Experience effects on brain development: possible contributions to psychopathology. J. Child Psychol. Psychiatry 44:33–63 Gunnar M. 2003. Integrating neuroscience and the psychological approaches in the study of early experience. Ann. NY Acad. Sci. 1008: 238–47 Hallmayer J, Glasson E, Bower C, Petterson B, Croen L, et al. 2002. On the twin risk in autism. Am. J. Hum. Genet. 71:941–46 Hanson D. 2004. Getting the bugs into our genetic theories of schizophrenia. See DiLalla 2004, pp. 205–16 Hanson D, Gottesman I. 1976. The genetics, if any, of infantile autism and childhood schizophrenia. J. Autism Child. Schizophr. 6: 209–33 Harder D, Zhang C, Gebremedhin D. 2002. Astrocytes function in matching blood flow to metabolic activity. News Physiol. Sci. 17:27– 31 Harrison R, Harel N, Panesar J, Mount R. 2002. Blood capillary distribution correlates with hemodynamic-based functional imaging in cerebral cortex. Cereb. Cortex 12:225–33 Haydon P. 2001. GLIA: listening and talking to the synapse. Nat. Rev. Neurosci. 2:185–93 Head E, Lott I. 2004. Down syndrome and beta-amyloid deposition. Curr. Opin. Neurol. 17:95–100 Hebb D. 1947. The effects of early experience on problem solving at maturity. Am. Psychol. 2:737–45 Hebb D. 1949. The Organization of Behavior. New York: Wiley Hodge C, Boakye M. 2001. Biological plasticity: the future of science in neurosurgery. Neurosurgery 48:2–16 Hodge S, Greenberg D, Betancur C, Gillberg C. 2002. Response to Visser. Am. J. Hum. Genet. 71:996–99

23 Dec 2004 19:46

AR

AR231-PS56-10.tex

AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN DEVELOPMENT Holland J. 1975. Adaptation in Natural and Artificial Systems. Ann Arbor: Univ. Mich. Press Ingram L, Kety S. 2000. Adoption studies in schizophrenia. Am. J. Med. Genet. (Semin. Med. Genet.) 97:18–22 Insel T, Fernald R. 2004. How the brain processes social information: searching for the social brain. Annu. Rev. Neurosci. 27:697– 722 Ivanov D, Kirov G, Norton N, Williams H, Williams N, et al. 2003. Chromosome 22q11 deletions, velo-cardi-facial syndrome and early-onset psychosis. Br. J. Psychiatry 183:409–13 Jablonka E, Lamb M. 2002. The changing concept of epigenetics. Ann. NY Acad. Sci. 981:82–96 Jaenisch R, Bird A. 2003. Epigenetic regulation of gene expression: how the genome integrates intrinsic and environmental signals. Nat. Genet. 33(Suppl.):245–54 Jessen K, Richardson W, eds. 2001. Glial Cell Development. London: Oxford Univ. Press Johansson B. 2000. Brain plasticity and stroke rehabilitation: the Willis Lecture. Stroke 31: 223–30 Jones P, Takai D. 2001. The role of DNA methylation in mammalian epigenetics. Science 263:1068–70 Kagan J. 2003. Biology, context, and developmental inquiry. Annu. Rev. Psychol. 54:1–23 Kamer A, Zohar A, Youngmann R, Diamond G, Inbar D, Senecky Y. 2004. A prevalence estimate of pervasive developmental disorder among immigrants to Israel and Israeli natives: a file review study. Soc. Psychiatry Psychiatr. Epidemiol. 39:141–45 Keating M, Atkinson D, Dunn C, Timothy K, Vincent G, Leppert M. 1991. Linkage of a cardiac arrythmia, the long QT syndrome, and the Harvey ras-1 gene. Science 252:704– 6 Keating M, Sanguinetti M. 2001. Molecular and cellular mechanisms of cardiac arrhythmias. Cell 104:569–80 Keller F, Persico A. 2003. The neurobiological context of autism. Mol. Neurobiol. 28:1–22

P1: IKH

283

Kennedy J, Lindsay A, Andreasen N, Mayeux R, St. George-Hyslop P. 2003. The genetics of adult-onset neuropsychiatric disease: complexities and conundra? Science 302:822–26 Kety S, Schmidt C. 1948. The nitrous oxide method for the quantitative determination of cerebral blood flow in man: theory, procedure and normal values. J. Clin. Invest. 27:476–83 Klin A, Jones W, Schultz R, Volkmar F, Cohen D. 2002. Defining and quantifying the social phenotype in autism. Am. J. Psychiatry 156:895–908 Kolb B, Gibb R, Robinson T. 2003. Brain plasticity and behavior. Curr. Dir. Psychol. Sci. 12:1–5 Kornberg J, Kawashima H, Pulst S, Allen L, Magenis E, Epstein C. 1990. Down syndrome: toward a molecular definition of the phenotype. Am. J. Med. Genet. 7(Suppl.):91– 97 Kuipers S, Trentani A, Den Boar J, Ter Host G. 2003. Molecular correlates of impaired prefrontal plasticity in response to chronic stress. J. Neurochem. 85:1312–23 Kurosinski P, Gotz J. 2002. Glial cells under physiologic and pathologic conditions. Arch. Neurol. 59:1524–28 Lainhart J. 2003. Increased rate of head growth during infancy in autism. JAMA 290:393–94 Lenox R, Gould T, Manji H. 2002. Endophenotypes in bipolar disorder. Am. J. Med. Genet. 114:391–406 Levine J, Reynolds R, Fawcett J. 2001. The oligodendrocyte precursor cell in health and disease. Trends Neurosci. 24:39–47 Lewis C, Levinson D, Wise L, DeLisi L, Straub R, Hovatta I. 2003. Genome scan metaanalysis of schizophrenia and bipolar disorder, part II: schizophrenia. Am. J. Hum. Genet. 73:34–48 Lewis D, Levitt P. 2002. Schizophrenia as a disorder of neurodevelopment. Annu. Rev. Neurosci. 25:409–32 Licinio J, Alvarado I, Wong M-L. 2002. Autoimmunity in autism. Mol. Psychiatry 7:329 Lingam R, Simmons A, Andrews N, Miller E, Stowe J, Taylor B. 2003. Prevalence of

23 Dec 2004 19:46

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

284

AR

AR231-PS56-10.tex

GOTTESMAN




AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

P1: IKH

HANSON

autism and parentally reported triggers in a northeast London population. Arch. Dis. Child 88:666–70 Lipkin W, Hornig M. 2003. Microbiology and immunology of autism spectrum disorders. Novartis Found. Symp. 251:129–43 Malhotra A, Kessler J, Mazzanti C, Bates J, Goldberg T, et al. 2002. A functional polymorphism in the COMT gene and performance on a test of prefrontal cognition. Am. J. Psychiatry 159:652–54 Manji H, Gottesman I, Gould T. 2003. Signal transduction and genes-to-behaviors pathways in psychiatric diseases. Sci. STKE 207: pe49 Maynard T, Haskell G, Peters A, Sikich L, Liberman J, LaMantia A. 2003. A comprehensive analysis of 22q11 gene expression in the developing adult brain. Proc. Natl. Acad. Sci. USA 100:14433–38 McGuffin P. 2004. Behavioral genomics: where molecular genetics is taking psychiatry and psychology. See DiLalla 2004, pp. 191– 216 Meaney M. 2001. Nature, nurture, and the disunity of knowledge. Ann. NY Acad. Sci. 935:50–61 Medhora M, Narayanan J, Harder D. 2001. Dual regulation of the cerebral microvasculature by epoxyeicosatrienoic acids. Trends Cardiovasc. Med. 11:38–42 Meehl P. 1962. Schizotaxia, schizotypy, schizophrenia. Am. Psychol. 17:827–38 Meehl P. 1977. Specific etiology and other forms of strong inference: some quantitative meanings. J. Med. Philos. 2:35–53 Mehler M, Kessler J. 1999. Progenitor cell biology: implications for neural regeneration. Arch. Neurol. 56:780–84 Merikangas K, Risch N. 2003. Will the genomics revolution revolutionize psychiatry? Am. J. Psychiatry 160:223–33 Moises H, Zoega T, Gottesman I. 2002. The glial growth factors deficiency and synaptic destabilization hypothesis of schizophrenia. BMC Psychiatry. www.biomed central.com/1471-244X/2/8 Moises H, Zoega T, Li L, Hood L. 2004. Genes

and neurodevelopment in schizophrenia. See DiLalla 2004, pp. 145–57 Monk C, Webb S, Nelson C. 2001. Prenatal neurobiological development: molecular mechanisms and anatomical change. Dev. Neuropsychol. 19:211–36 Morange M. 2002. The relations between genetics and epigenetics: a historical point of view. Ann. NY Acad. Sci. 981:50–60 Mowry B, Holmans P, Pulver A, Gejman P, Riley B, et al. 2004. Multicenter linkage study of schizophrenia loci on chromosome 22q. Mol. Psychiatry. In press Muhle R, Trentacoste S, Rapin I. 2004. The genetics of autism. Pediatrics 113:e472– 86 Murray R, Jones P, Susser E, Van Os J, Cannon M, eds. 2003. The Epidemiology of Schizophrenia. London: Cambridge Univ. Press Nadarajah B, Parnavelas J. 2002. Modes of neuronal migration in the developing cerebral cortex. Nat. Rev. Neurosci. 3:423–32 Nelson K, Willoughby R. 2000. Infection, inflammation and the risk of cerebral palsy. Curr. Opin. Neurol. 13:133–39 Nicholson R, Brookner F, Lenane M, Gochman P, Ingraham L, et al. 2003. Parental schizophrenia spectrum disorders in childhood-onset and adult-onset schizophrenia. Am. J. Psychiatry 160:490–95 Nijhout H. 2003. The importance of context in genetics. Am. Sci. 91:416–23 Opitz J, Gilbert-Barness E. 1990. Reflections on the pathogenesis of Down syndrome. Am. J. Hum. Genet. 7(Suppl.):38–51 Owen M, Williams N, O’Donovan M. 2004. The molecular genetics of schizophrenia: New findings promise new insights. Mol. Psychiatry 9:14–27 Pascual-Leone A. 2001. The brain that plays music and is changed by it. Ann. NY Acad. Sci. 930:315–29 Paulson O. 2002. Blood-brain barrier, brain metabolism and cerebral blood flow. Eur. Neuropsychopharmacol. 12:465–501 Pearson J, Kley I. 1957. On the application of genetic expectations as age specific rates in

23 Dec 2004 19:46

AR

AR231-PS56-10.tex

AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN DEVELOPMENT the study of human behavior disorders. Psychol. Bull. 54:406–20 Peretz I, Zatorre RJ. 2005. Brain organization for music processing. Annu. Rev. Psychol. 56:89–114 Petronis A, ed. 2003. Epigenetics: influence on behavioral disorders. In Encyclopedia of the Human Genome. London: Nature Publ. Petronis A. 2004. The origin of schizophrenia: genetic thesis, epigenetic antithesis, and resolving synthesis. Biol. Psychiatry. 55:142– 46 Petronis A, Gottesman I, Peixiang K, Kennedy J, Basile V, et al. 2003. Monozygotic twins exhibit numerous epigenetic differences: clues to twin discordance. Schizophr. Bull. 29:169–78 Petterson B, Nelson K, Watson L, Stanley F. 1993. Twins, triplets, and cerebral palsy in births in western Australia in the 1980’s. Br. Med. J. 307:1239–43 Plomin R, McGuffin P. 2003. Psychopathology in the postgenomic era. Annu. Rev. Psychol. 54:205–28 Pol H, Brans R, van Haren N, Schnack H, Langen M, et al. 2004. Gray and white matter volume abnormalities in monozygotic and same-gender dizygotic twins discordant for schizophrenia. Biol. Psychiatry 55:126–30 Preston S, Steart P. 2003. Capillary and arterial cerebral amyloid angiopathy in Alzheimer’s disease: defining the perivascular route for the elimination of amyloid (beta) from the human brain. Neuropath. Appl. Neurobiol. 29:106–17 Risau W, Esser S, Engelhardt B. 1998. Differentiation of blood-brain barrier endothelial cells. Pathol. Biol. 46:171–75 Rutter M, Silberg J. 2002. Gene-environment interplay in relation to emotional and behavioral disturbance. Annu. Rev. Psychol. 53:463–90 Sasaki R. 2003. Pleiotropic functions of erythropoietin. Intern. Med. 42:142–49 Schatzberg A, Nemeroff C. 2004. The American Psychiatric Publishing Textbook of Psychopharmacology. Washington, DC: Am. Psychiatr. Press. 3rd ed.

P1: IKH

285

Scher A, Petterson B, Blair E, Ellenberg J, Grether J, et al. 2002. The risk of mortality or cerebral palsy in twins: a collaborative population-based study. Pediatr. Res. 52:671–81 Schusta E, Boado R. 2002. Vascular genomics of the human brain. Cereb. Blood Flow Metab. 22:245–52 Shahbazian M, Huda Y. 2002. Rett syndrome and the MeCP2: linking epigenetics and neuronal function. Am. J. Hum. Genet. 71:1259– 72 Shonkoff J. 2003. From neurons to neighborhoods: old and new challenges for developmental and behavioral pediatrics. Dev. Behav. Pediatr. 24:70–76 Simonton D. 1999. Talent and its development: an emergenic and epigenetic model. Psychol. Rev. 106:435–57 Sing C, Stengard J, Kardia S. 2003. Genes, environment, and cardiovascular disease. Arterioscler. Thromb. Vasc. Biol. 23:1190–96 Sing C, Zerba K, Reilly S. 1994. Traversing the biological complexity in the hierarchy between genome and the CAD endpoints in the population at large. Clin. Genet. 46:6–14 Sklar P. 2002. Linkage analysis in psychiatric disorders: the emerging picture. Annu. Rev. Genomics Hum. Genet. 3:371–413 Sporn A, Addington A, Gogtay N, Ordonez A, Gornick M, et al. 2004. Pervasive developmental disorder and childhood-onset schizophrenia: comorbid disorder or a phenotypic variant of a very early onset illness? Biol. Psychiatry 55:989–94 Sporn A, Greenstein D, Gogtay N, Jeffries N, Lenane M, et al. 2003. Progressive brain volume loss during adolescence in childhood-onset schizophrenia. Am. J. Psychiatry 160:2181–89 Taylor C, Marta C, Claycomb R, Han D, Rasband M, et al. 2004. Proteomic mapping provides powerful insights into functional myelin biology. Proc. Natl. Acad. Sci. USA 101:4643–48 Thompson R, Nelson C. 2001. Developmental science and the media: early brain development. Am. Psychol. 56:5–15

23 Dec 2004 19:46

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

286

AR

AR231-PS56-10.tex

GOTTESMAN




AR231-PS56-10.sgm

LaTeX2e(2002/01/18)

P1: IKH

HANSON

Tienari P, Wynne L, Laksy K, Moring J, Nieminen P, et al. 2003. Genetic boundaries of the schizophrenia spectrum: evidence from the Finnish adoptive family study of schizophrenia. Am. J. Psychiatry 160:1587–94 Torrey E, Bowler A, Taylor E, Gottesman I. 1994. Schizophrenia and Manic-Depressive Disorder. New York: Basic Books Turkheimer E. 1998. Heritability and biological explanation. Psychol. Rev. 105:782–91 Turkheimer E. 2004. Spinach and ice cream: why social science is so difficult. See DiLalla 2004, pp. 161–89 Turkheimer E, Haley A, Waldron M, D’Onofrio B, Gottesman II. 2003. Socioeconomic status modifies heritability of IQ in young children. Psychol. Sci. 14:623–28 Ujike H, Takaki M, Kodama M, Kuroda S. 2002. Gene expression related to synaptogenesis, neuritogenesis, and MAP kinase in behavioral sensitization to psychostimulants. Ann. NY Acad. Sci. 965:55–67 van Erp T, Seleh P, Huttunen M, Lonnqvist J, Kaprio J, et al. 2004. Hippocampal volumes in schizophrenic twins. Arch. Gen. Psychiatry 61:346–53 Varmus H. 2002. Getting ready for gene-based medicine. N. Engl. J. Med. 347:1526–27 Varmuza S. 2003. Epigenetics and the renaissance of heresy. Genome 46:963–67 Veenstra-VanderWeele J, Christian SL, Cook EJ. 2004. Autism as a paradigmatic complex genetic disorder. Annu. Rev. Genomics Hum. Genet. 5:379–405 Virgintino D, Robertson D, Errede M, Benagiano V, Tauer U, et al. 2002. Expression of caveolin-1 in human brain microvessels. Neuroscience 115145–52 Visscher P. 2002. Increased rate of twins among affected sib pairs? Am. J. Hum. Genet. 71:995–96

Waddington C. 1957. The Strategy of the Genes. London: Allen & Unwin Walker E, Kestler L, Bollini A, Hochman K. 2004. Schizophrenia: etiology and course. Annu. Rev. Psychol. 55:401–30 Waterland R, Jirtle R. 2004. Early nutrition, epigenetic changes at transposons and imprinted genes, and enhanced susceptibility to adult chronic diseases. Nutrition 20:63–68 Weaver I, Grant R, Meany M. 2002. Maternal behavior regulates long-term hippocampal expression of BAX and apoptosis in the offspring. J. Neurochem. 82:998–1002 Webb S, Monk C, Nelson C. 2001. Mechanisms of postnatal neurobiological development: implications for human development. Dev. Neuropsychol. 19:147–71 White T, Andreasen N, Nopoulos P, Magnotta V. 2003. Gyrification abnormalities in childhood- and adolescent-onset schizophrenia. Biol. Psychiatry 54:418–26 Wing L, Potter D. 2002. The epidemiology of autistic spectrum disorders: Is the prevalence rising? Int. Retard. Dev. Disabil. Res. Rev. 8:151–61 Yoder E. 2002. Modifications in astrocyte morphology and calcium signaling induced by a brain capillary endothelial cell line. Glia 38:137–45 Zerba K, Ferrell R, Sing C. 2002. Complex adaptive systems and human health: the influence of common genotypes of the apolipoprotein E (ApoE) gene polymorphism and age on the relational order within a field of lipid metabolism traits. Hum. Genet. 107:466–75 Zonta M, Angulo M, Gobbo S, Rosengarten B, Hossmann K, et al. 2003. Neuron-toastrocyte signaling is central to dynamic control of brain microcirculation. Nat. Neurosci. 6:43–50

Gottesman.qxd

7/16/04

7:24 PM

Page 1

C-1

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

HUMAN DEVELOPMENT

Figure 1 Sketch of a systems biology approach toward explaining a complex behavior that incorporates dynamic interplay among candidate genes and gene regions, a sampling of endophenotypes (solid lines indicate those downstream of real genes), and the scope of preand postnatal environmental and epigenetic influences (harmful versus protective) over the course of development. Question marks indicate gaps in knowledge; p and q indicate regions of interest localized by research to specific numbered chromosomes; QTL = quantitative trait locus. Two planes intersect the reaction surface for the liability to developing schizophrenia over time, demarcating levels above which clinical diagnoses are warranted (cf. Gottesman & Gould 2003, Manji et al. 2003 for details). Copyright 2003 by I.I. Gottesman (used with permission).

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

Annual Review of Psychology Volume 56, 2005

CONTENTS

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Frontispiece—Richard F. Thompson

xviii

PREFATORY In Search of Memory Traces, Richard F. Thompson

1

DECISION MAKING Indeterminacy in Brain and Behavior, Paul W. Glimcher

25

BRAIN IMAGING/COGNITIVE NEUROSCIENCE Models of Brain Function in Neuroimaging, Karl J. Friston

57

MUSIC PERCEPTION Brain Organization for Music Processing, Isabelle Peretz and Robert J. Zatorre

89

SOMESTHETIC AND VESTIBULAR SENSES Vestibular, Proprioceptive, and Haptic Contributions to Spatial Orientation, James R. Lackner and Paul DiZio

115

CONCEPTS AND CATEGORIES Human Category Learning, F. Gregory Ashby and W. Todd Maddox

149

ANIMAL LEARNING AND BEHAVIOR: CLASSICAL Pavlovian Conditioning: A Functional Perspective, Michael Domjan

179

NEUROSCIENCE OF LEARNING The Neuroscience of Mammalian Associative Learning, Michael S. Fanselow and Andrew M. Poulos

207

HUMAN DEVELOPMENT: EMOTIONAL, SOCIAL, AND PERSONALITY Behavioral Inhibition: Linking Biology and Behavior Within a Developmental Framework, Nathan A. Fox, Heather A. Henderson, Peter J. Marshall, Kate E. Nichols, and Melissa A. Ghera

235

BIOLOGICAL AND GENETIC PROCESSES IN DEVELOPMENT Human Development: Biological and Genetic Processes, Irving I. Gottesman and Daniel R. Hanson

263 vii

P1: JRX

December 8, 2004

viii

12:13

Annual Reviews

AR231-FM

CONTENTS

SPECIAL TOPICS IN PSYCHOPATHOLOGY The Psychology and Neurobiology of Suicidal Behavior, Thomas E. Joiner Jr., Jessica S. Brown, and LaRicka R. Wingate

287

DISORDERS OF CHILDHOOD Autism in Infancy and Early Childhood, Fred Volkmar, Kasia Chawarska, and Ami Klin

315

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

CHILD/FAMILY THERAPY Youth Psychotherapy Outcome Research: A Review and Critique of the Evidence Base, John R. Weisz, Amanda Jensen Doss, and Kristin M. Hawley

337

ALTRUISM AND AGGRESSION Prosocial Behavior: Multilevel Perspectives, Louis A. Penner, John F. Dovidio, Jane A. Piliavin, and David A. Schroeder

365

INTERGROUP RELATIONS, STIGMA, STEREOTYPING, PREJUDICE, DISCRIMINATION The Social Psychology of Stigma, Brenda Major and Laurie T. O’Brien

393

PERSONALITY PROCESSES Personality Architecture: Within-Person Structures and Processes, Daniel Cervone

423

PERSONALITY DEVELOPMENT: STABILITY AND CHANGE Personality Development: Stability and Change, Avshalom Caspi, Brent W. Roberts, and Rebecca L. Shiner

453

WORK MOTIVATION Work Motivation Theory and Research at the Dawn of the Twenty-First Century, Gary P. Latham and Craig C. Pinder

485

GROUPS AND TEAMS Teams in Organizations: From Input-Process-Output Models to IMOI Models, Daniel R. Ilgen, John R. Hollenbeck, Michael Johnson, and Dustin Jundt

517

LEADERSHIP Presidential Leadership, George R. Goethals

545

PERSONNEL EVALUATION AND COMPENSATION Personnel Psychology: Performance Evaluation and Pay for Performance, Sara L. Rynes, Barry Gerhart, and Laura Parks

571

P1: JRX

December 8, 2004

12:13

Annual Reviews

AR231-FM

CONTENTS

ix

PSYCHOPHYSIOLOGICAL DISORDERS AND PSYCHOLOGICAL EFFECTS ON MEDICAL DISORDERS Psychological Approaches to Understanding and Treating Disease-Related Pain, Francis J. Keefe, Amy P. Abernethy, and Lisa C. Campbell

601

TIMELY TOPIC

Annu. Rev. Psychol. 2005.56:263-286. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

Psychological Evidence at the Dawn of the Law’s Scientific Age, David L. Faigman and John Monahan

631

INDEXES Subject Index Cumulative Index of Contributing Authors, Volumes 46–56 Cumulative Index of Chapter Titles, Volumes 46–56

ERRATA An online log of corrections to Annual Review of Psychology chapters may be found at http://psych.annualreviews.org/errata.shtml

661 695 700

18 Nov 2004 11:24

AR

AR231-PS56-11.tex

AR231-PS56-11.sgm

P1: IKH LaTeX2e(2002/01/18) 10.1146/annurev.psych.56.091103.070320

Annu. Rev. Psychol. 2005. 56:287–314 doi: 10.1146/annurev.psych.56.091103.070320 c 2005 by Annual Reviews. All rights reserved Copyright
First published online as a Review in Advance on September 10, 2004

Annu. Rev. Psychol. 2005.56:287-314. Downloaded from arjournals.annualreviews.org by Ball State University on 01/05/09. For personal use only.

THE PSYCHOLOGY AND NEUROBIOLOGY OF SUICIDAL BEHAVIOR Thomas E. Joiner Jr., Jessica S. Brown, and LaRicka R. Wingate Psychology Department, Florida State University, Tallahassee, Florida 32306-1270; email: [email protected], [email protected], [email protected]

Key Words genetics, serotonergic dysregulation, psychological risk ■ Abstract Suicide is a leading cause of death, but it is not well understood or well researched. Our purpose in this review is to summarize extant knowledge on neurobiological and psychological factors involved in suicide, with specific goals of identifying areas particularly