TIME, INTERNAL CLOCKS AND MOVEMENT
TIME, INTERNAL CLOCKS AND MOVEMENT
ADVANCES IN PSYCHOLOGY 115 Editors:
G. E. STE...
81 downloads
1058 Views
15MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
TIME, INTERNAL CLOCKS AND MOVEMENT
TIME, INTERNAL CLOCKS AND MOVEMENT
ADVANCES IN PSYCHOLOGY 115 Editors:
G. E. STELMACH E A. VROON
ELSEVIER Amsterdam
- Lausanne
- New York - Oxford
- Shannon
- Tokyo
TIME, INTERNAL CLOCKS AND MOVEMENT
Edited by Maria A. PASTOR and
Julio ARTIEDA Department of Neurology and Neurosurgery Universi~ Clinic University of Navarra Pamplona, Spain
1996
ELSEVIER Amsterdam
- Lausanne
- New York - Oxford
- Shannon
- Tokyo
NORTH-HOLLAND ELSEVIER SCIENCE B.V. Sara Burgerhartstraat 25 P.O. Box 211, 1000 AE Amsterdam, The Netherlands
ISBN: 0 444 82114 7 9 1996 Elsevier Science B.V. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior written permission of the publisher, Elsevier Science B.V., Copyright & Permissions Department, P.O. Box 521, 1000 AM Amsterdam, The Netherlands. Special regulations for readers in the U.S.A. - This publication has been registered with the Copyright Clearance Center Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923. Information can be obtained from the CCC about conditions under which photocopies of parts of this publication may be made in the U.S.A. All other copyright questions, including photocopying outside of the U.S.A., should be referred to the copyright owner, Elsevier Science B.V., unless otherwise specified. No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. This book is printed on acid-free paper. Transferred to digital printing 2005
PREFACE
Interest in the concept of time has a long history and has been a topic of study for a wide range of investigations. No change can take place without specification of time. While philosophers and physicists have been intrigued by the concept of subjective perception of time and its relationship to real time, natural scientists have been concerned mainly with investigating time as a factor in understanding the behaviour of animals from the migratory habits of birds to the periodical breeding cycles. The immense bulk of temporal perception studies, the variety of approaches, methods of measurement and even terminology has led to a difficulty, in reaching a global interpretation of the results. The aim of this book is to give an integrative approach to time sense and to focus the analysis to temporal factors in the processing of movement, trying to link temporal perception studies in the final common pathway, that is motion. To give some clues of human brain integrative processes at higher levels. And, finally, to clarify the neurophysiological substrate of these operations. The scheme will be based on a computer model for generation of voluntary movement, discussing its value at certain stages. This model is the basis of either animal and human studies although for the latter it serves deficient. The first step in order to make a movement is the decision to act, which escapes physiological analysis. This is followed by the "motor plan", which answers the questions "where?, when? and how?" to move. Planning the action involves the right spatial and temporal perception of either body, and environment, including the assembly of a range of instructions to execute the movement. These "motor programs" include not only simple sequences specifying the muscle activity of agonists, antagonists, synergists and postural fixators, but also complex subroutines. Finally, the plata is executed, which involves initiating the movement sequence, running the necesssary programs and controlling the course of the movement. The first chapter centres the attention in the hypothetical mechanisms of perception" is it perception a continuous function or is it fractioned?, and
vi
Preface
includes the discussion of the different approaches to study time perception and its participation on movement. From the second to file fifth chapters we tackle the basic knoxsaa mechanisms of temporal perception on animal research. The sixth chapter deals with the characteristics of time perception in humans. An integrative approach to the diverse types of perceptual mechanisms by human brain and their relationship with higher ~nctions is contemplated in the seventh chapter. This followed by a more general psychological approach from chapter eight, on the role of attention in time perception, to ten: a wide review of studies of human temporal perception. From chapter eleven, the study is centred in timing in motor tasks, starting from the role of central pattern generators and continuing with simple movements: why do we use such a long time reacting to a signal with a simple movement? In the last three chapters of the book we tackle the placement of the nervous system eentres responsible for time perception and integration with motor tasks. We think the book will be of enormous interest to neuroscientists especially those who are involved in temporal perception and motor control, giving an updated and at the same time integrative approach to the topic.
Mafia A. Pastor Julio Artieda Pamplona November 28, 1995
vii
TABLE OF CONTENTS
Neurophysiological Mechanisms of Temporal Perception .................... Julio Artieda and Mafia A. Pastor Processing of Temporal Information in The Brain ................................ Catherine E. Carr and Satoshi Amagai
27
Large-Scale Integration of Cortical Information Processing ................. Steven L. Bressler
53
Models of Neural Timing .................................................................... Christopher Miall
69
Neuronal Mechanisms of Biological Rhythms ...................................... Hugo Arechiga
95
Human Vs. Animal Time .................................................................... 115 Joim Campbell Time and Psycho-Physical Integration ......................................................... 127 Rafael Alvira The Role of Attention in Time Estimation Processes ............................ 143 Dan Zakay and Richard A. Block Reconstruction of Subjective Time on the Basis of a Hierarchically Organized Processing System ....................................... 165 Ernst P6ppel Time Perception Measurements in Neuropsychology ........................... 187 Paolo Nichelli The Development of Central Pattern Generators for Vertebrate Locomotion ........................................................................ 205 Keith T. Sillar
viii
Table of Contents
An Hierarchical Model of Motor Timing ............................................. 223 Brian L. Day Involvement of the Basal Ganglia in Timing Perceptual and Motor Tasks ................................................................................. 235 Mafia A. Pastor and Julio Artieda Exploring the Domain of the Cercbollar Timing System ....................... 257 Sean Clarke, Richard Ivry, Jack Grinband, Seth Roberts and Naomi Shimizu Timing in Perceptual and Motor Tasks After Disturbances o f the Brain ........................................................... 281 Nieole Von Steinbiiehel, Mare Wittmann and Ernst P6ppel Author Index ...................................................................................... 305 Subject Index ...................................................................................... 307
ix
LIST OF CONTRIBUTORS
Rafael Alvira Departamento de Filosofia Pr,'ictica Facultad de Filosofia Universidad de Navarra Pamplona, Spain Satoshi Amagai Department of Psychology, University of Maryland, College Park MD 20742-4415, USA Hugo Ar6chiga Divisi6n de Estudios de Posgrado e Investigaci6n Facultad de Medicina Universidad National Aut6noma de M6xico AP 14-740, 07000 M6xico DF, M6xico Julio Artieda Departamento de Neurologia y Neurocirugia Cliniea Universitaria Universidad de Navarra 31080 Pamplona, Spain Richard A. Block Department of Psychology, Montana State University Bozeman USA Steven L. Bressler Center for Complex Systems Florida Atlantic Univ. 777 Glades Road Boca Raton FL 33431 USA
List of Contributors
Catherine E. Carr Department of Zoology University of Maryland College Park Maryland 20742-4415, USA John Campbell Department of Physolophy New College Oxford OXI 3BN, U.K. Sean Clarke, Department of Psychology. University of California, Berkeley, CA 94720 USA Brian L. Day MRC Human Movement and Balance Unit The Institute of Neurology Queen Square London WCI 3BG, U.K. Jack Grinband, Department of Psychology. University of California, Berkeley, CA 94720 USA Richad Ivry Department of Psychology University of California Berkeley Berkeley California, U.S.A. R.C. Miall University Laboratory of Physiology Parks Road Oxford OXI 3PT, U.K.
List of Contributors
Paolo Nichelli Clinica Neurologica Universitfi di Modena Via del Poxxo, 71 41100 Modena, Italy Jos6 A. Pastor F.E.R.T. Inmaculada, 22 08017 Barcelona, Spain Maria A. Pastor Departamento de Neurologia y Neurocirugia Clinica Universitaria Universidad de Navarra 31080 Pamplona, Spain Ernst P6ppel Forschungszentrum Jiilich GmbH Postfach 19 13 D-5170 Jiilich, Germany Seth Roberts, Department of Psychology. University of California, Berkeley, CA 94720 USA Naomi Shimizu, Department of Psychology. University of California, Berkeley, CA 94720 USA Keith T. Sillar Gatty Marine Laboratory School of Biological and Medical Science The University of St. Andrews Fife KYI6 8LB Scoland
xi
eo
Xll
List of Contributors
Nicole Von SteinbQchel-Rheinwall Institut fiir Medizinische Psychologie Ludwig-Maximillians-Universit~t Goethestr. 31 W-8000 Munich 2, Germany Mare Wittmann Institut fiir Medizinische Psyclaologie Ludwig-Maximillians-Universit~t Goethestr. 31 W-8000 Munich 2, Germany Dan Zakay Department of Psychology Td-Aviv University 69978 Ramat-Aviv Israel
xiii THE MEMOIRS OF A SUNDIAL
Time kills, and we clocks are no exception to that. Some of us die by soRening, as if we were wax exposed to the sun, like in Dali's picture. Others expire from mechanical exhaustion. And some die of lack of sunlight. That is what is happening to me. Here in the village square, where I have marked time for my neighbours for two hundred years, I am now dying in the shade. I can see skyscrapers rising up that rob me of light, and my venerable iron needle is redundant. The tragic words carved around my face -"tempus fugit" - are no longer read. But I am still here, part of the grand west front of the church, as a silent witness to prompt the memory of the forgetful. Every hour that has swept across my face has been multiplied by thousands: my time, time without a voice, the cold time of stone, was different and vital for each and every one of the people around me. Now that my useful life can be counted by the seconds, I believe that time is above all simultaneity. I do not know much about these firings, but I believe tlmt I measured many lives at the same time: in one hour, all possible human experiences could be going on, together: suffering, pleasure, anxiety, happiness, hope, love, resentment, and many more. Every villager made what I thought was my time into his or her time, into currency of gold, silver, copper, clay or thin air, in exchange for what he or she thought would make for happiness. In one hour, the throb of a baby's heart in its mother's womb might be heard for the first time, but also the last feeble heartbeats of an old man on his deathbed. There might be a wedding, and at the same time someone might be betrayed; a hope might arise, or a prayer be answered. Now that the shadows are going to force my retirement, and I am going to join the sad category of things designated museum pieces, I realise that the time which I marked in my routine way is a gift, a treasure hidden in the hearts of human beings. Something that once happened in my beloved village comes nostalgically to mind. A large family was blessed with another child. The father announced to the children "An angel has brought you a little brother!" The children answered in chorus '%Ve want to see the angel!" If someone asks what time is, how time passes or what a clock has inside it, it would be more important for that person to appreciate the fruits of time. Because it is ha time, only in time, that everything that is good, worthy, beautiful and noble can be done. Time itself, which makes possible all human achievements, will flow into etemity as the fiver flows into the sea, and the sediments of work well done will remain for ever on the shores of the world of men.
xiv
J.A. Pastor
I shall end here. My face still says tempusfugit, but I prefer to say that it does not pass, but rather it transcends: time bears you humans onwards to culmination, carrying you along through your straggles to be yourselves. Time dies in what is circumstantial, but is resurrected in life itself. Michael Ende, for whom the bells on my tower rang recently, wrote in "Momo" that "time is life, and life dwells in the heart". The heart may stop beating, but it does not die; it travels up to the Always which God chose to furnish for the etemity of humankind in the felicity of living contemplation. When we clocks die, the khronos dies; but the kairos, which is life in its full maturity, is perpetuated. Time is beautiful, time is part of humanity's inheritance. These words come to you from a clock which is about to retire from its function as an assistant in your activities.
Jos~ Antonio Pastor Barcelona, September, 1995
Time, Internal Clocks and Movement M.A. Pastor and J. Artieda (Editors) 9 1996 Elsevier Science B.V. All fights reserved.
NEUROPHYSIOLOGICAL PERCEPTION
MECHANISMS
OF
TEMPORAL
JULIO ARTIEDA and MARIA A. PASTOR Clinical NeurophysJology Section. Department of Neuroh~gy and Neurosurgery. ClJnica UnJversJtarJa de Navarra. UnJversJdad de" Navarra. Pamplona. A'pain. ABSTRACT. Tinting is an essential component of perception, movemenl and behaviour, yet we still have liltle idea of how time is represented in the brain. In this chapter we review some aspecls of temporal perception: perception of simultaneity, temporal discrimination and perception of duration. In our opinion these categories are representative of different neural mechanisms. Some h.~1~otheses and models are discussed.
I. Introduction
The temporal analysis of infomlation is one of the basic mechanisms of cerebral function. The central nervous system continuously receives information which should be analysed both spatially and temporally for an adequate perception of the environment. Ever), cognitive process and every motor response should, in turn, be sequentially organised following temporospatial patterns. The physiological mechanisms of spatial processes are fairly well understood. The spatial or other qualities of a stimulus are codified spatially following a colunmar pattern. These columns are, in turn, distributed somatotopically, tonotopically etc. The mechanisms which allow the central nervous system to analyse and integrate the impulses from the different columns and areas in order to form a unitary perception of an object, at a temporal point of time, are poorly understood. One approach, to understand the way in which time is represented in the brain, has been the study of the perception of time using a psychophy,sical methodology and more exceptionally using a double psychophysical and neurophysical procedure. Temporal perception has been studied by breakingit do~al into the different features or levels of complexity:
2
J. Artieda and M.A. Pastor
perception of simultaneity or non simultaneity of two stimuli, temporal discrimination, discrimination of temporal succession and order, perception of duration, perception of the now and perception of sensation o f the time .flow. This chapter reviews some of these categories of the perception of time which are representative of the basic physiological and psychological mechanisms. Some of the facts, h)1~otheses and models which may help to understand more deeply the neurophysiological substrate of temporal perception are discussed.
2. Perception of simultaneity The threshold of simultaneity, is understood as the necessary minimal interval between two stimuli applied to different sites so that a particular subject is able to discriminate whether or not the stimuli have been applied simultaneously. The threshold of simultaneity, which is required for two auditory stimuli applied dichotically (one stimulus to each ear) to be perceived as not being simultaneous, is approximately 0.5 milliseconds. The interval needed for the subject to differentiate between the two stimuli increases from 3 to 20 milliseconds and it is only above these values that it is possible to discriminate the order of the two stimuli or, in other words, to discriminate which of the two stimuli was first applied (discrimination of the order of succession) (Hirsh and Sherrick, 1961). If two visual stimuli are applied to different areas of the visual field, with a small interval between them. the subject perceives movement of the stimulus. The temporal limit of this movement seems to be defined by Korte's Law (1915) and depends on the intensity and distance of the stimulus. With two visually adjacent stimuli on the fovea lasting 100 milliseconds, with an interval of 3 milliseconds between them. it is possible to perceive apparent movement (Westheimer and McKee, 1977). The perception of movement is optimal with an interval of 20 ms. (Kahneman and Wolman. 1970) . The same occurs in the tactile modality when two stimuli applied to different sites and not simultaneously, produce a sensation of movement (Klemm, 1925) at intervals at which it is still not possible to temporally discriminate the independence of the two stimuli (Nichelli, 1993 ). The perception of simultaneity is of great importance in daily life when spatially locating auditory stimuli. Auditory stimuli which originate in one side of the sound field reach the ear of that side of the body first and the
Neurophysiological Mechanisms of Temporal Perception
3
perception of non simultaneity between the two stimuli is perceived as a lateralization of the sound. In the visual and somaesthetic sphere the perception of non simultaneity of two stimuli may constitute a basic element in the codification and perception of movement. The thresholds of simultaneity have been studied in different animal species and in different sense modalities. Simos (1973, 1979) studied the problem of the detection of small temporal differences between stimuli by demonstrating that the bat can detect changes as small as 500 nanoseconds in the arrival time of jitter sonic echoes. This sensitivity in the detection of small time differences between stimuli can also be found in other animals and other sense modalities which include the discrimination of phase differences from signals in different parts of the body of electric fish (Carr et al., 1986; Rose and Heiligenberg, 1985) and the detection of binaural phase differences in the location of sounds in owls (Moiseff and Konishi, 1981), bats (Masters et al. 1985) and many other animals (Heffner and Heffner 1992). Jeffress in 1948 was the first to put forward a model to explain the enormous capacity animals have to detect the non simultaneity of stimuli applied at short time intervals. The model is based on the place theory, for the detection of interaural time differences where the simultaneity or non simultaneity of a stimulus and the interval between them is codified in the form of space. Figure 1 shows a diagram of the model as applied to the binaural auditory stimuli. The circuit comprises of two elements" delay lines and coincidence detectors. The delay lines are made up of axons and the amount of delay by means of the length of the axons. The coincidence detectors are neurones which respond when they receive simultaneous input from both ears. When two auditory, stimuli are applied to each one of the ears simultaneously, the neurone or coincidence detector which responds is that which receives the input from the ears simultaneously and therefore is the neurone in which the delay introduced by the length of the axons is similar. If the two stimuli are not simultaneous the neurone which responds is that in which the time interval between the two stimuli has been compensated for by the difference in the delay introduced by the length of both axons or delay lines. The Jeffress model explains not only how small time differences between the two stimuli (non simultaneity) can be detected but also how these small time differences can be codified. In this way a time interval is codified in the form of space.
J. Artieda and M.A. Pastor
Right
Delay line
I I I l t I I t-'
~
:~
= -.
=" q-
~"
~"
t"
"*
o
Left
Figure 1. Schematic diagram (Jeffress model) showing a circuit for the detection of small time differences between two stimuli applied to the right and left ears.
The Jeffress model or circuits which are very. similar have been demonstrated in different sensor~, organs of lower animals (Sullivan and Konishi, 1986: Carr and Konishi, 1990) and in the auditory system of both mammals (Smith et al., 1990) and birds (Young and Rubel, 1983" Rubcl and Parks, 1975; Overhoit et al., 1992). In these cases the axons of the cochlear nucleus act like delay lines and the laminaris nucleus or the neurones of the superior olive act like coincidence detectors. The circuit is thus able to measure and codify interaural time differences (Carr, 1993). Physiological studies of the auditor3, system in mammals have found responses consistent with this model. In the cat, cells Iocalised in the nucleus of the medial part of the superior olive respond to i,teraural time differences of between 0 and 400 microseconds (Yin and Chan, 1990). The localisation of neurones is organised in such a way that the neurones which respond to the small time differences are Iocalised in the anterior pole of the olive, while the neurones
Neurophysiological Mechanisms of Temporal Perception
5
which discharge at much larger time differences are localised more posteriorly in the nucleus. In this way a neuronal map is drawn up of the interaurai time differences. This system of delay lines, based on the length of the axons, allows the codification of very small time intervals but fails to explain the codification of longer time intervals and complex temporal sequences (see Carr and Amagai in this book).
3. Temporal discrimination Temporal discrimination is a perceptive temporal fi~nction whose threshold can be defined as a the n~inimum interval which is necessary between two successive stimuli with the same characteristics (somaesthetic, visual or auditory) and applied to the same site, for the stimuli to be perceived as separate in time (Artieda et al., 1990, 1992: Lacnlz et al., 1991). Temporal discrimination shares certain common characteristics with simultaneity and successive discrimination but differences between them do exist. The tern1 temporal discrimination is used when the characteristics of the stimulus are habitually similar and the stimulus is applied to the same site for which reason both stimuli share the same transmission pathways to the cerebral cortex. When referring to discrimination of simultaneity and of succession it is understood that the stimuli have different characteristics or that they are applied to different sites, for example both ears, different areas of the visual field, etc. Within the temporal discrimination threshold it is possible to distinguish a descending or fusion threshold and an ascending or discrimination threshold. The ascending or discrimination threshold is obtained by increasing the time intervals between the two stimuli until the patient begins to perceive them as separate in time. Conversely, the fi~sion threshold is the opposite and is determined by reducing the time interval between two stimuli until they become fused, being perceived by the subject being examined as one. Both thresholds establish a discriminative range or shadowland (Lacnlz et al., 1991; Artieda et al., 1992). One variety of the filsion threshold is the flicker fusion threshold which corresponds to the fusion threshold of a string of stimuli which the subject perceives as one continuous flow (Pieron, 1952). The thresholds of temporal and filsion discrimination vary depending on the sense modality. Artieda et al. in 1992 obtained mean values of temporal
6
J. Artieda and M.A. Pastor
J]E
~
....o
J] rL .. . . . . . . . . . ~ . .
o
""..~
~'i ~ f ~ t
9 -
.....
"
.9
T i , ~ n
. '. . .
t t t Vt
.O .... O ." ,'
....
,' .
."
o, C~xlificatim,."
,
_1 L ~ _ F I PL
O ,,.
J
.O ,.
,. ..
.,
:."
L~J]
'~
9 .o
I-L
Stimuli
A
A " 9 :
6 1
O
~
I
:
O
I
O
"
9
I
I
I: o
t
,,,9
E!
;
I'~
"
~
,
dr,.,,
~1
~
~
"
[I
Interstim
ulus
,
Interval
Figure 2. Schematic diagrams showing different models to explain the neural bases of temporal discrimination
discrimination (the mean between the discrimination threshold and finsion threshold) in three sense modalities: somaesthetic, auditory and visual 33.9 _+ 12 milliseconds, 18.1 +_ 7.2 milliseconds and 68.7 _+ 5.1 milliseconds respectively. These thresholds are in agreement with those referred to by other authors (Uttal, 1959: Green et al., 1961; Green, 1984).
Neurophysiological Mechanisms of Temporal Perception
7
The neurophysiological mechanisms which underlie the phenomenon of temporal discrimination are poorly understood. Three basic mechanisms are proposed as hypotheses, as illustrated in Figure 2. 3.1 TEMPORAL DISCRIMINATION AND THE PHYSIOLOGY OF SENSORY PATHWAYS Following the conduction of a stimulus there is a refractory period during which time the sensor3, pathway cannot conduct nerve impulses. This characteristic of all pathways is due to the behaviour of the nerve fibre membrane of both peripheral and central nerves, and to the excitability cycles of the relay nuclei (Shagass and Schwartz, 1964). This physiological characteristic of pathway,s may determine the discrimination thresholds and form the physiological basis for temporal discrimination as illustrated in Figure 2A. Artieda et al. (1992) and Lacruz et a1.(1991) demonstrated that there is no correlation between the thresholds of somaesthetic temporal discrimination and the excitability cycle of the sensory action potential of the median nerve or of the primary cortical components of the somatosensory evoked potential (Figure 3 ). The sensory nerve potential of the median nerve generated by the second stimulus was found to have completely recovered in the interstimulus intervals during which time no subject was able to discriminate between the two stimuli. A similar situation was observed when the potential was evoked somatosensorally: healthy subjects incapable of discriminating between the two stimuli during these interstimulus intervals presented the somatosensou, evoked potential as having completely recovered. In patients with Parkinson's disease this finding was more remarkable because of the higher temporal somaesthetic discrimination threshold associated to normal recovery, curves of the sensor3., nerve potentials and to normal somatosensorv evoked potentials (Figure3). When levodopa was administered an improvement in the temporal discrimination threshold was observed, parallel to the motor improvement, but no changes in the excitability cycles of the pathway or of the somatosensor 3, cortex (area 3b, mainly) were noted. These findings allow us to discard characteristics of nerve conduction as the basic mechanism of temporal discrimination. However, the excitability cycle of the pathway may have a limiting effect on temporal discrimination thresholds. Patients who suffer changes in the somatosensory pathway at the central or
J. Artieda and M.A. Pastor
peripheral level, such as polyneuropathy, lesions at the level of the sensory. thalamus or at the level of the primary somaesthetic cortex, show increases in the temporal discrimination thresholds. This increase in the threshold is produced by a change in the excitability cycles or at least by a despolarization volley dispersion or an increase in the background noise during the transmission of the corresponding sensor}, message. (Figure 2A and 3).
SEP cm~ T
~, 12D
SNAP P
~
Parklm~ i
] ......... I I
g
o 1(1) 80
."lO0
.( o
g
"U o o
"--"
0
LI
!
0
I
100
~
InlmllmJ I n m ~ n l
[,,- d . . .
0
0
J
i
100
ZD
I r W m n J I r m ~ ns
~ Sonmesthetictemporal disscrindnation threshold
Figure 3. Recovery curves for senson.' nerve action potential (SNAP) and cortical somatosensory evoked potential (SEP) following paired median nerve electrical stimulation in the wrist, in nomlal subjects and patients with Parkinson disease No differences in either curves were found but significant differences in somaesthetic temporal discrimination threshold were obtained between normal subjects and patients. 3.2 TEMPORAL DISCRIMINATION AND CODIFICATION IN SPACE. The second hypothesis explains temporal discrimination in terms of the codification of time in space, in a way similar to that previously mentioned
Neurophysiological Mechanisms of Temporal Perception
9
for the perception of simultaneity. Figure 2B illustrates one of the possible multiple neuronal networks. The first stimulus initiates and activates a neurone which via a collateral axon sequentially excites a string of neurones or coincidence detectors. The second stimulus, applied after a time interval with respect to the first, activates tl~e entire string of neurones with a similar delay in all of them. When there is a temporal convergence on one the neurones, from the second stimulus and the first impulse emanating from the collateral of the first neurone, a discharge is produced and the time interval between the two stimuli codified. Thus collaterals of the axon of the main neurone and of the chain of corresponding intemeurones act as delay lines as in the Jeffress model for the detection of non simultaneity. This system allows for a high resolution of temporal discrimination thus enabling the codification of intervals which last only a few milliseconds. However, physiological features of the pathways and relay nuclei derived from the refractory state and the excitability cycle, produce limitations within the system. In fact, the temporal discrimination thresholds in the sensory, visual and somatosensor)., modalities are higher than might have been expected with this model. 3.3 INTERNAL CLOCKS AND RHYTHMIC ACTIVITIES IN THE CENTRAL NERVOUS SYSTEM. A third possibility is the presence of internal clocks which permit the codification of time intervals. This system of internal clocks would be the only way to explain the mechanisms of the perception of the order of succession of two or more stimuli. One of the possible models is shox~.~ diagramatically in Figure 2, where the temporal codification of a double stimulus in the central nervous system resembles the analogue-digital conversion of computing systems. Treisman et ai. in 1990 calculated that the theoretical level at which the frequency of discharge of the pacemaker, in a similar model, would be found to be around 49.5 Hz. Popel suggested that the central nervous system works in a discontinuous way at intervals of 35 milliseconds based on the multimodal behaviour of choice reaction time histograms. This value is derived from the interval between neighbouring modes. Other authors had previously proposed a similar hypothesis (Stroud. 1956). An internal clock can be imagined as an oscillating circuit with a single neurone or group of pacemaker neurones, capable of generating rhythmical
10
J. Artieda and M.A. Pastor
electromagnetic activities in the brain which can be used in the temporal analysis of information. Rhythmical activities have been described since the beginning of clinical neurophysiology. Berger (1929), in his first descriptions of electroencephalographs of man, described alpha activity as a rhythmic activity with a frequency of about 10 Hz. Other rhythmic activities have been described in electroencephalographs such as the mu rhxehm in Rolando's area which disappears when movement takes place or sleep spindles (Hobson, 1988). Rhythmic activities in response to different sensory stimuli have also been described in man. Galambos in 1982 registered in the scalp an activity of 40 Hz which was evoked when an auditory or somatosensory stimuli of the same frequency was applied. More recent studies have confirmed these potentials at 40 Hz (Spidell, Pattee and Godie, 1985; Pantev et ai., 199 I, 1993). A rhythmic auditory, stimuli generates an activity of the same frequency on the cerebral cortex. At 40 Hz these potentials have a greater amplitude which may be explained by their entering into resonance with the frequency of the system, in this case the neurone groups of the auditor), cortex (Artieda and Pastor. 1995 unpublished findings). (Figure 4). There has been much debate with regard to the significance of these steady state potentials as well as their origin which seems to be demonstrated to be at the level of the auditory cortex, although they may be modulated by thalamocortical systems. (M~ikela and Hari, 1987; Firsching et al., 1987: Steriade et al., 1991 ). Not only does the application of a rhythmic stimulus generate responses at 40 Hz but an isolated stimulus such as a tone burst generates potentials at this frequency of 40 Hz, the so-called responses of the gamma band. These activities at 40 Hz (transient and steady) have been studied by both electrical and magnetic means and have demonstrated their origin at the level of the auditory cortex but in different areas (Pantev et al., 1991, 1993). Recently, the Singer group (Engel et al., 1991, 1992) has described, in the visual cortex of the cat, oscillatory activity at around 40 Hz in the discharge of neurones of a column when activated by a stimulus. Cross correlation studies show that the neurones of the visual cortex synchronise their response depending on how coherent the characteristics of the stimulus applied in the visual field are. If two columns are activated by one single stimulus or object, these columns synchronise their rhythmical discharge, however they do not if they are activated by different objects or stimuli(Engel et al., 1991, 1992). These studies indicate that the oscillator),
Neurophysiological Mechanisms of Temporal Perception
11
A 30 Hz
40 Hz
50 Hz
lOOm
I10 HZ
C
I 0"~ uv
~
0 t
!
,
1 ....
t
....
|
J
!
123 i
~. 0.8 0.6 i
0.4 0.2
! 20
,,
i
i
40
60
~
I 80
i O0
120
FretlM-K'pIHzl
Figure 4. A) Steady state potential evoked by right auditory stimulation (clicks 70db) at different rate. B) Fast Forrier transform (FFT) of the steady potential showed in A. 40 Hz stimulation evoked activity at same frequency of maximal amplitude as showed in A. B and C.
behaviour of these neurones to sensor), stimuli can be of great advantage when processing sensory information. The synchronisation between neurones of different columns activated by the same object or complex stimulus which codi~ different qualities can be a means of integration or solution to the "binding" problem. This synchronisation of the discharge of different columns may constitute a basic element in order to explain any perceptive phenomenon. Activity at 40 Hz is not exclusive to sensory cortical areas and has also been described in motor areas. In man, activity at 40 Hz precedes voluntary movements (Pfurtscheller, FIotzinger and Neuper, 1994). The oscillatory activity at 40 Hz, very often found in the central nervous system of
12
J. Artieda and M.A. Pastor
mammals (Basar et ai., 1987: Basar-Eroglu and Basar, 1991; Llinas, Grace and Yarom, 1991), can be thought to play an important role in the temporal codification of sensory infommtion and in the processing of motor and cognitive sequences.
4. Perception of duration The perception of duration is found at a higher hierarchical level than the perception of time, the discrimination of simultaneity, temporal discrimination or the perception of succession. The perception of duration is also a basic element in the perception of time. There are various methodological means of studying the perception of a time interval duration: the verbal estimation, temporal production, temporal reproduction and comparison of duration. All these methodological procedures involve mechanisms which are used in daily life in the perception of time. Perhaps the most typical and characteristic of all of these is the comparison of duration. Whenever we try. to evaluate an interval of time, we draw a comparison with a previous temporal experience. If what we are tr3,ing to perceive is an estimate of the duration of a time interval, by means of a verbal response, it is necessary., firstly, to convert by means of prior experience the duration of a second, and secondly to establish a comparison. This same procedure could be applied to the reproduction of a time interval or the production of an interval. Our perception of time is always relative to our experience of the duration of a temporal unit (a second, a minute etc.) Thus the comparison of two time intervals of differing duration, the threshold of discrimination of the duration of those tavo time intervals will be dependant on the duration of those intervals. It is easy to discriminate differences of two second intervals in the range of seconds but it would be impossible to distinguish the same differences between time intervals in the range of hours of duration. This relativity in the perception and estimation of the duration of time intervals allows us to put forward models based on accumulators or stores and comparators. Figure 5 proposes a modification of the model of duration judgements proposed by Allan and Gibbon in 1991 and modified by Nichclli (1993). The model proposes the necessity of a pacemaker or internal clock. This clock interacts with the infommtion which is codified and transmitted to
Neurophysiological Mechanisms of Temporal Perception
13
the cerebral cortex. This in turn is stored in a sensor3' memory or a working memory to be compared with successive experiences.
[Stimulus !
T ............. ,
Switch
,
i PacemakerI ~ IDurationencoding
i Cons ,ousness I _.
v
l a cc"mu'a'~ I
Comparator I
1 Decision
Figure 5. Schematic diagram of tile model for duration jugements modified from Allan and Gibbon (1991 ) and Nichclli (1993). All these levels present a neurophysiological correlation: the pacemaker or internal clock can be correlated with the oscillator3, activities demonstrated in animals and in man at frequencies in the range of 40 Hz already mentioned in the previous section, the transmission of precodified information with long latency evoked potentials and the process of storage and comparison with mismatch negativity. These neurophysiological correlations are discussed below. 4.1 CODIFICATION AND LONG LATENCY EVOKED POTENTIALS The application of an auditor 3, stimulus (tone burst of a particular frequency) generates a series of waves with a latency greater than 50 ms
14
J. Artieda and M.A. Pastor
which vary according to the frequency of the tone, the intensity, the interstimulus interval and the duration. With acoustic stimuli of more than 100 ms it is possible to differentiate between a complex generated by the onset of the stimulus, "onset potential", and a complex evoked by the end of the stimulus "offset potential" (Picton, Woods and Proux, 1978a, 1978b). Between these a negativity called the "sustained potential" (Picton, Woods and Proux, 1978a, 1978b) is maintained (Figure 6). 4.1.1 Onset Potential The onset potential is formed by a positive wave with a small amplitude or P 1, a large amplitude N I wave, and a positive wave P2 (Pieton et al., 1974; Davis and Zerlin, 1966). The amplitude in these potentials is greater at the vertex and therefore they are also known as vertex complexes. The voltage in electrodes at these locations oscillates around 5 uv. The latency at the peak of the N 1 wave oscillates between 75 and 100 ms and that of the P2 wave between 130 and 200 ms. The amplitude of the complex correlates exponentially and positively with the duration of the interstimulus interval. The increase in the intensity of the stimulus increases the amplitude, reaching a saturation level at high intensities which is accompanied by a reduction in latency. A feature of the stimulus which considerably determines the amplitude is the duration. Short duration stimuli generate waves of slight amplitude which grow quite considerably as the duration increases. Saturation level is reached with a duration of between 30 and 50 ms (Artieda, Moraes and Pastor, 1992). Other characteristics of the stimulus also condition variations in amplitude. Thus, stimuli with high frequencies generate waves with lower amplitudes and tones with brusque starts generate larger amplitudes (Picton, Woods and Proux, 1978b; Picton 1990). 4.1.2 Offset potential The offset potential presents latencies which correlate linearly and positively with the duration of the stimulus. They present a similar morphology to the stimulus of the onset potential differentiating themselves by an N I' wave and a P2' wave. The amplitude of the N 1' wave is usually less than that of the P2' wave, in contrast to the onset potential (Picton, Woods and Proux, 1978a, 1978b). The latency of the peak of the N 1' wave with regard to the end of the stimulus is 73+74 ms and that of the P2' wave 145+16 ms. The amplitude of the closing complex is less than that of the opening complex. The topographical distribution of the potentials are
Neurophysiological Mechanisms of Temporal Perception
15
I0 Hz
7.~ Hz
I O0 Hz
lOOms I i w
Tone Burst ! 1000 Hz) II I I I
! 000 Hz
I
~
~
I
Click's Train
Figalre 6 Long latency auditory evoked potentials recorded over verlex. On the left, evoked activity by tone burst stimulation of different duration are showed.On the right, clicks train stimulation at different rates were performed
very similar having a greater vertex amplitude with certain anterior diffusion. The changes of the closing complex with regard to the characteristics of the stimulus have not been closely studied. There is an clear increase in the amplitude with regard to the duration of the stimulus. Long duration stimuli generate potentials with greater amplitudes, especially the P2' wave. Anaesthesia reduces the amplitude proportionally more than for the opening complex.
4.1.3. Sustained Potential The sustained potential is a slow negative potential between the opening and closing complexes which sustains itself as long as the stimulus lasts (Picton, Woods, Proux, 1978a, 1978b). With
16
J. Artieda and M. A. Pastor
stimulus duration times which are longer than 1 second the potential decreases its amplitude. Its topographical distribution is very similar to that of other components its maximum amplitude being in the frontal and vertex regions (Picton, Woods and Proux, 1978a, 1978b ). The sustained potential behaves similarly but not identically to the opening complex. Its amplitude decreases proportionally with regard to the interstimulus interval. This reduction is less than for the opening complex. The intensity of the stimulus is directly correlated with the anaplitude. Attention to the stimulus especially in discriminatory, tasks of duration generates a marked increase in amplitude (Picton, Woods and Proux, 1978a, 1978b" Picton, 1990). The generators of the opening and closing complexes in addition to those of the sustained potential have been studied using intracraneal registers and magnetoencephalographs. Although it seems that they each have a rather different origin they do this in regions which are very close together and to be precise in the supratemporal cortex of both hemispheres (Hari et al., 1980: Hari et al., 1989; Sama et al., 1991). The effect of orientation of the dipoles and the addition of electrical fields produces a greater amplitude in the vertex than when conventional electrodes on the scalp are used. The function of these potentials is not kno~a with certainty but it seems logical to think that they express a codification of the characteristics of a stimulus and presumably fore1 the first step in the perceptive chain. The closing potentials and probably the sustained potential may play a very important role in the codification of the temporal characteristics (duration) of the stimulus.(Picton, 1990). 4.2 COMPARATOR AND MISMATCH NEGATIVITY If, during a sequence of stimulation in which a stimulus with constant characteristics at a particular frequency is being applied, a stimulus with slightly different characteristics is introduced unexpectedly and randomly, the potential evoked by the unexpected stimulus will have components which are somewhat different to the potential which the frequent stimulus evokes. The biggest difference will be an N I wave with a large duration latency and amplitude. This wave is clearly differentiated when the evoked potential is subtracted from the infrequent or unexpected stimulus. This negativity is known as "mismatch negativity" (MMN) (Butler, 1968; N/i/at~nen, 1988; N/i~t~en, Sipson and Loveless, 1982). In records without substraction MMN is masked and can be seen as an N I wave with an increased
Neurophysiological Mechanisms of Temporal Perception
17
amplitude, like a second negativity or a reduction in amplitude of the P2 wave. The MMN is registered independently of the attention which the patient pays to the stimulus. It is even recorded during sleep (Nielsen-Bohlman et al., 1991). Studies with magnetoencephalographs have suggested that MMN originates in the supratemporal cortex predominantly in the right hemisphere independently of the ear stimulated. Their origin seems to be somewhat more anterior than exogenous components (Sama et al., 1991" N~t/'inen et al., 1989; N/i~it/inen, Simpson and Reinikainen, 1982" Sams and Hari, 1991). The significance of MMN is not known with certainty but may reflect activity generated by the automatic, unconscious or preattentive comparison of different stimuli. The.,,, can be interpreted at a higher level in the codification of the stimulus, in the perceptive or discriminative chain (N~/it/inen et al., 1989: N/'i~it/inen. Simpson and Reinikainen, 1982" Sams and Hari, 1991). There are no studies which show or discard their clinical interest. Their use may be of interest in the evaluation of discrimination and sensory, memory. The neurophysiological mechanisms of the cerebral process of temporal perception are now coming to light as are new areas of study of quantifiable parameters which are related to the findings of psychophysical findings or physiological experiences. Their use in combined studies is beginning to shed light on various aspects of temporal perception which up to now were understood only superficially, almost exclusively at the level of mere description.
ACKNOWLEDGEMENTS. This work was supported in part by a grant from the Spanish. DGICYT PB92-1~713. The authors wish to thank Mrs Isabel Sanchez and Maria Lopez for their hclp in editing this chapter.
5. References Allan GA, Gibbon J. Human bisection at the geometric mean. Learn. Motivat. 1991 22: 39-58.
18
J. Artieda and M.A. Pastor
Artieda J, Pastor MA, Lacruz F, Obeso JA. Temporal discrimination and bradykinesia in Parkison's disease. In: Berardelli A, Benecke R, Manfredi M, Marsden CD, editors. Motor disturbances II. London: Academic Press, 1990: 175-80. Artieda J, Pastor MA, Lacruz F, Obeso JA. Temporal discrimination is abnormal in Parkinson disease. Brain. 1992; 115: 199-210. Artieda J, Moraes W, Pastor MA. Potenciales evocados de larga latencia. In: Ciges M, Artieda J, Sainz M, Stingl de Mendez M, editors. Potenciales evoeados auditivos, visuales y somatosensoriales. Sevilla: Omega, 1992: 155-69 Basar E, Rosen B, Basar-Eroglu C, Greitschus F. The association between 40 Hz-EEG and the middle latency response of the auditory evoked potential. Int. J. Neurosci. 1987: 33" 103-17. Basar-Erosglu C, Basar E. A compound P300-40 Hz response of the cat hippoeampus, int. J. Neurosci. 1991; 60: 227-37. Berger H. Uber das elecktroencephalogramm des Menschen. Arch. Psyehiat. Nervenkr. 1929; 87: 527-70. Bindra D, Waksberg H. Methods and terminology in the studies of time estimation. Psychol. Bull. 1956: 63: 155-59. Block RA. Experiencing and remembering time. Affordances, context, and cognition. In: Levin I, Zackay D, editors. Time and human cognittion. A life span perspective. Amsterdam: North-Holland, 1989. Butler RA. Effect of changes in stimulus frequency an intensity on habituation of the human vertex potential. J Acoust. Soc Am. 1968; 44: 94550. Carr CE. Processing of temporal information in the brain. Ann. Rev. Neurosei. 1993; 16: 223-43.
Neurophysiological Mechanisms of Temporal Perception
19
Carr CE, Heiligenberg W, Rose G. A time-comparison circuit in the electric fish midbrain. I Behavior and phy'siology. J. Neurosci. 1986a; 6: 107-19. Carr CE, Konishi M. A circuit for detection of interaural time differences in the brainstem of the barn owl. J. Neurosci. 1990: 10" 3227-46. Davis H, Zerlin S. Acoustic relations of human vertex potential. J. Acoust. Soc. Am. 1966: 39: 109-13. Efron R. The effect of handcdness on the perception of simultaneity and temporal order. Brain. 1963a: 86 216-84. Efron R. The effect of stimulus intensity on the perception of simultaneity in right and left handed subjects. Brain 1963b; 86: 285-94. Engel AK, K0nig P, Kreiter AK, Schiilen TB, Singer W. Temporal coding in the visual cortex: new vistas on integration in the nervous system. Trends Neurosci. 1992; 15' 218-26. Engel AK, KOnig P, Krciter AK, Singer W. Interhemispheric synchronization of oscillator3, neuronal responses in cat visual cortex. Science. 1991" 252:1177-79. Fetterman JG, Killleen PR: A conaponential analysis of pacemaker-counter systems. J. Exp. Psychol. Hum. Percept. Perform.. 1990: 16: 766-80. Firsching R, Luther J, Eidelberg E, Brown WE, Stor3' JL, Boop FA. 40 Hzmiddle latency auditory, evoked response in comatose patients. Electroencephalogr. Clin. Neurophysiol. 1987; 67:213-6. Fraisse P. Time and rhx~hm perception. In Carterette EC, Friedman MP, editors. Handbook of Perception. New York: Academic Press. 1978' 203-54. Fraisse P. Perception and estimation of time. Ann. Review Psychol. 1984: 35" 1-36. Freeman WJ, Skarda CA. Spatial EEG-pattemls, non-linear dynamics and perception" the neo-Sherringtonian view. Brain Res. Rev. 1985: 10" 147-75.
20
J. Artieda and M.A. Pastor
Galambos IL Makeig S, Talmachoff PJ. A 40 Hz auditory potential recorded from the human scalp. Proc. NatI.Acad. Sci. USA. 1981; 78: 264347. Green DM. Temporal factors in psychoacoustics. In: Goldman P.S, editor. Time Resolution in Auditor3.' Systems. Berlin" Springer-Verlag, 19 9122- ! 39. Green JB, Reese CL, Pegues JJ, Elliott FA. Ability to distinguish two cutaneous stimuli separated by a brief time interval. Neurology 1961" 11" 106-10. Harter R, White CT. Periodicity within reaction time distribution and electromiograms. Q. J. Exp. Psychol. 1968: 20: 157-66. Hari R, Aitoniemi K, J~rvinen ML, et al. Auditory evoked transient and sustained magnetic fields of the human brain. Exp. Brain Res. 1980: 40: 237-40. o
Haft R, H~imafil/iinen M, Kaukoranta E, et al. Selective listening modifies activity of human auditor3' cortex. Exp. Brain Res. 1989; 74: 463-70. Heffner RS, Heffiaer HE. Evolution of sound localization in mammals. In" Webster DB, Fay RR, P6pper AN, editors. The Evolutiona~' Biology of Hearing. New York: Springer-Verlag, 19 9691 Hirsh Jl, Sherrick C EJ. Perceived order in different sense modalities. J. Exp. Psychol. 1961" 62: 423-32. Hobson JA. The Dreaming Brain. New York' Basic Books. 1988" 336. Jeffress LA. A place theor3' of sound localization. J. Comp. Physiol. Psychol. 1948; 41" 35-9. Kahneman D, Wolman RE. Stroboscopic motion: Effects of duration and interval. Percep.Psychoph. 1970: 8" 161-64. Killeen PR, Fetterman JG. A behavioral theory of timing. Psychol. Review 1988" 95" 274-95.
Neurophysiological Mechanisms of Temporal Perception
21
Killeen PR, Weiss JG. Optimal timing and the Weber function. Psychol. Review. 1987; 94: 455-68. Klem O. l~Iber die Wirksamkcit kleinster Zeitunterschiede im Gebicte des Tastsinns. Arch. Ges. Psychol. 1925" 50: 205-20. Korte A. Knematoskophische Untersuchungen. Zeitschrift fiir Psychologie. 1915; 72" 193-96. Lacruz F, Artieda J, Pastor MA, Obeso JA. The anatomical basis os somaesthetic temporal discrimination in humans. J. Neurol. Neurosurg. Psychiatr. 1991" 54" 1077-81. Llinas R, Grace A, Yarom Y. In vitro neurons in mamnaaliann cortical laver 4 exhibit intrinsic oscillator3, activity in the 10 to 50 Hz frequency range. Proc. Natl. Acad. Sci. USA. 1991" 88:897-901. .
M~kela JP, Hari R.Evidence for cortical origin of the 40 Hz auditor), evoked response in man. Electroencephalogr. Clin. Neurophysiol. 1987" 66" 539-46. Masters WM, Moffat AJM, Simmons JA. Sonar tracking of horizontally moving targets by he big bro~.~l bat Eptesicus fitscus. Science. 1985" 228: 1331-33. Moiseff A, Konishi M. Neuronal and behavioral sensitivity to binaural time differences in the owl. J. Neurosci. 1981" 1' 40-8. N~,t~inen R. Implications of ERP data for psycological theories of attention. Biol. Psycol. 1988" 26" ! 17-63. N~i~it~inen R, Picton TW. The N! wave of the human electric and magnetic responsse to sound" A review and an analysis of the component stn~cture. Psychophysiology 1987: 24: 375-425. N~i~it~inen R, Paavilainen P, Rcinikainen K. Do event-related potentials to infrequent decrements in duration of auditory stimuli demonstrate a memor3' trace in man?. Neurosc. Lett. 1989: 107' 347-52.
22
J. Artieda and M.A. Pastor
N~t~nen R, Sipson M, Loveless NE. Stimulus deviance and evoked potentials. Biol. Psychol. 1982. 14" 53-98. N ~ t ~ e n R, Paavilainen P, Alho K et al. Do event-related potentials reveal the mechanism of auditory sensor3' memory in human brain?. Neurosc. Lett. 1989; 98" 217-21. Nielsen-Bohlman L, Knight RT, Woods DL, et al. Differential auditor3' processing continues during sleep. Electroencephalogr. Clin. Neurophysiol. 1991; 79:281-90. Nichelli P. The neuropsycology of human temporal information processing. Handbook of Neuropsycology. 1993' 8 Omstein RE, editor. On the experience of time. Harmondsworth: Penguin, 1969. Overholt T, Hyson R, Rubel EW. A dealy-line circuit for coding interaural time differences in the chick brain stem. J. Neurosci. 1992; 12" 1698-708. Pantev C, Elbert T, Makeig S, Hampson S, Eulitz C, Hoke M. Relationship of transient and steady-state auditor)., evoked fields. Electroencephalogr. Clin. Neurophysiol. 1993" 88: 389-96. Pantev C, Makeig S, Hokr M, Galambos R, Hampson S, Gallen C. Human audotiry evoked gamma-band magnetic fields. Proc. Natl. Acad. Sci. USA 1991 b; 88" 8996-9000. Picton TW. Auditory, evoked potentials. In: Daly DD, Pedley TA, editors. Current practice of clinical EIcctroencephalography. New York: Raven Press 1990: 625-78. Picton TW, Woods DL, Proux Gb. Human auditory sustained potentials. I. The nature of the response. Electroencephalogr. Cin. Neurophysiol. 1978a: 45" 186-97.
Neurophysiological Mechanisms of Temporal Perception
23
Picton TW, Woods DL, Proux Gb. Human auditory sustained potentials. II. StimulUs relationships. Electroencephalogr. Cin. Neurophysiol. 1978b" 45' 198-210. Picton Tw, Hillyard SA, Krausz SJ, et al. Human auditory evoked potentials. I. Evaluation of components. Electroencephalogr. Clin. Neurophysiol. 1974: 36:179-90. Pi6ron H. The sensations. New Haven:CT. Yale University Press, 1952. P~ippel E. Oszillatorische Komponenten Naturwissenschaflen. 1968" 55 449-50.
in
Rektionszeitcn.
P6ppei E. Time perception. I: Held R, Leibowitz W, Teuber H-L, editors. Handbook of Sensory Physiology. Heidelberg: Springer. 1978" 713-29. Pfurtcheller G., Fiotzinger D., Neuper Ch. Differentiaton between finger, toe and tongue movement in based on 40 Hz. EEG. Electroencephalogr. Clin Neurophysiol. 1994: 90: 456-60. Rose G, Heiligenberg W. Temporal hyperacuity in the electric sense of fish. Nature. 1985" 318:178-80. Rubel EW, Parks TN. Organization and development of brainstem auditory nuclei of the chicken: Tonotopic organization of N. magnocellularis and N. laminaris. J. Comp. Neurol. 1975" 164" 411-34. Sams M, Hari. Magnetoencephalography in the study of human auditory information processing. Ann. N.Y. Acad. Sci. 1991 620: 102-17. Sams M, Kaukoranta E, Hfimfilfiinen M, et al. Cortical activity elicited by changes in auditor3, stimuli: Different sources for the magnetic Nl00m and mismatch responses. Psycophysiology 1991' 28' 2 I-9. Shagass C, Schwartz M. Recover3.' fi~nctions of somatosensory peripheral nerve and cerebral evoked responses in man. Electroencephalogr. Clin. Neurophysiol. 10 1964:17'126-35
24
J. Artieda and M.A. Pastor
Simmons JA. The resolution of target range by echolocating bats. J. Acoust. Soc. Am. 1973; 54:157-73. Simmons JA. Perception of echophase infommtion in bat sonar. Science. 1979; 204: 1336-38. Smith PH, Joris PX, Yin TCT. Projections of spherical bushy cells to the MSO in the cat: evidence for delav lines. Neurosci. Abstr. 1990" 16" 723. .
Sullivan WE, Konishi M. Neural map of interaural phase difference in the owl's brainstem. Proc. Natl. Acad. Sci. USA. 1986: 83" 8400-4. Spydeli JD, Pattee G, Goldie WD. The 40 Hertz event-related potential normal values and effects of lesions. Electroencephalogr. Clin. Neurophysiol. 1985" 62" 193-202. Steriade M, Curro-Dossi R, Pareacue D. Oakson G. Fast oscillations (20-40 Hz) in thalamocortical systems and their potentiation by mesopontine eholinergic nuclei in the cat. Proc. Natl. Acad. Sci. USA. 1991" 88" 4396400. Stroud JM. The fine structure of psychological time. In" Quastler H editor. Information theoD' in psychology" Problems and methods. Glencoe, IL" Free Press, 1956:174-205. Treisman M, Faulkner A, Naish PLN, Brogan D. The internal clock. Evidence for a temporal oscillator underlying time perception with some estimates of its characteristic frequency. Perception. 1990:19: 705-43. Uttal WR. A comparison of neural and psychophysical responses in the somesthetic system. J. Comp. Ph.vsiol. Psychol. 1959: 52" 485-90. Wearden JH. Do humans possess an internal clock with scalar timing properties? Leam.Motivat. 199 l a. 22: 59-83. Wearden JH. Human perfomlance on an analogue of an interval bisection task. Quart. J. Exp. Psychol. 1991b: 43B" 59-81.
Neurophysiological Mechanisms of Temporal Perception
25
Westheimer G, McKedd S P. Perception of temporal order in adjacent visual stimuli. Vis. Res. 1977" 17" 887-92. Yin TCT, Chan JCK. Interaural time sensitivity in medial superior olive of cat. J. Neurophysiol. 1990: 64: 465-88. Young SR, Rubel EW. Frequency-specific projections of individual neurons in chick brainstem auditory nuclei. J. Neurosci. 1983: 7' 1373-78. o
This Page Intentionally Left Blank
Time, Internal Clocks and Movement M.A. Pastor and J. Artieda (Editors) 9 1996 Elsevier Science B.V. All rights reserved.
27
P R O C E S S I N G OF T E M P O R A L I N F O R M A T I O N IN THE BRAIN CATHERINE E. CARR ! and SATOSHI AMAGAI 2
1Dept. of Zoology. Universitj, o['Maryland, College Park MD 20742-4415. USA "Dept. of PsycholoKv. Universi~ o['Maryland, College Park MD 207424415. USA ABSTRACT. Studies of time coding in tile central nervous system have revealed circuits specialized for the encoding and processing of temporal information. Behavioral experiments show lha! animals can detect microsecond lime differences, while analysis of how temporal ilfformation is processed has uncovered many common principles. Despite different neural substrates, lime coding systems share similar fcalures and implement similar algorithms for lhe encoding of temporal informalion. Timing information is generally coded by phase-locked action potentials, and processed in a dedicated pathway in parallel with other stimulus variables. The elemenls of time-coding circuits have morphological and physiological features suited to their filnction.
1. Introduction
Many neural systems encode infommtion in the time domain with microsecond accuracy. These times would seem too small for single neurons to encode or resolve, because neural events occur on a millisecond, rather than a microsecond, time scale. Nevertheless, sensitivity to microsecond time differences is widespread. This review summarizes studies of time coding in the central nervous system (CNS) to show how temporal information is preserved and processed, and how the sensitivity to time differences arises. We describe behavioral sensitivity to time differences and neuronal specializations for the coding of temporal information and detection of time differences. The review concentrates on specialists such as weakly electric fish and barn owls because they have well developed abilities to detect time differences. Analysis of how animals process temporal information has
28
C.E. Cart and S. Amagai
uncovered many common principles; despite their different evolutionary origins, time coding systems implement similar algorithms for the encoding of temporal information. In time-coding systems, the stimulus is coded by phase-locked spikes and processed in a dedicated pathway in parallel with other stimulus variables such as amplitude. In electric fish, the separation between phase and amplitude coding begins at the receptor level, while in the auditory system the separation into phase and amplitude coding is derived within the central nervous system. A similar separation may be found in other sensory systems: visual information is processed in two channels that differ in their spatial and temporal acuity (Livingstone and Hubel, 1987; Maunsell and Newsome, 1987). Cells that encode temporal infonnation have morphological and physiological features suited to their function
2. Behavioral detection of small time or phase differences
Many animals have well developed abilities to detect fluctuations in temporal signals. Well knox~n examples are found in weakly electric fish, animals that employ biosonar signals, and animals with acute sound localization ability. 2.1 THE ELECTROSENSORY SYSTEM Weakly electric fish produce continuous electric organ discharges that foma an electric field around the fish. The field is detected by electroreceptors distributed on the body surface, and used for location of objects and for social communication. Many electric fish are nocturnal, and many live in black water habitats. Their electric sense presumably confers an advantage in these habitats because it can replace or supplement vision. Electroreception is an ancestral vertebrate trait found in most aquatic vertebrates (Northcutt. 1983). It appears to have been lost in the tclcost ancestor, and then independently reinvented in two teleost lineages, the South American gymnotifomas and the African momayrifomas. These two groups of fish have independently evolved electroreceptors and electric organs (Bullock. 1982). The electric organs are made of modified muscle cells which fire in synchron.v to produce the electric organ discharge (EOD) (Bennett, 1970). There are two types of discharge pattern, "'wave" or nearly sinusoidal discharges, and "'pulse" discharges. Each is thought to represent a
Processing of Temporal Information in the Brain
29
different strategy for characterizing the fish's environment, and both have evolved within each lineage (Dye and Meyer, 1986). 2.1.1 South American gymnot!formfish. Gynmotiform fish with wave-t3pe EODs have been the subject of intensive investigations centered on the Jamming Avoidance Response (JAR) (see Heiligenberg, 1977; Bullock and Heiligenberg, 1986). Each fish has a characteristic electric organ discharge, and when two fish encounter each other, their signals may add in the water and produce a beating signal which hinders electrolocation. In the JAR, each fish minimizes this interference by changing its electric organ frequency so as to increase the frequency difference between the two signals. Thus if its neighbor's frequency is slightly higher than its ox~al, the fish lowers its frequency, while if its neighbor's discharge frequency is lower, the fish raises its frequency. In order to perform this behavior, each fish must be able to determine whether its neighbor's electric organ discharge is higher or lower in frequency than its own. Behavioral studies have shown that it does this by evaluating changes in the amplitude and phase of the electric signal produced by the interaction of the fields (Heiligenberg et al., 1978). Since correct performance of the janmling avoidance response requires evaluation of differential phase, the behavior provides an assay for the minimum phase difference that can be perceived by the fish (Carr et al., 1986a: Rose and Heiligenberg, 1985). When fish are presented with stimuli that produce small phase differences between different parts of the body surface,, they show a correct JAR to these weak jamming stimuli when the phase differences are as small as 400 ns. Physiological recordings, however, show that the phase-locked responses of afferents are too jittery to permit such fine temporal resolution (see next section). H39eracuity is a perceptual tema to describe the phenomenon whereby sensory thresholds are lower than expected from the properties of individual receptors (Altes, 1989). Rose and Heiligenberg (1985) showed that this temporal hyperacuity must result from the convergence within the central nervous system of parallel phase-coding channels from sufficiently large areas of the body surface, because the ability to detect the small phase differences diminished when smaller numbers of receptors were stimulated. ~
2.1.2 African mormyriform .fish. Unlike their South American cousins, African Mormyriform fishes are predominantly pulse-type. Their EODs are pulses of short duration, ranging from hundreds of microseconds up to
30
C.E. Carr and $. Amagai
several milliseconds, discharged at irregular intervals interspersed with relatively long periods of silence (see Hopkins, 1986a for review). Each species exhibit a species-specific EOD waveform (Hopkins, 1981) which can be sexually dimorphic or show individual variation (Bass, 1986; Hopkins et al., 1990). Playback experiments have demonstrated that the fish can discriminate EODs using time-domain cues (Hopkins and Bass, 1981). The resolution of the system is at least in the sub-millisecond range and more recent studies with phase shiRed EODs hint at even greater resolution (vonder Emde and Zelick, 1994). In mormyriform fishes, the communication information is analyzed by the Knollenorgan system which remains distinct from the rest of the electrosensory system. The receptors respond to an outside positive voltage step with a single time-locked action potential. An externally generated EOD will cause two populations of receptors to fire at different times, depending on whether the current is entering or leaving the body where the individual receptors are located. It has been proposed that by comparing the times of occurrence of spikes from these two populations of receptors, pulsemormyriform fishes can recognize different waveforms (Hopkins, 1986a). The cues for electrosensorv communication are then similar to those of differential phase detection in Eigenmannia JAR. The only wave-type mormyriform, Gymnarchus, has a JAR that is, remarkably, identical to that of Eigenmannia in terms of sensitivity and computational algorithms (Kawasaki, 1993). Since the two groups are not closely related, this is thought to be an example of an extreme foma of convergent evolution. 2.2 THE AUDITORY SYSTEM Sound coming from one side of the body reaches one ear before the other, and the auditory, system uses these time differences to localize the sound source. Most animals actually encode the phase of the auditory signal, and then use interaural phase differences to compute sound location (Hef~er and Heffner, 1992; Fay, 1988). Animals with large heads have much larger time differences available to them, and conversely animals with small heads have to achieve much greater resolution of binaural time differences than a large animal in order to obtain the same degree of accuracy. Sound localization ability is not always related to head size, however. Instead, it is correlated with directing the attention of the other senses to the sound source. Animals
Processing of Temporal Information in the Brain
31
with narrow fields of best visual acuity require accurate sound localization to direct their gaze, while the echolocating mammals, bats and dolphins, use sound localization to direct their biosonar pulses (Heffner and Heffiler, 1992). The animals that use biosonar signals display exquisite sensitivity to temporal cues and only space prevents an exhaustive description. These animals emit orientation sounds and listen to the returning echoes. Comparisons of pulses and returning echoes can yield information about target structure, speed and range. The delay between the pulse and echo conveys the target range; for example, in the bat a 1 ms echo delay corresponds to a 17.3 cm target distance (Suga, 1990). Behavioral studies have shown that bats detect microsecond fluctuations in echo arrival time (Simmons, 1973; Simmons, 1979). 2.2.1 The Barn owl, Tyto alba. The barn owl's ability to detect time differences is acute. These owls can catch mice in total darkness on the basis of auditory cues alone (Payne, 1971: Konishi, 1973). A sound coming from one side of their body reaches one ear before the other, and the owl translates this interaural time differences into location in azimuth (the horizontal plane). The owl actually derives interaural time differences from the interaural phase differences present in the auditor3' stimulus. Behavioral experiments show that the owl uses the phase differences between the two ears rather than stimulus onset time to localize sound, because if sound is presented through earphones, and phase delayed in one ear with respect to the other ear, the owl will turn its head in the direction predicted by the phase difference (Moiseff and Konishi, 1981). Barn owls are very accurate (Konishi, 1973); Knudsen and Konishi (1979) determined the average error made in free field sound localization by having the owl turn its head towards a sound in the dark, and found that the owl's average error in turning its head to a sound was 1.5 ~ (about 1.5 Its intemuml time difference).
3. Encoding of timing information Discrimination of small time differences requires accurate transduction and processing of the original stimulus. Time coding arises in the peripheo,, but it is preserved and improved in the CNS. In addition to accurate responses from single neurons, the observed behavioral hyperacuity is assumed to
32
C.E. Carr and S. Amagai
involve averaging in large neural assembles or vector coding. Both strategies may be found in systems that have achieved fine temporal resolution. 3.1 THE ELECTROSENSORY SYSTEM The electric organ discharge is detected by sensory hair cells tem~ed eleetroreceptors. There are two kinds of electroreceptors, sensitive to phase or amplitude (Seheich et al., 1973" Szabo, 1965). Nerve fibers that innervate the phase-coding type of electroreceptors fire one spike on each cycle of the stimulus, phase-locked with little jitter to the zero-crossing of the stimulus. The degree of phase-locking to the stimulus is often quantified using a measure termed vector strength (Goldberg and Brown, 1969). Phase-locking must originate at the level of the receptor, although there are few studies of hair cells that address this issue. In mormyrid electric fish, the Knollenorgan electroreceptor produces a phase-locked spike in response to the electric stimulus (Bennett. 1970). These electroreceptors may be the only vertebrate hair cells that generate a spike. At the best frequency, about 1.1 kHz, the vector strength is high and it diminishes with increasing stimulus frequency (Hopkins, 1986b). These frequency dependent decreases in vector strength have also been found in the auditory system, and they have been interpreted as showing that the neurons" ability to phase-lock decreases with frequency (Weiss and Rose 1988). The unrelated mormyriform and gymnotiform electric fish appear to have independently converged upon very. similar algorithms for encoding the timing of the stimulus. In both groups, primary afferents convey phaselocked spikes to the lateral line lobe of the medulla, where phase and amplitude coding primary afferents terminate on different cell types (von der Emde and Bleckmann, 1992: Bell et al., 1989; Maler et al., 1981). Thus the segregation of phase and amplitude receptors in the skin is reinforced by the central connections formed in the medulla. The separation of phase and amplitude information into two parallel channels is a common feature of all time coding systems. The two channels not only have separate connections. but also distinct morphology. In gynmotiform fish, phase-coding afferents are specialized for maintenance of phase-locked spikes: they form large club terminals on large spherical cells that have few or no dendrites and thick axons. The spherical cells relay timing information directly to the midbrain toms. These connections are responsible for the fish's ability to detect small time
Processing of Temporal Information in the Brain
33
differences, and will be discussed below. The accuracy of phase-coding improves with the progression from receptors to primary afferents to spherical cells to giant cells in the midbrain torus (Carr et al., 1986a). The accuracy of the phase-coding was determined by measuring spikes from phase coders in the medulla and in the midbrain toms. The jitter of these spikes, defined as the standard deviation of the response time to the stimulus, decreased three-fold with the progression from medulla to midbrain. The basis for this improvement of accuracy may lie in the convergence from afferents to higher level neurons (but see below). The accuracy of even the best single neurons in these first stations of the time coding pathway does not match that of the behavior, however (Rose and Heiligenberg, 1985: Carr et al., 1986a). Electric fish can resolve temporal disparities in the submicrosecond range, a temporal resolution far superior to that observed in the primary electrosensory afferents. This hyperacuity is presumed to result from network processing. 3.2 THE AUDITORY SYSTEM The behavioral experiments described in the previous section have sho~n that most animals use interaural phase differences to localize sound. Evidence of how they do so comes from experiments on the encoding and processing of phase information in the central auditory system. Time coding systems appear to have changed little during anmiote phylogeny, and although barn owls are useful models of temporal coding in the auditor3, system, studies on time coding in other vertebrate sensory systems are equally instructive, and will be included. Recordings from auditor3, nerve fibers showed that spikes have a statistical tendency to phase-lock to the wavefoma of the acoustic stimulus (Kiang et al., 1965). Spikes occur most frequently at a particular phase of the tone, although not necessarily in every, tonal cycle. Thus the discharge pattern of a cochlear nerve fiber can encode the phase of a tone with a frequency above 1000 Hz even though the average discharge rate is low. The general assumption is that the modulating signal at the spike generator in the auditory nerve arises from components of the hair cell receptor potential, via the chemical synapse between the hair cell and the primary afferent. This timing information is degraded for high frequency sounds, presumably because of Iowpass filter effects in the hair cell (Kidd and Weiss, 1990).
34
C.E. Cart and S. Amagai
Cochlear hair cells encode and transmit both phase and amplitude information to the auditory, nerve. Work on the cellular basis of frequency tuning in turtle hair cells has shown that potassium channel kinetics and numbers vary in a systematic fashion along the basilar papilla. Recordings from single channels in turtle hair cells have shown that channel kinetics arc faster in those hair cells which oscillate at higher best frequencies (about 500 Hz) (Art et al., 1995). Thus the ability of cochlear hair cells to phase lock to different frequencies may be determined by their channel kinetics and distribution. The mechanism of high best frequency phase locking in other auditory systems is not yet well understood. Nevertheless, hair cells transmit phase information to auditory, nerve fibers which phase-lock to the stimulus, and encode amplitude by increases in spike rate. Thus there can be no predisposition towards coding for either amplitude or phase in the periphery, unlike electric fish. Despite the lack of specialized phase and amplitude receptors, the same parallel processing of phase and amplitude information that characterizes the electrosensory system is also found in the auditory. system. The segregation appears to begin with differences in auditor3' nerve terminals. In the bird, auditor3.' nerve afferents enter the brain and then divide into two. One branch ramifies in the dendritic field of the cochlear nucleus angularis, that codes for changes in amplitude, while the other branch terminates in the cochlear nucleus magnocellularis that codes for phase (Takahashi et al., 1984). The terminal in the nucleus magnocellularis forms a specialized ending termed an endbulb of Held (Brawer and Morest, 1974; Ryugo and Fekete, 1982). The endbulb synapse conveys the phase-locked discharge of the auditor3' nerve fibers to their postsynaptic targets in the nucleus magnocellularis. Thus the synaptic specializations in the auditory. nerve accomplish the same goal as the receptor specialization in electric fish. Each auditory, nerve endbulb has multiple sites of sxaaaptic contact on the soma to provide a substrate for the preservation of the phase-locked spikes between the auditory, nerve and the neurons of the nucleus magnocellularis. The endbulb is a secure and effective connection; physiological measures show that phase-locking is the same in the neurons of the nucleus magnocellularis than in the eighth nerve, while it is lost in the projection to the amplitude coding nucleus angularis. Phase-locked spikes encode the timing of the stimulus, and the CNS uses this code for the measurement of time disparities. Phase information is preserved and improved, and interaural phase differences are detected, in a circuit composed of the auditor3, nerve, the cochlear nucleus magnocellularis
Processing of Temporal Information in the Brain
35
and the nucleus laminaris (Figure 1). Many of the features of this circuit may represent specializations for the encoding of timing information. The neurons of the nucleus magnocellularis are morphologically and physiologically specialized for the encoding of temporal information. They have large round cell bodies, a thick axon and few medium length dendrites (Jhaveri and Morest, 1982). In the owl, magnocellular neurons in the high best frequency regions of the nucleus have fewer dendrites than the neurons in the low best frequency region (Carr and Boudreau, 1993). Similar changes in dendritic length are found in the nucleus laminaris in the chicken (Smith and Rubel, 1979). These reductions in dendritic area would decrease the total capacitance of the cell and improve the speed and accuracy of the phase-locked response to synaptic inputs (see also 3.3). The nucleus magnocellularis relays phase-locked spikes in a bilateral projection to the nucleus laminaris (Figure 1). The circuits that encode temporal information in the mammalian auditory system are similar to those in birds, and it is probable that the projection from auditory nerve fibers tO spherical bushy cells in the anteroventral cochlear nucleus is homologous to the projection described in birds and reptiles (Boord, 1968; Spzir et al., 1990; Carr, 1992). Mammalian bushy cells are morphologically and physiologically similar to magnocellular neurons of birds and reptiles. They project to the medial superior olive in a circuit responsible for encoding and detecting interaural phase differences. Like magnocellular neurons, bushy cells have large, round somata and thick axons, but they also have many more dendrites (Rhode et al., 1983; Wu and Oertel, 1984). Bushy cells are well suited to preserve the temporal firing pattern of auditory nerve inputs (Oertel, 1085). They fire only one or two spikes in response to electrical stimulation of the auditory nerve, and have non-linear current-voltage relationships around the resting potential. The effects of excitation are brief and do not summate in time (Wu and Oertei, 1984; Oertel, 1985). Similar physiological responses characterize phase-coding neurons in guinea pig ventral cochlear nucleus (Manis and Marx, 1991) and magnocellular neurons in chickens (Reyes et al., 1994). A rapidly activating and slowly inactivating potassium current appears to underlie the rapid repolarization and ability of these neurons to transmit well-timed events (Reyes et al., 1994).
36
C.E. Cart" and S. Amagai
A tu
IPSI NM
cn
ro
Cgl
f-
-0-0-
.
"0"
0
0
~
-0-.1
,~
i 1 . ' C~ =0 0 ~.~.
_,..| -,.| -,.@
-0"0" -0(/)
LAMINARIS
-.0-
,,-,~i~:'!..... ,. . . . . . . ,
-C)-
-0-
i.
CONTRA NM Figure 1" Jeffress model and schematic of brainstem auditory circuits for detection of interaural time differences in the barn owl.
Axons from the ipsilateral cochlear nucleus magnocellularis (IPSI NM) divide and enter the nucleus laminaris at several points along the dorsal surface. These axons act as delay lines within laminaris, interdigitating with inputs from the contralateral cochlear nucleus magnocellularis (CONTRA NM). Ill the cenler, the owl circuit has been modified to show the principles of the Jeffress model. Each binaural coincidence detector (A-E) fires maximally when inputs from the two sides arrive simultaneously. This call only occur when the imeraural phase differences are compensated for by an equal and opposite delay. For example, neuron A fires maximally when sound reaches the contralateral ear first and is delayed by the long path from the contralateral ear so as to arrive simultaneously with the input from the ipsilaterai ear. Thus this array forms a map of interaural time difference in the dorso-ventral dimension of the nucleus. In the owl. sound from the front is mapped towards Ihe ventral surface of the nucleus, and each nucleus appears to contain place maps of the contralateral and part of the ipsilaleral hemifieid (modified from Konishi. 1991). 3.3 M A I N T E N A N C E OF A C C U R A T E T E M P O R A L INFORMATION If the onset times of the action potentials were exact, it should be possible to encode the timing information by their times of occurrence despite the long duration of the action potentials themselves. However, a variety of factors such as the thermal noise, fluctuations in the state of the transduction mechanisms, ambient electrical noise and noise in the sensor3' channel itself
Processing of Temporal Information in the Brain
37
introduce a variability in response latencies (jitter). Behavioral evidence makes it clear that the sensory system can extract precise temporal information despite the jitter in the incoming information. One of the models often invoked to account for this temporal hyperacuity involves extensive convergence as a mechanism for reducing temporal jitter (Calvin, 1983; Carr et al., 1986a; Kawasaki et al., 1988; Heiligenberg, 1989). In a simplified version of this model, a signal occurring at time T causes a presynaptic neuron (N l) to fire an action potential with a jitter of variance cry'. Inputs from n presynaptic neurons converge onto a postsynaptic neuron (N2) where they are averaged. By simple probability, theory, the variance of the input to N2 is reduced by a factor of n, such that as n increases, jitter of the average of the inputs to N2 becomes smaller and smaller, allowing the system to recover T with increasing accuracy. The problem with this model is the unstated assumption that the intrinsic jitter of N2 (trT) is zero. In the ideal case of this model, an infinite number of N Is converging onto N2 would allow T to be extracted exactly. However, if the temporal jitter inherent in the spike generation mechanism were the same in N2 as in N I (that is, o'~2 = 0",2), then additional jitter of the same size as in an individual N I is introduced in N2. The jitter of the output from N2 in such a condition can then only be as small as the jitter of a single N 1 input, even with infinite convergence. Convergence on its ox~aa cannot therefore be viewed as the causal mechanism for reducing temporal jitter and generating temporal hyperacuity: it prevents jitter from getting worse than it already is. If convergence alone cannot lead to temporal h319eracuity, what else is necessary'? Under the averaging model, the variance of N2 output is o'~2/n + o'22. Temporal hyperacuity requires this to be less than the variance of N I (o'~2/n + o'22 < o'12). For this to be possible, N2 itself must have an intrinsically lower variance than N I (o"22 '(0"!2), otherwise o'~"/n + 0",.2 will always exceed crt2. In other words, N2 must be a more accurate neuron than N 1. In addition, through the use of convergence, the system needs to reduce the incoming jitter originating from N I (tO minimize o'~"/~) so that when the smaller jitter of N2 is added, the final variance does not exceed o'!2. It is the combination of these two factors that generates temporal hyperacuity, the degree of which depends on the extent of convergence as well as how much more accurate N2 is compared to N I. This observation has ramifications for models of brain organization. With the model solely relying on convergence, it is possible to visualize a hierarchical array of converging neurons and ascribe it the function of generating increasingly accurate temporal
38
C.E. Carr and S. Amagai
information. Under that scheme, there is no theoretical limit to the resolution that can be achieved and incredibly fine temporal resolutions found in behavioral data are possible just through sheer strength of numbers. It is important to realize that the model is missing an essential component of hyperacuity: neurons with intrinsically low jitter. Intrinsic properties of neurons should be reconsidered as a critical feature of hyperacuity. Improvement in intrinsic accuracy of N2 would presumably be achieved through anatomical and physiological specializations of both presynaptic and postsynaptic structures. Many of these features have previously been recognized in time coding systems and have been thought to be important in preserving temporal information. One general strategy is to make ever)ething large. Larger somata and axons are less vulnerable to noise caused by stray currents since their low input resistance and large current generating ability would keep voltage fluctuations small. Many of the known time-coding pathways include large cells: the spherical cells of the Gylrmotoid ELL (Maler et al., 1981), the cells of the Mormyrid nELL (Bell and Russell, 1978), nucleus magnocellularis in birds (Jhaveri and Morest, 1982). or the bushv cells of the anteroventral cochlear nucleus in mammals (Rhode et al., 1983). Further, the synaptic inputs to big cells would also have to be large and rapid so that the postsynaptic response has a fast rise time to minimize the influence of ambient voltage fluctuations. Since the postsynaptic neuron is large, this also demands that the synapse be able to generate a large current. One solution is to have large terminals that partially engulf the postsynaptic cell, presumably translating into massive release of neurotransmitter without depletion. These occur as the endbulbs of Held in birds (Brawer and Morest, 1974) and club endings in electric fish at numerous points in the time-coding pathway: Gynmotoid electrosensory lateral line lobes (Maler et al., 1981), Gymnotoid torus semicircularis (Carr et ai., 1986b), Mormyrid nucleus of the electrosensory lateral line lobe (Bell and Russell, 1978; Szabo et al., 1983) and Mormyrid nucleus extrolateralis anterior (Mugnaini and Maler. 1987). Another solution would be to eliminate or reduce the dendrites betaveen the synapse and the site of integration. Receiving inputs directly on the soma would minimize the attenuation of synaptie current, yielding fast rise times and reduction of the influence of stray currents. Lack of dendrites will also decrease the total capacitance of the cell through a reduction in the membrane surface area, yielding a faster rise time. This occurs in timecoding electric fish neurons and in the cells of the nucleus magnocellularis
Processing of Temporal Information in the Brain
39
and the nucleus laminaris in birds (Jhaveri and Morest, 1982; Smith and Rubel, 1979; Carr and Boudreau, 1993.). There is also evidence for physiological specializations in receptor channel kinetics. The chicken nucleus magnocellularis has been found to have glutamate receptors that have unusually fast kinetics and large conductance, which are characteristics that are well suited for use in time-coding pathways (Raman and Trussell, 1992). Given that the temporal accuracy of response would be ultimately limited by how small the intrinsic jitter can be made, it seems unlikely that accurate temporal representation of stimuli could be maintained to explain the extreme cases of temporal hyperacuity. A possible solution would be to recode the information so that it is no longer represented as the time of occurrence of spikes, but in an alternative format such as a system of labeled lines. Once this transformation takes place, higher processing areas can process the information using slower integration mechanisms without loss of resolution. 4. Detection of small time differences
Behavioral experiments have shown animals are capable of great accuracy in detecting time differences. The preferred model for detecting temporal disparities depends upon coincidence detection, and proposes that a neuron that detects time differences responds best to simultaneous inputs. The model was first proposed by Jeffress (1948) with his place theory for detection of interaural time differences. The model circuit is composed of two elements, delay lines and coincidence detectors (Figure 1). The delay lines are created by axons of varying lengths, and the coincidence detectors are neurons that respond maximally when they receive simultaneous inputs, i.e. when the time difference is exactly compensated for by the delay introduced by the inputs. The Jeffress model explains not only how time differences may be measured but also how they may be encoded. The circuit contains an array of neurons receiving input from afferent axons serving as delay lines. Because of its position in the array, each neuron responds only to sound coming from a particular direction, and thus the anatomical place of the neuron encodes the location of the sound (Figure 1). These neurons compute, time difference, a new variable, and transform the time code into a place code. The selectivity of all higher-order auditory neurons to time difference
40
(7.E. Cart and S. Amagai
derives from the "labeled-line" output of the place map (Konishi, 1986). Circuits in the fish midbrain and the auditory brainstem of barn owls. cats and dogs contain neurons tuned to particular time delays between events. These delay-tuned ensembles employ similar algorithms. The circuits in the auditory brainstem resemble the Jeffress model in all its particulars, while the circuits in fish midbrain differ from the Jeffress model because the output neurons do not appear to form a place map. 4.1 DETECTION OF PHASE DIFFERENCES IN ELECTRIC FISH The circuits for detection of small phase differences are quite similar in both mormyriform and g)~notiform electric fish, despite their independent evolutionary origins. Furthermore. within the African momayriformes, the wave type fish Gymnarchus appears to have independently constructed a medullary neural circuit for detection of phase differences (Kawasaki and Gao, in press) while the pulse t)q3e mormyrids have a phase comparison circuit within their midbrain torus (Amagai, 1994). It appears therefore that time detection circuits have evolved at least three times in electric fish. 4.1.1 Eigenmannia. When Eigenmannia is presented with sinusoidally var),ing electrical fields on two parts of its body, it can distinguish phase differences between these signals smaller than 1 las. The circuit for detection of these phase differences is constructed of phase-coding afferents from the medulla which project to two cell types of the midbrain toms, synapsing on giant cell bodies and on the dendrites of the small cells (Figure 2). These afferents form local connections that encode the phase of the electric organ discharge from one part of body surface. The giant cell's axons form horizontal connections that distribute this local phase information to small cells throughout the lamina, so that timing infommtion from one part of the body surface may be compared with an)' other part. Small cells compare information from one patch of the body surface, through afferent input onto their dendrites, with phase information from any other part of the body surface, through the giant cell input to their cell bodies (Carr et al., 1986b). Small cell responses encode either phase advance or phase delay (Heiligenberg and Rose. 1985). The small cell circuit allows the fish to perform all possible comparisons between different parts of the body surface, as required for correct performance of the januning avoidance response.
Processing of Temporal Information in the Brain
1
- -
i
.....
.
.
.
.
.
i
i
MIDBRAIN TORUS
differences
CELL
AFFERENT ELECTRORECEPTOR
IlJJl[
i
Detect phase ~v between ~/)D""'.:.,:.~,:parts of body
S.AU.L
II
,d,,elay lines ? Jib
GIANT CELL
SPHERICAL CELL
i
IlJ_lJl
41
surface
I!
phase information from one part of body
0 q| 0 MEDULLA
0000
0000
I'II'I0 O
0 0
SKIN
phase information from another part of body
Figure 2: Schematic circuit ill lhe electric fish midbrain for computation of phase differences between signals on any two parts of the body surface. Phase-coding electroreceptors converge on spherical cells in the medulla that in turn converge in a Iopographic projeclion onto giant cell bodies and small cell dendrites in the toms. Giant cells relay the phase-locked signal all over the toms, with Ihcir terminals synapsing on the cell bodies of small cells. Small cells are therefore able to compare phase information from different parts of the body surface. Open circles are used Io diagram Ihe approximate convergence belxveen differen! levels. (modified from Carr, 1993).
4.1.2 Gymnarchus. The Gymnarchus JAR is almost identical to that of Eigenmanma except that the phase computations takes place at the level of the medulla in the elcctrosensor 3' lateral line lobe (ELL). Just as in the midbrain of Eigenmannia, the afferents carn'ing phase infonuation temlinate on giant cells and smaller differential-phase-sensitive cells of ELL, though the latter connection has not yet been shm~x~ to be direct (Kawasaki and Guo, in press). Like Eigenmannia, the giant cells prqiect bilaterally and
42
(7.E. C,arr and S. Amagai
branch extensively in the same regions as the differential-phase-sensitive cells. The giant cells in Gymnarchus ELL therefore serve the same fimction as the giant cell of the ton~s of Eigenmanma, distributing local phase information available to all the phase computation neurons. 4.1.3 Pulse-Mormyri/brmes. The phase computation circuit in pulse mormyriformes is less well understood. Organizationally, it shares many of the features exhibited by Eigenmannia. The Knollenorgan afferents terminate on the large cells of nucleus of ELL, which project bilaterally to nucleus extrolateralis anterior (ELa) in the torus where they synapse oll small output cells and large GABAergic cells (Figure 3) (Mugnaini and Maler, 1987). The large cells have extensive arborizations within ELa and terminate on numerous small cells. They could therefore act as delay lines to create a basis for phase computations. The large cells, however, are probably inhibitory, so the computation mechanism is unlikely to be that of simple coincidence detection. The physiological characteristics of the nucleus that receives ELa output suggest that the phase computation is not complete within ELa (Amagai and Hopkins, 1990). 4.2 DETECTION OF INTERAURAL PHASE DIFFERENCES IN BIRDS AND
MAMMALS The circuits in the auditory brainstem that detect interaural phase differences resemble the Jeffress model. The cochlear nucleus axons act as delay lines, and the laminaris or olivary neurons act as coincidence detectors, to form a circuit that measures and encodes interaural time differences (Figure 1). Early physiological studies of the mammalian auditory system by Goldberg and Brown (1969) and others found neuronal responses consistent with the Jeffress model, and studies in the barn owl (Sullivan and Konishi, 1986; Carr and Konishi, 1990), cat (Yin and Chan, 1990: Smith et al., 1990) and chicken (Rubel and Parks. 1975; Young and Rubel, 1983; Overholt et al., 1992; Joseph and Hyson, 1993) have described circuits that conform to the model's requirements. There are three major parts to the Jeffress model" delay lines, coincidence detection and place coding of interaural phase difference. In the barn owl, magnocellular axons act as delay lines (Carr and Konishi, 1988; Carr and Konishi, 1990). They convey the phase of the
43
Processing of Temporal Information in the Brain
,,
. . . . . . . .
,
,
MIDBRAIN
Phase sensitive
.............
,
,,r
,,,,
,
,,
,
_
TORUS cells
~_
ELp
. . ..,
I
L A R G E I ... i,.i ' GABA till . ~ - - - - - - ] "
CELL
ELa
,
' ....:,, . . . . . ' . , delay line-? .~
CM S ELA LLL
.....
,
~T .....
9
ELECTROSENSORY LOBE NUCLEUS AFFERENTS
Knollenorgan
ELECTRORECEPTORS
phase information
Figure 3: Schematic circuit in the pulse mormyriforn~ electric fish midbrain for analysis of EOD waveforms. Phase-coding Knollenorgan inputs project to the cells of the nucleus of the electrosensory lateral line lobe which in turn project bilaterally Io the nucleus extrolateralis anterior (ELa) of the tonls and terminate on small output cells and large GABAergic cells. The large cell's axon branches extensively and terminate on many small cells, and is modeled here forming the delay lines. The small cells are no! shown as coincidence detectors, bul ralher send a process to the adjacent nucleus extrolateralis posterior (ELp) where the rest of tile phase computation is tllought to take place by an as yet unknown mechanism.
auditory stimulus in a bilateral projection to the nucleus laminaris such that axons from the ipsilateral nucleus magnocellularis enter the nucleus laminaris from the dorsal side. while axons from the contralateral nucleus magnocellularis enter from the ventral side. Thus these afferents interdigitate to innervate dorso-ventral arrays of neurons in laminaris in a sequential
44
C.E. Cart and S. Amagai
fashion (Figure 1). For each frequency band, recordings from these interdigitating ipsilateral and contralateral axons show regular changes in delay with depth in the nucleus laminaris (Carr and Konishi, 1990). These conduction delays are similar to the range of interaural time differences available to the barn owl (Moiseff, 1989). Delay line projections were first described for the chicken (Rubel and Parks, 1975; Young and Rubel. 1983) and are very similar to those described for the medial superior olive of the cat, where axons from the contralateral cochlear nucleus foma delay lines across the rostro-caudal axis of the nucleus, while the ipsilateral axons form a less organized projection (Smith et al., 1990). The other major requirements of the Jeffress model are that the targets of the delay lines act as coincidence detectors, and that these neurons encode ITD by their place within the nucleus. Goldberg and Brown (1969) showed that olivary neurons acted as coincidence detectors. The neurons of the nucleus laminaris and the medial superior olive phase-lock to both monaural and binaural stimuli, and they respond maximally when phase-locked spikes from each side arrive simultaneously, i.e. when the difference in the internal conduction delay is nullified by interaural time difference. Physiological responses from these coincidence detectors are similar (Goldberg and Brown, 1969: Yin and Chan, 1990: Sullivan and Konishi, 1984: Carr and Konishi, 1990). The output of these coincidence detectors appears to form a place code, as predicted by the Jeffress model (Smith et al., 1990). 5. Conclusions
Analysis of the encoding and processing of temporal information has uncovered many common features. These systems implement similar algorithms for the encoding and processing of temporal infonnation (Konishi, 1991). Time coding systems share numerous morphological and physiological adaptations to improve the time-coding of signal. Comparisons of circuits which measure time differences show that they depend on delay lines and coincidence detectors in some form. Short delays may be provided by axonal delay lines, with longer delays introduced by interposed synapses. Neural circuits that incorporate delays may be widespread in the CNS.
ACKNOWLEDGMENTS. Work supported by a grant from tile National Institutes of Health to CEC (DC 00436).
Processing of Temporal Information in the Brain
45
6. References
Altes RA. Ubiquity ofhy9eracuity. J Acoust. Soc. Am. 1989; 85" 943-52. Amagai S. Analysis of time-domain information by the Knollenorgan system ofmormyrid electric fish. Neurosci. Abstr.. 1994; 20: 370. Amagai S, Hopkins CD. Cells sensitive to time-domain infonnation in the midbrain of momayrid electric fish. Neurosci. Abstr. 1990" 16" 1325. Art JJ, Wu Y-C, Fettiplace R. The calcium-activated potassium channels of turtle hair cells. J Gcn Physiol. 1995" 105" 49-72. Bass AH. Electric organs revisitcd" Evolution of a vertebrate communication and orientation organ. In: Bullock TH, Heiligenberg W, editors. Electroreception. New York: J. Wiley and Sons, Inc. 1986' 13-70. Bell CC, Russell CJ. Termination of electroreceptor and mechanical lateral line afferents in the mormyrid acousticolateral area. J. Comp. Neurol 1978" 182" 367-82. Bell CC, Zakon H, Finger TE. Mormyromast electroreceptor organs and their afferent fibers in momayrid fish" I. Morphology. J. Comp. Neurol. 1989; 286:391-409. Bennett MVL. Comparative physiology: electric organs. Ann. Rev. Physiol. 1970; 32" 471-531. Boord RL. Ascending projections of the primary cochlear nuclei and nucleus laminaris in the pigeon. J. Comp. Neurol. 1968" 133' 523-42. Brawer JR, Morest DK. Relations between auditor3' nerve endings and cell types in the cat's anteroventral cochlear nucleus seen with Golgi method and Nomarski optics. J. Comp. Neurol. 1974; 160:491-506. Bullock TH. Electroreception. Ann. Rev. Neurosci. 1982: 5" 121-70.
46
C.E. Carr and S. Amagai
Bullock TH, Heiligenberg W, editors. Electroreception. New York: John Wiley and Sons. 1986. Calvin WH. A stone's throw and its launch window: Timing precision and its implications for language and hominid brains. J. Theor. Biol. 1983: 104: 121-35. Carr CE. Time coding in electric fish and barn owls. Brain Behav. Evol. 1986; 28: 122-33. Carr CE. The evolution of the central auditory system in reptiles and birds. In: Webster DB, Fay RR, Popper AN, editors. The evolutionary biology of heating. New York: Springer-Vcrlag. 1992" 511-44. Carr CE. Processing of temporal information in the brain. Ann. Rev. Neurosci. 1993; 16: 223-43. Carr CE, Boudreau RE. Organization of the nucleus magnoceilularis and the nucleus laminaris in the barn owl" encoding and measuring interaural time differences. J. Comp. Neurol. 1993: 16: 223-43. Carr CE, Konishi M. Axonal delay lines for time measurement in the owl's brainstem. Proc. Natl. Acad. Sci. 1988" 85" 8311-315. Carr CE, Konishi M. A circuit for detection of interaural time differences in the brainstem of the barn owl. J. Neurosci. 1990; 10: 3227-46. Carr CE, Heiligenberg W, Rose G. A time-comparison circuit in the electric fish midbrain. I. Behavior and physiology. J. Neurosci. 1986a; 6:107-19. Carr CE, Maler L, Taylor B. A time comparison circuit in the electric fish midbrain. II. Functional morphology. J. Neurosci. 1986b: 6: 1372-83. Dye JC, Meyer H. Central control of the electric organ discharge in weakly electric fish. In: Bullock TH, Heiligenberg W, editors. Electroreception. New York: J. Wiley and Sons, Inc. 1986:71-102.
Processing of Temporal Information in the Brain
47
Fay RR. Heating in vertebrates" A psychophysics databook. Winnetka, Illinois: Hill-Fay Associates. 1988. Goldberg JM, Brown PB. Response of binaural neurons of dog superior olivary complex to dichotic tonal stimuli: Some physiological mechanisms of sound localization. J. Neurophysiol. 1969; 32" 613-36. Hef~er RS, Heffner HE. Evolution of sound localization in mammals. In" Webster DB, Fay RR, Popper AN, editors. The evolutionary biology of hearing. New York: Springer-Verlag 1992:691-716. Heiligenberg W. Principles of electrolocation and jamming avoidance. In" Braitenberg V, editor. Studies of brain function, Vol. 1. New York: Springer-Verlag. 1977: 1-85. Heiligenberg W. Coding and processing of electrosensory information in gynmotiform fish. J. Exp. Biol. 1989: 146: 255-75. Heiligenberg W. Neural nets in electric fish. Cambridge, Massachusetts" MIT Press. 1991. Heiligenberg W, Rose G. Phase and amplitude computations in the midbrain of an electric fish: intracellular studies of neurons participating in the jamming avoidance response of Eigenmanma. J. Neurosci. 1985" 5" 515-31. Heiligenberg W, Baker C, Matsubara JA. The jamming avoidance response in Eigenmanma revisited: The structure of a neuronal democracy. J. Comp. Physiol. 1978; 127: 267-86. Hopkins CD. On the diversity of electric signals in a community of mormyrid electric fish in West Africa. Am Zool. 1981 21" 211-22. Hopkins CD. Behavior of mormyridae. In" Bullock TH, Heiligenberg W, editors. Electroreception. New York" J. Wiley Sons. Inc. 1986a: 527-76. Hopkins CD. Temporal structure of non-propagated electric communication signals. Brain Behav Evol. 1986b 31" 43-59.
48
C.E. Carr and S. Amagai
Hopkins CD and Bass AH Temporal coding of species recognition signals in an electric fish. Science. 1981" 212: 85-7. Hopkins CD, Comfort NC, Bastian J, Bass AH. Functional analysis of sexual dimorphism in an electric fish, Hypopomus pinnicaudatus, order Gynmotiformes. Brain Behav. Evol. 1990; 35" 350-67. Jeffress LA. A place theory of sound localization. J. Comp. Physiol. Psych. 1948; 41" 35-9. Jhaveri S, Morest K. Neuronal architecture in nucleus magnocellularis of the chicken auditory system with observations on nucleus laminaris: A light and electron microscope study. Neurosci. 1982' 7" 809-36. Joseph AW, Hyson RL. Coincidence detection by binaural neurons in the chick brain stem. J Neurophysiol. 1993" 69:1197-211. Kawasaki M. Independently evolved jamming avoidance responses employ identical computational algorithms: a behavioral study of the African electric fish, Gymnarchus niloticus. J. Comp. Physiol. 1993" 173" 9-22. Kawasaki M, Guo Y-X. Neuronal circuitry for comparison of timing in the electrosensory lateral line lobe of an African wave-type electric fish, Gymnarchus mloticus. J. Neurosci. 1995 (in press). Kawasaki M, Rose G, Heiligenberg W. Temporal hyperacuity in single neurons of electric fish. Nature. 1988: 336" 173-76. Kiang NYS, Watanabe T, Thomas EC, Clark EF. Discharge patterns of single fibers in the cat's auditor3, nerve. Cambridge, Massachusetts: MIT Press. 1965. Kidd RC, Weiss TF. Mechanisms that degrade timing information in the cochlea. Hearing Res. 1990; 49" 181-208. Knudsen El, Konishi M. Sound Localization by the Barn Owl (Tyto alba). J. Comp. Physiol. 1979; 133" 1-I I.
Processing of Temporal Information in the Brain
49
Konishi M. How the owl tracks its prey. Am. Sci. 1973; 61" 414-24. Konishi M. Centrally synthesized maps of sensory space. Trends Neurosci. 1986; 9: 163-68. Konishi M. Deciphering the brain's codes. Neural Computation. 1991" 3" 118. Livingstone ML, Hubel DH. Psychophysical evidence for separate channels for the perception of form, color, movement and depth. J. Neurosci. 1987: 7" 3416-68. Maler L, Sas E, Rogers J. The cytology of the posterior lateral line lobe of high frequency weakly electric fish (Gymnotoidei)" Dendritic differentiation and synaptic specificity in a simple cortex. J. Comp. Neurol. 1981" 195" 87140. Manis PB, Marx SO. Outward currents in isolated ventral cochlear nucleus neurons. J.Neurosci. 199 I' 11 2865-80. Maunsell JHR, Newsome WT. Visual processing in monkey extrastriate cortex. Ann. Rev. Neurosci. 1987" 10" 363-402. Moiseff A. Binaural disparity cues available to tile barn owl for sound localization. J. Comp. Physiol. 1989; 164: 629-36. Moiseff A., Konishi M. Neuronal and behavioral sensitivity to binaural time differences in the owl. J. Neurosci. 1981" 1" 40-8. Mugnaini E, Maler L. C3r and inmaunoc3r 3, of the nucleus extrolateralis anterior of the mormyrid brain" possible role of GABAergic synapses in temporal analysis. Anat. Embryol. 1987; 176" 313-36. Northcutt RG. Evolution of the optic tectum in ray-finned fishes. In: Davis R, Northcutt RG, editors. Fish Neurobiology, Voi. 2: Higher brain areas and functions. Ann Arbor: University of Michigan Press. 1983" 1-43.
50
C.E. Cart" and S. Amagai
Oertel D. Use of brain slices in the study of the auditory system: Spatial and temporal summation of synaptic inputs in cells in the anteroventral cochlear nucleus of the mouse. J. Acoust. Soc. Am. 1985; 78: 328-33. Overholt EM, Rubel EW, Hyson RL. A circuit for coding interaural time differences in the chick brain stem. J Neurosci 1992; 12: 1698-708. Payne RS. Acoustic localization of prey by barn owls (T~o alba). J. Exp. Biol. 1971" 54: 535-73. Raman IM, Trussell LO. The kinetics of the response to glutamate and kainate in neurons of the avian cochlear nucleus. Neuron. 1992; 9: 173-86. Reyes AD, Rubel EW, Spain WJ. Membrane properties underlying the firing of neurons in the avian cochlear nucleus. J. Neurosci. 1994:14" 535264. Rhode WS, Oertel D, Smith PH. Physiological response properties of cells labeled intracellularly with horseradish peroxidase in cat ventral cochlear nucleus. J. Comp. Neurol. 1983; 213: 448-63. Rose G, Heiligenberg W. Temporal hyperacuity in the electric sense of fish. Nature. 1985; 318:178-80. Rubel EW, Parks TN. Organization and development of brainstem auditor3, nuclei of the chicken: Tonotopic organization of N. magnocellularis and N. laminaris. J. Comp. Neurol. 1975" 164" 411-34. Ryugo DK, Fekete DM. Morphology of primary axosomatic endings in the anteroventral cochlear nucleus of the cat: A study of the endbulbs of Held. J Comp Neurol. 1982:210: 239-57. Scheich H, Bullock TH, Hamstra RH. Coding properties of two classes of afferent nerve fibers: High frequency electroreceptors in the electric fish Eigenmannia. J. Neurophysiol. 1973" 36: 39-60. Simmons JA. The resolution of target range by echolocating bats. J. Acoust. Soc. Am. 1973, 54:157-73.
Processing of Temporal Information in the Brain
51
Simmons JA. Perception of echo phase infommtion in bat sonar. Science. 1979; 204: 1336-38. Simmons JA, Ferragamo M, Moss CF, Stevenson SB, Aires RA. Discrimination of jittered sonar echoes by the echolocating bat, Eptesicus fitscus: The shape of target images in echolocation. J. Comp. Phy'siol. It)90: 167" 589-616. Smith PH, Joris PX, Yin TCT. Projections of spherical bushy cells to the MSO in the cat" evidence for delay lines. Neurosci. Abstr. 1990' 16" 723. .
Smith ZDJ, Rubel EW. Organization and development of brainstcm auditor3., nuclei of the chicken' Dendritic gradients in nucleus laminaris. J. Comp. Neurol. 1979; 186" 213-39. Spzir MR, Sento S, R3~go DK. The central projections of the cochlear nerve fibers in the alligator lizard. J. Comp. Neurol. 1990' 295' 530-47. Suga N. Cortical computational maps for auditory imaging. Neural. Networks. 1990: 3" 3-21. Sullivan WE, Konishi M Segregation of stimulus phase and intensity coding in the cochlear nucleus of the barn owl. J. Neurosci. 1984 4" 1787-99. Szabo T. Sense Organs of the lateral line system in some electric fish of the Gyxnnotidae, Momlyridac and Gyl~marchidae. J. Morphol. 1965" 117' 22950. Szabo T, Ravaillc M, Libouban S, Enger PS. The monm'rid rhombencephalon. I. Light and EM investigations on the structure and connections of the lateral line lobe nucleus with HRP labeling. Brain Rcs. 1983" 26" 1-19. .
Takahashi T, Moiseff A, Konishi M. Time and intensity cues are processed independently in the auditor 3, sy'stcm of the owl. J. Ncurosci. 1984" 4' 178186.
52
(7.E. Carr and S. Amagai
vonder Emde G, Bleckmann H. Extreme phase sensitivity of afferents which innervate mormyromast electroreceptors. Naturwissenschafien. 1992: 79: 131-33. von der Emde G, Zelick R. Behavioral test of the detection of synthetic signals imitating capacitive phase shifts in mormyrid weakly electric fish Neurosci. Abstr. 1994; 20: 370. Weiss TF, Rose C. A comparison of s3aachronization filters in different auditory, receptor organs. Hear Res. 1988- 33" 175-80. Wu SH. Oertel D. Intracellular injection with horseradish peroxidase of physiologically characterized stellate and bushy cells in slices of mouse anteroventral cochlear nucleus. J. Neurosci. 1984: 4: 1577-88. Yin TCT, Chan JCK. Interaural time sensitivity in medial superior olive of cat. J. Neurophysiol. 1990: 64: 465-88. Young SR, Rubel EW. Frequency-specific projections of individual neurons in chick brainstem auditory, nuclei. J. Neurosci. 1983' 7: 1373-78.
Time, Internal Clocks and Movement M.A. Pastor and J. Artieda (Editors) 9 1996 Elsevier Science B.V. All rights reserved.
53
L A R G E - S C A L E I N T E G R A T I O N OF CORTICAL I N F O R M A T I O N PROCESSING STEVEN L. BRESSLER Center for Complex A~vstems. Florida Atlantic Univ.. 777 Glades Road. Boca Raton. FL 33431 U.A'.A. ABSTRACT. An important problem in cognilive neuroscience is how information processing in widely distributed conical areas is integrated Io produce coherent perceptual and behavioral fimction. This question of large-scale integration of cortical informalion processing is considered in light of certain connectional properties of conical neuroanatomical organization. Computational principles are proposed that may allow the cortex to provide flexible adaptation to changing circumstances during goal-directed behavior. Evidence is also discussed that temporal s3.'nchrony of neuronal ensemble activity serves as the physiological basis for large-scale cortical fi~nctional integration.
I. Introduction
Integration is ubiquitous in neural fi~nction, taking place at every level of description. Effects at each level result from the concerted integration of lower-level events. For example, at the membrane level, action potentials represent the integration of events occurring at numerous ion channels, and graded potentials at the axon hillock arise from the integration of multiple synaptic events in the dendritic tree. Given the modular nature of cerebral cortical organization, it is reasonable to hypothesize that the operations of individual cortical areas are integrated in the production of coherent and coordinated perceptuomotor behavior. The richness and adaptability of mammalian behavior suggest that individual cortical zones work together in combined, flexibly configured assemblages at the systems level. The purpose of this chapter is to consider how integration may occur in the cerebral cortex at the systems level. It is proposed that the ability of cortex to function as a system in conjunction with subcortical structures depends on the coordinated interplay of its constituent parts, the individual
54
S.L Bressler
cortical areas. The fact that interareal pathways extend over a range of distances suggests that, besides interacting on a small spatial scale within local networks, cortical neurons also interact on a larger scale with neurons in other, widely distributed cortical areas. The large-scale cortical network is postulated to exist as a functional entity operating at the systems level by the integrated actions of distributed cortical zones. The primate cerebral cortex consists of a mosaic of tens, perhaps hundreds, of distinct functional areas extending throughout both hemispheres. Each cortical area is characterized by its intrinsic connections (its neurons being densely interconnected in local circuits) and extrinsic connections (its neurons having common input and output projection patterns). Functionally, it has been proposed that cortical areas are distinguished by unique elementary operations (Posner and Rothbart, 1994) presumably detemained by their particular patterns of connectivity rather than any atypical cell types or internal circuit properties. Although it is likely that each cortical area has a unique functional role. it is unlikely that any cortical area operates in isolation under normal behavioral conditions. One reason is that no behavioral function has been found to depend on just a single area. To the contrary, sensory and motor functions are performed by multiple cortical areas, more than twenty in the macaque monkey for vision alone (Felleman and Van Essen, 1991). Higher cognitive functions as well involve multiple areas rather than any single critical control center (Mesulam, 1990: Damasio and Damasio, 1994: Bressler, 1995). Another reason is that cortical areas project widely to other areas" on average, each area is estimated to send output to ten other areas (Felleman and Van Essen, 1991). so that it is to be expected that when one cortical area is active it will exert an activating influence on other areas over excitatory, projection pathways. A third reason is the multiplicity of afferent pathways to cortical areas (they also receive from roughly ten areas), which suggests that they normally perform under the influence of inputs from other areas. Thus cortical integration at the systems level appears to be a cooperative property that involves the interaction of multiple areas rather than the activation of a special integration center (Goldman-Rakic, 1988).
2. Principles of Cortical Integration The connectional organization of the cerebral cortex suggests some computational principles underlying the functional integration of cortical
Large-Scale Integration of Cortical Information Processing
55
areas into large-scale networks. The manner in which cortical areas interact is in large part determined by the architecture of the neuronal projections between them. An important feature of cortical architecture is the hierarchical organization of Cortical areas, as determined by the laminar origin and termination patterns of cortico-cortical connections seen in retrograde and anterograde labeling experiments (Felleman and Van Essen, 1991). Beginning with primary sensory and motor areas, other areas are assigned hierarchical positions in each modality depending on the pattern of ascending, lateral, and descending connections they make with other areas. At the highest hierarchical levels are the transmodal areas in frontal, parietal, and temporal lobes (Mesulam. 1985). A general fimctional correlate of this hierarchical arrangement is that each level processes infonuation of a more abstract nature than the levels below. Thus, for example, single neurons in visual area V I are selective for low-level visual features such as lines and gratings, those in a peristriate region such as area V4 arc selective for color, and those in parts of inferotemporal cortex are selective for composite visual patterns such as faces. A crucial question concerning the nature of computation in this arrangement is the dynamical relation between infommtion processing at different hierarchical levels. One possible interpretation of the cortical hierarchv is that infommtion at each level is more elaborated and refined than at the level below, with the most complete and specific perceptual and motor representations occurring at the highest levels. If one takes this view. then it might be expected that information processing would normally proceed in a serial manner along a cascade across hierarchical levels. Stimulus-related processing would begin in primary sensor3., areas and proceed sequentially from one area to another, with processing completed at each level and the resultant output being transferred onward before processing begins at the next level. Object recognition would occur when. after convergent activation of progressively higher levels, specific neurons representing the object were activated at the highest levels. Similarly. motor acts would occur as the result of the sequential elaboration at progressively lower levels of motor programs stored at the highest levels. This sequential activation model does not appear to be supported by single-unit studies of visual (Ashford and Fuster, 1985" Cobum et al., 1990: Dinse et al., 1991' McClurkin et al., 1991) and motor (Alexander and Crutcher, 1990a, b: Crutcher and Alexander, 1990: Kalaska and Crammond.
56
S.L. Bressler
1992) areas which show that stimulus- and movement-related activity occurs concurrently at multiple levels. Information processing, rather than proceeding sequentially from one hierarchical level to another, seems to involve an ongoing functional conjunction across levels. If so, then is it likely that information transmittal across the hierarchy, although not sequential, is nonetheless unidirectional, proceeding from lower to higher in sensory systems and from higher to lower in the motor system'? Another important feature of cortical organization to be considered in addressing this question is that interareal connections are largely bidirectional. Since cortical areas are almost exclusively connected by excitatory synapses between pyramidal cells (Braitenberg, 1989" Douglas and Martin, 1990). and since most connected areas are reciprocally coupled. bidirectional excitatory' interactions between areas presumably are an important feature of cortical network dynamics. The implication is that influences are exerted across hierarchical levels in a recurrent manner, back and forth between areas, rather than always in one direction from one area to another. The advantage of such recursive interactions between cortical areas may be to allow multiple local networks at different hierarchical levels to temporarily combine into larger processing units in order to accomplish complex fimctions that require contributions from them all (Mallot and Brittinger, 1989). In these large-scale networks, transmodai areas may play not only a receptive role, but also an organizational role by promoting interactions among areas in the specific sensory and motor areas (Damasio and Damasio, 1994: Mesulam, 1994). This type of binding mechanism may be used for the recall of a previous perceptual experience, with transmodal areas directing the re-interaction of the same areas that had interacted during the original act of perception (Damasio. 1989, 1990). The view that cortical infommtion processing involves the recursive interaction of local area networks at multiple hierarchical levels in a largescale network implies that activity in each local network evolves under the continuing influence of activity occurring concurrently in other local networks to which it is commcted (Zeki. 1990; Goldman-Rakic et al., 1992). Several potential computational advantages derive from this description of cortical information processing (Finkel and Edelman, 1989" Spores et al., 1991). The recursive sharing of information among local networks distributed across multiple hierarchical levels allows each local network to integrate discriminations made by multiple other local networks from which
Large-Scale Integration of Conical Information Processing
57
it receives input in its oxen1processing. Since the local network is constrained and directed by feedback from other local networks at different hierarchical levels, it has access to information from many levels, not just from one as in serial processing schemes. Furthermore, conflicts in the processing by different local networks may be eliminated through competition, and because outputs from a local network are returned to it recursively, it can iteratively synthesize responses to complex, ongoing stimuli. A further computational implication of reciprocal corticocortical pathways relates to spatial pattern analysis (Rolls, 1989: Ullman. 1994). Many cortical areas are known to process information in parallel across the two dimensions of the cortical surface, often represented by a topographic map (Van Essen and Maunsell. 1980). Although information in cortical areas receiving sensor3.' input is typically organized as a topographic map of the sensor3' receptor array (Udin and Fawcett, 1988), and topographic maps of the body musculature are found in several motor areas (Lemon, 1988), topographic maps need not be strictly sensor3' or motor, but may be computational in nature, and ma.v be found in higher-order association areas (Knudsen et al., 1987: Allman, 1990). Recent evidence suggests that infonnation is available in spatial patterns of the modulated amplitude of cortical local field potential activity (Freeman and Barrie, 1994). In this context, the process of recursive interareal interaction is viewed as a mechanism for iterative pattern matching (Mumford. 1992. 1994). Local networks in interacting areas engage in recursive operations which tend to reduce the difference between their activity, patterns. A hierarchically higher area backprqjects a "template" pattern to a lower area. enhancing features of the activity pattern in the lower area which match the template. The lower area projects its pattern forward to the higher area. at the same time modif3,ing the template.
3. Large-Scale Cortical Integration by Temporal Synchrony From the above discussion, simultaneous recordings of cortical activity from different areas may be expected to show signs of fimctional integration. In fact. studies of brain metabolism and blood flow do show consistent patterns of coactivated regions in relation to higher conical fitnctions (GoldmanRakic et al.. 1992: Friedman ct al.. 1994. Honvitz ct al.. 1995). Although these measures provide images of coactivated areas with great spatial
58
S.L. Bressler
resolution, their temporal resolution (at best on the order of seconds) is relatively poor because they depend on secondary effects which are delayed following changes in neuronal activity. Thus these techniques are not currently able to resolve changes in neuronal activity on the sub-second time scale necessary to observe the dynamics of functional integration. At a minimum, these dynamics are expected to take the form of patterns of synchronous, or correlated, activities of distributed neurons or neuronal groups (Tononi et al., 1994). Electrophysiologicai measures do have the requisite temporal resolution, although they typically have poor spatial resolution. Of these measures, single-unit recording has provided some evidence for the functional conjunction of cortical areas in the foma of correlated spike trains (Bullier et ai., 1992: Nelson et al., 1992: Roe and Ts'o, 1992: Nowak et al., 1994). However, these correlations tend to be weak. and require averaging over relatively long time periods to bring out the statistical effects. The correlated activity of individual cortical neurons may not be an effective means of observing long-range functional interactions between cortical areas (Eckhom. 1992: Engel et ai., 1992" Tononi et al.. 1992). On the fraction-ofa-second time scale of cognitive events, single cortical neurons typically generate only a few spikes (Abeles, 1991" Freeman and Barrie. 1994). Since very few contacts exist between individual cortical neurons (Gabbott et al., 1987: Braitenberg, 1989: Douglas and Martin, 1991), the joint firing of tens of cells is minimally required to alter the firing probability of a single target neuron (Abeles, 1991' Tononi et al.. 1992). Thus only a small percentage of cortical spikes can be accounted for by the activity of any other single neuron (Ts'o et al., 1986: Eckhom. 1991 ). Because cortical pyramidal cells are densely intercomlected with their neighbors, but only sparsely connected to any single neighbor, and because the excitator), drive to a pyramidal cell is provided by the convergent action of hundreds of others (Douglas and Martin. 1991), the collective activity of the local neuronal ensemble may be more appropriate than single-neuron activity for observing functional conjunctions between cortical areas (Tononi et al., 1992). Local ensemble activity is manifested both as the summed firing of action potentials, measured by the local multi-unit spike density, and as the sum of dendritic currents, measured by the Local Field Potential (LFP). A number of studies have reported correlated interareal cortical activity at the ensemble level between narrow-band oscillations in the gamma
Large-Scale Integration of Cortical Information Processing
59
frequency (20-90 Hz) range (Bressler, 1990; Singer, 1990, 1993, 1994). The correlation typically occurs in brief, fraction-of-a-second episodes of increased synchronization of the multi-unit spike density or LFP waveforms. Correlated multi-unit gamma oscillations have been reported in the cat between areas 17 and posteromedial lateral suprasylvian sulcus (Engel et al., 199 l a), and between area 17 of the iet~ and right hemispheres (Engel et al., 1991b). The LFP may be most advantageous for observing interareal conjunction because spatial averaging is an important characteristic of the communication between cortical areas (Freeman and Barrie, 1994). The activation of dendritic synapses on cortical pyramidal cells induces loop currents which flow through the neuronal interior from the s~lapse to the trigger zone. and externally, through the extracellular space. Because of the parallel orientation and elongated c)~oarchitecture of pyramidal cells, the loop currents generated by pyramidal cell synapses summate to form a local mean field which can be recorded with microelectrodes as the LFP. The spatial averaging manifested in the LFP thus may approximate that which occurs in the interactions betxveen areas as a result of axonal divergence and dendritic integration. Correlations of gamma LFP oscillations have been reported in relation to visual stimuli, in the cat among visual areas 17, 18. and 19 (Eckhom et al., 1988" Eckhom and Schanze, 1991" Eckhom, 1991. 1992), and in the monkey between visual areas V I and V2 (Frien et al., 1994), and, in relation to movements, between somatosensor3, and primar 3, motor areas (Murthy and Fetz, 1992) and between primary and nonprimau' motor areas (Sanes and Donoghue, 1993). The study by Bressler et al. (1993) has provided the most direct evidence for interareal fi~nctional coniunctions in the behaving monkey. Brief episodes of correlated LFP oscillations were observed in relation to perfommnce of a visual pattern discrimination task in complex patterns involving sites in widely distributed conical areas, including striate, prestriate, inferotemporal, parietal, motor, and frontal cortices. The frequency of the correlated oscillations, while including the ga~rmm band, was wider, covering the entire range between 0 and 100 Hz. The correlation of waves in the delta and theta ranges is consistent with earlier studies of human scalp-recorded event-related potentials, which also demonstrated correlations in complex patterns involving widely distributed sites (Gevins and Bressler, 1988' Gevins et al.. 1987, 1989a,b).
60
S.L. Bressler
4. Implications for the Understanding of Perceptual Unity Successful goal-directed behavior requires that an animal flexibly adapt to a continually changing set of plans, contingencies, and sensory cues. The mammalian cerebral cortex is remarkably effective at furnishing this adaptability. It allows the spatial and temporal integration of sensory events into perceptual structures and motor acts into behavior. It also provides coherence anaong sensor), modalities, between sensation and movement, between sensations from the internal milieu and the external environment. and between memor), and immediate perception. To accomplish these complex functions requires the integration of the elementary functions of widely distributed cortical areas. The idea that cortical areas reciprocally interact in large-scale networks may serve as a basis for understanding how this integration takes place. Each cortical area has the potential of entering a large number of different active states, each manifested as a spatially modulated activity distribution, as determined by its synaptic inputs and its internal synaptic weights (Freeman. 1975). Thus the number of possible combinations of states of an interconnected network of cortical areas is ver)' large. Considering that different operations may be accomplished by different networks, the total number of possible combinations is even larger. This set of available states constitutes a space which may be traversed by the organism as it interacts with its environment. Exteroceptive and interoceptive inputs to the cortex, as well as subcortical control structures, constrain the network configuration at each instant. Iterative pattern matching between multiple cortical areas may serve as a mechanism that leads to a global relaxation of the large-scale network into an optimal response to each momentary set of constraints, thereby affording the animal flexible adaptation to an ever-changing succession of behavioral situations. An important implication of this description is that information processing in the cortical hierarchy involves a continual interplay of topdown and bottom-up influences. Jackendoff (1994). from a strictly structural analysis, has reached a similar conclusion: that the semantic. syntactic, and phonological levels of linguistic processing are all engaged in continual correspondence matching during both speech comprehension and production. Finally. it has been noted that no single cortical area has been found whose destruction leads to the disintegration of all perceptual awareness
Large-Scale Integration of Conical Information Processing
61
(Damasio, 1990). Thus it does not appear that perceptual unity is equivalent to unity of brain location. However, disintegration of perception does occur to varying degrees in man3' fomas of mental illness, including schizophrenia. From the perspective of the current discussion, disruptions in the dynamics of functional integration of large-scale cortical networks may be responsible for this pathological breakdown of function. It is interesting in this regard that a recent study (Hoffman and McGlashan, 1993) found that schizophrenic symptoms could be simulated by reducing the connectivity between nodes, representing cortical areas, in a computer model of a largescale cortical network.
ACKNOWLEDGEMENTS. This work was supported by NIMH Grant MH42900 Io the Center for Complex Systems
5. References
Abeles M. Corticonics" Neural Circuits of the Cerebral Cortex. Cambridge: Cambridge Univ. Press. 1991. Alexander GE, Crutcher MD. Preparation for movement: neural representations of intended direction in three motor areas of the monkey. J. Neurophysiol. 1990a: 64" 133-50. Alexander GE, Crutcher MD. Neural representation of the target (goal) of visually guided am1 movements in three motor areas of the monkey. J. Neurophysiol. 1990b" 64' 164-78. Allman J. The origin of the neocortex. Semin. Neurosci. 1990.2" 257-62. Ashford JW, Fuster JM. Occipital and inferotemporai responses to visual signals in the monkey. Exp. Neurol. 1985" 90' 444-66. Braitenberg V. Some arguments for a theory of cell assemblies in the cerebral cortex. In: Nadel L, Cooper LA. Culicover P, Hamish RM, editors. Neural Connections, Mental Computation. Cambridge: MIT Press. 1989" 137-45.
62
S.L. Bressler
Bressler SL. The gamma wave: a cortical information carrier'? Trends Neurosci. 1990; 13" 161-62. Bressler SL. Large-scale cortical networks and cognition. Brain Res. Rev. 1995; 20: 288-304. Bressler S L, Coppola R, Nakamura R. Episodic multiregional cortical coherence at multiple frequencies during visual task performance. Nature 1993" 366: 153-56. Bullier J, Munk JHJ, Nowak LG. Synchronization of neuronal firing in areas V I and V2 of the monkey. Soc. Neurosci. Abstr 1992. 18" 11. .
Cobum KL, Ashford JW, Fuster JM. Visual response latencies in temporal lobe structures as a function of stimulus information load. Behav. Neurosci 1990: 104: 62-73. Crutcher MD, Alexander GE. Movement-related neuronal activity selectively coding either direction or muscle pattern in three motor areas of the monkey. J. Neurophysiol. 1990" 64" 151-63. Damasio A. The brain binds entities and events by multiregional activation from convergence zones. Neural Comp. 1989; 1" 123- 32. Damasio A. Time-locked multiregional retroactivation" a systems-level proposal for the neural substrates of recall and recognition. In: Eimas P, Galaburda A, editors. Neurobiology Cognition. Cambridge: MIT Press. 1990" 25-62. Damasio AR, Damasio H. Cortical systems for retrieval of concrete knowledge: the convergence zone framework. In: Koch C, Davis JL, editors. Large-Scale Neuronal Theories of the Brain. Cambridge: MIT Press. 1994" 61-74. Dinse HR, Kriiger K, Mallot, HA, Best J. Temporal structure of cortical information processing: cortical architecture, oscillations, and nonseparability of spatio-temporal receptive field organization. In: Kriiger J, editor. Neuronal Cooperativity. Berlin: Springer-Verlag. 1991: 68-104.
Large-Scale Integration of Cortical Information Processing
63
Douglas RJ, Martin KAC. Neocortex. In: Shepherd G, editor. The Synaptic Organization of the Brain. New York: Oxford Univ. Press. 1990: 389-438. Douglas RJ, Martin KAC. Opening the grey box. Trends Neurosci. 1991; 14: 286-93. Eckhom R. Stimulus-specific synchronizations in the visual cortex: linking of local features into global figures? In: Kriiger J, editor. Neuronal Cooperativity. Berlin: Springer-Verlag. 1991:184-224. Eckhorn R. Principles of global visual processing of local features can be investigated with parallel single-cell- and group-recordings from the visual cortex. In: Aertscn A, Braitenberg V, editors. Information Processing in the Cortex. Berlin: SPringer-Verlag. 1992: 385-420. Eckhom R, Schanze T. Possible neural mechanisms of feature linking in the visual system: stimulus-locked and stimulus-induced synchronizations. In: Babloyantz A, editor. Self-Organization, Emerging Properties, and Learning. New York: Plenum. 1991: 63-80. Eckhorn R, Bauer R, Jordan W, Brosch M, Kruse W, Munk M, Reitboeck HJ. Coherent oscillations: a mechanism of feature linking in the visual cortex'? Biol. Cvbern. 1988~ 60:121-30. Engel AK, Kreiter AK, K6nig P, Singer W. Synchronization of oscillatory neuronal responses between striate and extrastriate visual cortical areas of the cat. Proc. Natl. Acad. Sci. USA. 1991a; 88: 6048-52. Engel AK, K6nig P, Kreitcr AK, Singer W. Interhemispheric synchronization of oscillatory, neuronal responses in cat visual cortex. Science. 1991b: 252' 1177-79. Engel AK, K6nig P, Kreitcr A, Schillen TB, Singer W. Temporal coding in the visual cortex: new vistas on integration in the nervous system. Trends Neurosci. 1992: 15: 218-26. Felleman D, Van Essen D. Distributed hierarchical processing in the primate cerebral cortex. Cerebral Cortex. 1991: 1: 1-47.
64
S . L Bressler
Finkel L, Edelman G. Integration of distributed cortical systems by reentry: a computer simulation of interactive functionally segregated visual areas. J. Neurosci. 1989; 9:3188-208. Freeman WJ. Mass Action in the Nervous System. New York: Academic. 1975. Freeman WJ, Barrie JM. Chaotic oscillations and the genesis of meaning in cerebral cortex. In: Buzs~ki G. Llinas IL Singer W, Berthoz A, Christen Y, editors. Temporal Coding in the Brain. Berlin: Springer-Verlag. 1994: 1337. Friedman HR. Goldman-Rakic P S. Coactivation of prefrontal cortex and inferior parietal cortex in working memo~' tasks revealed by 2DG functional mapping in the rhesus monkey. J. Neurosci. 1994" 14" 2775-88. Frien A, Eckhom R, Bauer R. Woelbern T, Kehr H. Stimulus-specific fast oscillations at zero phase between visual areas V 1 and V2 of awake monkey. Neurorep. 1994: 5" 2273-77. Gabbott PLA, Martin KAC, Whitteridge D. Connections between pyramidal neurons in layer 5 of cat visual cortex (area 17). J. Comp. Neurol. 1987; 259: 364- 81. Gevins AS. Bressler S L. Functional topography of the human brain. In: Pfurtscheller G, Lopes da Silva F, editors. Functional Brain Imaging. Bern: Hans Huber. 1988: 99- I 16. Gevins AS, Morgan H, Bressler S, Cutillo B, White R, Illes, J, Greer D, Doyle J, Zeitlin G. Human neuroelectric patterns predict performance accuracy. Science. 1987" 235" 580-85. Gevins AS, Bressler S. Morgan H. Cutillo B, White R, Greer D, Illes J. Event-related eovariances during a bimanual visuomotor task. I. Methods and analysis of stimulus- and response-locked data. Electroenceph. Clin. Neurophysiol. 1989a: 74: 58-75.
Large-Scale Integration of Cortical Information Processing
65
Gevins AS, Cutilio B, Bressler S, Morgan H, White R, llles J, Greer D. Event-related covariances during a bimanual visuomotor task. II. Preparation and feedback. Electroenceph. Clin. Neurophysioi. 1989b" 74" 147-60. Goldman-Rakic P. Topography of cognition: parallel distributed networks in primate association cortex. Ann. Rev. Neurosci. 1988; 11" 137-56. Goldman-Rakic P, Chafee M, Friedman H. Allocations of function in distributed circuits. In Ono T, Squire L, Raichle M, Perrett D, Fukuda M, editors. Brain Mechanisms of Perception and Memory: From Neuron to Behavior. New York' Oxford Univ. Press. 1992: 445-56. Hoffman RE, McGlashan TH. Parallel distributed processing and the emergence of schizophrenic symptoms. Schizophrenia Bull. 1993" 19:11940. Horwitz B, Mclntosh AR, Haxby JV. Grady CL. Network analysis of brain cognitive filnction using metabolic and blood flow data. Behav. Brain Res. 1995" 66: 187-93. Jackendoff R. Consciousness and the Computational Mind. Cambridge: MIT Press. 1994. Kalaska JF, Crammond DJ. Cerebral cortical mechanisms of reaching movements. Science. 1992" 255" ! 517-23. Knudsen E, Du Lac S, Esterly S. Computational maps in the brain. Ann. Rev. Neurosci. 1987:10:41-65. Lemon R. The output map of tile primate motor cortex. Trends Neurosci. 1988; 11" 501-6. Mallot H, Brittinger R. Towards a network theory, of cortical areas. In: Cotterill R, editor. Models of Brain Function. Cambridge: Cambridge Univ. Press. 1989" 175-89.
66
S.L. Bressler
McClurkin JW, Optican LM, Richmond BJ, Gawne TJ. Concurrent processing and complexity of temporally encoded neuronal messages in visual perception. Science. 1991" 253: 675-77. Mesulam M. Patterns in behavioral neuroanatomy. In: Mesulam M, editor. Principles of Behavioral Neurology, Contemporary Neurology Series. Philadelphia: FA Davis. 1985" 1-70. Mesulam M. Large-scale neurocognitive networks and distributed processing for attention, language, and memory. Ann. Neurol. 1990; 28: 597-613. Mesulam M. Neurocognitive networks and processing. Rev. Neurol. 1994:150" 564-69.
selectively
distributed
Mumford D. On the computational architecture of the neocortex. II. The role of cortico-cortical loops. Biol. Cybcrn. 1992" 66" 241-51. Mumford D. Neuronal architectures for pattern-theoretic problems. In: Koch C, Davis JL, editors. Large-Scale Neuronal Theories of the Brain. Cambridge: MIT Press. ! 994" 125-52. Murthy VN, Fetz EE. Coherent 25- to 35-Hz oscillations in the sensorimotor cortex of awake behaving monkeys. Proc. Natl. Acad. Sci. USA. 1992; 89: 5670-74. Nelson Jl, Salin PA. Munk M, Arzi M, Bullier, J. Spatial and temporal coherence in cortico-cortical connections: a cross-correlation study in areas 17 and 18 in the cat. Vis. Neurosci. 1992" 9" 21-37. Nowak LG, Munk MHJ. Chounlamountri N, Bullier J. Temporal aspects of information processing in areas V I and V2 of the macaque monkey. In" Pantev C, Elbert T, Liitkenhrner B. editors. Oscillatory, Event-Related Brain Dynamics. New York: Plenum. 1994" 85-98. Posner MI, Rothbart MK. Constructing neuronal theories of mind. In: Koch C, Davis JL, editors. Large-Scale Neuronal Theories of the Brain. Cambridge: MIT Press. 1994" 183-99.
Large-Scale Integration of Cortical Information Processing
67
Roe AW, Ts'o DY. Functional cotmectivity between V I and V2 in the primate. Soc. Neurosci. Abstr. 1992: 18: 11. Rolls ET. Functions of neuronal networks in the hippocampus and neocortex in memory. In: Byme JH, Berry WO, editors. Neural Models of Plasticity. San Diego: Academic Press. 1989: 240-65. Sanes JN, Donoghue JP. Oscillations in local field potentials of the primate motor cortex during voluntary movement, Proc. Natl. Acad. Sci. USA. 1993" 90" 4470-74. Singer W. Search for coherence: a basic principle self-organization. Concepts Neurosci. 1990: 1" 1-26.
of
cortical
Singer W. Synchronization of cortical activity and its putative role in information processing and learning. Ann. Rev. Physiol. 1993" 55' 349-74. Singer W. Putative functions of temporal correlations in neocortical processing. In: Koch C, Davis JL, editors. Large-Scale Neuronal Theories of the Brain. Cambridge: MIT Press. 1994" 201-37. Sporns O, Tononi G, Edelman G. Dynamic interactions of neuronal groups and the problem of cortical integration. In" Schuster H, editor. Nonlinear Dynamics and Neuronal Networks. Weinheim: VCH Verlagsgesellschatt. 199 I" 205-40. Tononi G, Sporns O, Edelman GM. Reentry and the problem of integrating multiple cortical areas simulation of dynamic integration in the visual system. Cerebral Cortex. 1992: 2 310-35. Tononi G, Sporns S, Edelman GM. A measure for brain complexity" relating functional segregation and integration in the nervous system. Proc. Natl. Acad. Sci. USA. 1994: 91" 5033-37. Ts'o DY, Gilbert CD, Wiesel TN. Relationships between horizontal interactions and functional architectt.re as revealed by cross-correlation analysis in cat striate cortex. J. Neurosci. 1986: 6' 1160-70.
68
S.L. Bressler
Udin SB, Fawcett JW. Fommtion of topographic maps. Ann. Rev. Neurosci. 1988; 11: 289-327. Ullman S. Sequence seeking and counterstreams: a model for bidirectional information flow in the cortex. In: Koch C, Davis, JL, editors. Large-Scale Neuronal Theories of the Brain. Cambridge: MIT Press. 1994: 257-70. Van Essen DC, Maunsell JHR. Two-dimensional maps of the cerebral cortex. J. Comp. Neurol. 1980:191:255-81. Zeki S. Parallelism and fimctional specialization in human visual cortex. Cold Spring Harbor Syrup. Quant. Biol. 1990; 55: 651-61.
Time, Internal Clocks and Movement M.A. Pastor and J. Artieda (Editors) 9 1996 Elsevier Science B.V. All rights reserved.
69
MODELS OF NEURAL TIMING CHRISTOPHER MIALL
University Laboratory of Physiology, Parks Road, Oxford, OX1 3PT, UK ABSTRACT. That the centr~li nervous s.vstcm is capable of accurate encoding of time is obvious, as numerous human and animal experiments can demonstrate. But it is not at all clear how the phenomena of storage ;Ind processing of temporal informalion are achieved: it is in f~lct r~lther difficult to see how neurons can operale accurately over time sc~llcs of seconds or minules, which must be required for controlling everyday bclmviours. In this chapter I want to first describe whal I consider to be the Inain difficulties in the neural encoding of time, then list some of the ways in which neurons, as we currently understand them. might actually be used to encode temporal information. Two models will then be described in more detail, that could be used for timing. Both are 'network' models in which a large population of neurons combines to encode a temporal interval. I will also briefly and selectively review some anifici;ll neur~il network models that deal will time. The ideas presented here lmve been tested m~linly through computer simulations. and so it remains to discover which, if any. of these methods are used in biological systems.
1. Introduction The question of how to encode time in neural systems should perhaps be split into three parts. First there is the problem of the generation of temporal control signals for movement. Second is the problem of processing continuously variable signals, and extracting their temporal structure. Third, there is the problem of detecting, storing, and recalling discrete time intervals. The first problem, of generating temporal control signals, for example to produce a finely timed motor response such as speech or a rhythmic voluntary limb movement, has being tackled successfully in several instances, especially in our understanding of central pattern generating circuits (Harris-Watavick et al.. 1992" Selverston, 1993). It seems that intrinsic pacemaking ncurons, complex membrane dynamics and reverbatory
70
C. Miall
circuits are common to all well understood examples: it may well prove to be the same for less obviously 'patterned' outputs like voluntaD' limb movement. So the mechanisms by which such temporal activity are generated, while not yet certain, have at least been identified and can realistically be extrapolated from those well documented examples. The second problem was of dealing with continuously variable signals and using the temporal structure of these signals to extract, for example, image motion or to predict the path of a moving target. I emphasise prediction because this seems vital to overcome the delays associated with action. We know rather little of these sorts of processes. There are, of course, some good theories of how visual motion detectors could be constn~cted and the cortical areas concerned with visual motion processing have been Iocalised. But, as far as I know, neither the mechanism nor the site of a predictor has yet been found. Third. there was the problem of how to detect, store, and recall discrete intervals of time bounded by marking events. In this category would fall. for example, estimating the time between two hand claps or the problem of timing an interval between presentation of a stimulus and the delayed response for a reward. This is the processing of time as an distinct quantity, separated from the physical or perceptual characteristics of the stimuli. It is this aspect of timing that we are most in the dark about, and this aspect that I believe is most difficult to propose a neural mechanism for. This is largely' what I will discuss in this chapter. 1.1. BEI IAVIOIIIUM_,AND NEIJRAI, TIME SCALES
The nervous system operates over the range of microseconds (in the encoding and separation of sensory signals) to years (in memor3., and learning). At the finest scale, tinting is dependent on and limited by the biophysical nature of neurons. Thus the quantal nature of membrane ion channels, the stochastic diffusion of neurotransmitters, the integration of post-sxxmptic events and the conduction velocity of active membranes set lower bounds to the time scale on which events can be accurately measured. processed or distinguished. When people are asked to estimate time over longer scales, perhaps tens of seconds, they tr3., to count. We can in fact estimate well even if forced not to count, by being asked to complete some other task during the interval. However, the choice of a counting strategy implies that we are more accurate in estimating multiple stnall intervals than in estimating one long interval: hence use of a counting strategy may help us
Models of Neural Timing
71
avoid the difficulty of pure time estimation by relying instead on the sorts of rh3ethmicai pattern generators to which I alluded earlier. Over much longer scales of hours, days and years, we probably resort of estimating time with reference to other k n o ~ time signals: the occurrence of an event relative to a mealtime, perhaps, or relative to some other remembered ('time-stamped') event. Hence at this range, time is not being handled as an explicit time var3'ing neural signal, but as non-temporal, relational, knowledge about different episodes. However. we do routinely estimate intervals or durations ranging from hundreds of milliseconds up to several seconds, and can do so without resorting to tricks like counting in our heads (Macar et al., 1992). The mechanism or mechanisms responsible for this range of timing ability seem to be most difficult to resolve in tenus of individual neurons or neural circuits. There are two reasons why time keeping over these ranges is difficult. First, the intrinsic temporal range of many neurons is limited. A typical estimate for the membrane time constant is about 5 to 20 ms: synaptic and axonal delays within the vertebrate brain may add another 10-20 ms. Given this 'basic' operating range tens of milliseconds, it becomes difficult to extend the temporal scale to hundreds of milliseconds. There are of course a wide variety of complex membrane properties (with slow exponential decay of membrane potential, oscillatory or rebound behaviour, bistable membrane potentials and the like), which might be extend the time horizon. I will return to these possibilities fi~rther latcr on. However, I think it fair to say that those neurons within vertebrate circuits that are known to be important in temporal processing (e.g. prefrontal cortex or cerebellum) do not display, obviously complex d.~ammics. The second difficulty is that neurons are inaccurate (although this does not necessarily mean that they are unreliable" cf. Bailck and Rieke. 1992). They operate stochastically, both at the level of the membrane and the s)ampse (Holden, 1976). We do not kalow the range over which synaptic weights nom~aily var3.' with long term potentiation or depression, nor the precision with which they are set, nor the precision with which they transmit single impulses. Hoxvever, the resolution of sy~aptic weights is probably quite low, perhaps 4-5 bits or 16-32 separable levels. There is also a limit to how accurately neurons can encode a quantity as a change in membrane potential, because of the signal-to-noise ratio of the signal: estimates of 4-7 bits (or 16-128 separable levels) are sometimes given (Laughlin, 1989" Attwell and Tessier-Lavigne, 1989). Spiking axons can only encode quantities as inter-spike intervals, and this sets a limit to the
72
C. Miall
accuracy of reading the code. To read an interval with high precision requires some time to assess the mean firing rate, and also requires little or no temporal jitter at the synaptic junctions. Lastly, most neurons adapt to a steady input, and their firing rate falls with time. Together these factors mean that quantity cannot be represented with any great precision, either as the activation level of a single cell nor as a synaptic connection between cells. These arguments mean that single units may not be able to accurately encode long time intervals: there is an intrinsic trade-off between accuracy and firing duration or oscillation period, such that we can expect neurons to be most reliable at short intervals. Cells with vet3' long oscillation periods would have slowly changing membrane potentials, and hence one would expect some scatter of the moment at which spiking initiates in each cycle. These arguments also imply some problems with networks of neurons, if high levels of accuracy are required for their combined processing. For example, if a recurrent network of neurons is to have stable oscillatory. behaviour, the synaptic weights may need to be set with unrealistically high precision. Thus we should ask what mechanisms are feasible for single neurons, circuits or networks to encode time. We can therefore ask which mechanisms seem more likely (or perhaps less unlikely), and also compare these schemes with the techniques used in artificial computer models of timing. Neural network models have been criticised because of their 'biological implausibility'. But among the many different models there are useful insights into how large numbers of neurons can co-operatively solve tasks, tasks that pose ahnost impossible problems for small assemblies of neurons. Finally, it would be useful to develop guidelines that would allow us to experimentally distinguish between these various ideas and discover which mechanisms are used in real biological systems. The last aim is probably beyond the scope of this chapter.
2. Potential time codes
There have been many different mechanisms proposed for neural coding of time. In the Table I have loosely grouped the theories or mechanisms into those based on single cells (or at least serial cotmections between cells): those dealing with small circuits (where the individual members of the circuit may have rather specific roles): and those dealing with populations or
Table I Possible neztral mechanisms for encoding time. A selection of neural properties are given, which may endow neurons \vith the ability t o encode teniporal information. Theorertical proposals are then sorted in terms of both the numbers of neurons that niiglit be in\.ol\.ed, and how spcciific is the role of individual neurons in the coding Mechanism Intrinsic membrane d>namics' Temporal integration Adaptation Conduction delay S?.naptic delay Recurrent connections Activation levels
Single cells pacemaker cells counter tinier delay line
counter
Small neural circuits oscillators; pattern generators pattern generators pattern generators coincidence detector tapped delay line pattern gelierators re-entrant loop models
Distributed population
reentrant loop models
neural integrator; population activity model correlation models Pattern codes coincidence dctcctor Spatial codes labelled lines shifting activity wave (I Dynamic nic~iibranepropcnics niay also underlie temporal integration and adaptation. but are taken here to mean more complcs time varying beha\.iour such as spontancous oscillations or plateau potentials.
74
C. Miall
networks of cells, where it is assumed that the individual neurons are less specific to the particular process. The first mechanism is that based on single pacemaker neurons, which are certainly 'biologically feasible', as they arc found in many different neural circuits. There is howevcr a general problem in using intrinsic pacemakers to time arbitrar), time intervals. One either needs a mechanism to adjust the period of the pacemaker, so that it can take on any periodicity (Torras, 1985, 1986). or one needs a large pool of different pacemakers from which one can select thc appropriate individual. The alternative seems to be to have a fast pacemaker acting as an 'internal clock', and then solve the problem of counting the ticks of the clock. This brings me to the second mechanism, which is to use temporal intcgration (for example in the postsynaptic membrane) to sum up a temporal sequence of events. Here the post-synaptic potential represents the integrated quantity, until the spiking threshold is reached and the cell is activatcd. This 'integrate and fire' model is in fact widely used for a variety of models, and is again quite feasible. It also faces difficulties beco.use to set up a versatile timer, one must either control the integration time of the membrane or the period of the internal clock, so that the post-synaptic cell reaches its firing threshold at the appropriate moment. A second possibility, based on the same principle, is that the neuron integrates supra-threshold activity: in other words the firing rate of the neuron reflects the integrated quantity. Again, this may be problematic for single neurons, as it poses high accuracy demands on the maintained firing rate, and on its interpretation by some other target neuron. I will discuss a population version of this scheme later, which largely avoids these difficulties. A third possibility is that a neuron fires a prolonged burst of action potentials, and the duration of the burst is self-terminated due to an adaptation mechanism. This seems a feasible mechanism, and makes no demands on the accurate decoding of the action potential bursts. The neuron might be 'switched on' with a supra-threshold activation, and a behaviour could easily be triggered at the end of the burst through disinhibition. To be flexible, one would need to adjust the adaptation rate of the neuron, or allow selection among many neurons. Serial chains of neurons acting as delay lines might be possible (Licklidcr. 1951) with each synapse adding a small delay to the signal, but few neurons in the vertebrate brain have synapses powcrfi~l enough to ensure signal transmission down a single neural chain. Another idea was to make use of the conduction delays in axons, to spatially map time as distance along an
Models of Neural Timing
75
axon (Jeffress, 1948), and this scheme is known to be used within the mcdial superior ovilary nucleus for detecting sub-millisecond interaural time dclays (Carr and Konishi, 1988). Re-entrant (recurrent) pathways have also be proposed to form 'delay lines' (e.g. Longuet-Higgins, 1968), but because of the limits on neural accuracy mentioned before these schemes seem unlikely to me. The demands on maintai.aing an accurate representation of the signal as it repeatedly passes through many synapses seem too great. 2.1.
OSCILI~AT()RSAS TIMERS
An alternative possibility is to make use of the interactions between groups of oscillator3, pacemaker neurons. The idea of using oscillators to store an arbitrary temporal sequence was fundamental to Longuet-Higgins' holophone (1968). Church and Broadbctat (1990a.b) used a related scllcme to model animals' time estimation in delayed-reward experiments. A fourth type of model invokes neurons with a broad range of membrane dynamics (Tank and Hopfield, 1987" Grossbcrg and Schmajuk. 1989), so that time can be mapped onto neurons with appropriate d.xammics. In all thcse schemes, the oscillators have a broad fi'cqucncy range, or even a regular hierarchy of oscillation periods. For example, in Church and Broadbcnt's model the pacemaker periods increased in powers of two (1, 2, 4 .... 256 seconds). In Longuet-Higgins' holophonc a broad bank of neural filters was required, one tuned to each frequency. Thus they require that the population of pacemakers or the membrane time constants include at least one neuron with a duration as long or lo,~gcr than the interval to be stored. This is not impossible in neural terms, and hierarchies of pacemakers have bccn suggested to underlie other long duration processes, for example, diurnal rhyttuns. A somewhat different scheme relied on the phenomenon of'beating' amongst large networks of oscillating neurons (Miali 1989a, 1992). This model was shown to be robust, and made no unreasonable physiological demands, and ! will briefly review it here. I originally proposed a scheme which relied on a large population of pacemakers with only a narrow distribution of oscillation periods and which used a simple Hebbian learning rule to store long time intervals (Miall, 1989a). In this model a unique group of pacemakers could be selected that had the appropriate beat frequency to store any particular time interval. Consider a group of oscillators (pacemaker neurons), each with a slightly different frequency of oscillation, and each spiking for a brief part of each
76
C. Miall
cycle (Figure 1). The beat frequency of any pair of these oscillators is then the frequency at which they spike simultaneously. Thus their beat frequency is much lower than their intrinsic oscillation frequency; it is given by the difference between the frequencies of the two cells. For a population of oscillators the beat frequency is given by the lowest common multiple of the periods of their oscillations. A group of a few hundred pacemaker cells, even with quite similar oscillation frequencies, could encode a wide range of time intervals and could recall the intera,al at a later time.
A
B Oscillators
"r'~r~
,,
IIIIIIIIII
-
=
IIIIIIII|IIIIIIIIIIIIIII
II
.
z
Output
= o
.
IIIIIIIIIIIIIIIIIIIIIIIIIIIII
I
I
,. I I I I I I I I I I I I I I I I I I I I I I I . I I I I I I I I I I I IUl I I I I I I I I I I I I I I I I I I III I II I II I
I I
unit te
Interval to be ~ m d
t~
|=
9
Figure !- Storing time with oscillating neurons. A" The network" a heterogeneou population of oscillators mutually excite an output neuron, which sums incomin activity and fires when a threshold is reached. B: A schematic diagram of aclivil in 6 oscillalors, indicated by short vertical bars. The interval t0-tl can be encode by selection of those few oscillators active both at I{) and at t l (oscillators 1.2.6) their beat frequency matches Ihc lesl inlen'al. Modified from Miall (1989a). To demonstrate the basic model, the mechanism indicated in Figure 1 was simulated on a computer (Miail. 1989a}. A population of up to 500 pacemaking units was defined, with oscillation frequencies between 5-15 Hz chosen using a random number generator to give an average frequency of 10 Hz (Llinas, 1988) and a standard deviation about the mean of 1.6 Hz. The output of each pacemaker was zero for 90% of each cycle, and 1 for the remaining 10%. All pacemaking units synapsed onto a single output unit via Hebbian synapses taking values of 1 or zero. The total input from the pacemakers was summed, and displayed as time histograms (Figure 2). Now. to store any specific interval, h)-tl, first all pacemaker units were synchronised at tO (as in Figure I B). Those active at time tl were noted, and the strength of their synapses onto the output unit set to value 1. To test the
Models of Neural Tindng
77
specificity with which the selected units stored each interval, all the pacemakers were again re-synchronised, and the activity of the output unit monitored (Figure 2). The model has been tested with a population of between 10 and 500 pacemakers, with intervals ranging from 200 milliseconds to 10 seconds, and with the percentage of each cycle that a oscillator was considered to be active ranging from 50-99.9%. The shortest interval that could be stored was set by the shortest period of any pacemaker in the population. The longest interval that could be stored was difficult to specif)'. It depends on the number of pacemakers, their activity thresholds, and the precise distribution of oscillation frequencies within the group. In simulations of 250 or 500 pacemakers, the upper time limit seemed to be at least 20 seconds.
a
'
b
o i1
n1:70
'
"-t
~
'
t
t
'
'
t
t
t'
t
n4:55
d
n2"80
'
n3.,4
'
ttt'
t
Figure 2: Records of the output unit's activation during recall of multiple random intervals. The horizontal line indicates the threshold for the output neuron to fire, set equal to the mean number of oscillator selected for each inlerval. A: Recall of the interval 3.5s: the threshold is crossed only after the target interval (arrow) BD: Recall of two. three or four intervals (arrows) by the same network. As the number of stored intervals increases, the activation level approaches the threshold. The nelwork's capacity was about 4 intcra,als. From Miall (1989a).
78
C. MiaU
The basic mechanism can also be used to store more than one time interval simply be selecting additional pacemakers that encode for each additional time (Figure 2B). In the 1989 paper, I demonstrated how the simple Hebbian network could learn in a single trial but had relatively low storage capacity, so that a network of 500 or more oscillators would be required to accurately store a pattern of say 5 or 10 arbitrary time intervals. That model used synapses taking only integer values of 0 and 1, or 0, +1 and -1. In a biological framework it might be more realistic to assume that synapses could take values between 0 and I. and would gradually move between these two limits on repeated presentations of a particular interval. In other words, a learning rule might only gradually increase the strength of selected synapses relative to the others. Thus in 1992 I extended my original scheme to a description of a temporal network using the perceptron learning nde to set its synaptic weights. Using a rule which allowed the synaptic weights to take on any integer value (which is filnctionally equivalent to allowing different levels between 0 and !. a network of just 100 units could store at least 20 time intervals (Miall, 1992). With the optimum value of the activity threshold of the oscillators, about 70 cycles through the input set where required to learn such a sequence. 2.1.1. Discussion o f the oscillation model. It is possible that the nervous system could use the basic mechanism proposed here in the perception of time, or in the production of long sequences of neuronal activity. The oscillatory components could be realiscd as individual pacemaker neurons. as entrained groups of pacemakers, or as reverbator3' circuits containing several neurons. In the latter cases, any one cell in the circuit could make a synaptic connection to the output neuron. Each neuron or reverbator 3, circuit of neurons could have a relatively short oscillation period, as is frequently found in nervous tissue, while the whole ensemble could co-operate in behaviours with time courses hundreds or thousands of times longer. Evidence for such a scheme would be a group of rhy'thmically active cells forming synapses upon output units that sum activity over some selected subgroup of the ensemble. A strong requirement would be that the oscillators could be started s.x~chronously, or phase-adjusted at the start of each input sequence. Pacemaking units are found in a number of brain sites, and the figure of 10 Hz chosen for these simulations is biologically reasonable (Llinas, 1988).
79
Models of Neural Timing
However, three wcaknesses of the scheme were apparent as a model of biological time-keeping. The most sevcre is that tile nctwork is intolerant of even ven' minor changes in the period of the whole population. If each oscillator's period was shifted by just 0.5 %, then recall was poor; if all the periods changed in the same direction then tolerance was improved but still low. Only if the units randomly fluctuated about a ver)' stable mean period would the network be able to toleratc noise. In other words, if cach unit showed some variation in activity threshold, so that for any one cycle there was uncertainly about its activity but it remained oscillating at the same period, then the system would be quite robust. The networks can be madc Icss sensitive by a protocol of training them in the face of small random noise, as then the network continues to learn until it reachcs a safety margin greater than the noise level. Then minor fluctuations in the oscillators have less effect during recall. A second failure to mimic biology is the relationship between interval duration and accuracy: The networks as modelled were either accurate or they failed. There was no distribution of responses about the desired time that might lead to the t),9ical Wcbcr's Law relationship between errors and duration. Again, by training with noisy oscillators, this would no longer be the case. In fact, Church and Broadbent (1990a,b) do just this to ensure that their models mimic the distribution of time-estimates shown in dclavcdreward experiments. Their model is quite closely rclatcd to mine, in that it selects active oscillators from a population to encode an interval. The difference is that their schcme rcsults in a binary code. as the oscillator periods increase in powers of two. requiring at least one oscillator with a period as long as the test interval. The remaining difficulty with the scheme presented here is that the group of selected units encoding a particular time interval or sequence nceds to bc synchronously reset to allow recall of the stored interval. This is possible, but would require some powerfill rcsct signal to reach the entire group of oscillators. One feature of the proposcd model, which I originally thought of as a flaw, is that individual neurons within the group of oscillators, and indeed the output unit itself, would show little of no evidence of being related to the task in hand, In other words, it would be very difficult to detect such a scheme by single unit recording tcchniqucs (Figure 2. and also Miall. 1980). It is therefore interesting to scc recent data on correlated high frequency oscillators in between pairs of cortical neurons (Gray et al., 1989: Engcl ct al., 1992). A rccent rcport (Vaadia et al., 1095) has demonstrated that o
80
C. MiaU
frontal neurons may show no overall ch,'mge in firing rate, while the correlation between the pairs activity changes significantly and can differ between go and no-go trials in a delayed response task. This is just the sort of difficult detection problem that I had originally worried about, and its solution implies that one might indeed be able to detect the mechanism that I had proposed. 2.2. POPIH~ATIONIN'~GRATI()N FOR TIMING Having described the oscillator3.' models. 1 want to turn now to a rather different aspect of timing. There is considerable psychophysical and behavioural evidence for some fonn of counter, used in timing tasks, so that any given interval can be estimated by accumulating or integrating the ticks of a internal clock, and compared with a reference value (Gibbon and Allan. 1984). Neurons in frontal areas of the cortex show a gradual increase or decrease in their averaged activity during delayed response tasks, as expected of cells functioning as an integrator or counter (Niki and Watenabe, 1979). As mentioned previously, I would not expect any single cell to be able to integrate accurately over long periods, and in general the activation of these frontal neurons can be quite erratic even within any one trial. A population of neurons may be much more robust, however. Thus a possible integration mechanism is to simply consider the total activity within a population of noisy neurons, none of which can individually accumulate or integrate (Miall. 1993). Consider a group of neurons, all receiving the inputs from an internal clock which is periodically emitting a pulse of activity. Imagine that each neuron has only a low probability of being activated by any one clock tick. but once switched on, it remains on. Further, we can also add a small probability that each active unit will switch itself off at any moment. Hence, the trial by trial activation of individual neurons will be erratic, and show no clear increase or decrease in activity with time. However, the total activity within this population monitored, for example by assuming that each of these neurons projects an excitator)., input to another neuron or network of neurons, could represent the accumulated measure of time. Obviously this measure of total activity should rise or fall monotonically with the number of clock ticks, i consider here only the case of monotonic increase. On the first tick, a small number of units will be activated, and some will be inactivated. At the second tick. more of the inactive units will be activated, .
.
.,
Models of Neural Timing
81
and a few active ones turn off, and so on. As the number of time steps increases, the number of active units asymptotes towards a fixed limit. Hence the network will have an upper boundary' on the number of time steps it can encode. Figure 3A shows the averaged results from 10 runs of a simple computer simulation. The net had either 250, 500 or 1000 units: the probability of each unit switching on was 0.05 per clock tick, and the probability of each one turning off was 0.0001 per iteration of the model. Since the clock ticks arrived every. !00 iterations, po.[l"= 0.25pon. The total activity did indeed increase with time, but the variance also rises over the first 6-10 time steps. So the fidelity of the system depends on the incremental increase in activity with each clock tick (Figure 3B), but this must be with respect to the variance at each step. Figure 3C plots the ratio of the change in mean to the standard deviation of the mean. A cut-off threshold of 1.96 is shox~.~ which represents the 95% confidence limit for accurate detection of the increase in activity per time step. For the largest net, this limit is reached after about 13 clock ticks, whereas the smaller nets fail earlier, after 9 or 5 clock ticks. 2.2.1. Discltssion o./'the integrator model. It is clear that a population of 'low quality' neurons can foma the accumulator required for an internal timer or to count randomly timed events. In this simple simulation the individual units were either on or off. all synaptic weights were fixed (and uniform). and all heurons were random both in their activation and inactivation. One could easy add more power to the network, for example allowing low resolution changes in activation level' even if only two or three separable activity levels were allowed (for example, low medium or high activity), this would have the effect of increasing the population size 2-3 fold. Note that it is only the source of the input that determines whether the network output reflects time or quantity, and thus the same network could be used to time or to count (Figure 3 C,F). It also means that timing is feasible with irregular clock pulses, as allowed for b.v scalar timing theory (Church and Gibbon. 1982) or behavioural timing theory (Killccn and Fcttcnnan. 1988). The total activity in the net correlates with the passing time or number of events. However, no obvious clue to the timing ability of the network would be gained by recording from any one neuron on any one trial. I should try to describe what fonu of input (clock) or output (threshold device) this mechanism would use. As a timing device, a Synchronous pulse is required at all neurons in the net: for counting the net should rcccive
82
C. Miall
discrete pulses from the sensor3.' system detecting the events to be counted. But only a fraction of the population of neurons should be activated by each input pulse. One possibility would be to require coincident input from the timer and from another independent input. For example, each neuron might receive an independent series of action potentials uncorrelated with the common timer input, then only those neurons with coincident arrival of both inputs would become activated. Alternatively, one might imagine that the whole population of neurons has asynchronous sub-threshold oscillator3., behaviour. Then the additional cxcitator)." timing input arriving in-phase with this oscillation could switch them to their active state: again, if the population oscillations were uncorrelated, the selected group would be just those few at the peak phase of their oscillation period. Hence this has links with the mechanism suggested in Section 2.1. for the selection of a subset of oscillators amongst a larger population. The output side requires a threshold mechanism. The accumulating activity within the network should only trigger the delayed response when the correct number of timing pulses have been integrated: this threshold device should receive an equally weighted input from every member of the integrator population (to avoid biasing the statistics of the count). One possibility, which would avoid difficulties raised earl), regarding the limits to accurate threshold setting in single neurons (Section 1.1 ), would be for the accumulating activity to switch on - or o f f another group of neurons. If these were mutually excitatory,, then the threshold for switching within the whole group could be robust, despite individual neurons behaving erratically. The principle here, as in the main integrator population, is of averaging across many neurons. A mutually excitatory group would therefore switch rather abruptly from low to high activity, as the input drive from the accumulating counter reached the threshold for explosive positive-feedback driven excitation. In principle, this threshold feature could even co-exist within the integrative population. One could have a number of excitatory interconnections which, when the total activity reached some critical point, would cause rapid mutual firing of the whole group. To vary. the threshold value, one would add slight inhibition or excitation from another source. Hence the neural mechanics of this whole scheme are quite easy to imagine. The net displays initially increasing variance with increasing events or time, and therefore approximates the ps.vchophysical data on behavioural
Models of Neural Timing
83
.p.o.i
~o
I
9p = o . o s
i::! / /
p9 ,, 0.02
300
I./
tOO
1
,
I
~
0
Z
4
O
O
I0
12
14
B
HI
oil
20
.
.
.
0
N
.
2
.
.
J,,
4
I 1
q•
,ns 1000 on: SO0
II g
.......
gO
r'-"
t~l t2
e
14
~l
.p.0.1
U"-
p9 . O.OS
9n - 2 S 0 40
|. 2O
r
..... :[
Jl.. . . . 4
| I!
.
4
,_ II
_ 10
12
14
16
8
10
12
14
16
F
2
I
0
2
4
II
II
10
12
TImeSlep
14
|11
III
10
~[
0
2
4
t
Ev~
Figure 3' Timing and counling. A-C" Timing. A: Mean number (4- SD) of active units against number of lime steps. The input is from a regularly ticking clock. B: The incremenl in mean activity per clock tick. C: The ratio of incremental increase to standard deviation. The horizontal line is a! 1.96 (95% confidence limit). D-F: Counting. D" Mean activi.ly against event number. A 500 unit network received, irregular inputs arriving with a probability of 0.005 per iteration. The probability thal each unit would be aclivalcd by an event was varied from 0.1. 0.02 to ().1)5. E: Incremental activity: F: Incremental increase over standard deviation.
84
C. MiaU
timing; Killeen (1992) reviews evidence that the counter may indeed be inaccurate in accumulating clock ticks. The asymptotic curves in Figure 3A and 3D imply that there would be systematic errors of timing and counting with underestimation of long intervals or high counts and overestimation of short intervals and low counts, features which are typical of time psychophysics (see again Killeen, 1992; Church and Gibbon, 1982). Further, the ratio of differential increase in activity to the variance (Figures 3C and 3F) implies that there is an upper limit to the number of accumulated clock intervals or the number of events that can be distinguished. This again approximates the human behavioural evidence (Kristofferson, 1980), while Gallistel (1990) reviews evidence that animals may also count their og~ behavioural responses, although with considerable variance. 2.3. OTHER CODINGSCI-~MES One can of course invent an infinite variety of temporal coding schemes, and if the two schemes described here have any merit, I suggest that it is that they are both fundamentally simple. They make limited demands on the properties of the individual neurons, and require only a simple algorithm to select or train the appropriate member or synapses within the population. Before finishing, therefore. I will briefly speculate about one or two remaining possibilities (Table 1). I have discussed the possibility of delay lines; while a simple delay line does not seem to be feasible, except for very short time scales, the principle of the spatial mapping of time does have its attractions. Would it be possible to spatially map time in a cortical sheet? There seem to be two possibilities. First, time might be mapped locally, such that each neuron might represent a particular time or interval. Individual neurons might have more or less sharply defined 'temporal receptive fields', and there could be considerable overlap between neighbouring neurons. If so, the passing of time would reflect the movement of neural activity across this cortical sheet, until some critical location was reached, and the time taken to traverse the map would be the target interval. This is in essence the same idea as a delay line, but extrapolated to two dimensions. It has an interesting parallel in the recent work of Abbott and Blum (1995). They have explored the spatial mapping of dynamic stimuli (for example the passage of a visual stimulus across a retinotopically organised visual map), and have shown through computer
Models of Neural Timing
85
simulations that temporal as3~nmetry in the rules governing Hebbian learning between neighbouring neurons can lead to a temporal shift in the retinotopic map. This is analogous to a short term prediction of the motion of the stimulus (cf. Montague and Sejnowski, 1994) The second alternative is that the wave of activity over the cortex has no fixed path, and may even pass the same areas of cortex more than once, but the overall spatio-temporal pattern of neural activity has a particular structure. For instance, it is thought that cortical activity waves may behave in a very complex (Wright et al., 1992), even chaotic manner, so that after some initial stimulus a complex reverberation or activity wave may pass backwards and forwards over a set if interconnected neurons (Milton, Chu and Cowan, 1993). It might be possible to derive timing schemes based on this model, if the activity wave were reproducible. If one could select just those neurons that were active at each particular moment after the initial event, one would then have a distributed, overlapping set of neurons coding for each time interval.
3. Timing in artificial neural networks In this final section, I will very briefly review coding techniques employed in artificial neural networks. It is first worth mentioning how digital computer models can deal with the time dimension at all. and then look at those neural network models that have sought to represent time, limiting my choice to a few networks that explicitly aim to encode temporal events. There are a large class of 'neural mimics', which are the most physiologically accurate models of individual neurons, modelling membrane properties with cable equations or with compartmental models, and thus accurately modelling the dynamics of the membrane potential of the neurons (review: Koch and Segev, 1989). These are computationally demanding, and there have been very few models of large networks of neurons. Few have been tested on the sorts of learning tasks for which discrete artificial neural networks are so popular (Buhmann and Schulten. 1986: Pearson, Finkcl and Edelman, 1987' see also Koch and Segev, 1989). At a simpler level, neurons can be simulated with sets of continuous differential equations (Pineda. 1987; Pearlmutter. 1089). These models still have dynamic, time-dependent properties and can represent the behaviour of neurons as if they were non-spiking devices. They have continuously -
86
(7. Miall
variable activation values, with d.~amics analogous to the membrane, synaptic and adaptation conductances (Amari, 1978), but have no spatial dimensions. There are also discontinuous versions which add an instantaneous impulse representing an action potential. The action potential is triggered by, and resets the membrane potential, while synapses transmit the action potential (Perkel, 1964: Miall, 1989b; Amit, Evans and Abclcs, 1990; Judd and Aihara. 1993). Finally, at the most simple level, there are 'digital' models in which a computer simulation is iterativcly calculated, and at each calculation the activity or synaptic weights of the elements are changed (Rumelhart and McClelland, 1986). These models encode time as a discrete sequence of states governed by the iteration rate of the model. As an approximation to neurobiology, activity within the network can be treated as representing average neural firing rates, measured over large enough units of time that the individual action potentials and membrane time constants can be ignored (Hopfield, 1984).These models cope rather poorly with the temporal dimension, as the encoding is largely forced into the iterated pattern of inputs or outputs. The most simple way to extend a network from a static to a temporal representation is therefore to map sequential samples from the inputs into a 'spatial pattern' and thus train the network to recognise certain regular features in this sequence of changing spatial patterns. The vast majority of published examples of artificial networks are simulated with this form of iterative process. These networks can therefore be powerful tools. when combined with a pre-processor to converts the incoming temporal signal into a discrete set of inputs which can then be treated as a spatial pattern. A common form of pre-processor, which may also be used on connections within a network, is the tapped delay line. Imagine a set of regular samples from a time series, perhaps speech, each passed into a delay line which delays the signal by a different amount. If the early samples are delayed the most, and the later samples delayed the least, they can exit the delay line simultaneously, and provide a spatially distributed snapshot of the time series (Lang, Waibel and Hinton. 1990). Typically a tapped delay line. from which the delayed signal can be read off at many different points, or multiple delays lines of various lengths are employed. More complex varieties of delay line are worth mentioning - dispersive, Gaussian and adaptive delays. In dispersive and Gaussian delay lines, the output is a timeweighted version of the input. So. if the input is a discrete pulse, the output
Models of Neural Timing
87
may be a temporally blurred version whose output rises and falls around some median time delay (review: de Vries and Principe, 1992). Adaptive delays have also been used, in which the delay time or the parameters of the dispersion are modified, again avoiding prior knowledge of the optimal form (review" Mozer, 1993). And lastly, nets have been designed with 'slow' and 'fast' s3alapses, the slow ones delaying signals between neurons (Sompolinksy and Kantor, 1986: Kleinfeld and Sompolinksy, 1988). The overall architecture of the neural network is also of great importance. and one can distinguish between 'feedforward' networks, with infomaation flowing in only, one direction, and 'recurrent' networks in which connections within the network feed activity back from outputs to inputs. This gives recurrent networks dy~mmic scope, as activity within the net is no longer governed simply by the inputs (whether or not delayed) but can be sclfsustaining. Imagine a unit with a self-excitatory connection - if the strength of this connection is high enough then activity in the unit in response to a pulse input will feedback and re-excite itself, and this activity will be maintained long after the input pulse has ended. Thus a recurrent network can have behaviour similar to that of a dispersive delay line, but can also have more complex dynamic bchaviours. The timing of the network is again often implicit in the iteration rate of the model (Jordan, 1986). Thcse recurrent networks have been used widely for signal processing, motor control, sequence production and the like: veu, many applications can be found in Neural Infommtion Processing Systems (Touretzky et al. It9891995). A few final properties of artificial neural networks should be mentioned. There are nets that incorporate dynamic activation thresholds (Horn and Usher, 1989: Heskes and Giclcn. 1992), so that their responses vary with constant inputs. The parameters controlling the threshold can endow the network with very rich bchaviour. A trivial example might be an oscillatory threshold, so that the unit acts much like a pacemaker: another could be a bistable membrane, switching from one state to another. Alternatively. the synapses may adapt over time (Dong and Hopfield. 1992) or with activity from 'modulators' (Dahaene, Changeux and Nadal, 1987). The level of activity within one set of synapses thus sets the moment-to-moment wcight of another set of connections (Dahaene et al., 1987: Schmidhuber, 1992). In summar3,, artificial neural networks (those other than neural mimics of the biophysical properties of neurons) encode time in three basic ways. They use 'hard-wired' devices such as dclay lines or spatio-temporal pro-
88
C. Miall
processors; they endow the networks with a variety of explicit temporal features; or they use the dynamics of recurrent connections within the network. The first and second of these three have biological counterparts, but also have some problems when mapped onto real neurons. The third, based on dynamic recurrent behaviour, may also have biological correspondence, but may suffer from its demand for high resolution in activation levels or in the precision of synaptic weights.
4. Conclusions The aim of this chapter has been to address what I perceive to be the difficulties of mapping time onto biological neural networks. I have reviewed a number of possible encoding schemes, but described just two of these in any detail. Both are very. simple models and how closely they approach reality is far from clear. However, they do highlight some interesting points. The first is that networks of neurons can operate on time scales very different from the time scale of their constituent parts. Hence the mismatch between the temporal horizon of individual neurons and the timing of ever).,day behaviour may be overcome by considering population based models of timing. Second. neither scheme would display an obvious 'timing function' if accidentally overheard with a recording electrode. So the fact that we are still ignorant of how the brain times events may be revealing in itself. It implies that a distributed code is used. that can only be decoded by observing many different neurons simultaneously. Third, both schemes operate with what might be considered 'low quality' neurons, and this emphasises the power of neural networks. Many very simple units can contribute to a powerful and robust network, the whole definitely exceeds the sum of the parts.
ACKNOWLEDGEMENTS: ! thank tile Weilcome Trust for their support.
5. References Abbott LF and Blum KI. Functional significance of long-tern1 potentiation for sequence learning and prediction. Preprint. 1995.
Models of Neural Timing
89
Amari S. Mathematical theory, of neural networks. Tokyo: Sangyo-tosyo. 1978. Amit DJ, Evans MR and Abeles M. Attractor neural networks with biological probe records. Network 1990: 1' 381-405. Attwell D and Tessier-Lavigne M. Designing synaptic connections in the retina. In: RM Durbin, RC Miall and GJ Mitchison, editors, The computing neuron. Wokingham, UK' Addison-Wesley. 1989' 337-354. Bailek W and Ricke F. Reliability and information transmission in spiking neurons. Trends Neurosci. 1992 15,428- 434. Buhmann J and Schulten K. Associative recognition and storage in model networks of physiological neurons. Biol. Cybern. 1986: 54, 319-335. Carr CE and Konishi M. Axonal delay lines for time measurement in the owl's brainstem Proc Nat AcadScii USA. 1988" 85' 8311-8316. Church RM and Broadbent HA. A connectionist model of timing. In" ML Commons, S Grossberg and JER Staddon, editors, Quantitative models of behaviour' Neural networks and conditioning. Hillsdale, NJ" Lawrence Erlbaum Associates. 1990a: 225-40. Church RM and Broadbent HA. Alternative representations of time, number and rate. Cognition. 1990b" 37" 55-81. Church RM and Gibbon J. Tc,nporal gcneralisation. J Exp Psychol' Animal Behav. Proc. 1982' 8" 165-86. Dahaene D, Changeux JP and Nadal JP. Neural networks that learn temporal sequences by selection. Proc. Nat. Acad. Sci. USA. 1987: 84: 2727-31. de Vries B and Principe JC. The gamma model - a new neural model for temporal processing. Neural Networks. 1992: 5: 565-76. Dong DW and Hopfield JJ. Dynamic properties of ,leural networks with adapting s3xmpses Network. 19t)2' 3" 267-83.
90
C. Miall
Engei AK, Koenig P, Kreiter AK, Schillen TB and Singer W. Temporal coding in the visual cortex: New vistas on integration in the nervous system Trends. Neurosci. 1992; 15:218-26. Gallistel CR. The organisation of learning. Cambridge, MA: M.I.T. Press. 1990. Gibbon J and Allan LG, editors . Time and Time Perception. Ann. N.Y. Acad. Sci. 1984: vol. 423. Gray CM, Konig P, Engel AK and Singer W. Oscillatory responses in cat visual cortex exhibit inter- columnar synchronization which reflects global stimulus properties. Nature. 1989: 338: 334-37. Grossberg S and Schmajuk NA. Neural dynamics of adaptive timing and temporal discrimination during associative learning Neural. Networks. 1989; 2: 79-102. Harris-Warwick RM, Marder E, Seiverston AI and Moulins M, editors. Dynamic biological networks: The stomatogastric nervous system. Cambridge, MA: M.I.T. Press. 1992. Heskes TM and Gielen S. Retrieval of pattern sequences at variable speeds in a neural network with delays. Neural Networks 1992: 5:145-52 o
Holden AV. Models of stochastic activity of neurons. Berlin, Springer. 1976. Hopfield JJ. Neurons with graded responses have collective computational properties like those of two-state neurons. Proc. Nat. Acad. Sci. USA. 1984: 81: 3088-92. Horn D and Usher M. Neural networks with dynamical thresholds. Physical. Rev. A. 1989: 40: 1036-44. Jeffress LA. A place theory of sound localization J. Comp. Physiol. Psychol. 1948:4 I" 35-9.
Models of Neural Timing
91
Jordan MI. Attractor dynamics and parallelism in a connectionist sequential machine. Proc 8th Ann Conf Cogn Sci. Hiilsdale, NJ' Lawrence Erlbaum Associates. 1986; 531-46. Judd KT and Aihara K. Pulse propagation networks" A neural network model that used temporal coding by action potentials. Neural Networks. 1993" 6, 203-15. Killeen PR and Fettemmn JG. A behavioural theory of timing. Psychol. Rev. 1988: 95,274-95. Killeen PR. Counting the minutes. In: F Macar. V Pouthas and WJ Friedman, editors, Time, action and cognition: Towards bridging the gap. NATO ASI Series D Vol. 66, Dordrecht' Kluwer. 1992" 203-14. Kleinfeld D and Sompolinsky H. Associative neural network model for the generation of temporal patterns. Biophysical J. 1988" 54, 1039-51. Koch C and Segev I, editors. Methods in neuronal modelling: From s3~apses to networks. Cambridge, MA: M.I.T. Press. 1989. Kristofferson AB. A quantal step fimction in duration discrimination. Percept. Psychophys. 1980; 27, 300-6. Lang K, Waibei AH and Hinton GE. A time-delay neural network architecture for isolated word recognition. Neural Networks. 1990' 3, 23-44. Laughlin S. The reliability of single neurons and circuit design" A case study. In: RM Durbin, RC Miall and GJ Mitchison, editors. The computing neuron. Wokingham, UK: Addison-Wolsey: 1989: 322-36. Licklider JCR. A duplex theor)., of pitch perception. Experientia. 1951" 7: 128-34. Llinas RR. The intrinsic electrophysiological properties of manunalian neurons: Insights into central nervous system fimction Science. 1988" 242" 1654-64.
92
C. MiaU
Longuet-Higgins HC. Holographic model of temporal recall. Nature. 1968; 217, 104. Macar F, Pouthas V and Friedman WJ, editors. Time, action and cognition: Towards bridging the gap. NATO ASl Series D, Vol. 66, Dordrecht: Kluwer Academic. 1992 Miall RC. Oscillators, predictions and time. In: F Macar, V Pouthas and WJ Friedman, editors. Time, action and cognition" Towards bridging the gap. NATO ASI Series D. Vol. 66, Dordrecht: Kluwer Academic. 1992: 215-27. Miall RC. The storage of time intervals using oscillating neurons Neural Comp 1989a: 1" 359-71. Miall RC. The diversity of neuronal properties. In: RM Durbin, RC Miali and GJ Mitchison, editors, The computing neuron. Wokingham, UK" Addison-Weisey. 1989b: 1!-34. Miall RC. Neural networks and the representation of time. Psychol. Belg. 1993: 33" 255-69. Milton JG, Chu PH and Cowan JD. Spiral waves in integrate-and-fire neural networks. In: SJ Hanson. JD Cowan and CL Giles, editors, Advances in neural infommtion processing systems, San Mateo, CA" MorganKaufmann. 1993" 5" 1001-06. Montague PR and Sejnowski TJ. The predictive brain: temporal coincidence and temporal order in synpatic learning mechanisms Learn. Memory. 1994: 1" 1-33. Mozer M. Neural net architectures for temporal sequence processing. In" A Weigend and N Gershenfeld, editors, Predicting the future and understanding the past Redwood City, CA" Addison-Welsey. 1993. Niki H and Watenabe M. Prefrontal and cingulate unit activity during timing behavior in the monkey. Brain.Res. 1979; 171, 213-24.
Models of Neural Timing
93
Pearlmutter BA. Learning state space trajectories in recurrent neural networks. Neural. Comp. 1989' 1" 263-69. Pearson JC, Finkel LH and Edelman GM. Plasticity in the organisation of adult cerebral cortical maps: A computer simulation based on neuronal group selection. J. Neurosci. 1987" 7" 4209-23. Perkel DM. A digital computer model of nerve cell functioning. Tech. Rep. RM-4132-NIH, Rand Corp., CA. 1t~64. Pineda F. Generalisation of backpropagation to recurrent neural networks. Physical. Rev. Lctt. 1987" 19" 2229-32. Rumelhart DE and McCIclland JL. Parallel distributed processing. Cambridge, MA: M.1.T. Press. 1986. Schmidhuber J. Learning to control fast-weight memories: An alternative to dynamic recurrent networks. Neural. Comp. 1992: 4, 131-39. Selverston AI. Modeling of neural circuits: What have we learned? Ann. Rev. Neurosci. 1993, 16:531-46. Sompolinsky H and Kanter I. Temporal association in asynunetric neural networks. Phvsica. Rev. Lett. 1986: 57' 2861-64. Tank DW and Hopfield JJ Neural computation by concentrating infomlation in time Proc. Nat. Acad. Sci. USA. 1987: 84: 1896-900. Torras CIG. Pacemaker neuron model with plastic firing rate: Entrainment and learning ranges Bioi Cybern. 1985" 52' 79-91. Torras CIG. Neural network model with rlay~hm-assinlilation capacity IEEE Trans Svst Man Cvbemet: 1986: 16 680-93. Touretzky DS, Moody JE, Hanson SJ, Lippma~m RP, Cowan JD, CL Giles, Tesauro G and Alspector J, editors. Advances in neural information processing systems, Vol. I-6. San Mateo, CA: Morgan-Kaufmann: 1989-95.
94
C. MiaU
Vaadia E, Haalman I, Abeles M, Bergman H, Prut Y, Slovin H and Aertsen A. Dynamics of neuronal interactions in monkey cortex in relation to behavioural events. Nature. 1995, 3 73: 515-18. Wright JJ, Sergejew AA and Liley DTJ. Computer simulation of electrocortical activity at millimetric scale. Electroenceph. Clin. Neurophysiol. 1994, 90:365-75
Time, Internal Clocks and Movement M.A. Pastor and J. Artieda (Editors) 9 1996 Elsevier Science B.V. All fights reserved.
95
NEURONAL MECHANISMS OF BIOLOGICAL RHYTHMS HUGO ARECHIGA Divisi6n de Estudios de Posgrado e Investigaci6n, Facultad de Medicina. UNA~., M~xico, D.F. ABSTRACT. Rh)~mficity is an inherent property in many biological s),slems. Regular variations are known to occur in the level of physiological fimctions ,and the nervous system plays a key role in the generation and integration of the various physiological dlythnts. Neuronal ensambles are capable of generating dlytlmfic activit3., in a wide rmlge of frequencies from 1 KHz in fast firing neurons to circalmual dlythms in neurosecretory activity. Various mecha~fisms at the cellular and molec~dar level underlay the generation of fllese rh.xlhms: a) altemanc3' of io~fic currents restdting in rhythnfic fluctuation of membrane potential in endogenously pacemaking neurons, b) interactions between elements of neuronal networks, being rhytlmficity an emergent property of the network, c) rhythmic changes in the bios~a~thesis of neurotransmillers. neurohonnones, secondary messengers and membrane proleins. A given d~)~un may be the end product of file interplay of different mechmfisms. Endogenous rhyttuns are also under entraining itffluences, either of mutual nature betaveen different oscillators. or from external sources.
1. Introduction The various biological functions undergo rh)r changes within a wide range of frequencies, from nearly 1 KHz, as in the discharge of action potentials in some neurons, to yearly periods in seasonal rh3r The various rhythms in an organism maintain constant phase relations among them. This has led to the notion of an integrated system of biochronometry; actually, the precise temporal coupling of the body functions is a requirement for a healthy condition. The medical implications of biochronometry are a well knox~aa and still growing field. Susceptibility to physical or chemical hazards, or the effectiveness of therapeutic manipulations, vary depending on the hour of day,, and seasonal differences in propensity to some diseases are k n o ~ since Hippocratic times. The
96
H. Arechiga
disrupting effects on human performance, resulting from travelling across time zones, or work shifts, are among the many instances in which the proper knowledge of physiological rhythms attains special medical importance (Reinberg and Smolensky, 1993" Halberg et al., 1977). With still debatable exceptions of species dwelling in caves or in deep underwater environments, all animals and plants, and even unicellulars are known to express circadian rh3ahms. The adaptive value of the time-keeping systems is well established: rh.~hms with periods ranging from 20 to 28 hours are known as circadian" in resemblance to the nomenclature of electromagnetic radiations, those rhyduns with periods shorter than 20 hours and longer than 28 hours are designated respectively as ultradian, and infradian.A given physiological fimction may display rhytlunicity in more than one domain; quite often in all three of them, with definite interactions between the various frequencies, thus suggesting phase-coupling mechanisms. Although these rhythms are normally entrained by external events, most notably related to geophysical cycles, they are capable of persisting under constant environmental conditions, thus revealing the existence of physiological time-keeping mechanisms, which are generically kno~aa as
biological clocks. During the last three decades, a vast wealth of evidence has been gathered on the phenomenology of biological rh3eduns. One remarkable feature is the commonality of properties of the rh3.~lmas integrated in the biochronometry system in the various biological species. This has suggested the likelihood of common cellular and molecular mechanisms underlying the generation of rhythms and it is in the nervous system, where both the generation of time signals and the entrainment of rh~luns take place. This chapter will review some basic information on the generation of biological rhytmicity. Since it has been more amply and thoroughly documented for circadian rhythms, this is the frequency range that will be specifically presented, with particular emphasis on the following aspects, a) general properties of rhythms, b) basic functional organization of the circadian system, and c) cellular and molecular mechanisms underlying rh.~lunicity.
2. General features of circadian rhythmicity
As seen in Table I, circadian rh.~hms, and for what is known, also ultradian
Neuronal Mechanisms o f Biological Rhythms
97
and infradian rhythms, share some common properties, most notably the following: TABLE I
OVERT PROPERTIES OF CIRCADIAN RHYTHMS ' 1. Persistence under constant environmental con(fitions
Biological period differs from geophysical period Dampening out while free nmning 2. Entrainment b.v environmental clues
Phase dependency of entrainment Multiple entraining clues are effective Transient period during entrainment 3. Homeostatic maintenance
Temperature independence of period Internal synchronization
2.1. PERSISTENCE CONDITIONS
UNDER
CONSTANT
ENVIRONMENTAL
This issue was the subject of ample discussion decades ago (Aschoff, 1960: Moore-Ede et al., 1984). In as much as "enviromnental constancy" can be secured, circadian rhythms persist in secluded or underground facilities, isolated from daily cycles of energy flux on the earth surface, or at the magnetic poles of the planet, and in space modules. However, under these "free running" conditions, rhythmicity differs from its normal expression in some important aspects" a) The duration of the circadian period departs from 24 hours. The vcr3., denomination of circadian (circa, around) alludes to this deviation from the geophysical period, b) the rh~1hm amplitude gradually decreases, until disappearing after a number of circadian cycles depending on the nature of the rhythm and the conditions under which it is expressed. The existence of spontaneous rhytlunicity was among the first demonstrations of behavioral patterns executed independently from external input. It contributed to change the concept of a reflexely driven central nervous system, and to establish the notion of endogenous programs capable of being expressed in the absence of external conmlands.
98
H. Arechiga
2.2. ENTRAINMENT BY ENVIRONMENTAL CLUES This is indeed, the normal situation. Various external factors, most notably light and darkness, but also sound, temperature and other physical and chemical influences are knms~a to entrain circadian rhythmicity. Some internal factors, such as food. and homeostatic agents also appear to possess entraining influence on rhythms. The range of entrainment differs from one rhythm to another. The response to entraining influences depends on the phase relation between the entraining input and the endogenous cycle. The phase-response curve is a distinctive feature of a circadian rhythm (Pavlidis. 1973). Setting the phase is not the only way in which environment influences rhythmicity. Both period and amplitude are also modulated by environmental influences. For instance, the duration of the circadian period of locomotion, in normally night-active animals, is shorter than 24 hours under continuous darkness, and longer under constant light. For day-active animals, the reverse is true. ("Aschoff's rule", Aschoff, 1960). Some rhythms persist longer, and the amplitude of the circadian cycle is higher under constant light, while others fare better under darkness. Entrainment does not require a prolonged action, in fact, one usual tool in the study of circadian rhythms is the "skeleton periods", in which the entraining influence takes the shape of discrete, short pulses, applied only at the transition times between day and night (Pittendrigh. 1974). In the same vein, a full-fledged circadian rhythm can be elicited by a single environmental input, applied after the overt rhythm has faded away under constant conditions. This suggests that external signals not only modulate, but also trigger the expression of endogenous rhythmicity. A general feature of entrainment is the existence of transient periods. The change from one phase to another usually takes several cycles. 2.3 HOMEOSTATIC MAINTENANCE OF RHYTHMICITY One clear evidence of the importance of rh3~hmicit3, in the physiological fabric is the strong homeostatic devices involved in its maintenance (Pittendrigh, 1974). For instance, a distinctive property of the rh)r partaking of the time-keeping s.vstem in an organism, is that the circadian period is time-compensated, i.e., its Q~0 is 1.0. The independence of period from temperature is a fundamental property of any system, biological or
Neuronal Mechanisms of Biological Rhythms
99
physical, capable of keeping time. Since the rate of metabolic fimctions is affected by temperature, this remarkable property of circadian rhythms indicates a complex system of biochemical compensator 3, reactions underlying rhythmicity. As a matter of fact, overcompensation has been identified in some rhythms with a Qt0 lower than 1.0. Temperature compensation affects only the period, the amplitude of rhythms does va~' with temperature. Also, not every, endogenous rhythm is temperature compensated. Fast rhythms in neurons or in the heart, do change their period with temperature. Tiffs indicates that only some rlkvthms are an integral part of the biochrometry system. Another example of the powerful homeostatic maintenance of circadian rhytmicity is the strength of the phase relations between circadian rhyahms. Even under constant conditions, there is a net tendency to maintain the time relations between rhythms" however internal desychronization is a common consequence of the suppression of entraining influences. These features are common to all circadian rhythms and are the basis for the conception, as illustrated in 1, of the three basic levels of analysis of the neurobiological substrate of rh.x.~hmicity, (Arechiga, 1993a) namely a) the identification of circadian pacemakers and the analy,sis of the mechanisms underlying rh3~thm generation, b) the identification of the receptors and pathways conveying the entraining influences, and the analy,sis of the neurobiological mechanisms of entrainment, and c) the study of the mechanisms by which time signals from the circadian pacemakers are capable of s~xlchronizing the rhythms in the various body fimctions. These topics will be discussed below.
3. The identification and analysis of circadian pacemakers For several decades, a search for the anatomical substrate of rhxlhm generation has been conducted. More recently, as some cellular conglomerates have been idcntified as capable of producing circadian rhytmicity, the molecular mechanisms underlying rh31hm generation are being explored. 3.1. LOCALIZATION OF CIRCADIAN PACEMAKERS. As in the location of putative centers of neural control of other physiological
100
H. Arechiga
functions in the body, various approaches have been used to identif3, the structures responsible for generating circadian rhythms. 3.1.1. Lesions and ablations. This time-honored procedure has been thoroughly applied to the location of structures in the body, and particularly in the central nervous system, necessary for the maintenance of circadian rhythmicity. It is as a result of this type of experiments, that the destruction of the optic peduncle was found to interrupt locomotor rhythmicity in insects (Nishitsutsuji-Uwo et al., 1991), and lesions of the suprachiasmatic nucleus in the mammalian hypothalamus have been s h o ~ to abolish the circadian rhythms of functions such as locomotion, food and water ingestion, and others (Klein et al., 1991). Similarly, pinealectomy has been shown to suppress the iocomotory ryhthm in reptiles and birds (Falcon et al., 1989: Takahashi et al., 1989). Selective lesions in the supraoesophageal ganglion and excision of the eyestalk have been shown to interrupt circadian rh.~hms of locomotion and sensory input in crustaceans (Arechiga et al., 1993b). The usefulness of this approach depends on the anatomical discreteness of the structures controlling a given rhythm. The more distributed the control is in different structures, the less effective are lesions or ablations in suppressing such rh)ethm. Indeed, it is remarkable that some rhythms are actually abolished after singular anatomical disruptions, thus suggesting the existence of discrete neuronal networks, selectively endowed with the property of generating circadian signals i.e. behaving as "biological clocks ". It is as yet unclear whether all rhx~hms in a given organism are controlled by a single "master clock". From the available evidence, it seems likely that the control of the circadian fabric is achieved by more than one pacemaker. 3.1.2. implants. This procedure is complementary, of the previous approach, i.e., it aims at exploring whether the structures whose lesion or ablation suppresses a given rhythm, are capable of restoring it when implanted in the organism. There are good examples of restoration of rhythmicity after implanting a putative "circadian clock". For instance, implants of embryonic hypothalamus containing the suprachiasmatic nucleus are capable of restoring some circadian rhythms in adult rats, lost after ablation of the nucleus (Drucker-Colin et al., 1984; Ralph and Lelunan, 1991). Pineal implants have been shown to restore locomotor3' rh3~hms lost after pinealectomy, (Zimmerman and Menaker, 1979) and a similar result has been reported after implants of ganglia in insects. In some cases, the implant
Neuronal Mechanisms of Biological Rhythms
SOURCES OF ENTRAINMENT
ENTRAINMENT PATliWAYS A N D PA( ' E M A K E R S
()UTPtIT
101
( )VERT RHYTHMS
i
E n
d i
a s
t
a
e
r
n
a
C I'
I1
a c
o
m
c
P
i
U
Figure 1. Basic levels of analysis of circadian rhythmicit),. (See text).
has been shown not only to restore a lost rhythmicity, but to impose on the host, the phase of the donor's rhythm, (Truman. 1974). This is a singular instance in neurobiology, whereby a neuronal network program may be transferred from one individual to another. Again, these results indicate the high degree of robustness of the timekeeping mechanisms, which are capable of self-adjustment to assimilate a foreign source of circadian rhythmicity. 3.1.3. Persistence t!f rhythmicity in vitro. Another powerful criterion to identify, circadian pacemakers, is to assess whether they are capable of maintaining circadian rhytmicity after explanted from tile body and kept in culture. Putative circadian clocks, such as the suprachiasmatic nucleus of the h31~othalamus, the pineal gland in reptiles and birds, the gastropod eye, and the crustacean eyestalk have been shown to maintain rhxlhmicitv in vitro. In some cases, such as the eve of Aplysia and Bulla, basic properties
102
I-1. Arechiga
of rhythmicity such as entrainment and phase-response curves are the same in the isolated system, and in the whole animal (Eskin, 1971" Block et al., 1993). As will be discussed below, this approach is proving extremely valuable to analyse the cellular and molecular mechanisms of rhythm generation. 3.2. CELLULAR AND MOLECULAR MECHANISMS UNDERLYING RHYTHM GENERATION.
3.2.1. Celhdar substrate o/" rhythmicity. As with other neurobiological functions, there are two basic tenets in the conception of how rh3~hms are generated in a neuronal network, a) the circadian daytlun is an emergent property of a system composed by non-circadian elements, and b) the rhythm is generated by cellular circadian pacemakers, capable of communicating rhythmicity to the rest of the sy,stem. In the first case, the circadian rh~r could be synthesized by a demultiplicative interaction between short period rhythms, linked in various ways, such as sequences of activation and inhibition, feedback loops, and selective convergence onto common targets (Block et al., 1993). The generation of a given frequency in a neuronal network, either from non-rhythmic elements, or from those with intrinsic rhythms different from that of the final output, has been described in various natural systems and there is a great number of theoretical models generating rhythmic outputs. As can be seen in Figure 2, systems as diverse as those generating automatic escape swimming movements in the mollusk Tritonia (Getting, 1989) by monosynaptic connections among intemeurons (A) and the pyloric rh.~r in the lobster, with chemical and electrical specific connections B (Selverston et al., 1976) bear great resemblances in organization. The same is true for other neuronal networks generating a rhythmic output, such as the one controlling the respiratory, movements in the cat (Richter, 1994), bear great resemblances in organization. All three have reciprocal synaptic interactions. However, important differences exist. Whereas the pyloric rhythm is generated as an emergent property of a network composed by neurons devoid of instrinsic pacemaking activity, specific pacemaking units have been identified in Tritonia. All these are ultradian rhythms. So far, no neuronal network has been shox~ capable of generating a circadian output, from interactions among non-circadian units.
Neuronal Mechanisms of Biological Rhythms
103
B
Figure 2. Diagramatic representation of two neuronal networks generating rh),ahmic outputs. A. Escape swimming system of Tritonia (From Getting. 1989). DSI. dorsal swimming interneurons: VSI. ventral swimming interneurons: Y and C2. interneurons with these denominations. --} . excitation: ---o inhibition. B. Pyloric system of lobster. AB. anterior burster neuron: VD. ventricular dilator neuron; IC. ixfferior cardiac neuron. PD. pyloric dilator neuron. LP. lateral pyloric neuron. ---o inhibitory chemical synapses: ~ electrical synapses. - . . . . weak synapses: strong synapses. (From Selverston el al.. 1976).
At any rate, a single neuron may display a variety of rhythms, the circadian or infradian components acting as an envelope for higher frequencies. However, important differences exist. The existence of a definite interaction between circadian and infradian rhythms has been established from experiments showing that photoperiodic manipulation may affect infradian reproductive cycles, and likewise, manipulations on ultradian rhythms, may influence the circadian periodicity, (Pittendrigh, 1974, 1991). Again. from lesion and ablation experiments, the same organs appear to be responsible for the gcneration of rhythms in the three frequency domains. One approach to explore a possible relation between circadian and ultradian rh~hmicity, has been to divide organs known to display a circadian rhythm and seek for possible changes in the period, as the number of cellular units becomes smaller, or even to abolish the rhytlun altogether, once key elements in the network are destroyed. So far, the results have
104
H. Arechiga
indicated a constancy of circadian period even in small conglomerates of neurons. For instance, brain slices containing portions of suprachiasmatic nucleus appear to maintain a circadian rhythm of electrical activity (Klein et al., 1991), and small parts of the eye of Aplysia and Bulla continue displaying circadian rh3etmicity in the firing of compound action potentials (Koumenis and Eskin, 1993). The anatomical organization of networks displaying circadian rhtyhmicity is an important and as yet not completely understood substrate of rhythm generation. For instance, electrical coupling has been described between cells in the circadian network of the gastropod eye (Block et al, 1993; Colwell et al., 1992), and among the neurosecretory cells showing circadian rhythmicity in the crustacean eyestalk (Alvarado-Alvarez et al., 1993). A synctitial organization would be similar to the substrate of rhythmicity in organs such as the heart and in some ganglia. An arrangement of this sort renders the gradual cutting approach irrelevant, since any part of a synctitium would maintain the rhythmicity, and imposes the need of isolating the individual components of the network.In fact, individual cellular pacemakers have been postulated in the gastropod eye (Michel et al., 1993) in pinealocytes (Deguchi, 1979), and in cultured neurons from the suprachiasmatic nucleus (Murakami et al., 1991). The ionic basis of fast rh.~hms in nerve cells has been amply characterized. The alternancy between opening and closing of specific ion channels is capable of generating the necessary fluctuations of membrane potential in order to maintain a rh3~hmic discharge. Various models of ion current alternancies have been produced to account for rhythmicity in specific neurons, with periods ranging from a few milliseconds to several seconds, basically depending on the number of currents sequentially involved and their kinetics. A variety of models of rh)r generation, are based on the assumption of interactions between ion currents, with selective sequences and feed-back loops, (Adams and Benson, 1985" Friesen and Block. 1984). Figure 3 illustrates a model of ionic current arrangements generating rhythmicity in a single neuron. R 15 in the abdominal ganglion of Apl.l,sia. As proposed by Adams and Benson (Adams and Benson, 1985), the interaction of membrane capacitance (C=), leak conduct,'mce (Gt). three fast currents generating action potentials represented as conductances for sodium (GN,), calcium (Gc~) and potassium (Gk), and their respective voltages (ENa, Ec,, Ek) and very importantly, two slow transient currents ID and IH. By ascribing numerical values to these components of the circuit model, a
Neuronal Mechanisms of Biological Rhythms
105
rhythmic output is obtained, as bursts of action potentials. (Adams and Benson, 1985). These fast rh)~hms may show some of the features of circadian rhythmicity, such as entrainment and phase-response curves, thus providing helpful models to probe general properties of ryhthms.
......
,,
+
~ t111111 G i
G.
Gee
III!111
9 lliilll '-IllIlll i
i ....
,
T,I
.L.
T
s
Eco I
T : -/+'2.---/+"t +0
. S,
. e
"
. +0 |Ira
+-
. + ,+'J .~-+~S
B
Figure 3. A, electrical circuit model (left) of tile membrane properties generaiing a bursting paliern in the neuron R 15 in the abdominal ganglion of Aplysia. B. Rhytmic oulput obtained by adjusling numerical values Io.ille model. (Modified from Adams and Benson. 1985)
3.2.2. Molecular mechanisms of rhythm generation.
Should the existence of cellular circadian pacemakers be assumed, endogenous, molecular mechanisms must then be envisaged. On the one hand, circadian rh.x,~hmicity is dependent on protein s)~thcsis. Amplitude, period, and most importantly, phase, can be altered by protein synthesis blocking agents, thus suggesting that synthesis of specific proteins underlies the maintainance of rh.~hmicity. On the other hand, from experiments in genetics, a locus in chromosome X in Drosophila has been identified as altered in mutants lacking circadian rhythmicity (Dunlap, 1993" Hall, 1995). The responsible gene has been cloned, and its structure is now known for several animal species. The per gene has been expressed, and by means of antibodies raised against its associated protein, the location of
106
H. Arechiga
putative per-positive cells has been identified in the central nervous system of various species. A good coincidence has been found between the location of immunopositive cells and that of cellular elements involved in the generation of circadian rhythmicity. Immunopositivity has been located in the cell nucleus. The per protein content has been found to va~" in a circadian manner, and in transfection studies, arrhythmic mutants have been transformed into rhythmic after receiving the per gene. The rhythm of per protein appears to be self-generated, since the protein has been found capable of inhibiting its own transcription, thus suggesting that a negative feedback loop is the key element generating the circadian rh.~hmicity. One interesting supporting evidence is that the per protein rh3.~hm appears to be temperature-compensated. Some intriguing observations have also been made, though, such as the widespread distribution of the per immunopositivity in the nervous system and other parts of the body, as well as the positive reaction found in glial cells. So far, no direct effects have been shox~aafor per protein in modulating membrane functions, which is a necessary step to explain its role as a time signal generating a circadian output, as illustrated in Figure 4. Other genes also appear to be involved in the generation of circadian rh~lmaicity, and further studies are necessary to complete a model to explain the molecular genesis of rhythmicity. The aforementioned results are encouraging in supporting the notion of cellular circadian pacemakers. Let us also bear in mind that more than one mechanism of circadian rh~lun generation may be operating.
3.2.3. Coupling o/" circadian pacemakers. One feature colrunon to the various rhythms is the coupling between circadian pacemakers in a given organism. With the exception of the pineal gland, all organs knox~ to generate circadian rhythmicity in vitro, are paired structures and although each is fully capable of maintaining a circadian rhythm, they are normally coupled with their contralateral partners, mostly through s3aaaptic interactions. Mutual coupling has a reinforcing action on rhytlmaicity, as can be inferred from experiments showing that after uncoupling, the rhyChms in the uncoupled pacemakers soon drift out of phase among each other (Page and Nalovic, 1992). The nature of such interactions is quite similar to that of entrainment, which will be described below.
Neuronal Mechanisms of Biological Rhythms
107
4. The entrainment of circadian rhytmicity The analysis of entrainment has led to the identification of specific receptors and neural pathways subserving it. and of neurotransmitters and modulators conveying time signals. MEDIATORS OF ENTRAINMENT PATHWAYS
RECEPTORS I-T1 m
//t ,
iNTRACELLUL IMESSENGER s
rTl rT1 r'l-! r ' n I--i-1
iRiS,I t.~, ~~ t.oI t_~~, ~ 'i ,' }
:
,
,:
:
:i
,:
I :
; i
: '
I I
, ',
, I
I-'L~
i i
;
, I
---. . . . .
x
~--
t'NTe~U/txArEsT/Io e NES
h
~.=
.
.
.
.
.
.
.
REGULATORY PROTEINS
.
I I
!
I
II
r
t
I
RNA' S
/
" " " N_~PJH_pS_~_P_RQ "----'~'~_T_ E I
I
MEMBRANE PROTEINS F/~'//A Y/~.~'t F//,dA
Figure 4. Molecular events in the generation and entrainment of a circadian rhythm in a single cell. (See text). (Modified from Arechiga. 1993a)
4.1. THE ENTRAINING INFLUENCES. As mentioned above, light is the most powerful agent entraining circadian rhythms, it is also quite amenable to experimental manipulation and
108
1-1. Arechiga
consequently, it has been the most thoroughly studied. The entraining phase may be either the onset or the suppression of light, depending on the rhythm. Light is also the common entraining agent for seasonal rhythms. The photoperiod length appears to be the efficient link between circadian and seasonal rhythms (Withrow and Withrow, 1959). However, light is not the only entraining agent; temperature, sound, hydrostatic pressure, chemical signals, and a variety of social clues are known to be effective in entraining circadian rhythms (Aschoff, ! 960; Moore-Ede et al, 1984: Pittendrigh. 1974). Although much less studied, some internal signals may also pla.v a role in entrainment, as is the case of food (Stephan, 1992) and homeostatic influences. There is also a long-held view. that the more sophisticated is the central nervous system, the more important become social factors in entraining circadian rhythmicity (Aschoff, 1965) Entrainment may be exerted by cyclical influences, but also by single. powerful, non-periodic inputs, as may be the case of food delivery.. The relationship between light intensity and entrainment has been amply studied. In fact, one complementary, aspect of Aschoffs rule, mentioned above, is that the period length under constant light, is a logarithmic function of light intensity. Several cycles are usually necessary to attain full entrainment, depending on the magnitude of the phase angle between the ongoing rhytlun and the entraining signal, i.e., the greater the angle, the longer the transient period of entrainment. This is of particular relevance when detemlining "jet-lag" or work-shift effects (Aschoff, 1960: Moore-Ede et al, 1984). The physiological mechanisms underlying phase relationships are still unclear. and may differ from one system to another. 4.2. RECEPTORS AND PATHWAYS MEDIATING ENTRAINMENT. Specific pathways of circadian rh.~hm entrainment have been identified in different species. In mammals, photic entrainment appears to be initiated in the retinal receptors, and conveyed to the suprachiasmatic nucleus by a specific retino-hypothalamic pathway. The severance of this tract suppresses entrainability of various circadian rhythms, such as locomotion and water intake, which remain in a free-running condition aftenvards (Johnson et al.. 1988). There is no precise information as to the pathways conveying information on non-photic entrainment, such as the one mediated by food. In avians, lower vertebrates and in invertebrates, the receptors mediating
Neuronal Mechanisms of Biological Rhythms
109
circadian photic entrainment arc not in the retina. It is interesting to note that the effect of light in entraining biological rh)r is phylogenetically older than the development of visual systems. Various locations have been identified for extra-retinal photoreceptors in the hypothalamus, the pineal gland, and other brain areas. Various photopigments responsible for the transduction have been described in extraretinal photoreceptors. The functional organization of the retino-hypothalamic entrainment pathway differs from the visual pathway in some important aspects i.e.a.) only tonically active, slow conducting ganglion cell axons have been identified, b) the neuronal networks both at the retina and hypothalamus are interspersed in a meshy way, thus contrasting with the geometrical order characteristic of the visual networks (Tessonneaud et al., 1994). The existence of specific pathways for photic entrainment, different from the primary visual pathwa.vs has added an interesting dimension to the stud), of afferent input to the circadian pacemakers. The precise way in which environmental signals are encoded in these pathways requires a more ample anal.vsis. 4.3. NEUROTRANSMITTERS MEDIATING ENTRAINMENT. The search for neurotransmitters mediating entraimnent has led to the characterization of several substances performing this function. The first candidate was serotonin, found to mediate entraining influences on the eye of Aplysia (Koumenis and Eskin, 1993). Various putative neurotransmitters have been postulated for entrainment in the supraehiasmatie nucleus (Russak and Bina, 1990). The entraining substances may also be hormones, since melatonin has been shown to entrain circadian rhvthmicitv in avians and humans. (Russak and Bina, 1990: Takahashi et al., 1989: Wurtman and Waldhause, 1985). From the available evidence, it is unlikely that a single substance mediates the action of different entrainment pathways, in fact, colocalization of neurotransmitters in supraehiasmatic neurons has been amply documented (Arechiga, 1993a). The secondary messengers and immediate early-genes involved in the cascade of molecular events resulting in entrainment are presently a subject of intensive study (Arechiga, 1903a: Takahashi et al., 1989). Figure 4 illustrates the likelihood of a variety of receptors, intracellular messengers and genes, active in the entrainment of a molecular loop generating a circadian rh.~,~hm.
110
H. Arechiga
5. Internal synchronization
There is only limited infomlation on the physiological mechanisms conveying the time signals from the circadian pacemakers to the effectors displaying the overt rhythmicity. The output of the circadian network may be either neural, i.e., synaptic, or hormonal in nature. In some cases, a single mechanism may account for the expression of the "clock". For instance, in birds, melatonin may reproduce the phase modulation of the circadian rhythm of locomotion obtained by pineal implants, thus suggesting that melatonin is sufficient to convey the s).~chronizing output of the pineal rhythm. In other instances, the s3~chronizing action of an implant does require that further outgros~,th from it actually innervates the target organ, thus indicating a synaptically mediated output from the circadian pacemaker (Page, 1982). The interactions between different time signals onto the effectors displaying the overt rh3r are quite varied. A given effector may be under the influence of more than one physiological signal, varying with different periods, phases and coupling strengths. This richness of inter-modulating influences is a feature of the circadian system that only recently is being explored (Sanchez de la Pefia, 1993).
6. References
Adams WB,. Benson JA. The generation and modulation of endogenous rhythmicity in the Aplysia bursting pacelnaker neuron R 15. Prog. Biophys. Molec. Biol. 1985" 46: 1-49. Alvarado-Alvarez R. Garcia U, Ar6chiga H. Electrotonic coupling between neurosecretory cells in the crayfish eyestalk. Brain Res. 1993 613" 43-8. Ar6chiga, H. Circadian rh~1hms. Current Opinion in Neurobiology. 1993" 3: 1005-10. Ar~chiga. H., Femfindez-Quir6z. F., Fem,-indez de Miguel, F, and Rodriguez-Sosa, L., The circadian system of crustaceans. Chronobiol. Int. 1993; 10: 1-19.
Neuronal Mechanisms of Biological Rhythms
111
Aschoff, J. Exogenous and endogenous components in circadian rhy.thms. Cold Spring Harbor Symp. Quant. Biol. 1960; 25' 11-28. Aschoff, J. Circadian rhythms in man. Science 1965; 148" 1427-32. Block GD. Khalsa SBS., McMahon L., Michel S., Geusz M. Biological docks in the retina: Cellular mechanisms of biological timekeeping. Int. Rev. Cytol. 1993" 146: 83-144. Colwell CS., Khalsa SBS. and Block GD. Cellular mechanisms of entrainment. Chronobiol. Int. 1992" 9" 163-79. Deguchi AT. Circadian oscillation in cultured cells of chicken pineal gland. Nature. 1979: 282: 94-5. Drucker-Colin IL Aguilar-Roblero R.. Garcia Hem.5.ndez F., Fern,'indezCancino F., Bermfdez-Rattoni F. Fetal suprachiasmatic nucleus transplants" diurnal rhythm recovery, of lesioned rats. Brain Res. 1984" 311' 353-57. Dunlap JC. Genetic analysis of circadian clocks. Annu. Rev. Physiol. 1993" 55" 683-728. Eskin A. Properties of the Aplysia visual system: in vitro entrainment of the circadian rhythm and centrifilgal regulation of the eye. Z. Vergl. Physiol. 1971" 74: 353-37. Falc6n J., Marmill6n JB., Claustrat B, and Collin JP., Regulation of melatonin secretion in a photoreceptive pineal organ an in vitro study' in thc pike. J. Neurosci. 1989, 9' 1943-50. Friesen WO. and Block GD. What is a biological oscillator'? Am. J. Physiol. 1984; 246:R847-851. Getting PA. Emerging principles governing the operation of neural circuits. Annu. Rev. Neurosc. 1989. Halberg F., Carendente F. Comelissen G. and Katinas GS. Glossary of chronobiology. Chronobiologia. 1977: 4. Suppl. 1.
112
H. Arechiga
Hall JC., Tripping along the trail to the molecular mechanisms of biological clocks. Trends Neurosci.. 1995" 18: 230-240. Inouye SYT. Kawamura H. Characteristics of a circadian pacemaker in the suprachiasmatic nucleus. J.Comp. Physiol. 1982; 146: 153-60. Johnson RF., Morin LP. and Moore RY., Loss of entrainment and anatomical plasticity after lesions of the hamster retino-hypothalamic tract. Brain Res. 1988" 460:279-315. Klein DC., Moore RY. and Reppert SM. ( eds. ) Suprachiasmatic Nucleus: The mind's clock. New York, Oxford University Press. 1991. Koumenis C. and Eskin A. The hunt for mechanisms of circadian timing in the eye of Aplysia. Chronobiol. Int. 1993" 9:201-21. Michel S., Geusz M., Zarisky JJ. and Block GD., Circadian rhythm in membrane conductance expressed in isolated neurons. Science 1993" 259: 239-41. Moore-Ede MC., Sulzman FM. and Fuller CA. The clocks that time us. Cambridge, Mass. Harvard Universit3.' Press. 1984. Murakami N., Takamura M., Takahashi K., Utomiya K., Kuroda H. and Etoh T., Long-Term cultured neurons fom rat suprachiasmatic nucleus retain the capacity of circadian oscillation of vasopressin release. Brain Res. 1991" 545: 347-50. Nishitsutsuji-Uwo J., Petropulos SF, and Pittendrigh CS., Central nervous system control of circadian rh.x~hmicity in the cockroach. I. Role of the pars intercerebralis. Biol. Bull. 1967: 133" 679-696. Page TL. Transplantation of the cockroach circadian pacemaker. Science 1982; 216: 73-5. Page TL. and Nalovic KG., Properties of mutual coupling between the two circadian pacemakers in the eye of the mollusc Bulla gouidiana. J. Biol. Rhythm. 1992; 7: 255-68.
Neuronal Mechanisms of Biological Rhythms
113
Pavlidis T. Biological oscillators: Their mathematical analysis. New York, Academic Press. 1973. Pittendrigh CS. Circadian Oscillations in cells and the circadian organization of multicellular systems. In: The Neurosciences, Third Study Program. F.O. Schmitt, F.G. Worden.editors Cambridge, Mass., MIT Press. 1974" 437-58. Pittendrigh CS. Temporal organization: reflections of a Darwinian clockwatcher. Annu. Rev. Physiol. 1993 55" 17-54. Ralph MR., and Lehman MN. Transplantation" a new tool in the analysis of the mammalian h.vpothalanaic circadian pacemaker. Trends Neurosci. 1991" 14: 362-66. Reinberg A., and Smolensky M. editors, Biological Rhythms and Medicine. New York, Springer-Verlag. 1993. Richter DR. Central control of respiratory movements. In' D. Jordan, editor Central control of the autonomic nervous system. Chur, Switzerland. Harvard Academic. 1994:18-29. Russak B., and Bina KG., Neurotransmitters in the mammalian circadian system. Annu. Rev. Neurosci. 1990" 13 387-401. Sfinchez de la Pefia S., The feed-sideward cephalo-adrenal immune interactions. Chronobiologia. 1993" 20" 1-52. Selverston AI., Russell DF., Miller JP., and King D. The stomatogastrie nervous system" Structure and fi~nction of a small neural network. Prog. Neurobiol. 1976" 7" 215-290. Stephan FK., Resetting of a circadian clock by food pulses. Physiol. Behav. 1992: 52: 997-1008. Takahashi JS., Murakami N, Nikaido SS., Pratt BL., and Robertson LM. The avian pineal, a vertebrate model system of the circadian oscillator: cellular regulation of the circadian rhythms by light, second messengers, and maeromolecular synthesis. Rec.Prog. Horm. Res. 1989: 45" 279-352.
114
H. Arechiga
Takahashi JS. Molecular neurobiology and genetics of circadian rhythms in mammals. Annu. Rev. Neurosci. 1995; 18:531-53. Tessonneaud A., Cooper HM., Caldani M., Locatelli A., Viguier-Martinez MC. The suprachiasmatic nucleus in the sheep: retinal projections and c3r organization.Cell Tissue Res. 1994; 278: 65-84. Truman J. Circadian rhx~hmicitv in an insect brain. In: F.O. Schmitt, F. G. Worden editors. The Neurosciences, Third Study Program. Cambridge, Mass, M.I.T. Press. 1974: 525-29. Withrow AR., Withrow R. eds. Photoperiodism and related phenomena in plants and animals. Washington, D.C., Am. Assoc. Adv. Sci. 1959. Wurtman RJ., and Waldhause R. editors. Melatonin in humans. Cambridge, Mass, MIT Press. 1985. Zimmerman NH., and Menaker M. Tile pineal gland: a pacemaker within the circadian system of the house sparrow. Proc. Natl. Acad. Sci. USA. 1979; 76: 999-1003.
Time, Internal Clocks and Movement M.A. Pastor and J. Artieda (Editors) 9 1996 Elsevier Science B.V. All rights reserved.
115
HUMAN VS. ANIMAL TIME JOHN CAMPBELL New College, Oxford. OXI 3BN ABSTRACT. A distinction is drawn between different kinds of representation of time, as phase and as particular time. This distinction is related to the kinds of causal significance given to time representations by animals and humans.
1. Temporal Orientation Suppose you are lost and finally pull out a map and try to get your bearings, an ordinary Ordnance Survey map, say. What you have to do is to bring the information on the map into some kind of congn~ence with the spatial information that you have from perception. The map uses one frame of reference, the Ordnance Survey reference grid, to speci~ spatial relations. Through vision you know that some things are near, some fi~rther away. some up and to the left, some in front and some behind. What you have to do is bring the two systems into congruence, so that you can say" 'This is the telephone box here, the church is a good way be|find us, and that is Wheatley over there: so the path must be off to the left down here." When you manage that, you have oriented yourself with the frame of reference used in the map. Can we make sense of a similar procedure in the ease of time? Can we make sense of the idea of there being temporal frames of reference with which we orient ourselves'? We can. Take, for example, the question so often hotly disputed by clergymen, 'Which day of the week is it'?" (Some affect indifference, but it is rarely genuine.) Here the problem is exactly to orient oneself with respect to another frame of reference, so that one can say, 'Yesterday was Tuesday. so tomorrow is Thursday and in four days it will be the Sabbath.' One uses terms such as 'today" and 'tomorrow" in orienting oneself using the temporal frame of reference, just as, in the case of spatial orientation, one uses 'here" and 'over there" in orienting oneself using
116
J. Campbell
the map. There are also parallels in the strategies one uses to keep oneself oriented - for example, one may keep track of landmark days, such as the Sabbath, or use methods which enable one to tell directly whether this is Wednesday (Friedman 1990). We have to mark a fundamental distinction between the two t3~pes of temporal orientation, both of which use tensed temporal indicators. One is what I will call 'temporal orientation with respect to phase" The other is what I will call 'temporal orientation with respect to particular time'. Temporal orientation with respect to phase is much more primitive. Suppose we have an animal which hibernates, and through the part of the year for which it is awake regulates its activity depending on the season. Such an animal certainly has a use for temporal orientation. It can recognise that it is now late spring, perhaps by kecping track of how long it has been since winter, and realise that it will soon be summer. But it may not have the conception of the seasons as particular times - it may be incapable of differentiating between the autumn of one 3'car and the autumn of another. It simply has no use for the conception of a particular autumn, as opposed to the general idea of the season. So while this animal is capable of orientation with respect to phases, it is not capable of orientation with respect to particular times. In contrast, our ordinar3., keeping track of days of the week really is orientation with respect to particular times. We think in terms not just of yesterday having been Tuesday, but it also having been Tuesday three weeks ago. We think in terms of particular Tuesdays as well as having the general conception of that day of the week. We can, for example, think of what happened on one Tuesday as causally affecting what happened on a later Tuesday - whereas it may be quite impossible for our hibernating animal to think of what happened one autumn as causally affecting what happened on a later autumn. The question I want to address in this chapter is what makes the difference betaveen these two ways of representing time; what makes it the case that many animals seem at best able to represent time as phase whereas humans represent time as linear. How does it come about that humans are capable of orientation with respect to particular times?
2. Types of Tense There must be a certain internal structure in a system of representation if what we have is to be a representation of time at all. The conductor does not
Human Vs. Animal Time
117
represent time to the orchestra. We need the distinction that is sometimes drawn between biological time and cognitive time. Not every, biological mechanism that keeps an organism in step with the rhythms of its environment can be described as representing time for the organism. Most obviously, an animal's reproductive cycle may be determined by the time of year, without the animal having any ability to represent the time of 3'ear. In contrast, an animal which uses an interval timer to control its foraging behaviour may engage in quite complex computations to learn from experience in determining which temporal intervals in foraging give it the best rate of return. Here we certainly do seem to have cognitive rather than merely biological time. If an animal is representing time at all, it must have some analogue of the use of tense and temporal indexicals. An animal which has a circadian clock may use it to identify, the time at which food appears at a particular place. For example, on encountering food at a particular place at noon, the animal may form the hypothesis that in general there will be food at that place at noon, so that at subsequent noons it expects food to be at that place. But for the clock to be serving that fimction, it must in effect be saying 'it is now noon" when the food is first spotted and when it is subsequently expected. And it may be that we can have a somewhat more sophisticated use of the clock, so that the animal can represent, 'it will soon be noon', or 'noon is just over'. Certainly, without some use of indexicals or tense, we would not have the use of a clock to represent time at all. Just so, our ordinary use of conventional clocks depends on our grasp of tense and indexicals in the public language. If we could not understand the clock as saying 'it is noon now" the best we could manage would be to use the clock as a direct, nonrepresentational spur to action. There is a background point to be made here about the use of tense and temporal indexicais. Use of a term such as 'now" is always governed by the token-reflexive rule" Any token of'now" refers to the time at which it was produced. Of course, the rule has to be applied with some care to representations of an animal, since it does not literally use tokens of the English word 'now'. But as I said, there must be something analogous in its System of representation if the animal is to be said to be using its clock to represent time of day at all. This token-reflexive rule is not enough to specifi,, the meaning of the indexical. For we need also to know something about the structure of the
118
J. Campbell
underlying domain of times. If we are considering the ordinary English use of 'now', then we would naturally take the topology of the underlying domain of times to be that it is linearly ordered by 'earlier than" (that is. that it is irreflexive, connected and transitive). If, however, we are considering the use of a circadian clock by an animal, the underlying domain of times may have to be taken to be cyclical rather than linear. We could put the point by saying that the animal is representing times as phases, phases which may recur. The animal is not representing times as unrepeatable. The animal represents 'happening at noon" as a feature that a token event may have, on a par with representation of features such as colour or place of occurrence. The animal has no use for the distinction between noon on one day and noon on another. It simply formulates hypotheses about 'what happens at noon', hypotheses which are confirmed or disconfirmed by events, and acted on as appropriate. It is consistent with this cyclical conception of time that the creature should think in terms of individual events. The creature may, for example. use a data pool about events to compute the probability of this or that happening at noon, and that use of probability requires it to be thinking in terms of token events. It is just that the particular unrepeatable time at which an event happened is not a notion for which the creature has an)' use. This creature may still be using 'now" in accordance with the nile, 'Any token of "now" refers to the time at which it was produced" But the underlying domain here is a domain of phases of the day, rather than a domain of particular times. In view of this, it would be quite possible for an animal to be using the past tense or past-tensed temporal indicators, without having our ordinar), conception of the past. An animal might have some use for memory, of what happens in the morning, though registering that it is currently noon, and use the fact that morning is earlier than noon in planning its future actions. But this way of using tensed temporal notions might still relate to phases rather than to particular times.
3. Interval Timing What then is needed, for one to be thinking in terms of linear time rather than in terms of time as phase? One answer is that we have to look at the dynamics of temporal thinking: we have to look at the way in which one's temporal thinking changes over time, at the cognitive dynamics of someone
Human Vs. Animal Time
119
who is thinking about the past. So, for example, we have to consider the move from thinking, 'It is snowing today', said on one day, to 'Yesterday it snowed', said on the following day. When we have a thinker who can make that kind of transition effortlessly, then, you might say, we have the conception of time as linear. The problem with this is that the cognitive dynamics of a human can in effect be mimicked by an animal which has not only a circadian clock but an interval timer, an internal stopwatch which uses to measure elapsed intervals. An animal might form the intention to forage in a particular area. and then leave after a pre-determined interval. When it leaves after that interval, having used its timer, that is it - it has kept faith with its intention and has no fi~rther commitments. So this looks quite like human cognitive dynamics, rather than the d3~amics of a creature which only has a circadian clock. But mere possession of an interval timer is not enough for grasp of linear time. This does suggest, though, a further use for tense in animal representation. So far we have remarked on the need to suppose that an animal using a circadian clock ma.v be using tense in orienting itself with respect to phase. But it ma.v also be that an animal using an interval timer is using tense in connection with elapsed intervals, and that the underlying domain here cannot in general be taken to be phases. This use of tense would be explained in tenus of the interval from the use of the token to the time of some designated event - when the interval is over, that is. So, for example, this use of tense might include, 'a day hence', or 'twenty minutes hence', where the tensed indicator is designating the interval from production of the token to the target event. In this system, "now" would simply be a limiting case, in which the interval from the production of the token to the target event is zero. This use of tense plainly ought not to be explained in terms of phase, and that is particularly evident if the interval in question is longer than the period of the animal's oscillator. But does it involve representation of time as linear'? There seems to be a sense in which we do not have designations of particular times at all here. All xvc have being represented are the temporal relations between designated events - the representation and the target event. And even if we do suppose that we are dealing here with particular times, we are at best dealing with a series of unconnected islands of time. temporal intervals within which various times are temporally related to one another, but there is no registration of the temporal relations between times in different timed intervals. Moreover. it seems to be entirely arbitrary whether
12o
J. Campbell
we regard the interval timer as employing the future tense or the past tense, as representing, 'three hours till lunch time', or as representing 'two and a half hours since I set off the stopwatch'.
4. The Causal Significance of Assignations of Time I suggest that one way to put the distinction between an animal thinking in terms of cyclic time and a human thinking in terms of linear time is to think about the difference between the types of physics that are being used by the animal and the human. The animal certainly has some kind of physics of its environment, which shows up in how the animal interacts with it. xxhich kinds of interaction the animal takes to be possible, what it will try to do and what it will not try. to do. It is a practical physics of its environment, one e,'xhausted by its implications for the animal's actions. We can look at the ways in which the notions of time are given causal significance by the creature using them. It is not that we are looking for definitions or reductions of the temporal notions" all that we want is to see how they figure in the creature's primitive physics of its environment. If an animal is to be said to be using the notion of two things being simultaneous, for example, it is certainly legitimate to ask what causal meaning it assigns to the relation what use its physics makes of it. Many concepts are causally significant. One way to think of this is as a matter of their bearing upon our ability to make explicitly causal judgements of the form, 'x caused y'. We think of ourselves as building up a detailed, reflective picture of the causal relations holding in the world we inhabit, and of our own place in that causal nexus. But there are cases in which one's grasp of the causal significance of a notion has to do not with any detached picture, but rather consists in one's practical grasp of its implications for one's oxx~ actions. We can contrast a theoretical understanding of the causal properties of particular types of wood, for example, or different metals, such as iron or silver, with the understanding possessed by the carpenter or metalworker. The artisan's grasp of causal properties is not a matter of having a detached picture of them. It has to do rather with the structure of his practical skills" the particular way in which he deals with various ty'pes of wood. or how he uses different metals. The detached theorist need not have these skills. It is in characterising the propositional knowledge of the carpenter or metalworker that we have to use
Human Vs. Animal Time
121
working concepts. The subject's grasp of such a notion has to do with his practical grasp of its implications for his own actions. It is because the physics that an animal uses has this exclusively practical character that it can be said at best to have a conception of the phascs of a cyclical time. A grasp of time as linear cannot be put to use by a purely practical physics. The physics can give significance only to the conception of repeatable phases, which can be relied upon to reappear so that the animal can act with respect to them: it cannot through action give meaning to assignations of unique unrepeatable times to events. We can drive home the point by considering an animal with a circadian oscillator and asking how could we provide the animal with the capacity to tell one day from another, in such a way that it acquires the ability to represent time as linear. We could include oscillators with a longer period than circadian - the animal could use circalunar or circannual oscillators, for example. But if we are considering a system which uses only filrther oscillators, such as circalunar or circannual oscillators, then there will still come a point at which the system does not differentiate between events which happen at the same phase of the longest oscillator, but at diffcrcnt times - for example, events which happen at the summer equinox of different years. This need by no means be a deficiency in the system - the whole point may be that the system is dcsigncd to record circannual phases, and has no further interest in putting a date on phenomena. On the other hand. suppose we consider the use of a decay or accumulation process to differentiate between events which happen at the same circadian phase on different days. Then there will be no question about the system continuing to confound events which happen at the same phase of some longer-tern1 oscillation. But all that we have here is a way of differentiating between events that happen at the same circadian phase on different day,s. This would be of some use to the animal in remarking, for example, that on days xvhcn it was F at one time. it was G at some later time. and that it was G only on thosc days. There is no need to take the fi~rthcr step of requiring that the animal has a way of recording on which day thc event occurred. There is a dilemma here. In the case in which the animal is using a circalunar or circannual oscillator, we can take it that it really is being used to record times, because of the use to which the recorded infommtion is put when that phase of the circalunar or circamutal oscillator comes around again. This is how the animal's practical grasp of the physical significance ,
o
.,
122
J. Campbell
of the temporal data operates. But in the very nature of the case, the same factor that makes it so evident that we are here dealing with a record of time also means that we are bound to have the problem of confounding the times of events which happen at different times but at the same phase of the oscillator; the whole point is that the information can be put to use when the same phase comes around again; that is what makes us confident that we are here dealing with a record of time. On the other hand, suppose we are considering an animal which is using a decay or accumulation process. Here there is no such thing as the 'phase" of the process, if that is taken to mean a point in an oscillation, for there is no oscillation here at all. This means that there is no problem of confounding events which happen at different times but in the same phase. It also means, however, that there is no prospect of the animal acting upon the recorded information when it comes to the same point of the cycle again, just because there is no cycle. So we have no reason to describe the animal as recording which day this or that happened on; all that matters to the animal is that the confound between days has been removed, and merely removing the confound between days does not of itself require that one should have a way of saying which day this or that happened on. The animal's practical grasp of the significance of the use of the decay or accumulation process amounts only to removal of the confound. It might be that the use of the decay or accumulation process does more than simply remove the confound: it might be that in forming its expectations, the animal assigns progressively more weight to what happened on later than on earlier days, for example. So it is not just that there is an existential statement about there having been various days on which various events occurred: the various days are all put in some kind of order by this progressive assignation of order to them. But this in itself does not require that we should yet think of the animal as recording on which day the event happened. Let us go back to the hypothesis of an animal which is using an entrained oscillator to record circadian phases. This ability to record the phases of particular happenings enables one to home in on each of several circadian phases as they come around again. But it does nothing to show that one grasps the temporal relations of these phases to one another. More generally, use of a circadian oscillator does not mean that the animal has any conception of the connectedness of time - the fact that ever)' time is temporally related to ever)' other time. The animal has only an ability to reidentify each phase when it appears.
123
Human Vs. Animal Time
The general point I have been making is that what gives us the right to regard an animal as using the oscillator to record the times of various events is the use that the animal makes of the information on future occasions when that phase of the cycle comes around again. This means the animal cannot be ascribed representation of anything more than phase. It could do nothing with the further information that the event happened at a particular past time, one among many particular past times all temporally related to each other, and which may or may not be occurring at the same phase of the cycle. Information about particular past times has no role in the animal's future engagements with its world. A grasp of time which is ultimately exhausted by its significance for the demands of action can be at most an ability to represent phase, since the same phase can be re-encountered and that can affect one's actions. It cannot be a representation of particular time. What is required for the ability to represent linear time, I shall suggest, is the capacity not just to intcract with one's surroundings but to rcflcct on those interactions' to represent the causal relations between oneself and one's surroundings over time.
5. Causation in Narrative Memory The way in which humans give physical significance to linear time is through the construction of narratives detailing the causal relationships between various events. You might, for example, think in tenns of a whole network of plans and projects for the future, a whole career trajector)'. Musing on these topics, one can construct a whole narrative for one's fi~ture life, detailing various cumulative changes that one plans to make. using a grasp of the causal connections between various envisaged future events. You can have a whole narrative structure concerning your projected fi~ture. in which the various actions you plan to perfonn and the various events you expect to occur are all integrated into a single coherent stor3' as to how things will go. And this can exploit one's grasp of a linear time. One is thinking of the various things that one will do and the various consequences they will have as all occurring at unique unrepeatable times. Most starkly. preparation for death, when one makes out a will or plans finally to score off one's enemies before it is too late. involves thinking of the linear time-series. rather than of the time of death as a phase which may recur. .
124
3. Campbell
Narrative construction does not take its principal or even its most characteristic form in relation to intention and the future; it also figures in our thought about the past. To see the role of narrative construction in our thought about the past, it helps to consider one model of the development of episodic memory. Children of three or four years old can easily say 'what happens when" various recurrent events happen, like having lunch or going to school; they can provide 'scripts" which list the main features of these events in the right order. This laying doxs~ of general scripts presupposes a grasp of temporal order - you have to grasp the idea that when you make tea, you first of all boil the water and then put it into the teapot, for example. And it may also use a background conception of cyclical time. It may put to work one's grasp of the phases of the day, for example: one's grasp of the script for breakfast might include the information that it happens in the morning. But so far there is only the conception of the cycle of phases. There is no need or use for the idea of specific past times, no need or use for the conception of time as linear. Now one simple model of the development of memory in children supposes that this kind of script memory comes first, and that episodic memo~, depends upon having this script memory. Up until three or four, children spend their ener~, on accumulating general scripts" after that, they start to recall particular events in terms of the ways in which they deviate from the standard scripts they have already laid do~aa (cf. Nelson, 1988. for discussion). Whether or not this model is correct, we can ask" what has to be added, to possession of general scripts using a background of cyclical time. for the child to be thinking in terms of a linear time? One proposal is that the cn~ciai shift is when the memor3' is memory not of t)q3es of event - "generalised events" - but of individual, token events, specific happenings. But straight off this does not seem to be correct. An individual event has many characteristics - it may involve objects of various colours, be noisy or quiet, and so on. The phase of day at which it occurs may recorded as simply one among many of the features that the event has. There is no reason why one should have any more specific information about the time of the event than that, even if one is thinking about a particular token event, rather than a t~qge. You might object that if one does not even have the conception of linear time, and of particular times rather than phases, then one has no use for the idea of a particular past event. But even someone with only cyclical time might well have a use for talk about individual events. In particular, they might compute probabilities, or
Human Vs. Animal Time
125
quantify over token events in establishing the existence of correlations between types of event. If this is correct, it means that we could have episodic memory., in the sense of 'memory of a particular event', as opposed to the laying doxx~ or reinforcement of a script, even though we had cyclical rather than linear time. What kind of memory then would demand the conception of time as linear? It seems evident that so long as we focus on memory, of an individual events alone, even if it is a token event rather than a generalised event-type. memory will not demand the conception of time as linear. We might then consider memor3.' of the temporal relations between events. But so long as we are considering memoq' only of the temporal relations between events, as earlier than or later than one another, it will be possible that this time order is being imposed on a domain of cyclical phases rather than a linearly ordered collection of particular times. It would be as if someone were to travel around the earth, along the equator, remarking as he or she went along which places were to the east or west of which. This memory, of order would not be enough to make the memory into a memory of a linearly ordered set of places: the places remain stubbornly, ordered in a cycle. In seeming to remember that noon on one day preceded noon on another day, I might simply' be making the same mistake as someone who. having travelled around the globe, concludes that London is west of London. The picture changes, though, when we consider memoq, of the causal relations between events. We have to give a central place to the notion of a narrative detailing the causal relations between various events. If I remember that what happened at noon on one day causally affected what happened at noon on another, then I must be thinking o f the first event as having preceded the second. Cause must precede effect. And my, grasp of the causal relations among the various events I remember might have considerable sta~cturc. It may be controlled by an intuitive physics governing the kinds of changes of changes that can happen around me. It may also be informed by, an understanding of causation in socialpsychological relations. Indccd. when we think of the role that this kind of narrative memor3.' plays in our lives ever3, day. it seems likely that a grasp of social-psychological relations will be of central importance. This kind of narrative construction is. of course, not important just for its role in making it possible for us to think in temls of particular times. It is through possession and use of the ability to constn~ct narratives of our lives that we
126
J. Campbell
are able to criticise and corrcct our most basic motivations, and to form a conception of how to live.
6. References
Friedman, William J. About Time. Cambridge, Mass: MIT. 1990. Gallistel. Charles R. The Organisation of Learning. Cambridge Mass: MIT. 1990. Nelson, Katherine. 'The Ontogcny of Memory, for Real Events'. In Remembering Reconsidered, U. Ncisser and E. Winograd. editors. Cambridge: CUP. 1988.
Time, Internal Clocks and Movement M.A. Pastor and J. Artieda (Editors) 9 1996 Elsevier Science B.V. All rights reserved.
127
TIME AND PSYCHO-PHYSICAL INTEGRATION RAFAEL ALVIRA
Departamento de Filosq/ia l'rdctica, Facultad de FilosolTa 3, Letras, UnJversidad de" Navarra. PamphJna. Spain ABSTRACT. The first thesis of this paper is tirol time is flaree-dimensio~ml. Wlml we understand by time is a realit), which has fllree vectors, a) the dynamism ofbeing (time as origin): b) the measure of this dynamism (time as mediation): c) file duration of being (time as end). The second thesis is flint time is not a penuanent passm.e b.v. which woldd imply the dominant presence of negativi(v, but that time is really ahva.vs a "'wholeness", someflling "complete". a pemmnent s)alfl~esis of past, present and fi~lure. The third thesis is that the simultnneib, of time is space, and the variabili .ty of space is time. The fourth fl~esis is tim! there are flu'ee ~diseases" of time. corresponding to its three dimensions, the dise~lse in the origin is ageing: file disease in file mediation is madness and lack ofrh.vthm: the disease in the end is weakne~:~.
I. The proposal I -am going to make in tiffs article is at once a synthesis and a project. It is a s3althesis of reflections on the meaning of thne, and a scheme suggesting lines along which this study could be followed up. I shall dispense with the conventional critical apparatus, ,as the references would be limitless. such is the interest flaat this subject has generated over fl~e centuries. I shall first discuss time considered "m itself". I shall then turn to fl~e "levels" of time, and finish by examining time in the humma being. II. It is a commonplace to consider time in conjunction with movement. Either the former belongs in some way to the latter, or both lie witlfin the same sphere. Time is also very, often presentcxi as behag involved in space, be this because the former depends on the latter, because they are opposite realities, or because the latter is made to depend on the fonner. The tendencv of modem research is to "free" time from its spatial implications, or rather, as in certain presentations of the theory of relativity, to speak of a 67~ace-tt'me in which time predonfinates. Similarly, time is also placed in a relationship with "matter". Even if the latter concept, knowledge of which is generally taken for granted, is highly
128
R. Alvira
problematic - what do we mean by matter'? -it is undoubtable that discussions of time in the westem tradition have explicitly referred to matter, what is material is temporal, and time is the timc of the material world. My first thesis is therefore a straightforward one. what we understand by time is relatively different according to whether it is linked to matter, space or movement. How is it then possible to connect these three concepts or realities in such a way that we can obtain a coherent image of time? We can call on the basic end or aim. This scheme is in itself tripartite scheme, origin, mediatir inclusive of the four concepts listed. From the origin we move to the end, and the relationship between origin and end is a space sketched out in time. The concrete connection we have made can be presented as follows. a) Origin. time mad dynamism (matter). b) Mediation. thne and measure (basic space). c) End or aim. time and duration (movement). The first requirement for the existence of real time is that something should be real. This realit),, now taking the expression in the bold. genetic sense, is an emergent act or activity,, a "burst of energy". If it is considered to be different from time, then time is its measnre, insofar as it undergoes a certain cha/lge, or moves, and it is also the duration of this activity. However, if time, as a measure and duration, is different from the regality which it measures and whose movement it expresses, then it is abstract with respect to this reality, which means that there are three possibilities. Either it is another reality facing the first reality, which is absurd: or it is merely ideal, and exists only in the mind (this is in turn understood as being different from realit),, which is problematic), as Aristotle says when he maintains that time as such is only in the soul, or Hegel means when he calls it temporality; or it is accidental with respect to a reality, xxhich is then automatically viewed as being substantial. The interpretation of time as ,an accident, however, presents many difficulties. as the studies bv N. Grhnaldi have clearly shovm. If time is not substantial, first of all this enables us to envisage a non-temporal substance, ~ i d ~ is ~Kat is traditiottally h~o~aa as God. The other substances. however, are very p r o b L ~ ' c , given tlmt ~hile they move and change, the}' are measured by a time which does not n~-asure tl~naa pfimari'ly as" substances ~.
! In the aristotelian philosophy it is said tirol there is no lime without movement and that movement is the act of a being in potency insofar as being in potency, not as being in act. Tha! is Io say, insofar as in act it is not measured by time.
Time and Psycho-Physical Integration
129
It is quite clear that the discussion is complex at this point for many reasons. Not all file philosophers who have placed time on the accidental plmle have done so in the same way. For example, Aristotle sees the accident as being not so much time as being in time. Moreover, we can rightly maintain that accidents are very important for the subst-m~ce, and that they affect it. In the final analysis, however, it is difficult to see how what is secondary - the accidental - can be a determinant of the life of what is primary, that is, substantial. For this reason, attempts are made to invert this reasoning and maintain the thesis that time is what is "subst,'mtial", now ~4thout accidents. This flaesis alone -it is said- can guarantee the true reality of thne -and the temporal, by making way for an understanding of time ,as the principle of never-ending newness which is proper to the existence of a being whose life is changhlg. If fllere is no real n e ~ e s s , time is irrelev,'mt and existence is purely the "tiresome" permanence of substantial reality. This would happen ~ifla the basic matter or stuff, because pure matter is always und![tbrentiated, mad hence not affected by time, mad even more on the very. largest scale, xxith the divine substance, which, being eternal and not temporal, would have no problem other titan its ox~11 eternal boredom. enclosed in its solitude and its lack of an)~hing new. Given all this, even fllough material reality -understood in this m mmer- would not experience boredom in the perpetual ne~aaess of its life. it would suffer the evil of a perpetual lack of an ending, of hlstability. The above thesis, according to which thne and reality are either different or the same, has serious flaws, in my view, they cml only be resolved if we take into account the fact that "realit3.'" is not uniform, but that it presents diverse. progressively complex forms, that time is not the same for a stone as for an animal as for a human being. The follm~4ng discussion of tiffs point will, I hope. provide a suggestion as to how this problem c-an be solved. For now, it is enough to say that no fimdamental reason can be seen for relating time to the "medium" and the "end", and denying its hwolvement in the "origin". There should also be ,an "'original time". Let us move on to look at the second moment, the "mediation", "thne and measure". We can, of course, conceive the time as a continuum, but not only ,as a conanuum. In fact, if this were the case, we should find it very difficult to or ql?er ~ithin time, as no internal understand how we can distinguish hr break occurs from which we c,-m verifi:, such a distinction. If there is a certain variation, then, even though this might be continuous, such a break would necessarily exist. The "absolute time" of reality, postulat~ by Ne~on, together with the "relative time", which mechmlics supposes, an infuute, homogeneous o
130
R. Alvira
time, is indiscernible as time. it is non-time. It is hardly surprising that Nex~on's followers should have v,ithdrax~ this idea from circulation. If, then, in time there are necessarily breaks, these are constitutive, to use an image, time proceeds stroke by stroke, step by step. Time can be thought of as being like a beating heart or a clock. The idea of a stroke or step is not straightforward as it appears here. On the one hand, it is typically spatial, however much the intention is to avoid tiffs. The step goes.from one point to another. We can try to refilte this observation by saying that it reintroduces space, placing it on the same level as time, which was not the intention. It could then be said tlmt the stroke of time alwavs opens or creates a space, but in the first place it is the temporal dynamism which does this. and in the second, this is a "space of time", a space only for the mind that connects it. It would be the mind that would introduce the relationship of simultaneity, which is characteristic of all true space. But in reality, no such simultaneity would ever occur. In other words, between the beginning of one beat and the beginning of the next there would be no intervening space, precisely because there is no simultaneit3,, the one follows on from the other. It is my view that this last thesis reflects no more than a part of the truth. Of course, by time we understand a reality" whose elements are not simultaneous, while .space is the opposite, it is formed only if its extremes are simultaneous, otherwise it never truly exists. Time and space are therefore, in a certain sense, opposites. Pure space cancels out time, and tmre time cancels out space. Classical thought, which insisted on the reality of God and the eternal, thus very clearly gave space prec~ence over time, sttbordinatmg the latter and implicitly converting it into space or mininfizing its mode of being. More recent thought, however, emphasizes the material, evolutionary, nature of realit3', and so it gives priority, to time and subordinates space, turning it into a particular dimension of what is temporal. To my mind, even though this view might seem to be a compromise or an oversimplification, the two stances outlined above are in fact unilateral. Or. in other words, it is h~correct to say that fl~ere exists a pure space which can cancel out time, or a pure time that can cancel out space. We must in turn avoid any compromise position which favours one or other of these two realities, be this, to take the most frequently quoted examples, Aristotelianism, which defends the idea of"subordinate" time. or Einsteinian relativity., which proclaims that "space" is subordinate.
Time and Psycho-Physical Integration
131
The stance of authors in the idealist tradition seems to me also to be inappropriate. According to this view, which has various different versions, time and eternity (in other words, space) are two sides of the same coin. In my view this solution does not draw a clear disthlction between the roles of sight, which enables me to perceive space and is ecstaac, that is, non-temporal, and hearing, which gives me access to time, and is dark, that is, non-spatial. I ~ us now go back to the starting point of the above reflections, flaat is, that the stroke or step is unintelligible x~4thout spatial relations. This is not to say that the progression from one stroke to the next cannot be thought of except as a space (which is what actuall.v happens), as tiffs fllesis introduces space only as a place in our thoughts. What this &Jes mean is that the movement from one stroke to another is a spatial relationshq~. In fact, the newness that is proper to genuine time is thinkable only as a going out from where we were or where we dwelt, but this is completely impossiblc if there is no outside, ,and this outside is space. If it is maintained that the d3~aamism of time or the temporal gives us the otttside, the answer is that the negation which is the reason why one moment is not the next (mid without which time is not real), or, if we prefer, because the notion of m s ~ t is reiected, the negation which gives us a before ,and at~er, or a was .and is not, this negation constitutes a space betaveen the before and at~er. It is generally k n m ~ by these very. words, as a space ol'time. This space of time is a unit. mad therefore a logos: like all units, it is a measure. It is thne as a mea.vure mid as a medium. There is no measurement without a medium. The medium here is the very. form of the interval. The interval is a negation, and it is not just m~y negation, but the one which has the specific form marked by the beat or d3rmmic stroke. Absolute negation is thinkable only in relation to an absolute being. It would have to exist inside this being, as it could not remain outside, and it would be simple, an absolute nothingness. The other negations have the foma proper to the relations that determine them. The interesting point ,as far ,as time is concerned is that ever3' haterval, ever3.' space of time, ever3, beat is, to a certain extent, equal, but also in some way did, rent from the one which prec~es ,'rod the one which follows it. If these intervals were not equal, then time would be a nonsense at the very root. unintelligible both in itself and for the human mind, as we cml easily see. This provides us with another focus, another ~pect, which reinforces the thesis defended here. Along these lines, time possesses a spatial quality, equality is here spatiality. On the other hand. however, if there were no differences it would
132
R. Alvira
not be time, but just pure space. Diversity also occurs in space, but this must therefore be a dynamic diversit),. Each beat is different in terms of dynamism or energy from the one which comes before it, even though its form is essentially identical. To use classical philosophical terms, the essence of time as a measure is successive repetition; its existence is the difference in energy of each stroke or beat. Time as a measure is real. intelligible time. It is not intelligible time, which is time m general inasmuch .as this exists in the soul or in thought, but rather, as we have just stated, it is real intelligible time. In fact, time as an origin, that is, time as radical dynamism which transfom~s itself, is indql~nite in its very" origin. and time as a mere duration is also indefinite. Only firings which have a unit can be grasped or conceptualized - in this case, time as a concrete measure. Sensory perception would not be able to occur, either, without this dimension of time. A continuous sound completely lacking in rh)~m and melodic variation would not be noticed with regard to tempomlit3.'. We would not be aware. through our senses, that time was passing. Lastly, in its relationship to the etad or aim, time appears as a duration. Duration is not the first, objective, "conscious" end towards which all realit), is moving. Nor is duration, in its action of keeping in being, the end of real or vital development. All flint this means is that duration is the end and the conclusion of time. Time needs to last mad to "make flaings last" in order to exist as time. If each beat or stroke, on the one broad, were completely new, and on the other they totally obliterated those which had gone before, there would be no such thing as time. In fact, someflfing completely new, and something that obliterates its predecessors totally, are two dimensions of flae same thing, that is, of something that cannot exist. What is new, of course, is something dilTerent from what is past, but if flae past were not retained, flae new would not be new. Here, too, we can assert that this thesis is only true g~th respect to thought which compares past and present, but this is not the case. Experience offers endless proof, on all levels, flint retention is real. The past is not something that is only in the memory, whether flais be conscious or intelligible: it is in reality, in a way' that is not at all abstract. The past is not, like the "new" present, the latest burst of energy: it is what is available for this prc~ent, it is its reserve. It is the body.
Time and Psycho-Physical Integration
133
Conservation of time is a further indication that "spatiality" cannot be annihilated or subordinated. If the past is maintained, as it most certainly, is, then the identity of all reality exists. This identity is "spatiality". To speak in coarse terms, what reality., especially life, wants is to be, to live. But locked within this living is the fact that it has to carry on living, to endure. This is an end which we cannot do ~ithout. The celebrated controversy' over the primacy of the principles of conservation, or growth and innovation proves irrelevant here: the first st,-mce turns out to be "mechanicist", the second. "vitalist". Naturally, life implies gro~,lh" and the maximum life, love. always wants more, as Nietzsche say's. But it is impossible to interpret this more, however much newness this implies as opposed to conservation. III. After this attempt at an outline of what "time in itself' is, we can now go on to approach the subject of the "levels of time". This can be done from at least two viewpoints, one which starts out from the spatial situation, and another which sets off from the temporal side. In other words, time is the time of reality, but the latter cma in rum be regarded mainly under one or other of these perspectives. "Spatially" speaking, reality makes itself manifest in degrees or "steps". In the neo-Platonic philosophical tradition, one fundamental principle is the existence of the so-called "orders of being". In the last few hundr~ years, especially within the modem tradition, the expression "realms of nature" has often been used. Inanimate nature cma be distinguishcxl from the plant kingdom, as the latter can from the anhnal kingdom. The activity of living beings is much more complex, versatile and m,'my-faceted th,'m that of hmnimate beings. To the extent that time is written hato dyammism, living beings have a time which is different from that of inanimate beings. :rod which varies among them according to each one's individual vital structure. Their energy is different in each case, and so, therefore, is the "origha" of their time, m~d their "clock" or measure. Regarding duration, living beings evince a dialectic which is frequently found in nature. strength is matched x~ith weakness. ,and vice versa. On the one hand, they arc much richer in terms of energ3' th,-m are in,-mimate beings, but on the other, they die. They store up their past and plan out their future with increasing intensity. that is, they make time flaeir ox~11,they, somehow shape it. They "move" through time: they do not restrict themselves to letting time be something that "happens" to them. Nonetheless, time takes its revenge on these relative, domhmted beings by abandoning them.
134
R. Alvira
If one of the main dimensions of what we call intelligence is being conscious, "taking charge" or "making oneself the master" of the situation, this consciousness consists first and foremost of the ability to synthesize time in a more intense way. Being conscious means understanding the past as the past and relating it to the plans for a future seen as a future, and doing all flais in a way that is centred and made present with reference to the subject or, if we x~4sh, even though the expression is ,'unbiguous, the self Synthesizing time is not the same as creating it. The most usual thesis is that, given that all reality has a certain unity, whereas time, like movement, consists of someflaing passing, fllen "time as a whole" (the unity, of past. present and future) only exists in flae mind (or in the soul, as Aristotle wrote); in other words, it is a creation of the mind. The time of real reality, however, would never be "whole", which would show. depending on which of these opposing viewpoints is adopted, either that material reality, is paradoxical of its ver}, essence, and its mah exists only in the spirit, or flint the only truth is that of temporal beconfing, and fl~ought is a mere abstraction. What I want to say is flint "time as a whole" is always there, in all the kingdoms of nature, and that what file consciousness does is intensi~' the synthesis by a procedure of taking possession of it, more or less according to the biological level of the being in question. As far as the discussion of this issue is concemed, the thesis can also be formulated as follows, scholars who contradict the classical tradition by claiming that time is primary, in character are right, but in my opinion flaey are wrong if they also maintain that time is only whole in the mind. and that a real but not "whole" time is the only radical. What would be the consequences of negating the wholeness of time in all its natural realities'? The answer is the primacy of negativity. Nothing can rest or be maintained, nothing can lay claim to a certain permanent identity, and flaerefore ever}.~ing is tainted, riddled with fl~is negativity, which is continually pushing outwards. Primary time would be thus what is real to the same extent that what is real is total poverty,, it is as positive as it is poor. But if time is whole. "'complete", flaen there is no need to accept these consequences. Moreover. time is not the only reality, the only principle. We must now tum once more to this subject, which we examined above in our discussion of space. On the sensory level we find "spatial" and "temporal" senses. Touch and, above all, sight are spatial. The characteristic feature of sight is precisely that it entrances me, that is, it obliterates time for me on a subjective level. Smell and
Time and Psycho-Physical Integration
135
hearing are temporal senses. Hearing goes inwards, it interiorizes me ,and uncovers my past; smell goes out, opening me to the future. Last of all. taste is the sense which belongs to the present, it holds present spatio-tcmpomi awareness within itself. This equilibrium of sensor3' perception must have a real meaning. What right does the intellect have to deny this? It is not only time that exists, space exists too. Therefore we do not live in variation alone, but also in ecstasy, time and eternity. In classical thought, which subordinates time, etemit3' sometimes assimilates the changeable, "ncxv" nature of time by asserting that etemit3' is not only ecstasy but also living wilhtml growing old, a pemmnent nev~ness. Hcre, two distinct experiences are being merged, one being time, the other, etemit3'. We shall shortly return to the subject of a~eing, but first we must look at the present-day conmaonplace notion of the plurality o/" ames, and flleir supposedly un~,ielding nature. This entails the view, to my mind a correct one, that ever3' reality has its o~sal time. But it then goes on to say that flfs implies that simultanei~ must be ruled out. For simultaneity to be tree. an absolute time would have to be real, which is ,'m untenable h)19ofllesis. The problem would not be just that it would be impossible to observe simultaneity because the obscrvcr would necessarily have to have a viex~point: the problem would be that there is no simultaneity at all. This thesis is genuinely hard. It does not prove easy to reconcile it with common sense, whether on a sensory or an intellectual level. It is said to be a problem of the sensor 3' consciousness of the recipient, or of the difficulty of accepting somethhag as real which we are incapable of imaghaing. Science has left too far behind the straightforward view of the world based on common. everyday experience. It is quite clear that science leaves this experience far behind at many points, but the idea that science refomas the.lhndamental experiences is not so obvious. It would come as a genuine surprise to find that the times of the mother giving birth and the child who is being bona are not simultaneous even at that moment. It is strange, too, that someone can hold a conversation xsith another person in a complete lack of simult,aneit3'. Of course, the accusation could be made that these are trivial examples which do not take account of recent theoretical advances. But it is also true that the theory, once it has been thought out, should be brought back to the fundamental experience. How can we understand life ~sithout simultaneit),? Doubtless. as we have seen. eveq, reality has its o~1~ time. Indeed, file hypofllesis of absolute time
136
R. Alvira
-the "extemar' measure of all time - is hard to maintain. But it is equally difficult to accept that there is no simultaneity. On these grounds, then. we could propose that simultaneity does exist, and that it is, as we saw above, .space itself. Pure space cancels out pure time. ,and vice versa. Given that this cancelling out is essential, that is, according to definition and form, but not existential, what actually exists is simple, the simultaneity of ame is space. ,and the variability of space is time. This thesis can be understood as a simplification, but it mav well be no different from fundamental experience ,and common sense. The reflections set out above are rooted in the mode of being of all consciousness, be it sensory or intellectual, which intensifies the synthesis of time. The time has now come to tackle a subject that has already been mentioned, and which is relat~ in a special way to both consciousness and any form of life: the question of ageing. In a vcr3.' striking way. tiffs would seem to be a t3~ical phenomenon belonging to the temporal nature of living beings. How are we to understand this phenomenon? A simple approach could be expressed as follows, ageing is the change in the vital fimction of the past. The past ceases to be perfectly integratcxl into the present, stops behaving as an active reserve injecth~g the present with new life. and starts to reduplicate itself, that is. becomes the past as such. There is no call for confusion, even though this statement might seem to be a play on words. The past as such is also present, but it is present as the past. not as a life-giving sap. This is thus a situation in which a distance is established. and also something that seems to be a backward tum, but that really is not. as the "before" we have lived through never existed as" a past. Along these lines, ageing is also something new. This is because the reserve of energy which the past constitutes - the very body of the living being is its past - no longer receives or possesses the strength to produce something different, somet~lg new. When what is new is the lack of new things, the living being beghas to cease to live. to move away from the realms of the vital, he or she starts to die. Death is the moment at which the present of a being is only that being's past as such. Now nothing can change. Fomlerly flae force of the integration and .synthesis of time which make up life made it possible to remake one's own past. This is no longer the case. For this reason, death in this sense cannot be.followed by a judgement: death itself is the judgement. Up to this point we cannot say that "this life is like this"; but we can say tiffs at the moment of death.
Time and Psycho-Physical Integration
137
It is generally held that the death of living beings is one of the clearest indices if not actually the clearest - of temporality. Yet it would really be truer to say that the opposite is the case. as death is not a synthesis - the "time as a whole" we looked at above - but a summarv, and therefore it is a patent demonstration that time can be converted into "space", as the is of the judgement definitively places that being in space. The summary is the curvature of time which is no longer conanuing. This is why death has always been thought of in connection ~ith eternity, to the extent that the latter means ,an ecstatic or "spatial" consideration. Thus even when etemit3.' has to be regarded as "eternal life", it must be this not ,as a negation of time but ,as a negation of ageing. Thne flint does not grow old is consistent ~ith eternity. Ageing implies a kind of negation or negativity that attahls some power over vital energy: it resists it. Resistance dist,'mces living beings from their oxsrl lives. Life must be outgoing, must ahvays express itself outwards, it means going to or towards another place. The classical tradition maintahas an opposite vicw, which is that life is integration, energy feedback and, therefore, interiorization and autonomy. But this is a half-tntth, as it is not possible to interiorize without exteriorizing: in order to get inside onesel.fone at least has to be able also to go out. This going out means going to where one is not. But this not is only ageing if it is converted hlto rcsist,'mce which overcomes the original energy. There may thus be thr~ "diseases" of time, which correspond to its three moments of constitution. First of all. there may be an energy deficit, a low intensity, either initially or later, accompanied by a decrease or reduction in energy, and the being starts to fail to cope with the resistances he meets ~ith (ageing and death). Secondly there may be a poor fommlization of the impnL,'es, of the beats of time. Loss of the "measure", of the rh3~hm, drives the temporal being to the disorder because of the lack of regularity (madness mid lack of r h ~ m ) . Thirdly, the "mechanism" of relention of the past may fail, and if this happens the organism first of all loses force ,and later dies. even if it is still young and has potential energ3', because it can no longer endttre (weakness). This means, to return to the subject mentioned at the beghmhlg of this paper, that every, material being. ,and particularly every living being, should bear ~ithin it three e/.]bctive and divers'e dimensions, which are respectively responsible for the intensity o['energv, for its mea.wtre or rh3~amic fonu, ,and for its capacit3; for retention or duration. Time is fllree-dimensional. To end this section on the "levels" of time, we must refer to these levels from the "temporal" point of view. The point was made at the outset flint the "levels" .
138
R. Alvira
of time could be approached from the angle of the "orders of being", which is the spatial consideration that has been made up until now, and from the point of view of the passing of time, which is that of evolution. If we look at the material cosmos, a "stratified" world unfolds before our eyes, but we can only "read" this spatially (a giraffe is in a different stratum from a pine tree, and this in tum is in a different layer from, say, a cn.,stal of quartz - and because each one is different, it has a different time) or in its temporal appearance (one stratum emerges after another, even though -,and tiffs is important- all of them exist together). The evolutionary, progressive presentation of the forms of life permits the disappearance of certain kinds of being (such as the dinosaurs), but not the disappearance of flae "kingdoms of nature" which gradually appear. We can draw various conclusions from this. On the one hand. it confirms that duration is true. On the other, it also reinforces the truth of.space, as every new "reahn of nature" is a new space which in turn forms space with other realms. One may conjecture, moreover, that every. "realm" is a part of the universal whole, that is. that it is not self-sufficient, and that therefore it would only be able to die if the universal whole did so too. The reality of evolution may also confinn another thesis flint has been implicit in this study from file beginning, that is. that time is a qualitative entity, not merely a quantitative one. In fact. the time in which one "jmnps" from one realm of nature to another, or from one species to another, is not qualitatively equal to the other times proper to ever)., living being. IV. The human being possesses the highest level of consciousness of all the beings that populate this world. This enables them not only to know time and be able to say something about it. as I have tried to do here, but also to have ,an especially rich, modulated sense of#me. Feeling time is already a certain form of consciousness which many animals possess. Having a mental awarem'ss of this is more than a mere addition because, apart from flae qualitative jump this means, the intelligible consciousness assumes the sensor)' consciousness. Being aware of time means, even for the sensory consciousness, being above time. Being aware of the distortions in the sensory consciousness is even more: it shows a mastery of time on the part of a being - the human being - which is nonetheless temporal. This means that our underst,'mding of time is due to wlmt the sensor)., consciousness tells us about it. ,and it is coloured by the information this
Time and P~:ycho-Physical Integration
139
provides. But the mental consciousness can in turn influence the sensory consciousness, from the point of view both of medical action and of moral action. The purpose of these tavo kinds of action is to achieve better balance and integraaon in human life, that is, to find and generate the appropriate measure for the latter. The concept of measure is radical and cannot be superseded. From the medical point of view this tends to be expressed in terms of the notion of equilibrium, and from the moral one, in terms of integration. The moral-psychological meaning of time c-an also be seen in the dimensions described above. Regarding the origin, we live with a feeling of greater or lesser intensity, which we ourselves cml heighten or reduce. We are also aware of the more or less ordered rh)'thm, expressexl in one way or another, and of the more or less rapid cadence. Perhaps the most interesting aspect of all this lies here in our awareness of duration, which ranges from an apparent absence of the passing of time to its massive presence, these are the taro opposite poles betaveen wlfich lie numerous intervening stages. The apparent absence occurs in the happy lilb, which is always described as being a state of a certain su.[liciency ,and integration. The person who f~ls happy, while he is feeling like this. does not notice the passing of time, he is not aware of the duration of time, even though he is certain q/'Jt by connotation, .as being happy is living a fiilly human life, -and living also means the duration of
living. The massive presence of time occurs in the experience of boredom, which is the opposite of happiness. When we are bored, we become aware of the passing of a time which never passes and in which nothing happens, flint is. we experience flae pure duration of pure time. Thus as in happiness we have the sensation of e.,dstential lighmess, in boredom life seems like a weight. The human being hesitates when faced with this choice. Choosing happiness means losing the awareness of temporal duration and thereby ceashag to be its master. But choosing bor~om means ab,'mdoning what is best in life. For this reason, the most usual option is to choose a middle course in life -neither blissfully happy nor excessively boring- or to rake the road of oblivion. This can be achieved by the time-honour~ means, work, mnusement, drink, and SO 011.
Neither the via media nor the road of oblivion have enough power to integrate, and boredom is a weight because it exposes mina's life to a form of duration which is unsuit~ to it. Human time is richer in content and form.
140
R. Alvira
Failing to reach the happy life, therefore means, the distortion of the human being's last key to his ox~aatemporal nature. This will be followed by temporal disarray, the loss of vital intensit3.', and disorder in the physical tone. This is a general, relative ageing subsequent to the lack of happiness, or in other words, the absence of youth.
References
Agustine. Opera. In: P. Kn611. Editor. New York: Johnson reprint. 1986 Agustine. The confessions. Garden Cit3": Doubleday. 1960. Aristotelis. Opera.. 1. Bckker. Editor. Berlin: De Gruyter. 1960. Aristotelis. The complete works. J. Bames, Editor. Princenton: Princenton University Press. 1985. Bergson H. Oeuvres. Paris: Edition du Centenaire. 1959. Grimaldi N.Ontologie du temps. Paris: P.U.F. 1993. Heidegger M. Beitr~e zur Philosophic (Vom Ereignis). Gesamtausgabe, Band 65. V. Frankfurt a. M: Klostennann. 1971. Heidegger M. Sein und Zeit.. Gesamtausgabe, Band 2. V. Klostermann. Frankfurt a. M., 1977. Hegel G.W.F. Werke. Ed. Eva Moldenhauer and K.M. Michel. Frankfurt a. M: Suhrkamp Verlag. 1971. Jankelevitch V. L'aventure. rennui, le srrieux. Paris: Montaigne. 1963. Kant I. Werke. Hrgb. von dcr Krniglich Preussischen Akademie der Wissenschafien. Berlin. 1912. Llinas R.R. and Par6 D. Commentar3, of dreanfing and wakefulness. Neurosci. 1991; 44.
Time and Psycho-Physical Integration
141
Nichelli P. The neuropsycholog3, of human temporal information processing. In Handbook of Neuropsycholog3.'. F. BoUer and J. Grafmaneditos. 1993; 8. Plato. The dialogues. Yale University Press. New Haven. 1990. P6ppel E. Time perception. In Handbook of sensory Physiology. Berlin: Springer. 1978: 8. POppel E. Grenzen des Bex~llsstseins. Uber Wirklichkeit und Welterfahnmg. Stuttgart: Deutsche Verlags-Anstalt. 1988
This Page Intentionally Left Blank
Time, Internal Clocks and Movement M.A. Pastor and J. Artieda (Editors) 9 1996 Elsevier Science B.V. All fights reserved.
THE ROLE PROCESSES
OF
ATTENTION
IN
143
TIME
ESTIMATION
D A N Z A K A Y 1 and RICHARD A. BLOCK 2
IDept. of Psycholo,eo~, Tel-Aviv l lniversi~ 2Dept. of Psychology. Montana State l/niversi~-Bozeman ABSTRACT. Several cognitive models of time estimation have been proposed. We discuss the different role of attenl.ion in prospective and retrospective time estimation processes, the empirical evidences and methodological problems in the stud.x, of attention, and the relationship beBveen attention, temporal uncertailfly and lemporal relevance-toward a general theoretical framework for understanding time estimalion processes.
1. Introduction
Although the time dimension plays an important role in human life (Miehon, 1985), no single sense organ or perceptual system mediates psychological time. This is especially puzzling regarding durations in the range of milliseconds, seconds, and minutes, since the perception of short durations is essential in enabling the organism to develop a faithful representation of the immediate external environment and to respond to incoming stimuli. For example, a basketball player must accurately perceive a 3-s duration while waiting under the basket, because standing in that zone for more than 3s is a violation of the rules. The referee must also perceive the 3-s duration in order to determine if the player has violated the rule. Neither can use a watch, because they must simultaneously observe other players. More common everyday situations that involve perception of short durations are driving (e.g., while passing another car) and crossing a street in heavy traffic. Animals are also capable of responding to short durational information. (In this chapter, we use the term animals to refer to nonhuman animals). However, researchers have not identified a time-sensing organ in animals, and it is not clear what infonnation animals may use to make time estimates.
144
D. Zakay and R.A. Block
This state of affairs has led researchers to look for relationships between various biological rhythms (e.g., circadian rhythms, alpha brain waves, heart rate, body temperature) and duration estimates. Studies of this type, however, have not revealed any systematic relationship between biological cycles and short duration estimation of up to an hour. Aschoff (1984, 1993) found that people's estimates of intervals in the range of hours are related to their sleep-wake cycles, but he concluded on the basis of studies using a temporal isolation paradigm that their estimates of short time intervals in the range of seconds "represent a "personal tempo' which, apart from variations within the circadian cycle, is independent of the duration of wake time. The production of short time intervals remains completely unaffected by internal desynchronization" ( 1993. p. 155). Of the various physiological correlates investigated, only body temperature appears to influence duration judgments (Campbell and Bimbaum, 1994), albeit in somewhat inconsistent ways that are not clearly understood (Hancock, 1993). Of course, overall body temperature probably influences the functioning of many different brain areas. Left without a plausible biological clock that might mediate animal or human timing behavior, time researchers started to look for neurological or cognitive processes that could provide an adequate explanation. In this chapter, we review cognitive models of duration timing and offer an integrative model with more explanatory, power than existing models. We start by describing findings on human and animal duration timing. A section describing models of human and aninml duration timing follows. In the next two sections, we evaluate the role of attentional mechanisms in prospective duration estimation and review biopsychological evidence on the role of attention in duration timing. Finally, we propose an attentional-gate model of duration estimation.
2. Timing Behavior 2.1. HUMAN DURATION TIMING Duration estimation involves cognitive processes sensitive to the contextual conditions under which the estimation task is peffomled (e.g., Block. 1989: Zakay, 1990). Block argued that "'a complete understanding of any kind of temporal experience is possible only if we consider complex interactions
The Role of Attention in Time Estimation Processes
145
among all of these factors" (1989, p. 334). One of the most influential contextual factors is the duration-judgment paradigm. Subjects make estimates of short durations (in the range of seconds and minutes) prospectively when they know before a target interval Starts that its duration is to be estimated, or retrospectively when they do not know this until the target duration is terminated. In addition to the duration-judgment paradigm, other influential contextual factors are the estimation method (e.g., comparison, verbal estimation, reproduction, or production), the environmental conditions (e.g., external tempo in the form of light or noise), the physical context (e.g., the experimental room or time of experiment), the cognitive context (e.g., the type of cognitive processes required during a target interval), and the stimulus infommtion processing load during the interval. The importance and relevance of time for a subject is another contextual factor that must be considered (Block, 1989: Hicks, Miller. and Kinsboume, 1976" Zakay, 1990). The structure of events filling a target interval (e.g., a coherent homogenous or heterogeneous structure) and subject's expectation for an early or a late ending of the target interval also influence subjective duration (Jones and Boltz, 1989). Emotional states of a person involved in time estimation (e.g., anxiety) also have an effect Another factor that influences time estimation is the delay between the termination of an estimated interval and the start of the estimation process (Zakay and Fallach, 1984). Most phenomena that reflect the influences of these factors can be traced to the fimdamental distinction between prospective and retrospective time estimation. A recent recta-analysis (Block and Zakay, 1994) reveals a general tendency for prospective estimates to be longer than retrospective estimates. Since prospective and retrospective estimates often differ, they must involve somewhat different processes (Block. 1990, 1992: Brox~, 1985; Grondin and Macar, 1992 Hicks, Miller, and Kinsboume, 1976: Macar, Grondin, and Casini. 1994; McClain, 1983" Zakay, 1989. 1990. 1992a). One phenomenon that strongly reflects this distinction involves the relationship between estimated durations and information processing load required during the interval. In the context of retrospective estimation, this relationship is positive (Omstein, 1969" Zakay, 1989). However, in the context of prospective estimation, this relationship is negative (Curton and Lordhal, 1974: Fortin and Rousseau. 1987). The filled-duration illusion also demonstrates this distinction. In retrospective studies, estimated
146
D. Zakay and R.A. Block
durations of intervals filled with some nontemporal task that demands information processing are longer than estimates of empty intervals during which subjects were not required to perform any task (Coren, Ward, and Enns, 1993; Omstein, 1969). Other investigators have reported the opposite finding in studies using a prospective paradigm (Zakay, Nitzan, and Glicksohn, 1983). Another fundamental distinction between prospective and retrospective time estimates relates to the impact on duration timing of contextual changes and segmentation occurring during the time period. The more contextual changes occur, the longer are retrospective duration judgments (Block, 1978, 1982, 1989, 1990; Block and Reed, 1978). Similarly, retrospective duration judgments are longer to the extent that high priority events segment the duration (Poynter, 1983, 1989). In contrast, Zakay, Tsal, Moses and Shahar (1994) found that degree of segmentation does not influence prospective estimates. Thus, people are not necessarily accurate estimators of short intervals, and contextual factors influence their estimates. Nevertheless, there is a positive monotonic relationship between subjective and objective time (Allan, 1979), although the exponent of the psychophysical power function in humans may be slightly less than 1.00 (Eisler, 1976). In animals, children, and some abnormal adults, the exponent of the power function is lower than it is in nomaal human adults, perhaps as small as .5 (Eisler, 1980, 1984, 1995). The focus of attention, which can be diverted to or from the time dimension, has a major influence on timing behavior. For example, in a "watched-pot" experiment (Block, George, and Reed, 1980; Cahoon and Edmonds, 1980), subject's attention focuses on the passage of time, resulting in lengthening of subjective durations. This finding was obtained even under retrospective conditions, indicating that under some conditions attention is allocated to time even in a retrospective paradigm. 2.2. ANIMAL DURATION TIMING Contemporary behavioral psychologists typically explore timing by animals by investigating pigeons and rats during relatively short time periods of seconds and minutes. The general finding is that animals are sensitive to different stimulus durations and time-based schedules of reinforcement. One of the more common methods to explore animal timing is the peak procedure (Roberts, 1981), which uses a modified discrete-trials fixed-interval (FI)
The Role of Attention in Time Estimation Processes
147
schedule. A relatively long, variable interval separates each trial. At the start of each trial a stimulus, such as a light, is turned on. On most trials, the first response occurring after a FI (e.g., 30 s) has elapsed since the onset of the light is reinforced" then the light is turned off. On other trials, no reinforcement is delivered, and the light is turned off only a~er a relatively long period, usually at least twice the FI (e.g., 60 s). Averaged across many of the latter kind of trial, the typical response rate is an approximately bellshaped function of time since the start of the FI period. This timing behavior reveals a scalar property" Regardless of FI length, the average response rate at any time, expressed as a proportion of the peak rate. is a function of the proportion of the total duration that has elapsed. In other words, the normalized response-rate curve does not vary much from one FI length to another. The bisection paradigm also reveals the scalar property of animal timing behavior. In this paradigm, an animal is first exposed to two stimuli, one of which is shorter (S) in duration than the other, which is longer (L). Later. the animal is presented probe stimuli of variable durations, t, and has to perform a discrimination task. Responses to these probe stimuli tell us whether the animal perceives the duration of t as being closer to S or to L. A bisection point, defined as the duration that is equally likely to be perceived as "short" or as "'long." can be calculated. A typical finding is that the bisection point lies at the geometric mean of the S and L durations (Church and Deluty, 1977). 2.3. COMPARING HUMAN AND ANIMAL TIMING A direct comparison of human and animal timing behavior is not easy, because the paradigms and methods of measurement are different. One major difference is that humans can respond directly in time units, whereas in the case of animals only indirect inferences based on nonverbal responses are possible. One paradigm that enables a direct comparison is the bisection paradigm. Wearden (1991) conducted two experiments, using normal adult humans in an analogue of the bisection task. He found that the bisection point was at a duration just shorter than the arithmetic mean of the standard short and long durations. Allan and Gibbon (1991), however, reported that normal adults, like animals, bisect at the geometric mean. The picture is even more complicated, since Rodriguez-Girones and Kacelnik (1994) found that the
148
D. Zakay and R.A. Block
bisection point in humans lies somewhere between the arithmetic and geometric means. These findings indicate that human timing behavior is less rigid than that of animals and is more context-dependent. Indeed, Allan and Gibbon (1991), used longer, filled intervals, whereas Wearden (1991 ) used shorter, empty intervals. Nevertheless, Wearden (1993) claimed that human timing behavior exhibits appropriate scalar properties, as was indicated in experiments on interval production with concurrent ehronometric counting (Wearden and McShane, 1988). The scalar properties in humans probably do not depend on the characteristics of a reference memory store (RodriguezGirones and Kacelnik, 1994). Human timing behavior is undoubtedly more varied than animal timing behavior. A major factor responsible for this variability is attentional allocation. 2.4. MODELS OF TIMING BEHAVIOR Models of timing behavior can be categorized into two major categories: timing-with-a-timer models and timing-without-a-timer models (Ivry and Hazeltine, 1992). Timing-with-a-timer models propose a pacemaker mechanism. Timing-without-a-timer models propose that subjects construct psychological time from processed and stored information. Block (1990) discussed two timing-with-a-tinier modeis--chronobiological and internalclock models--and three kinds of timing-without-a-timer models--attentional, memory-storage, and memor),-change models. Treisman (1963) proposed one of the first formal timing-with-a-timer models, which asserted that an internal clock underlies human duration judgment. In this model, a pacemaker produces a regular series of pulses. The pulse rate varies as a fitnction of input from an organism's specific arousal center. Specific arousal is influenced by external events, in contrast to general arousal, which depends on internal factors such as those that underlie circadian rh31hms. A counter records the total pulse count, which is then transferred into a store and into a comparator mechanism. A verbal selective mechanism assists in retrieving infommtion from the store. This is a long-term memory, store containing knowledge of correspondences between total pulses and conventional verbal labels for time periods. Another timing-with-a-timer model is the one that underlies scalar timing theory,, or scalar expectancy theory. Many animal researchers have adopted this model, since it provides an excellent account of a wide variety of
The Role of Attention in Time Estimation Processes
149
evidence (Church, 1989). This model accounts for duration perception and production by proposing an internal clock, memory stores, and a decision mechanism. The internal clock consists of a pacemaker, a switch, and an accumulator. As in Treisman's model, the pacemaker generates more or less regularly spaced pulses. When the organism perceives an external timing signal indicating the beginning of a time period, the switch allows pulses to pass into the accumulator. The accumulator holds the total pulse count during the time period. Perceived duration is a monotonic function of the total number of pulses transferred into the accumulator. The contents of the accumulator are transferred into a working memory, store for comparison with the contents of the reference memory store. The latter contains longterm memor), representations of the approximate number of pulses that were accumulated in similar situations in the past. This number is then transferred to the comparator, which compare the total pulse count of the two stores. The major problem with the timing-with-a-timer models proposed by, Treisman and by scalar-timing theorists is that they do not acknowledge the major role of attentional processes. However, several other theorists have proposed models of psychological time in which attention to time, or temporal information processing, plays a major explanatory role. Thomas and his colleagues (Thomas and Cantor, 1975" Thomas and Weaver, 1975) developed and tested one of the most explicit of these, a mathematical model in which attentional allocation influences duration judgments. Although Thomas and his colleagues studied only human duration judgments of stimuli presented for less than 100 ms, the model is potentially a general model of temporal infommtion processing, even involving longer time periods (Michon, 1985). The model proposes that the perceived duration of an interval containing certain information is a monotonic function of the weighted average of the amount of information encoded by two processors, a temporal information processor and a nontemporal information processor. The organism divides attention between the two, which operate in parallcl. Perceived duration is weighted to optimize the reliability of the information that each processor encodes, because as more attention is allocated to one processor, the other becomes more unreliable. If little or no stimulus information occurs during the to-be-judged duration, the person allocates more attention to temporal information. If a task demands considerable information processing, the person allocates more attention to this nontemporal information.
150
D. Zakay and R.A. Block
3. The Role Of Attention In Prospective Time Estimation As we suggested earlier, only an attentional model that assumes a cognitive counter can provide a satisfactory explanation for prospective timing behavior in humans and for its distinctiveness from retrospective timing. In this section, we review empirical evidence supporting this claim. The typical paradigm that reveals characteristics of prospective duration estimation requires subjects to estimate the duration of either a complex or a simple task. If clock time is constant, people give longer prospective estimates of the duration of simple, rather than complex tasks. Simple tasks presumably require less processing effort and hence fewer attentional resources. A similar paradigm involves asking subjects to produce given durations while they concurrently perfoma either a simple or a complex task. Attentional models predict a positive relationship between length of produced durations and degree of task complexity. This prediction has been supported empirically (Zakay. 1993). However. such evidence does not provide a direct test of the assumed role of attention, since attentional allocation is not necessarily the only variable influenced by the taskcomplexity manipulation. Zakay (1989: Zakay and Block. 1995) manipulated attentional allocation more directly. One manipulation involved the secondary-task method that is commonly used to measure mental workload (e.g., Gopher and Kimchi. 1989). Differences in primary-task resource demands are assumed to be reflected in secondary, task performance (Wickens. 1984). Time estimation may be defined either as the primar3' or the secondary' task. If subjects are instructed to treat duration estimation as the primary task. their prospective duration estimates increase as compared to those for which duration estimation is the secondary task. presumably because they allocate more resources for that purpose. On the other hand. if subjects are instructed to treat duration estimation as a secondary task. their prospective duration estimates decrease. Support for the claim that attentional resource allocation was manipulated comes from the finding that when duration estimation was the primary task. performance of a simultaneous nontemporal secondary task was impaired compared with the performance of that same nontemporal task when it was the pri,nar3' task and duration estimation was the secondary task.
The Role of Attention in Time Estimation Processes
151
A second paradigm that manipulates even more directly the allocation of attention involves the latent inhibition (LI) method. LI is measured by the decremental effect on subsequent learning of nonreinforced preexposure of the to-be-conditioned stimulus (Lubow, 1973, 1989; Lubow and Gewirtz, 1995). For example, if a child is exposed to a stimulus several times, the subsequent efficiency of that stimulus as a conditioning stimulus decreases significantly. LI is a robust phenomenon that occurs in a variety of learning tasks and species, including various animals as well as human children and adults. Lubow (1973, Iq89) argued that a model of LI should include both attentional and learning constructs. Conditioned inattention theo~' (CAT) is such a model (Lubow, Wcincr and Schnur, 1981). CAT states that nonreinforced preexposure to a stimulus retards subsequent conditioning to that stimulus because during such preexposure the animal learns not to attend to it. A basic assumption of CAT is that with repeated stimulus presentation, the attcntional response inevitably declines and is replaced by a conditioned inattentional response. Indeed, prospective estimates of the duration of familiar stimuli arc longer than those of unfamiliar stimuli (Avant, Lyman, and Antes. 1975: Kowal, 1987). We may explain this by proposing that subjects allocate fewer attentional resources for processing a familiar stimulus than a novel stimulus. When a familiar stimulus occurs. more attentional resources arc available to be allocated for temporal information processing during the exposure interval, especially during the interval following identification of the name of the stimulus. A masking procedure was developed for inducing LI in adults. In this procedure, a stimulus is exposed many times together with another task demanding most of a subject's attcntional resources. The stimulus is mcrclv presented" the subject is not required to respond to it. The distracting task apparently initiates a process leading to conditioned inattention to the stimulus. Experiments clearly demonstrate the efficiency of the masking method for creating a very powerfi~l LI effect in human adults (Lubow and Gewirtz, 1995). This use of the LI paradigm in the frame of timing behavior relies on the following rationale: If a stimulus, previously conditioned to inattention in a masking task, is exposed and subjects are asked prospectively to estimate its exposure duration, attentional resources should not be used for analyzing the stimulus. Thus. more attentional resources will be available to be allocated to the duration estimation task rather than to the stimulus information processing task. If so, the prospective estinaate should be longer than that of
152
D. Zakay and R.A. Block
an identical objective exposure time of the same stimulus that was not conditioned to inattention, in the latter case, attentional resources should be divided between the processing of the stimulus and the duration estimation task. These predictions have been empirically supported, both for auditory and for visual stimuli, and with two different t~ges of masking tasks (Zakay 1989; Zakay and Block, 1995). Thus, it seems that the finding generalizes to different modalities and masking tasks. The role of attention in prospective duration estimation is apparently similar in children and in adults. Zakay (1992b) tested 7- to 9-year-old children's time estimation, using an attentional-distraction procedure. The children were required to reproduce the duration of illumination of a light bulb. When their attention was attracted away from the estimation task (by a jumping frog toy), prospective durations became shorter, although the illumination time was constant. When the distraction was eliminated, prospective estimates returned to their original length. These findings support the h.~9otheses stemming from the attentional model regarding the role of attention in prospective time estimation. They also show that it is useful to assume the existence of a functional cognitive timer requiring mental resources for its operation. Berlyne (1966) argued that upon the presentation of any event, the count in the timer is increased, and time estimation is associated with the value of it at the moment of estimation. However, concepts like "'focusing attention on the passage of time" or "temporal infommtion processing," which are used in current attentional models, need to be clarified. CONTROLLED 3.1 AUTOMATIC AND TEMPORAL INFORMATION
PROCESSING
OF
Attention to time has been a central construct of some time estimation research. Here, we consider attention to time to mean attention to temporal information. Some temporal infomlation is processed automatically. For example, temporal information about the relationship between two occurrences of a single stimulus, or that between occurrences of related stimuli, appears to be processed and stored automatically. Such infomaation is part of the memorial record of an event, even under incidental learning conditions (Hintzman, Summers, and Block, 1975). It is also likely that temporal information about well-learned motor sequences is used relatively automatically. The structure of an event also has been shown to affect
The Role of Attention in Time Estimation Processes
153
dynamic attending processes. These effects are most plausibly automatic. When an event has a coherent structure, attending to its temporal properties is facilitated, as compared with irregularly-timed events (Boltz, 1991. 19q2). In many other situations, however, people must deploy attention in order to process temporal infomlation (Jackson, 1985). In these situations. attending to time appears to demand access to the same attentional systems as does attending to stimulus information. If a person must divide attention between temporal infomaation about several events or intervals, the person's ability to keep track of this information suffers in proportion to the number of events or intervals involved (Brown and West, 1990" Macar, Grondin, and Casini, 1994). 3.2. PSYCHOBIOLOGICAL ATTENTION
EVIDENCE
ON
THE
ROLE
OF
Psychobiological evidence from both animal and human experiments clarifies the kinds of models discussed and proposed here (Block, in press: Church, 1989). Brain modules or areas that subserve the timer or internalclock mechanism are separate from, but interconnected with, those modules or areas that subserve the proposed role of attentional processes. Functioning of the internal clock or cognitive timer seems to rel.v mainly on the frontal lobes of the brain, especially the dorsolateral prefrontal cortex. Converging evidence from psychophannacological, electrophysiological, and lesion studies seemingly isolates the timer to this brain region. For example, researchers that have administered drugs to animals trained on FI schedules suggest that the timer may be subserved by dopaminergic neurons, which the prefrontal cortex is known to contain. It is unclear exactly which areas of the brain subserve attention to time. Studies using positron emission tomography suggest that scvcral anatomically separate areas of the brain, including the thalamus, the parictal lobes, and the anterior cingulate gyrus, are critically involved in various attentional aspects of task performance (for a review, see Posncr and Raichle, 1994). These areas also appear to play somewhat different roles. The likely candidate for an area that may involve the allocation of attention to external events or to time is the anterior cingulate cortex. This area sccms to be the central component of an executive attention network, which may directly control or influcnce the working-memory filnctions of the
154
D. Zakay and R.A. Block
dorsolateral prefrontal cortex. As such, it may subserve the role of attention in timing stimuli and durations.
4. Attentional-Gate Model
As stated earlier, an improved model of prospective time estimation should have scalar properties and should contain elements that can explain both automatic and controlled attentional influences on prospective duration estimation. This model should also explain other context-dependent timing behavior. In order to achieve these goals, we propose a model that combines features of Treisman's (1963) model and the scalar-timing model (Church, 1984: Gibbon, 1984). as well as Thomas's (Thomas and Cantor. 1975; Thomas and Weaver. 1975) model. We call this model the attentional-gate model (Block and Zakay, 1995: Zakay and Block, 1994). A major element of this model is a gate. The gate is a cognitive mechanism controlled by the allocation of attention to time. As an organism allocates more attention to time, the gate opens wider or more frequently. The complete model consists of a pacemaker, a gate, a switch, and a cognitive counter (see Figure 1). .;:-..-.-::Possible Connection Arouse'
Attenttonal Ailocstion to
['.-L2
c'c'"*"'*") i "I
Temporal Id caning
Time
.,,'.,,0..' I~ ~.,,~ *'"
I t
"Cognitive L...bI-Reference [ co.nt.,
j "[
.
L...~
Figure 1. Tile AttenlionaI-Gate Model.
.
.
.
Vq
uomo,y
.
Y,
.,,.~ 1
The Role o f Attention in Time Estimation Processes
155
The pacemaker produces pulses at a rate influenced by both general (e.g., circadian) arousal and specific (e.g., stimulus-induced) arousal. Treisman and his colleagues (Treisman, Faulkner, Naish, and Brogan. 1990: Treisman, 1993) have proposed adding to the pacemaker a calibration unit in order to handle the influences of arousal while keeping the rate of pulses emitted by the pacemaker constant. The need for such a calibrator has yet to be resolved by empirical data. On each occasion on which an organism attends to time, as opposed to external stimulus events, the attcntional gate opens more frequently, thereby allowing more pulses to bc transferred from the pacemaker to the cognitive counter. This is the case when time is important and relevant (Zakay, 1992a), especially under prospective conditions. When time is not relevant or when duration estimates occur under retrospective conditions, the gate narrows, allowing fewer pulses to pass through it. In this model, duration judgments involve counting the total pulses accumulated in the cognitive counter, a process that probably also requires attentional resources. There is still a need for a switch that opens or closes the pathway to the counter. The switch operates in an all-or-none fashion according to the temporal meaning of stimuli. When a stimulus signaling the beginning of a relevant interval is perceived, the switch opens the pathway, the counter is set to zero, and the flow of pulses begins to be accumulated. The switch closes the pathway when the organism perceives a signal indicating the temaination of a relevant interval, preventing additional pulses from entering the counter. When an estimation or a response is needed, the count is transferred to short-term memory'. The switch mechanism is a cognitive element needed to explain some of the contextual dependency of timing behavior. Since the switch is governed by the meaning system, the same signals can influence duration estimation in different contexts in a different way, according to the meaning assigned to the stimuli in the different contexts. Humans, unlike animals, are probably not dependent on reference memory, in producing temporally based responses (Rodriguez-Girones and Kacelnik, 1994). but can make use of it whenever necessary. The attentional-gate model contains important modifications to extant internal-clock models. It includes and elaborates the notion that a subjcct may divide attentional resources between attending to external events and attending to time, and it specifies the consequences of each. Attending to time is necessar3, for pulses to be transmitted to the cognitive counter. The number of pulses that are transmitted during a time period is therefore a
156
D. Zakay and R.A. Block
function of two factors" the pulse rate, which is influenced by general and specific arousal (of. Kahneman. 1973); and the proportion of time that the gate is open, which is detemlined by the amount of attention allocated to time. Thus, the influences of arousal and of attention are treated independently. The proposed model contains scalar-timing properties and explains complex human timing behavior. There are still many open questions regarding the structure of the model, such as the question of the exact role of the switch. The attentional-gate model must still be tested and validated by empirical studies.
5. Conclusions
In this chapter, we analyzed the role of attention in human duration timing. Empirical findings strongly indicate that attentional allocation influences judgments of short duration and that the role of attention is different under prospective and retrospective conditions. The role of attention is a core process under prospective conditions. In this paradigm, attention is divided or shared between temporal information processing and any nontemporai information processing that subjects need to perform during a to-beestimated duration. A previously unanswered issue concerns the functional meaning of focusing attention on time--that is, temporal information processing. We addressed this by proposing an attentional-gate model. The attentional-gate model relies on general models of attentional resource sharing to provide a cognitive mechanism that can explain the nature of the link between attention and time. Nevertheless, many important questions remain. For example, does attctation play a similar role in animal timing behavior'? A possible method for studying this question involves presenting ancillary stimulus information during the time period. Under retrospective conditions, subjects usually do not attend much to time unless the nontemporal information processing task is relatively easy or boring. Instead, the amount of attention that subjects devote to the nontemporal task during the target interval determines the amount of stimulus information perceived and encoded. When subjects later make a retrospective duration judgment, it is mainly influenced by the number of contextual changes that are associated and retrieved with this nontemporal
The Role of Attention in Time Estimation Processes
157
information. Thus, attention influences retrospective duration judgments more indirectly than it does prospective judgments. An important, as yet unresolved, issue concerns whether it is possible to formulate an integrative model for prospective and retrospective duration judgment. The attentional-gate model seems to need a pulse generator to signal the passing of time. Thus, in its present form, the attentional-gate model contains an internal clock. However, the attentional-gate notion is a stand-alone concept which does not necessarily depend on any specific source of time-related information, such as an internal clock. The source of time-related information may instead be a process that generates certain kinds of contextual associations (see Block and Zakay, 1995). Cognitive neuroSci, research, along with additional cognitive studies of duration judgment, may ultimately resolve this issue.
ACKNOWLEDGEMENTS. Preparation of this chapler was supported b.va grant from the United Slales-lsracl Binalional Sci. Foundalion.
6. References
Allan LG. The perception of time. Percept. Psychophys. 1979' 26: 340-54. Allan LG, Gibbon J. Human bisection at the geometric mean. Learn. Mot. 1991" 22: 39-58. AschoffJ. Circadian timing. Ann. N. Y. Acad. Sci. 1984" 423:442-68. Aschoff J. On the passage of subjective time in temporal isolation. Psychol. Belgie.. 1993" 33(2)' 147-58. Avant LL, Lyman PJ and Antes JR. Effect of stimulus familiarity upon judged visual duration. Percept. Psychophys. 1975 17' 253-62. Berlyne DE. Effects of spatial order and inter-item interval on recall of temporal order. Psychonomic Sci. 1966: 6: 375-76.
158
D. Zakay and R.A. Block
Block RA. Remembered duration: Effects of event and sequence complexity. Mem. Cognit. 1978; 6" 320-26. Block RA. Temporal judgments and contextual change. J. Experimental Psychol." Learn. Mem. Cognit. 1982; 8: 530-44. Block RA. Experiencing and remembering time: Affordances, context, and cognition. In: Levin I, Zakay D, editors. Time and human cognition: A lifespan perspective. Amsterdam" North Holland. 1989" 333-63. Block RA. Models of psychological time. In: Block RA, editor. Cognitive models of psychological time. Hiilsdale, N J" Erlbaum. 1990: 1-35. Block RA. Prospective and retrospective duration judgment: The role of information processing and memory. In: Macar F, Pouthas V, and Friedman, WJ, editors. Tilne, action and cognition: Towards bridging the gap. Dordrecht: Kluwer Acad. 1992" 141-52. Block RA. Psychological time and memory systems of the brain. In: Fraser JT, Soulsby MP, editors. Dimensions of time and life: The study of time, VIII. Madison, CT: International Universities Press, in press. Block RA, George EJ and Reed MA. A watched pot sometimes boils" A study of duration experience. A. Psychologica. 1980: 46:81-94. Block RA, Reed MA. Remembered duration" Evidence for a contextualchange hypothesis. J. of Experimental Ps.x'chol." Human Learning and Memory. 1978.4: 656-65. Block RA, Zakay D. Prospective and retrospective duration judgment: A meta-analysis. Poster presented at the meeting of the Western Psychological Association, Kailua-Kona, HI. 1994. Block RA, Zakay D. Cognitive models of time revisited. Manuscript in preparation. 1995. Boltz M. Time estimation and attentional perspective. Percept. Psychophys. 1991" 49(5): 422-33.
The Role of Attention in Time Estimation Processes
159
Boltz M. The incidental learning and remembering of event durations. In' Maear F, Pouthas V and Friedman WJ, editors. Time, action and cognition. Dordrecht: Kluwer Acad. 1992 153-63 Brown SW. Time perception and attention" The effects of prospective versus retrospective paradigms and task demands on perceived duration. Precept. Psychophys.. 1985 38 i 15-24. Brown SW, West AN. Multiple timing and the allocation of attention. Acta Psychologica. 1990: 75' 103-2 I. Cahoon D, Edmonds EM. The watched pot still won't boil" Expectancy as a variable in estimating the passage of time. Bull. Psychon. Soc. 1980: 16' 115-16. Campbell SS, Bimbatm~ JM. Time flies when you're cool" Relationship between body temperature and estimated interval duration. Sleep Res.. 1904' 23" 483. Church RM. Properties of the internal clock. Ann. N.Y. Acad. Sci. 1984" 423" 566-82. Church RM. Theories of timing behavior. In: Klein SB, Mowrer RR, editors. Contemporary learning theories Instrumental conditioning theory and the impact of biological constraints on learning. Hillsdale, NJ: Edbaum. 1989:41-71. Church RM, Deluty MZ. Bisection of temporal intervals. J. Exp. Psychol.' Animal Behavior Processes. 1977' 3" 216-28. Coren S, Ward LM and En.ls JT. Sensation and Perception, 4th edition. New York: Harcourt Brace Collcge. 1993. Curton ED, Lordahl DS. Effects of attentional focus and arousal on time estimation. J. Exp. Psychol. 1974: 103 861-67. Eisler, H. Experiments on subjective duration 1868-1975" A collection of power function exponents. Psychol. Bull. 1976; 83(6)" 1154-71.
160
D. Zakay and R.A. Block
Eisler H. Psychophysical similarities between rats and humans. Bull. Psychon. Soc. 1980:16:125-27. Eisler H. Subjective duration in rats: The psychophysical function. Ann. N.Y. Acad. Sci. 1984; 423:43-51. Eisler H. Time perception from a psychophysicist's perspective. Manuscript in preparation. 1995. Fortin C, Rousseau R. Time estimation as an index of processing demand in memory search. Precept. Psychophys.. 1987: 42(4): 377-82. Gibbon J. Scalar timing in memory. Ann. N.Y. Acad. Sci. 1984" 423" 5277. Gopher D, Kimchi R. Engineering Psychol. An. Review Psychol. 1989. 40" 431-55. Grondin S, Macar F. Dividing attention between temporal and nontemporal tasks" A performance operating characteristic - POC - analysis. In: Macar F, Pouthas V and Friedman WJ, editors. Time, action and cognition. Dordrecht: Kluwer Acad. ! 992" I 19-28. Hancock PA. Body temperature influence on time perception. J. Gen. Psychol. 1993:120:197-216. Hicks RE, Miller GW and Kinsboume M. Prospective and retrospective judgments of time as a function of amount of information processed. American J. Psychol. 1976' 89" 719-30. Hintzman DL, Summers JJ and Block RA. Spacing judgments as an index of study-phase retrieval. J. of Experimental Psychol." Hum. Learn. Mem.1975: 104(1): 31-40. Ivry RB, Hazeltine RE. Models of timing-with-a-timer. In: Macar F. Pouthas V, and Friedman WJ. editors. Time, action and cognition" Towards bridging the gap. Dordrecht, Netherlands: Kluwer Acad. 1992" 183-89.
161
The Role o f Attention in Time Estimation Processes
Jackson JL. Is the processing of temporal information automatic or controlled? In" Michon JA, Jackson JL, editors. Time, mind and behavior. Berlin: Springer-Verlag. 1985" 179-90. Jones MR, Boitz M. Dynamic attending and responses to time. Review 1989; 96(3): 459-91. Kahneman D. Attention and effort. 1973.
Psvchol. .
Englewood Cliffs, NJ" Prentice-Hall.
Kowal KH. Apparent duration and numerosity as a function of melodic familiarity. Precept. Psychoph.vs.. 1987: 42(2): 122-31. Lubow RE. Latent i~fllibition. Psvchol. Bull.. 1973" 79: 398-407. Lubow RE. Latent inhibition and conditioned attention theory. Cambridge, UK" Cambridge University Press. 1989. Lubow RE, Gewirtz JC. Latent inhibition in humans: Data, theory, and implications for schizophrenia. Psychol. Bull.. 1995" 117(!): 87-103. .
Lubow RE, Weiner I and Schnur P. Conditioned attention theory. In" Bower G, editor. The Psychol. of learning and motivation, vol. 15. New York" Acad. Press. 1981. Macar F, Grondin S and Casini L. Controlled attention sharing influences time estimation. Mem. Cognit. 1994" 22(6): 673-86. McClain L. Interval estimation' Effect of processing demands on prospective and retrospective reports. Precept. Psychophys.. 1983" 34(2): 185-89. Michon JA. The compleat time experiencer. In: Michon JA, Jackson JL, editors. Time, mind and behavior. Berlin: Springer-Verlag. 1985" 20-52. Omstein RE. On the experience of time. Penguin. 1969.
Hamlondsworth, England:
162
D. Zakay and R. A. Block
Posner MI, Raichle ME. hnages of mind. San Francisco: WH Freeman. 1994. Poynter WD. Duration judgment and the segmentation of experience. Mcm. Cognit. 1983" 11(1): 77-82. Poynter WD. Judging the duration of time intervals: A process of remembering segments of experience. Inferring time's passage. In: Levin I, Zakay D, editors. Time and human cognition" A life-span perspective. Amsterdam: North-Holland. 1989' 305-31. Roberts S. Isolation of an internal clock. J. Exp. Psychol." Animal Behavior Processes. 1981" 7:242-68. Rodriguez-Girones MA, Kacelnik A. Interval bisection with and without reference memory. Proceedings: Time and the Dynamic Control of Behavior. Li6ge, Belgium. 1994: 1-24. Thomas EAC, Cantor NE. On the duality of simultaneous time and size perception. Precept. Psychophys.. 1975" 18" 44-48. Thomas EAC, Weaver WB. Cognitive processing and time perception. Precept. Psychophys.. 1975: 17: 363-67. Treisman M. Temporal discrimination and the indifference interval" Implications for a model of the "internal clock". Psychological Monographs. 1963" 77 (13, whole No. 576): 1-13. Treisman M. On the structure of the temporal sensory system. Psychologica Belgie. 1993" 33" 271-83. Treisman M, Faulkner A. Naish PL and Brogan D. The internal clock: Evidence for a temporal oscillator underlying time perception with some estimates of its characteristic frequency. Perception. 1990; 19" 705-43. Wearden JH. Human perfommnce on an analogue of an interval bisection task. Q. J. Exp. Psychol. 199 I" 43B: 59-81.
The Role o f Attention in Time Estimation Processes
163
Wearden JH. Decisions and memories in human timing. Psychologica Belg. 1993" 33(2)" 241-53. Wearden JH, McShane B. Interval production as an analogue of the peak procedure" Evidence for similarity of human and animal timing processes. Q. J. Exp. Psychol. 1988" 40B: 363-75. Wickens CD. Engineering Psychol. and human performance. Columbus" Charles E. Merrill. 1984. Zakay D. Subjective time and attentional resource allocation: An integrated model of time estimation. In: Levin I, Zakay D, editors. Time and human cognition: A life-span perspective. Amsterdam' North Holland. 1989: 36597. Zakay D. The evasive art of subjective time measurement" Some methodological dilcnunas. In" Block RA, editor. Cognitive models of psychological time. Hilisdalc. NJ' Eribaum. 1990: 59-84. Zakay D. On prospective time estimation, temporal relevance and temporal uncertainty. In' Macar F. Pouthas V., and Friedman WJ., editors. Time, action and cognition. Dordrccht: Kluwer Acad. 1992a: 109-18. Zakay D. The role of attention in children's time perception. J. of Exp. Child Psychol. 1992b: 54:355-71. Zakay D. Time estinaation methods. Do they influence prospective duration estimates'? Perception. ! 993" 22" 9 I- 101. Zakay D, Block RA. A functional model of the cognitive timer. Paper presented at the meeting on Time and the Dynamic Control of Behavior. Li6ge, Belgium. 1994. Zakay D, Block RA. The role of attention in prospective duration estimation. Manuscript in preparation. 1995. Zakay D, Fallach E. Immediate and remote time estimation: A comparison. Acta Psychologica. 1984 57 69-81.
164
D. Zakay and R.A. Block
Zakay D, Nitzan D and Glicksohn J. The influence of task difficulty and external tempo on subjective time estimation. Precept. Psychophys.. 1983" 34(5)" 451-56. Zakay D, Tsal Y, Moses M and Shahar I. The role of segmentation in prospective and retrospective time estimation processes. Mem. Cognit. 1994: 22(3): 344-51.
Time, Internal Clocks and Movement M.A. Pastor and J. Artieda (Editors) O 1996 Elsevier Science B.V. All fights reserved.
165
RECONSTRUCTION OF SUBJECTIVE TIME ON THE BASIS OF H I E R A R C H I C A L L Y ORGANIZED PROCESSING SYSTEM ERNST POPPEL
Institut far Medizinische Psychologie, Ludwig-Maximilians-Universit~it, Goethestr. 31/1. 80336 Manchen and Helmholtzzentrum JtTlich, 52425 Jfilich. Germany ABSTRACT. Two distinct levels of temporal processing have to be distinguished. A high-frequency mechanism is responsible for systems states with a duration of approximately 30 milliseconds. Theses systems states are apparently implemented by neuronal oscillations. A low-frequency mechanism is operative in the domain of 3 seconds. Automatic and presemantic temporal binding operations define these longer-lasting systems states, which can be used to define states of being conscious (STOBCON). Evidence for these two distinct and hierarchically linked processing s3'stems comes from ps3chophysical,', neuropsychological, and neurophysiological experiments.
1. Introduction
The continuity of subjective time is one of the most misleading concepts in the cognitive and neurosciences. This concept probably goes back to Isaac Nexs~on and his overwhelming influence on the development of physics and, thereby, also on psychophysics. One underlying idea in psychophysics is that subjective reality is a direct reflection of objective reality. By understanding the transformation rules between objective and subjective reality as expressed in the psychophysical laws, one can easily reconstruct subjective reality (e.g., Stevens, 1951). Regarding subjective time, we have to argue accordingly, and we should go back directly to Newton, who stated in the beginning of his Philosophiae Naturalis Principia Mathematica under Definitions:
166
E. Pi)ppel
"I do not define time, space, place, and motion, as being well known to all. Only I must observe, that the common people conceive those quantities under no other notions but from the relation they bear to sensible objects. And thence arise certain prejudices, for the removing of which it will be convenient to distinguish them into absolute and relative, true and apparent, mathematical and common. I. Absohtte, true, and mathematical ame, o f itself and from its own nature, flows equably without relation to anything external, and by another name is called duration: relative, apparent, and common time, is some sensible and external (whether accurate or unequable) measure of duration by the means of motion, which is commonly used instead of true time; such as an hour, a day, a month, a year." The central statement (in italics) refers to the general nature of time, continuity being one of the main features of physical processes as observed on the macroscopic level ("time flows equably"). Subjective time in the psychophysical tradition has erroneously also been considered a continuous phenomenon. However, if one analyses subjective phenomena in different temporal domains, one is impressed by the great number of experimental observations on discontinuous information processing. Apparent continuity of time is a secondary phenomenon - actually an illusion - which is made possible by discrete information processing on different temporal levels. Experimental evidence suggests the existence of at least two independent processing systems that are characterized by discrete time sampling and that are hierarchically linked with one another. Both these systems are thought to be fundamental for perceptual acts, cognitive control, and motor planning. First, I shall present observations that suggest the existence of a highfrequency processing system generating discrete time quanta in the domain of approximately 30 milliseconds. Then I shall address a low-frequency processing system, which is operative in the domain of 2 to 3 seconds. Evidence for the two temporal processing domains derives mainly from psychophysical, neurops.x'chological and neurophysiological experiments. Thus, arguments are based on a broad perspective taking into consideration qualitatively different experimental paradigms.
2. The domain of 30 milliseconds
Evidence for a distinct high-frequency mechanism derives from studies on temporal order thresholds (Hirsh and Sherrick, 1961; von Steinbiichel,
Reconstruction of Subjective Time
167
1995). If subjects are asked to indicate the sequence of two sensory events, temporal order thresholds in the domain of approximately 30 ms are observed independent of the sensory modality. To measure auditory order thresholds, two click stimuli are presented binaurally to a subject. If the stimuli are presented simultaneously, the subject will fuse the stimuli perceptually so that only one stimulus is heard in the center of the head (Lackner and Teuber, 1973). A delay of one stimulus against the other results in hearing the two clicks separately in each ear if the interval between the two is 2 to 3 milliseconds. Although the subject hears two clicks and might even know they are no longer simultaneous, he will not be able to indicate their temporal order correctly. The delay between the two clicks must be approximately 30ms before a subject can reliably indicate the correct sequence. Interestingly, similar values are observed both for the visual and the tactile modality in analogous experimental situations. This similarity of temporal order thresholds in the different sensory systems is particularly noteworthy because it indicates that the temporal order threshold is probably based on a different neuronal process than the temporal fusion threshold. I suggest that fusion thresholds are a direct reflection of the peripheral transduction processes of the different sensory systems, which are characterized by different time constants. The transduction process in the auditory, system is shorter (up to 1 ms) compared to the transduction processes in the visual or tactile system. Because of this temporal difference in transduction, the auditory fusion threshold (a few milliseconds) is considerably shorter than the critical flicker-fusion threshold in the visual modality (tens of milliseconds) (Pbppel et al, 1975). It is obvious that temporal order thresholds must be longer than fusion thresholds because independent representations of stimuli are a necessary condition to define their sequence. Interestingly, such an independent representation is not sufficient to allow the mental construction of temporal order, as experiments on auditory order threshold indicate. The auditory order threshold is one order of magnitude longer than the auditory fusion threshold. It seems that temporal order is made possible by certain operating characteristics in the thalamo-cortical network. Because of the structural similarity between cortico-thalamic pathways of the different sensory systems and because a circular network with inhibitory and excitatory connections always shows oscillatory characteristics, it can be hypothesized that the visual, auditory and tactile modality not only share the same
168
E. POppel
temporal order threshold, but that these might be based on neuronal oscillations. Order thresholds indicate directly that temporal processing is discontinuous. Different stimuli which are processed within a temporal window of approximately 30 ms are treated as co-temporal, i.e., a temporal relationship with respect to the before-after dimension cannot be established for such stimuli. Information gathered within a temporal window of 30 ms is treated as a-temporal, i.e., there is no temporal continuity defined and definable for stimuli that follow each other within such intervals. Time in the Ncwtonian sense (see above) does not exist on an experiential level for intervals shorter than appro• 30 milliseconds. This statement does not imply that the central nervous system cannot process information with smaller intervals than 30 ms (for example, the localization of objects in auditory space requires a much higher temporal resolution). However, to establish distinct events that are related to each other such that their temporal order can be indicated, the shortest temporal interval is set in the domain of approximately 30 milliseconds. Before turning to a discussion of neuronal oscillations, which are thought to be essential for structuring subjective time in a discrete fashion, some other observations from psychophysical experiments supporting the notion of temporally distinct processing stages are appropriate. Interesting evidence comes from a certain class of experiments on reaction time. It has been shown that, under stationa~ conditions, response distributions of reaction time show multimodal characteristics (Hatter and White, 1968, P6ppcl, 1970). This has been observed in choice reaction time, when subjects have to selectively press a response button when either a visual or an auditory stimulus is presented. The distinct modes in the response distributions are separated by 30 to 40 milliseconds (llmberger, 1986, Jokeit, 1990). It has been hypothesized that these distinct modes represent successive and discrete decision-making steps being processed in central neuronal populations (V6ppel, 1994). Multimodal response distributions have also been observed in measurements on response latencies of saccadic and pursuit eye movements. If a moving visual stimulus initiates pursuit eye movements, the latency of such eye movements has a strong tendency to be multimodally distributed with 30 to 40 ms intervals between ncighbouring modes (P6ppel and Logothetis, 1986). Similar observations on multimodal latency distributions
Reconstruction of Subjective Time
169
have been made on saccadic eye movements for human subjects (Frost and P6ppel, 1976; Ruhnau and Haase, 1993) and other primates (Fuchs, 1967). It should be pointed out that multimodalities as referred to here are not always observed in standard experimental situations. There are many reasons why this is so. They can all be summarized under one heading sta#onarity. Practically all psychological variables underlie massive effects resulting from fatigue, lack of concentration, learning, diurnal variation, etc. These modulating variables are a rich source for instaaonari~ of experimental observations. As the distinct temporal effects referred to above are in the millisecond domain, these non-systematic effects might easily overshadow the temporal processing proper unless stationari~ conditions are meticulously controlled. Other instationarities may come from the statistics usually employed for the interstimulus intervals (P6ppel et al, 1990 a; Ruhnau and P6ppel, 1996). If one wants to present successive stimuli in a random fashion, a rectangular distribution of stimulus occurrence for predefined interstimulus intervals is often chosen (e.g., a stimulus is presented randomly with equal probability either after 2, 3 or 4 seconds). The statistics leads to a systematic change in subjective probabilities for stimuli occurring late in an interstimulus interval. If a stimulus has not occurred early on, its probability of occurring later necessarily has to increase. As probability of stimulus occurrence is negatively correlated with reaction time, one necessarily introduces insmtionarity because of the rectangular distribution of stimulus presentation. This effect can be overcome by choosing an exponential decay of stimulus presentation, which results in a constant subjective probability throughout the interstimulus interval, and, thus, introduces the necessary
stationarity. The multimodal response distributions of reaction time for eye movements and the observations on temporal order thresholds can be explained on the basis of excitability cycles or neuronal oscillations (POppd 1970, POppel et al, 1990b). According to this modal, if a supra-threshold stimulus comes into contact with the sensory surface, a neuronal oscillation with a period of approximately 30 ms is initiated. This oscillation is (technically speaking) conceived of being a relaxation oscillation and not a pendulum oscillation (Wovor, 1965). Rdaxation oscillations can be triggered almost instantaneously by external stimuli, or their phase can be reset with only minor delay. Furthermore, relaxation oscillations fade out after several periods unless they are rotriggered. This fading can be observed oxperi-
170
E. Pi~ppel
mentally in a progressive reduction in response amplitude. In the extreme case, only two or three successive periods are observed because of high damping constants characterizing the system. In addition, the period of relaxation oscillations is not stable as in pendulum oscillations, i.e., the period can vary depending on the special properties of the system or due to environmental conditions. Thus, if I refer to processing periods of 30 ms, this does not mean a fixed number, like a physical constant, but indicates a temporal domain of periods in the range of 20 - 50 milliseconds. These processing periods set up by neuronal oscillations represent temporal quanta within which - as mentioned before - information is treated as co-temporal and, therefore, a-temporal. I believe that this interpretation is best exemplified by the experiments on order threshold mentioned above. The domain of perceptual successiveness can only be entered beyond 30 ms, because only a~er approximately 30 ms is it possible to determine temporal order. This means that information within this temporal limit is still in the domain of nonsuccessiveness. This interpretation is supported by choice reaction-time studies. In a choice situation, neuronal processes of the brain apparently have to go through discrete temporal phases within which new constellations are set up as reflected in decisions. These are then fed into the motor response system. In 1966, Sternberg observed that the exhaustive scanning process through short-term memory is discontinuous, with approximate step durations of 30 to 40 ms. Up to now, I have mentioned observations from psychophysical experiments that provide a basis for the theoretical argument, but there are also other observations deriving from neurophysiological experi- ments supporting the notion of discrete temporal processing. Experiments on the auditory evoked potential demonstrate that the midlatency response shows a clear oscillatory component in the first 100 ms after stimulus onset (Galambos et al, 1981; M~kel~i and Hari, 1987). After an initial phase of brain-stem activity, an oscillation sets in with high interindividual stability. Similar oscillatory, responses can be seen in the other sensory modalities (Eckhom et al, 1988; Gray et al, 1989; Murthy and Fetz, 1992). It has been suggested that oscillations in the 40-Hz domain are essential for cognitive mecha- nisms (Llinas and Ribar3,, 1993), and it has been demonstrated that oscillatory components in the midlatency region of the auditory evoked potential are sensitive markers for wakefulness in contrast to anaesthesia (Madler and P/Sppel, 1987). When patients receive general anaesthetics, one can still observe the brain-stem response, but the mid-latency oscillatory
Reconstruction of Subjective Time
171
activity disappears. If oscillator3, activity is preserved under anaesthesia patients still can process sensory information, although ot~en only implicitly. It can be hypothesized that general anaesthetics selectively destroy the temporal coherence within neuronal populations in the cortico-thalamic pathway. It is as if a temporal glue between neurones were lacking, which is a necessary condition for reliably processing sensory information. Patients going through this physiological state when the oscillatory responses have disappeared have been subjectively in a phase of a-temporali~. Such patients oRen ask aRer an operation, "When will the operation start?", which indicates that, unlike sleep, the brain is processing no information at all. Nothing has happened, and no information has been made available to central processing mechanisms of the brain. (It should be stressed that the disappearance of the oscillatory components of the midlatency response is only observed aRer general anaesthesia. Some receptor-specific anaesthetics even lead to an increase in oscillator3, activity; patients in such a state oRen implicitly process information (Schwender et al, 1994). Oscillator3, activities can also be observed on the level of single cells. Such stimulus-entrained activities have been described for the visual, auditory, and somato-sensory system. In particular, work on the visual system by the Russian scientist Podvigin and his group during the last decades should be mentioned (for an overview see: Podvigin and P6ppel, 1994). It was demonstrated that single units in the lateral geniculate body of the thalamus show oscillatory activities, which is not surprising if these oscillations are a network property, of a thalamo-cortical pathway. Presumably, oscillatory activity represents an intrinsic property of cortical neurones, as has been demonstrated for a certain class of units (Llinas et al, 1991). However, oscillatory activity may be a general property of neurones altogether. After firing an action potential, conductance changes are also observed in the dendritic tree, which results in a transient alteration of the computational properties for afferent information (Stuart and Sakmann, 1994). Because of the fixed morphology of the cell and its dendritic tree, such a rapid modulation of sensitivity could be temporally constant. Thus, an action potential could introduce sensiti- vity changes in the cellular activity, resulting in a tendency toward oscillatory activity. This overview provides some factual information about discrete time sampling in the 40-Hz domain. There also exist theoretical considerations about the practicability and even the biological neces- sity of processing neuronal information discontinuously. I wish to present two such theoretical
172
E. POppel
arguments in support of this conclusion, one derived from sensory analyses, and one based on the central mode of functional representation. As already mentioned, transduction processes in the different sensory systems are characterize~ by different time constants. This biophysical fact leads to a logistical problem for the brain if information from the different sensory channels is to be integrated. An object in space might for instance be defined by visual and auditory information - like a person to whom one is talking. As transduction in the auditory system takes much less time than in the visual system, central availability of auditory information is earlier than visual information. A difference of this kind would, however, only apply to near objects, when the time the sound-wave takes from its source can be neglected. At a distance of approximately l0 meters, the transduction time in the retina corresponds to the time a sound-wave takes (P6ppcl ct al, 1990 b). Beyond this horizon of simultaneity, the central availability of visual stimuli will be prior to that of auditory stimuli. For an object moving in space and changing its distance in relation to the observer, the central availability of stimuli in the two sensory modalities is highly unpredictable. Both the physics (light vs. sound velocity) and the biophysics (differential transduction times) result in permanent shifts of central temporal availability in the two sensory systems. With respect to sensory, integration allowing central fusion of two sources of information, such unpredictability could be overcome logistically by introducing temporally neutral zones within which information is treated as cotemporaneous. Such zones could be set up by neuronal oscillations, one period representing an interval within which information from various sources is collected indiscriminately from its source. An important argument for such a collective mechanism comes from the demands of the motor system. To program a movement, information from different sensory channels often has to be integrated. This can be done most efficiently if temporally neutral zones are introduced in which information coming from various sources is integrated, defining a distinct physiological state. Especially ballistic movements, whose neuronal program has been completed before the beginning of the movement, would profit from such a mechanism. I believe that the different experimental results described above reflect this mechanism, which fulfils theoretical considerations. A qualitatively similar argument suggesting the necessity of discrete processing states comes from observations on the localization of functions. One essential conclusion that can be drawn on the basis of anatomical
Reconstruction of Subjective Time
173
studies and neurological observations is that elementary functions are locally represented. Anatomical studies on the visual system for instance have clearly demonstrated a spatial segregation of functions (Zeki, 1978). Similarly, neuropsychological observations strongly support the notion of a spatially distributed representation of elementary functions (e.g., P6ppel, 1989). It has to be stressed that each functional state is characterized by the simultaneous activity of different neuronal modules, as has been demonstrated by PET-studies on pain perception (Talbot et al, 1991). Distributed and simultaneous activity require a strict temporal regimen to guarantee functional competence. As in the co-ordination of sensory input, systems states could serve a useful function in temporal co-ordination of distributed activities, and such system states could be implemented by neuronal oscillations. Simulation studies based on theoretical considerations have actually demonstrated the essential role oscillations could play in functional co-ordination (Sporns et al, 1989; Tomoni et al, 1992). Neuronal oscillations in the 40-Hz domain may actually represent spatial binding of distributed activities. In this case, oscillations represent binding, i.e., equal frequency and identical phase of an oscillatory activity is the expression of binding of spatially-distributed activities (Gray et al, 1989). The model suggested here is different: Neuronal oscillations in the 40-Hz domain are triggered by exogenous and endogenous events and provide a temporal framework within which binding operations are implemented by distinct neuronal algorithms. These oscillations thereby represent a necessary condition for binding and not the binding operation itself.
3. The domain of 3 seconds
Whereas the high-frequency mechanism discussed above is thought to logistically organize distributed neuronal activities and to implement primordial events, a low-frequency mechanism appears to integrate successive events within a temporal window of approximately 2 to 3 seconds. One of the most convincing experiments on temporal integration comes from studies on the reproduction of visual and auditory stimuli of different durations (P6ppel 1978; Elbert et al, 1991). This psychophysical technique is nonverbal, i.e., stimuli are presented and have to be reproduced independent of a verbal reference. This language independence of the reproduction technique is shared by the psychophysical technique of
174
E. POppel
comparison, in which stimuli of different durations have to be compared with respect to their duration (Hornstein and Rotter, 1969). Often in studies on time estimation the psychophysical techniques of production and verbal estimation are employed (e.g., Aschoff, 1992; Rubia et al, 1996). These exercises require knowledge of temporal units (like seconds) and might sometimes suffer from their verbal reference because this reference can contribute nonsystematically to an increase of the observed variance. If one wants to understand the data-generating mechanism of central timing, it is therefore often preferable to use the nonverbal techniques, like reproduction of temporal intervals. If visual or auditory stimuli of different durations are presented, subjects reproduce these stimuli veridieally up to intervals of approximately 3 seconds. Longer stimuli are reproduced shorter and with much higher variability. Often stimuli of up to about 3 seconds are reproduced slightly longer than the stimulus, probably due to a reaction-time component. The longer reproduction of short intervals and the short reproduction of long intervals results in a transition from over- to underestimation. The transition point has been referred to as indifference point. It has been argued that the indifference point is an experimental artefact caused by the great number of reproductions required by a subject in a t)gical experimental session. On the basis of the adaptation-level theory (Helson, 1964), it has been suggested that the indifference point corresponds to the geometric mean of all stimulus durations which are processed by a subject during an experimental session. Because this idea implies that the indifference point might be explained as an artefact due to the procedure employed, experiments have been performed in which subjects were required to reproduce only one stimulus duration, and this was done in two different temporal domains, namely between 0.5 and 5 seconds and between 10.5 and 15 seconds (P0ppel, 1978). Results demonstrated a clear interindividual indifference point at approximately 3 seconds' stimulus duration. Such an indifference point was not observed for the experi- mental group that had to reproduce the intervals between 10.5 and 15 seconds. This suggests that temporal reproduction experiments are sensitive to qualitatively different mechanisms operating in two different temporal domains, namely below and above approximately 3 seconds. The observation of a veridical reproduction of temporal intervals up to approximately 3 seconds suggests the existence of a specific temporal integration mechanism in this domain. Further experimental evidence
Reconstruction of Subjective Time
175
suggests that this temporal integration is automatic and presemantic, i.e., it is not driven by content, but occurs prior to a semantic evaluation of the information processed. The veridical reproduction of temporal intervals within a defined temporal window is presumably just one expression of the well-known time-order error (K6hler, 1923), which can be observed for all prosthetic continua in the different sensory systems (Stevens 195 l; P6ppel 1978). If two stimuli have to be compared with respect to their subjective intensity, they have to be presented within a temporal window of up to approximately 3 seconds. If the temporal interval between the two stimuli is longer than this critical interval, the second stimulus will be overestimated with respect to intensity compared to the first stimulus. The brain seems to provide an operative basis in which successive events are represented veridically and can, therefore, be compared appropriatdy only within a limited span. Another expression for the temporal integration up to approximately 3 seconds comes from psychophysical experiments studying Weber's Law in time perception (Getty, 1975). It turns out that Weber's Law is valid only up to temporal intervals of 2 to 3 seconds if temporal intervals have to be compared. For longer intervals, the difference between temporal intervals has to be increased dramatically to allow their distinct perception. This also demonstrates that temporal proces- sing up to approximately 3 seconds obeys qualitatively different mechanisms than for longer intervals. There is another class of observations supporting the idea of automatic temporal integration in which introspective co-operation of the subjects is required. One such class includes experiments on the spontaneous alteration rates in perception using ambiguous figures or sound sequences as stimuli. If stimuli that can be perceived under two perspectives or with two different meanings are presented, there is an automatic change of perceptual content atter an average of 3 secondS. The Necker cube, which can be seen under two different perspectives, has been used for visual stimulation, and a phoneme sequence (like CU-BA-CU-BA, where one hears either CUBA or BACU) has been used for auditory stimulation (e.g., Radilova et al, 1990). The spontaneous alteration rates in the visual and the auditor), modality can be interpreted as follows" automatically after an exhaust period, atten- tional mechanisms may be elicited that open sensory, channels for new information. If the information of the perceptual stimulus is simply the other alternative of an ambiguous figure, this alternate perspective or meaning will gain control in conscious representation. It is as if the brain asks, "What is new in
176
E. POppel
the world?" every 3 seconds, and, under such unusual stimulus conditions, the temporal eigen-operations of the brain are unmasked. A similar spontaneous alteration of perception has also been observ~ in experiments on binocular rivalry. (Gomez et al, 1996). If vertical and horizontal gratings with different colors are superimposed and viewed through glasses with corresponding colors, either the vertical or the horizontal grating will be perceived. In this situation, a sponta- neous fluctuation between the vertical and horizontal grating is seen, and the alteration rate is, again, about 3 seconds. Neurological observations have s h o ~ that brain lesions often result in a slowing down of processing (e.g., P6ppel et al, 1975; von Steinbiichei et al, 1996). It could be demonstrated that this effect is also true for the alteration rate in binocular rivalry (POppd r al, 1978). In one such case, an interesting observation could be made with respect to the dynamics of the alteration process itself. Because of the extreme deceleration of neuronal processing, the patient was able to describe how the vertical grating when dominant in his perception was gradually pushed away by a slow movement of the horizontal grating, as if a travelling wave were carrying the horizontal grating. Similarly, when the horizontal grating was dominant, it was gradually removed by the vertical grating until it became dominant. Observations like these fortify the hypothesis that the neuronal shift mechanism resulting in a new percept is not an irregular process which occurs in randomly selected positions in the neuronal net, but that neurones in local networks interact with each other and that their synchronized activity spreads sequentially in a topographically organized way (like in a cristallization process) to gain new territory.. Other evidence for an automatic temporal integration process comes from a simple experiment in which a subject is required to mentally structure auditory sequences. If one listens to the beats of a metro- home, one is automatically d r a ~ into the perceptual habit of accenting each second or third beat, thereby structuring the continuous metronome beats subjectively into a rhythm. By positing a subjective accent to every second beat, for instance, two successive beats are perceived as a unit (Szdag r al, 1996). The question then arises what the temporal interval can be up to which such units can be formed. It turns out that two beats cannot lie further apart than 2 to 3 seconds to allow subjective accentuation. Beyond this interval, it is no longer possible to mentally connect the second to the first beat, i.e. the first beat has then already disappeared in a perceptually not directly available
Reconstruction of Subjective Time
177
past. This is one reason why one can refer to such integration intervals also as subjective present (P6ppel, 1988; Edelman, 1989). A temporal grouping for an auditory chain of stimuli as suggested by the metronome experiment can also be observed in the temporal segmentation of spontaneous speech. Independent of the language spoken or of age, speakers have the tendency to construct closed verbal utterances up to about 2 to 3 seconds (Kowal r al, 1975; P6ppel, 1988; Vollrath et al, 1992). The expression of such a lingu- istic unit is usually followed a pause, presumably used to mentally prepare the next linguistic unit and load it into an output system. It is important to stress that temporal segmentation as observed in spontaneous speech does not necessarily apply to reading out loud. Written language is characterized by a different s)~tactic structure, and the reader apparently uses different neuronal processes from the speech domain to transport verbal utterances. It may not be accidental that, in many languages, a line in classical verse most often has a duration of approximately 3 seconds, as if poets had an implicit knowledge of the temporal segmentation process governing our brain (Turner and P6ppr 1988; but see also Kien and Kemp, 1994). It has also been demonstrated that worla'ng,memory is similarly limited to a few seconds if rehearsal is not allowed. In now classical experiments on short-term memory (Peterson and Peterson, 1959), sequences of letters were presented, and the subject's task was to reproduce them correctly. If the subject is asked to perform another task immediately after stimulus presentation, the access to the presented information prior to the disruptive task is heavily disturbed. Only within a temporal window of approximately 3 seconds can information be retained veridically. The temporal limitation of working memory is experienced in ever3,day life, e.g., when one wants to write down a new telephone number and is disturbed before doing so. Temporal segmentation in the domain of approximately 3 seconds has also been observed in studies on movements, as demonstrated by a simple experiment on sensory-motor s)~chronization. A subject is requested to synchronize finger taps to a regular sequence of short clicks, which are presented by earphone. It has been shown that click occurrence is anticipated with a motor response by some tens of milliseconds when the interstimulus interval is, for instance, 700 milliseconds (e.g., Radii et al, 1991" Vos et al, 1995). This negative asynchrony, as the anticipatory response has been referred to, has not been resolved yet, but the following hypothesis seems plausible to
17 8
E. P~pp el
explain the effect. The subject is attempting to synchronize the anticipatezl perceptual representation of the next auditory stimulus with the anticipated movement corresponding to this stimulus. Both these systems, namely the auditory and the tactile or muscle-spindle system (which informs the perceptual center about the accuracy of the movement), arc characterized by different delays between the periphery (i.e., the beginning of the transduction process) and the perceptual center. Because of this temporal difference, a phenomenal synchrony of the anticipated events must be asynchronous on the observer level. Thus, it is actually incorrect to refer to stimulus anticipation; we actually observe the expression of proper internal synchronization masked by the different delays in the two sensory, systems. This negative asynchrony is, however, only observed for interstimulus intervals up to approximately 3 seconds (Mates et al, 1994). A subject can program a properly synchronized motor response corresponding to a sensory event only within this temporal window. If the next sensory event lies too far in the future, the subject is no longer capable of programming the corresponding response. If the interstimulus interval is too long, synchronization, if attempted, shows an extremely high variability. Subjects often try to synchronize their movements with the sensory stimuli by simply reacting to them, which would express positive asynchrony. In this situation, we no longer deal with active anticipation, but with passive reaction to the sensory event. Temporal segmentation in the domain of 3 seconds has also been observed in studies on the duration of spontaneous intentional acts. It could be shown that there is a strong tendency of such movements to be structurally embedded in a temporal window of approximately 3 seconds" duration. Homologous movements have been studie,d in different cultures, and it was found that, independent of the state of acculturation, the duration of such movement patterns is identical (Schleidt et al, 1987). Cultures studied were those of the Yanomani Indians, Trobriand Islanders, Kalahari Bushmen, and Europeans. As a cultural transfer between these groups is extremely unlikely one is forced to conclude that a universal temporal constant dominates a certain class of human motor behavior. It has been demonstrated recently that t~ical movement patterns of different mammalian species also have a strong tendency to last on average 3 seconds. This observation suggests a universal temporal mechanism transcending human behavior (Gerstner and Fazio, 1995). It can be argued that, for all higher mammals, movement patterns are controlled by a
Reconstruction of Subjective Time
179
homologous neuronal mechanism that automatically integrates information up to approximately 3 seconds. In a recent study on mismatch negativi~ (Sams et al, 1993) using MEGtechnology (magnetencephalography), it was observed that interstimulus intervals of 3 seconds elicited by far the greatest amplitudes of the MMN (magnetic mismatch negativity). The mismatch negativity is thought to be an expression of increased cortical activity. In the study mentioned, the MMN was recorded for the auditory, cortex. An explanation for the resulting preference for approximately 3s is that the auditory system opens up its channels to pick up new information. This suggests that approximately every 3 s the "closed mind" opens up for new infor- mation coming from the environment. The above-mentioned experiments on temporal segmentation of 3seconds' duration covered perception, memory, cognition, and movement control. These different experiments suggest that temporal segmentation is an underlying principle of all higher data-generating mechanisms. Evidence from neuropsychology suggests that the different temporal segmentation processes might each be implemented in the corresponding neuronal domains. Nevertheless, we must conclude that there is a general temporal segmentation mechanism which is automatic and presematic. It provides an operative basis for precepts, memories, and volitional control. Because of the omnipresence of this mechanism, the single states of 3-second segments could be referred to as "states of being conscious" (STOBCON). Each STOBCON represents a mental island of activity distinctly separated from the temporally neighboring ones. How, then, does temporal continuity arise? It is suggested that each STOBCON, being implemented in a 3-second window, is semantically linked with a previous and with a following one. Thus, the continuity of experience is an illusion because distinct STOBCONs follow each other. Continuity arises because of a specific mechanism linking the contents of each temporal window to the next one. However, temporal segmentation in the two domains described, namely for the domain of 30 milliseconds and for the domain of 3 seconds, is a necessary prerequisite for the construction of subjective continuity. Paradoxically, continuity is made possible by discontinuous information processing in the brain, being most prominently represented in a subjective domain by each STOBCON.
180
E. PlJppel
4. References Aschoff J. On the dilatability of subjective time. Perspectives in Biology and Medicine, 1992; 53" 276-280. Eckhom, R, Bauer R, Jordan W, Brosch M, Kruse W, Munk M, Reitboeek HJ. Coherent oscillations: A mechanism of feature linking in the visual cortex? Biol Cybernetics 1988; 60: 121-130. Edelman GM. The Remembered Present: A Biological Theory of Consciousness. New York: Basic Books, 1989. Elbert T, Ulrich R, Rockstroh B, Lutzenberger W. The processing of temporal intervals reflected by CNV-like brain potentials. Psychophysiology 1991; 28: 648-655. Frost D, P6ppel E. Different programming modes of human saccadic eye movements as a function of stimulus eccentricity: Indications of a functional subdivision of the visual field. Biol Cybernetics 1976; 23: 39-48. Fuchs AF. Saccadic and smooth pursuit eye movements in the monkey. J. Physiol. 1967; 191" 609-631. Galambos R, Makeig S, Talmachoff PJ. A 40-Hz auditory potential recorded from the human scalp. Pror Natl. Acad. Sci. USA. 1981; 78: 2643-2647. Gerstner GE, Fazio VA. Evidence of a universal perceptual unit in mammals. Ethology 1995:101: 89-100. Getty DJ. Discrimination of short temporal intervals: a comparison of two models. Percept. Psychophys. 1975; 18: 1-8. Gomez C, Argandona ED, Solier RG, Angulo JC, Vazquez M. Timing and competition in networks representing ambiguous figures. Brain Cogni. 1996 (in press).
Reconstruction of Subjective Time
181
Gray C, K6nig P, Engel AK, Singer W. Oscillatory responses in cat visual cortex exhibit inter-columnar synchronization which reflects global stimulus properties. Nature 1989; 338" 334-337. Harter R, White CT. Periodicity within reaction time distributions and electromyograms. Quart. J. Exp. Psychol. 1968; 20: 157-166. Helson H. Adaptation-Level Theory. New York: Harper and Row, 1964. Hirsh IJ, Sherrick CE. Perceived order in different sense modalities. J. Exp. Psychol. 1961; 62: 423-432. Homstein AD, Rotter GS. Research methodology in temporal perception. J. Exp. Psychol. 1969; 79: 561-564. llmberger J. Auditor3, excitability cycles in choice reaction time and order threshold. Naturwissenschaften. 1986; 73" 743-744. Jokeit H. Analysis of periodicities Naturwissenschaflen. 1990; 77" 289-291.
in
human
reaction
times.
Kien J, Kemp A. Is speech temporally segmented? Comparison with temporal segmentation in behavior. Brain Lang. 1994; 46: 662-682. K6hler W. Zur Theorie des Sukzessivvergleichs und der Zeitfehler. Psychol. Forschung 1923; 4:115-175. Kowal S, O'Connell DC, Sabin EJ. Development of temporal patterning and vocal hesitations in spontaneous narratives. J Psy,cholinguistic Res 1975; 4: 195-207. Lackner JR, Teuber H-L. Alterations in auditory, fusion thresholds after cerebral injury in man. Neuropsychologia 1973; 11:409-415. Llinas R, Grace AA, Yarom Y. In vitro neurons in mammalian cortical layer 4 exhibit intrinsic oscillatory activi~, in the 10- to 50-Hz frequency range. Proc. Natl. Acad. Sci. USA 1991" 88" 897-901.
182
E. Pi~ppel
Llinas R, Ribary U. Coherent 40-Hz oscillation characterizes dream state in humans. Proe. Natl. Aead. Sci. USA 1993; 90:2078-2081. Madler C, P6ppel E. Auditory evoked potentials indicate the loss of neuronal oscillations during general anaesthesia. Naturwissenschatten 1987; 74" 4243. M~ikel~i JP, Haft R. Evidence for cortical origin of the 40 Hz auditory evoked response in man. Electroencephal. Clin. Neurophysiol. 1987; 66: 539-546. Mates J, MOiler U, Radii T, P6ppel E. Temporal integration in sensorimotor synchronization. J. Cogn. Neurosei. 1994; 6" 332-340. Murthy VN, Fetz EE. Coherent 25- to 35-Hz oscillations in the sensorimotor cortex of awake behaving monkeys. Proc. Natl. Acad. Sci USA 1992; 89: 5670-5674. Newton I. Philosophiae Naturalis Principia Mathematica, London 1687. Peterson LB, Peterson MJ. Short-term retention of individual items. J. Exp. Psychol. 1959; 58: 193-198. Podvigin ND, P6ppel E. Characteristics and functional meaning of oscillatory processes in the retina and lateral geniculate body. Sensory Systems 1994; 8: 97-103. POppel E. Excitability cycles in central intermittency. Psychol. Forseh 1970; 34:1-9. POppel E. Time Perception. In: Held R, Leibowitz HW, Teuber H-L, editors. Handbook of Sensory Physiology Vol 8: Perception. Berlin: Springer, 1978: 713-729. P6ppel E. Mindworks. Time and Conscious Experience. Boston, Harcourt Brace Jovanovich, 1988.
Reconstruction of Subjective Time
183
P6ppel E. Taxonomy of the subjective: An evolutionary perspective. In Brown JW, editor. Neuropsychology of Visual Perception. Hillsdale: Erlbaum, 1989:219-232. P6ppel E. Temporal mechanisms in perception. Int. Rev. Neurobiol. 1994; 37: 123-129. P6ppel E, Brinkmann R, von Cramon D, Singer W. Association and dissociation of visual functions in a case of bilateral occipital lobe infarction. Arch. Psychiat. Nervenkr. 1978; 225" 1-21. P6ppel E, von Cramon D, Backmund H. Eccentricity,-specific dissociation of visual functions in patients with lesions of the central visual pathways. Nature 1975; 256" 489-490. P6ppel E, Logothetis N. Neuronal oscillations in the brain. Discontinuous initiations of pursuit eye movements indicate a 30-Hz temporal framework for visual information processing. Natunvissenschaften 1986; 73" 267-268. P6ppel E, Schill K, von Steinbiichel N. Multistable states in intrahemispheric learning of a sensorimotor task. Neuroreport 1990 a: 1" 6972. P6ppel E, Schill K, von Steinbiichel, N. Sensory, integration within temporally neutral system states: a h39othesis. Naturwissenschaften 1990 b; 77: 89-91. Radii T, Mates J, llmberger J, P6ppcl E. Stimulus anticipation in following rh)~hmic acoustical patterns by tapping. Experientia 1990; 46" 762-763. Radilova J, P6ppel E, llmberger J., Auditor3' reversal timing. Act. Nerv. Super. 1990; 32: 137-138. Rubia K, Schuri U, von Cramon DY, P6ppel E. Counting seconds: Time estimation as a network property. 1996 (manuscript submitted). Ruhnau E, Haase VG. Parallel distributed processing and integration by oscillations. Behav. Brain. Sci. 1993" 16" 587-588.
184
E. POppel
Ruhnau E, P6ppel E. Prediction paradox and interstimulus interval bridging a gap between logic and experiment. Nature, manuscript submitted 1995. Sams M, Haft R, Rif J, Knuutila J. The human auditory, sensory memory trace persists about 10 see: Neuromagnetie evidence. J. Cogn. Neurosci. 1993; 5: 363-370. Schleidt, M, EibI-Eibesfeldt I, P6ppel E. A universal constant in temporal segmentation of human short-term behaviour. Naturwissenschat~en 1987; 74:289-290. Schwender D, Madler C, Klasing S, Peter K, P6ppel E. Anaesthetic control of 40-Hz brain activity and implicit memory. Conscious. Cognit. 1994; 3: 129-147. Spores O, Gaily JA, Reeke GN, Edelman GM. Reentrant signalling among simulated neuronal groups leads to coherency in their oscillatory activity. Proc. Natl. Acad. Sci. USA 1989; 86: 7265-7269. Von Steinbiiehel N. Temporal system states in speech processing. In Herrmann HJ, Wolf DE, P6ppel E, editors. Supercomputing in Brain Research: From Tomography to Neural Networks. Singapore: World Scientific. 1995" 75-81. Von Steinbiichel N, Wittmann M, P6ppel E. Timing in perceptual and motor tasks after disturbances of the brain. This volume. Steinberg S. High-speed scanning in human memory. Science 1966; 153: 652-654. Stevens SS. Mathematics, measurement, and psychophysics. In Stevens SS. Handbook of Experimental Psychology. New York: Wiley, 1951: 1-49. Stuart GJ, Sakmann B. Active propagation of somatic action potentials into neocortical pyramidal cell dendrites. Nature 1994; 367: 69-72.
Reconstruction of Subjective Time
185
Szclag E, von Steinbiichel N, Rciser M, de Langen E G, P6ppel E. Temporal constraints in processing nonverbal rh31hmic patterns. Acta Neurobiologiae Experimentalis, 1996; 56" (in press). Talbot JD, Marett S, Evans AC, Meyer E, Bushnell MC, Duncan GH. Multiple representations of pain in human cerebral cortex. Science 1991; 251" 1355-1357. Tononi G, Spores O, Edelman GM. Reentry and the problem of integrating multiple cortical areas: Simulation of dy~aamic integration in the visual system. Cereb. Cortex 1992" 2"310-335. Turner F, POppel E. Metered poetry, the brain, and time. In Rentschler I, Herzberger B, Epstein D. Beauty and the Brain. Biological Aspects of Aesthetics. Basel: Birkh~,user, 1988" 71-90. Vollrath M, Kazenwadel J, Kriiger H-P. A universal constant in temporal segmentation of human speech. Naturwissenschaften 1992; 79: 479-480. Vos P, Mates J, van Kruysbergen NW. The perceptual center of a stimulus as the cue for s~achronization to a metronome: Evidence from asvnchronies. Quart J. Exp. Psychol. 1995; 48A" 1024-1040. Wever R. Pendulum versus relaxation oscillation. In Aschoff J, editor. Circadian Clocks. Amsterdam" North-Holland Publ. 1965" 74-83. Zeki S. Functional specialisation in the visual cortex of the rhesus monkey. Nature 1978; 274: 423-428.
This Page Intentionally Left Blank
Time, Internal Clocks and Movement M.A. Pastor and J. Artieda (Editors) 9 1996 Elsevier Science B.V. All fights reserved.
187
TIME P E R C E P T I O N MEASUREMENTS IN NEUROPSYCHOLOGY
PAOLO NICHELLI
Sezione di Neurologia, Dip. cfi Patologia Neuropsicosensoriale. Universitdt degli Studi. Largo Del l'ozzo. 71. 41100 Modena. Italy.
ABSTRACT. This chapter examines the different paradigms used for measuring time perception and the factors that can affect the measurements. The resulls obtained with different methods in different groups of brain-damaged patienls are reported and interpreted in the framework of a model derived from the Scalar Timing Theo .ry (Gibbon. 1990).
I. Introduction
Man devised mechanical clocks because subjective time is not precise enough to serve all the needs of society. However, we still base our actions and most of our behaviors on a subjective dimension of time. Navon (1978) suggested that "our conception of the world (or of stimuli in the world) is a hierarchy of dimensions, in which time occupies the first level" (p. 227). Given its importance, it is not surprising that researchers have tried to understand the basis of information processing responsible for subjective time. In the last few years there have been several attempts to built up information processing theories of time perception. Yet, relatively little research effort has been devoted to understanding its neural basis. The reason for the lack of such an effort is probably due to the difficulty in incorporating all factors affecting subjective time measurement in a single model. As a result, time perception is probably the topic with the largest gap between theories developed from normal subjects' performance and those based on neuropsychological studies. In this chapter I will describe the different paradigms that have been used for measuring time perception and I will examine the factors that can affect
188
P. Nichelli
the measurements. I will also describe how different methods can influence the interpretation of results generated from studying brain-damaged patients. I hope to give the reader a sense of what has been achieved so far and the challenges facing us in order to achieve a better understanding of the neural basis of time perception.
2. Distinguishing within subjective durations Fraisse (1984) made a distinction between estimation and perception of duration. According to his view, "'estimation" takes place when memor3' is used either to li~tk two past events or to associate a moment in the past with a moment in the present. "'Perception" only involves the psychological present and concerns events lasting between 100 msec and 5 sec. Intervals lasting less than 100 msec seem instantaneous (i.e., without duration), while those lasting more than 5 sec involve a long-term episodic memor 3' mechanism. Although the validity of the boundaries of Fraisse's categories might be questioned (see Allan. 1979 on this point), the methodological implications are straightforward: in programming an experiment one should use appropriate durations to explore the hypothesized temporal processes. In human timing experiments it is also important to realize that with intervals longer than 1000 mscc. humans might use chronometric counting. Shorter intervals make counting, while possible, not very useful and not spontaneously used. Indeed. some of the discrepancies found between human and animal timing are due to the failure of recognizing that chronometrical counting was involved (Wearden and Lejune, 1993). A further important distinction is between prospective and retrospective time estimates. Ever3,'body is aware that our experience of passing time (experienced duration.) is influenced by what we are doing. Indeed, many studies have corroborated the view that in prospective timing, time appears to pass more rapidly if one is engaged in a higher level of behavioral activity (Vroon, 1970, Hicks et al.. 1977). On the contrary, a low level of behavioral activity results in the impression that time is dragging, which causes the experience of boredom. Yet. these effects are reversed in retrospect (remembered duration). A period devoid of interesting events is estimated as having had a very short duration if one looks back on it, while the retrospective temporal estimate of interesting events is that of a long duration (Omstein. 1969. Vroon. 1970).
Time Perception Measurements
189
Reliable experimental testing of remembered duration in a neuropsychologicai setting is limited by the fact that retrospective judgments can be investigated only once. Indeed, retrospective estimates can only be obtained if subjects are unaware that they will be requested to tell the elapsed time. Repeating retrospective time estimates would influence the subjects to attend to the passing time, thereby transforming a retrospective in a prospective estimate. The distinction between prospective and retrospective time judgment is dependent on the different methods of examining time perception. Verbal estimation and comparison can be uscd either with a prospective or with a retrospective paradigm. On the contrary, time production and reproduction can be only requested in a prospective paradigm, since the experimenter must inform the subjects about the task before the subject can produce (or reproduce) the required duration.
3. Methods for measuring time perception Allan (1979) classified the different methods of measuring time perception into two major categories: duration scaling and duration discrimination. In duration scaling tasks, subjects are asked about the perceived duration of a set of easily discriminable temporal intervals, while in duration discrimination tasks, they are requested to distinguish among a set of highly confusable intervals. Table 1 presents a list of the main methods used in time perception studies. This list is based on that proposed by' Allan but takes into account some experimental procedures that have been introduced more recently. The distinction between duration scaling and duration discrimination tasks is a ver?' usefill one since similar procedures can call for very. different cognitive components depending on whether or not intervals are highly confusable among each other. 3. !. DURATION SCALING TASKS Duration scaling tasks include ten_~oral production, synchronization, magnitude estimation, temporal reproduction, and c a t i o n ' rating. Subjects' performances with production, reproduction, and estimation procedures need to be separately analyzed in tcnus of accuracy and precision. "Accuracy" refers to the extcnt to which produced or estimatcd durations resemble real values, whereas the tenu of "'precision" (or
P. Nichelli
190
Table I. Methods Used in Time Perception Studies Duration Scaling
1. Temporal Production 2. Synchronization 3. Magnitude Estimation - Verbal Estimation - Estimation by Analogical Comparison 4. Categor>' Rating - Time Bisection (*) - Temporal Generalizalion 5. Ratio-Setting - Temporal Reproduction Duration Discrimination
1. Comparison 2. Single Stimulus 3. Time Biseclion (*) (*) Time bisection tasks can be considered a discrimination task whenever the ratio between the long and the short standard does not exceed 2. With long/short ration > 4 it can be considered a calcgor}.' rating task
"consistency") is used to indicate the closeness with which measurements agree with one another, i.e. they are consistent with any specific bias (Topping, 1955). As an example, if durations are measured with a clock that is running faster than the conventional one, measures will be inaccurate but precise (i.e., consistent). On the contrary, if durations are measured with different clocks, each running at a different rate, results will be both inaccurate and imprecise. Accuracy on production, reproduction, and estimation tasks can be measured by the ratio between estimated and true time. Values of this ratio larger than one are indicative of temporal overestimation, whereas a ratio smaller than one indicates temporal underestimation. Thus overestimation is when verbal estimation is too long and time production or time reproduction is too short. As an example, imagine that a subject is required to produce a 10 see interval and that he stops the clock at 9 see. Since the ratio between estimated time (10 see) and true time (9 sec) is greater than one, his temporal production will be considered indicative of temporal overestimation. In other words, when
Time Perception Measurements
191
subjects overestimate durations they verbally estimate intervals longer than the true ones but they produce intervals shorter than the real ones. As a consequence, provided that the subject's bias is constant, the correlation between production and estimation performances will be negative. This is particularly evident when the same subjects are asked to use both methods in the same experimental session (Carlson and Feinberg, 1970). Temporal production experiments require the production of an interval of specified duration (e.g., 0.4 or 10 sec) with the interval being defined by two responses or by the duration of a single response such as a button press. After each production, feedback may be given. Different sorts of feedback may be provided. In some studies specific information about the clock time produced is given (Wearden and McShane, 1988), while in other experiments feedback only specifies that some minimum required time value has been exceeded (Zeiler et al.. 1987). Two separate measures are obtained by a temporal production experiment: the "mean produced interval" and the "coefficient of variation". The "'mean produced interval" is an index of the subjects' accuracy of performance, while the "coefficient of variation" (standard deviation/mean) is a measure of their precision (i.e., consistency). Precision appears to be remarkably improved if subjects are allowed to use ehronometric counting. Several e::periments (Wearden, 1991a) have shox~ that short interval production (not depending on chronometric counting) has coefficients of variation ranging between 0.10 to 0.16, while counting-based interval production e.'xhibits ver)., small coefficients of variation (ranging between 0.04 and 0.08). Temporal production tasks have been used to study the effect of concurrent nontemporal processing on time estimation (Fortin and Rousseau. 1987, Fortin et al., 1993. Fortin and Breton, 1995). In this kind of experiments subjects are first trained to produce an interval of a fixed duration (e.g., 2 see). Then they are asked to provide the answer for a concurrent task (e.g., recognizing a probe) when the trained delay has elapsed. In this way, various experimental manipulations of the concurrent task can help in identif3,ing the specific processing demands that can affect time perception. Synchronization is a variant of temporal production. The experimenter presents a standard duration or a sequence of intervals and the subjects is requested either to respond in s3aachrony with the interval's termination or to produce a response at the same frequency. Rhythmic tapping is a particular instance of synchronization. Here, subjects are initially presented with a sequence of brief tones with constant spacing, and are required to tap a
192
P. Nichelli
response key in synchrony with the tones. Then, the tones are discontinued and subjects make a series of self-paced taps. Coefficients of variation derived from intertap intervals are close to 0.05, which is much less than the best possible coefficients obtained in an interval production task (usually in the range between 0.10 an 0.16). According to Wing and Kristofferson (1973a, 1973b) two processes are involved in timing of intertap intervals: a "timekeeper system" which determines when a response should be emitted and an "implementation system" which executes the command. The variability of the intertap intervals arises from the variability within these two processes. Wing and Kristofferson's model postulates that these two sources of variability are completely independent and that the variability of the implementation system can be obtained from the lag 1 autocovariance of a series of intertap intervals. It follows that, given a series of intertap intervals, it is possible to obtain separate estimates of the functioning of the timekeeper (internal clock) and of the implementation system. Taking advantage of this method. Ivry and Keele (1989) determined that increased variability in perfomling rhythmic tapping by cerebellar patients is due to increased timekeeper variability. Duchek et al. (1994) demonstrated that the same hold true for mild Alzheimer disease patients. Accordingly, we may conclude that cerebellar and Alzheimer disease patients, despite their different lesion pattern, have both an impaired internal clock mechanism. In magnitude estimation tasks, a standard reference duration is presented at the beginning of each session and subjects are requested to assign to each duration a number to represent the magnitude of perceived durations. Instead of giving the response with a number, the experimenter may ask subjects to draw a line (estimation by analogical comparison). Verbal estimation is a kind of magnitude esti.nation task where subjects are requested to judge durations in terms of conventional time units. Verbal estimation can be used with retrospective paradigms. However. this method is probably prone to a response bias of reporting the estimated durations in round numbers (Homik, 1981, Zackay. 1990). Furthermore, it is also prone to cognitive biases of representativeness and availability (Twersky and Kahneman, 1974) as subjects probably give the estimation on the basis of similarity of the task they performed during the target interval to a category of tasks whose durations they knew from before (Zackay, 1990). On the other hand. estimation by analogical comparison with a line's length is probably prone to perceptual biases in estimating line length (Hartley, 1977) It also has the potential danger of an anchor effect: a short interval presented at the
Time Perception Measurements
193
beginning of the session might pull judgments down, while a long interval might pull it up. Allan (1979) regards temporal reproduction as a particular instance of a ratio-setting task. In ratio-setting tasks, the examiner presents a temporal interval and the subject is requested to generate a specified proportion of that interval (Eisler, 1975, Allan, 1978) Reproduction is a ratio-setting task in which the proportion of the interval to be generated is equal to 1. In case subjects are requested to generate proportions less or greater than i of a given interval, the corresponding ratio-setting task is referred to respectively as fractionation or multiplication. Fractionation and multiplication have little, if any, practical interest. Shaw and Aggelton (1994) with both a verbal estimation and a reproduction task, demonstrated that Korsakoffs amnesics may be impaired in their ability to car~' out temporal judgments but that their impairment does not depend on the presence of memory deficits. Nichelli et al. (1993) showed that Alzheimer disease patients are both inaccurate and imprecise in their verbal estimates of durations ranging from 5 to 40 sec, while amnesic (non-demented) patients are inaccurate but often show nomaal precision in their estimates. Catego~,-rating requires subjects to locate the perceived duration in one of "n" ordered categories. The commonest instance of category rating tasks is time bisection. In time bisection tasks, subjects are first trained to discriminate a standard short and a standard long stimulus. When the3' are consistently able to do so, the two standard durations are presented along with intervals with intemlediatc duration. Subjects are requested to classif3, each interval as more similar to the short or to the long standard. The proportion of intervals judged to be long is then plotted as a fitnction of interval duration. An S-shaped curve usually is obtained (see Fig. l).This allows the experimenter to compute: (1) the bisection point (i.e., the duration classified as "long" on 50% of trials). In Fig. 1 the bisection point is the value on the abscissa corresponding to P(L)= 0.5. Lower bisection points indicate that the subject classified short stimuli as more similar to the long rather than to the short standard, while larger bisection points suggest the opposite tendency. (2) the difference limen (i.e., half the difference of the duration classified as "long" on 75% of trials and that classified as "long" on 25% of trials). In Fig. 1 the difference limen (DL) is half of the difference between values on the abscissa corresponding to P(L) = 0.75 and to P(L) = 0.25; (3) the Weber ratio (i.e., the difference limen divided by the bisection point). According to Weber's law. the difference limen is expected to var 3, as
194
P. Nichelli
a function of the bisection point. Thus, the ability of the subject to discriminate time intervals is better measured by the value of the Weber ratio than by the difference limen. Finally, from the proportion of unexplained variance, one can measure (4) the subject's precision (i.e., evaluate the consistency over repeated trials in adopting the same decisional criterion). Note that time bisection can be considered an instance of category rating only when the standard short and the standard long intervals are very distinct from each other, i.e., with ratios between the long and short standard of 4.0 or more. In this case human subjects classify durations in terms of their similarity to the short and the long standard. On the contrary, with longshort ratios ranging from 1.2 to 2.0, durations are so close to each other that subjects identify intem~ediate durations as being either the short or the long standards, i.e. they are forced to use a discrimination rather than a classification procedure. i
1.00'
A
i
. e,~.. e o.p ~ ~, ,.lllF o9. ,v
-
-
~
/ o"
0.75 ................................
i. el i
I ! i
o 9 . ,
~ ,,,,-%
,,,,,,I El.
V
0.50
!
.............
o
-,~ e e I
;0
:.
t 9
9 o .
r
..s
,,
bisection poi nt
i i i
9
,
:
i
o/ ~176176176176
. , .
"
/ 0.25 .......
9
I
,,
9
a a i
: : ;
I
I
I
I
100 200 300 400 500 600 700 800 900 Interval
Fig. 1. Plotting of an individual lime bisection performance. Tile standard short and long intervals were respectively I()0and 900 msec.
Time Perception Measurements
195
Time bisection tasks have been widely used to study timing mechanisms in animals. In animal studies (Church and Deluty, 1977, Maricq et al.. 198 I. Meek, i983) the bisection point occurs at the geometrical mean of the two standard intervals rather than, for example, at their arithmetic mean. For instance, in case of standard intervals of 2 and 8 see, the bisection at the geometrical mean would be at 4 see, while the arithmetical mean is at 5 see. A bisection point at the geometrical mean is predicted by the Scalar Timing Theory (Church and Deluty, 1977, Gibbon, 1986) as a result of a direct comparison between intervals stored in reference memory, and those in working memory (Gibbon et al., 1984) In humans a bisection at the geometrical mean has been only obtained (Allan and Gibbon, 1991) when subjects are forced to use a discrimination procedure. Whenever subjects use a classification rather than a discrimination procedure their bisection point is located somewhere between the geometrical and the arithmetical mean Wearden, 1991 b, Nichelli et al., 1994). A further very special instance of category rating is the so callcd temporal generalization procedure. In this task there is only one category. which is defined by a particular standard interval. The method is an analog of a procedure used by Church and Gibbon (1982) with rats. Subjects are first repeatedly presented with a standard interval. Then, after training, the standard interval is presented with a number of durations spaced in equal steps around it. For each interval's presentation, the subject has to decide whether or not that duration was equal to the standard. The proportion of "yes" responses are then plotted as a function of the stimulus duration. This method has been mostly used as a tool to verify, some assumptions that derive from Scalar Timing Model (Wearden, 199 l a, Wearden. 1992, Wearden and Towse, 1994) However. it can also be used to compare decision processes about tcmporal representations in different groups of patients. 3.2. DURATION DISCRIMINATION TASKS Duration discrimination tasks include the method of comparison, the single stimulus method, and the time bisection procedure. In the method of Comparison two durations are presented sequentially and the subject is requested to decide whether they are same or different. There are two possible versions of this tasks: in one version one of the two stimuli is held constant and is referred to as the standard stimulus. In the second
196
P. Nichelli
version (roving-standard teclmique) both stimuli can vary form trial to trial (Allan, 1977). A task requiring duration discrimination with a standard interval was used by Ivt3' and Keele (1989) to evaluate time discrimination in eerebellar patients. These authors used the Parameter Estimation by Sequential Testing (PEST) procedure (Taylor and Creelman, 1967, Pentland, 1980) to determine, after each subject's response, the upper and the lower threshold point. Based on this threshold, the PEST procedure can select, on each trial, the stimulus to associate to the standard one. Further examples of the use of this method in brain damaged adults and in clumsy children can be found in Keele et al. (1985) and in Williams et al. (1992). In the single stimulus method, on each trial the subject is presented with one of two possible durations and the subject is requested to decide whether it was the short or the longer value. A particular modification of the single stimulus method is the time bisection procedure. Time bisection can be considered a discrimination task whenever the ratio between the standard long and the standard short stimuli does not exceed 2.0. In this case, as in experiments with animals, normal subjects' bisection is at the geometrical mean. A common interpretation (Wearden, 1991b) of this finding is that, when the standard short and the standard long are close enough to each other, probe intervals of intenncdiate duration are attributed to either the short or the long standard via a discrimination rather than a classification procedure. Examples of this kind of time bisection by normal subjects can be found in a study by Allan and Gibbon (1991). However, there are no studies on brain damaged patients that have used stimuli within a range of durations that restricts the duration of long standard to no more than twice the short standard.
4. Factors affecting timing performance Various aspects of the content of the interval might influence its estimated length. The effect of task complexity and its different interactions with prospective and retrospective time estimates have been reported earlier. In the same vein, many authors have made a distinction between "filled" and "empty" time. The common implication is that "filled" durations are estimated as longer in comparison with equal "empty" durations (Coren et al., 1984). However, as for task complexity, this distinction needs to be considered as interacting with the estimation paradigm (prospective as opposed to retrospective). Also. the whole notion of "empty" time is
Time Perception Measurements
197
problematic because the experimenter has no control on what the subject is doing during an empty interval. Empty time intervals are particularly prone to subjects using counting or other strategies to improve their temporal estimates. However, counting might be utilized with filled intervals as well. Jackson (1985) demonstrated that individual differences in controlled strategies can influence temporal information processing. Controlling the use of estimation strategies is a goal difficult to achieve (Cahoon and Edmonds, 1980) but important to pursue. In any case the possibility that subjects have employed estimation strategies should always be considered when interpreting time-estimation experiments. A further aspect of the interval which is capable of affecting duration judgment is its segmentation. Indeed, several authors (Fraisse, 1963, Block and Reed, 1978, Block, 1982, Poynter, 1989) have argued that change is the basis of experienced duration. The existence of environmental tempo, like a metronome beat, might attract subjects' attention to the passage of time. Zackay et al. (1983) found a positive correlation between the frequency of flickering of a white bulb and reproduced time estimation. The methodological implication of these findings are that an estimated interval might be segmented by, stimuli or influenced by external tempo without the awareness of the experimenter. Another possible source of error in interpreting temporal judgments is failure to realize that the order of stimulus presentation (e.g., in comparison tasks) influences performance. This phenomenon is otten referred to as "time-order error". An extensive discussion of this topic can be found in Allan (1979) Time-order errors are positive when the first of two intervals is judged to be longer, whereas in a negative time-order error, the second of two intervals is judged to be longer. This methodological problem is usually dealt with counterbalancing the presentation order. However, since the two intervals usually ma.v not have the same objective duration, this remcdv might not be enough to balance the magnitude of the error. o
5. Interpreting subjective duration measurements. Duration judgments can be interpreted as the result of several processing mechanisms. Fig. 2 shows a schematic model of the different processes that are involved in time perception. I propose it as a tool for better understanding the results of time perception experiments in a neuropsychological setting. Tile right part of the model is derived from the
198
P. Nichelli
information processing account of the Scalar Timing Theory (Gibbon. 1990) According to this model prospective timing is accomplished via an internal clock mechanism. The internal clock consists of a Pacemaker, a Switch. an Accumulator, a Current Time Working Memory, and a Reference Memory system. The Pacemaker emits pulses at a regular rate. A signal activates the Switch. Pulses accumulate and are counted. The result of this count is transferred to Working Memor)., and compared with values in Reference Memory. Errors in timing may results from dysfunction at any of these levels. I hypothesize that pacemakers are reverberatory circuits containing several neurons with an oscillating neural activity. Neurons that belong to these reverberatory circuits might be part of the perceptual processing stream. If this is the case, one could argue that under resting conditions, the oscillation (and hence pacemaker rate) stabilizes at a certain rate. Incoming infommtion, by requiring extra computation, could slow the oscillation and reduce the pacemaker rate. This would cause time underestinaation. Attentional mechanisms or a specific brain lesion could influence the Switch. a device that gates stimuli from the Pacemaker to the Accumulator. Disorders at this level would cause an impairment of precision of time estimates. Frontal lobe patients might be impaired at this level because damage to large scale knowledge units might fail to automatically trigger the switch (Grafinan, 1989, p. 125). Working Memory, is the mechanism by which we register the passage of time. Any leakage of information at this level would cause underestimation of time, both in verbal estimation and in production tasks. But, in spite of their inaccuracy, subjective times would be as precise as those of normal subjects (i.e., consistent with a systematic bias). Pure amnesic patients might have a problem at this level, as demonstrated by the fact that they tend to underestimate prospective durations (Richards, 1973, Williams et al., 1989) However, not all amnesic subjects underestimate durations with a systematic bias. Values stored in Working Memor2,' need to be compared with those in Reference Memory. If Reference Memo~, is also impaired, both under- or over-estimation can occur and precision of the estimates may be affected (Nichelli et al., 1993) I hypothesize that different mechanisms ere involved in retrospective time judgments. In this case infommtion stored by some kind of non temporal processor is retrieved and transformed in units that can be compared with the Reference Memory system. According to this model, attention allocated to the nontemporal attributes of the stimulus would yield poor prospective
199
Time Perception Measurements
INCOMING INFORMATION
I lll ],l l I l.I I I ~~:~~
.......:i.::i:!i:!~:.!i~~!::i:~:i~!i~!i~i~i!:.!~i!ii~~i:~i~,i::ii~~~":"i~:~:i~;
+
ItN~Processor
" 4Fi!!ii~ii i~ ii~ii .~:!i:~i~i,i:~;(~i,~;i~i:i~!~iii~iiiiiiii~i~ i!i
;
J
Retrieval-Based L Counter
ttentional
Mechanisms
+
[Pacemaker1~! sw,tc.~!-"1 Accumu'a'~ I
)
.......
Reference ~ Memory,
CurrentTime Working Memory
RESPONSE
Fig.2. Schematic representation of the model for duration processing.
time judgments but veridical estimates of duration in retrospect. On the contrary., devoting attention to processing temporal cues would allow excellent prospective timing, but poor judgment of remembered time (Michon, 1985). The Reference Memor)' mechanism stores time intervals to be compared with those in working memor3' or with those obtained by the retrieval-based counter. I hypothesize that an impairment at this level would affect both prospective and retrospective duration judgments. Reference Memor3, could be evaluated by asking the subject to estimate duration of routine actions. Shaw and Aggleton (1994) recently demonstrated that the ability of Korsakoffs patients to estimate time intervals correlates with their performance at a cognitive estimation task. Korsakoffs patients (and in
200
P. Nichelli
general, patients with a frontal lobe lesion) might have an impairment at this level. We are just at the beginning of a long path towards a better understanding of the neural structures that subserve time perception. The model I am proposing is aimed to help generating hypotheses regarding the effects of lesions to different functional components and to argue about the putative role of different structures of the brain in timing performances. This could lead to new constraints to our current knowledge of the processes that are involved in time perception and eventually to a refinement of the model itself. ACKNOWLEDGMENTS. I want to thank Annalena Venneri for her carefid revision of the paper and Jordan Grafinan for his helpfid suggestions and for his very friendly and warm support.
6. References
Allan GA, Gibbon J. Human bisection at the geometric mean. Learn. Motiv. 1991:22: 39-58. Allan LG. The time-order error in judgments of duration. Can J. Psychol. 1977;31" 24-31. Allan LG. Comments on current ratio-setting models for time perception. Percept. Psychophys. 1978:24: 444-450. Allan LG. The perception of time. Percept. Psychophys. 1979:26" 340-354. Block RA. Temporal judgement and contextual change. J Exp Psychol" Learn Mem Cognit 1982.8" 530-544. Block RA, Reed MA. Remembered duration: Evidence for a contextualchange hypothesis. J. Exp. Psy'chol: Human Learn Mem 1978:4: 656-665. Cahoon D, Edmonds EM. The watched pot still won't boil" Expectancy as a variable in estimating the passage of time. Bull. Psychon. Soc. 1980:16" 115-116.
Time Perception Measurements
201
Carlson VR, Feinberg I. Time judgment as a function of method, practice, and sex. J. Exp. Psycho. 1970"85" ! 71-180. Church RM, Deluty MZ. Bisection of temporal intervals. J. Exp. Psychol" Animal Behav. Processes 1977"3' 216-228. Church RM, Gibbon J. Temporal generalization. J. Exp. Psychol: Animal Behav. Processes 1982:8 165-186. Coren S, Porac C, Ward LM. Sensation and Perception. (2nd ed. ed.) New York" Academic Press. 1984 Duchek JM, Balota DA, Ferraro FR. Component analysis of a rhythmic tapping task in indivisuals with senile dementia of the Alzheimer Tyepe and in individuals with Parkinson's disease. Neuropsychology 1994"8" 218-226. Eisler H. Subjective duration and psychophysics. Psychol. Rev. 1975:82: 429-450. Fortin C, Breton R. Temporal interval production and proceszsing in working memory.. Percept. Psychophys. 1995:57(2): 203-215. Fortin C, Rousseau R. Time estinaation as an index of processing demand in memory search. Percept. Psychophys. 1987;42(4): 377-382. Fortin C, Rousseau R, Bourque P, Kirouac E. Time estinaation and concurrent nontemporal processing: specific interference from short-termmemory demands. Percept. Psy'chophys. 1993;53(5): 536-548. Fraisse P. The Psychology, of Time.New York: Harper, 1963 Fraisse P. Perception and estimation of time. Annual Rev. Psychol. 1984"35 1-36. Gibbon J. The structure of subjective time: How time flies. In: The Psychology of Learning and Motivation. New York" Academic Press, 1086 105-135. Gibbon J. Representation of time. Cognition 1990;37: 23-54.
202
P. Nichelli
Gibbon J, Church RM, Meek WH. Scalar timing in memory. In: Gibbon J, Allan L. Ann New York Acad Sci vol. 423" Timing and Time Perception. New York" The New York Academy of Sciences, 1984: 52-77. Grafman J. Plans, actions, and mental sets" Managerial Knowledge Units in the frontal lobes. In: Perecman E. Integrating theory and practice in clinical neuropsychology. Hillsdale, N J" Lawrence Erlbaum Associates, 1989" 93138. Hartley AA. Mental measurement in the magnitude estimation of length. J Exp Psychol" Hum Percept. Perform. 1977:3" 622-628. Hicks RE, Miller GW, Gaes G, Biennan K. Concurrent processing demands and the experience of time-in-passing. Am. J. Psychol. 1977:90:719-730. Homik J. Time cue and time perception effect on response to mail survey. Journal of Marketing Research 1981"18" 243-248. Ivry RL, Keele SW. Timing filnctions of the cerebellum. J. Cogn. Neurosci. 1989" 1(2): 136-152. Jackson JL. Is the processing of temporal infomlation automatic or controlled. In" Michon JA, Jackson JL. Time, Mind. and Behavior. Berlin: Springer-Verlag, 1985" 179-190. Keele S, Pokomy R, Corcos D. Ivr3.' R. Do perception and motor production share common timing mechanisms. A correlational analysis. Acta Psychol 1985:60:173-191. Maricq AV, Roberts S, Church RM. Mcthamphetamine and time estimation. J Exp Psychol: Animal Bchav. Processes 1981"7" 18-30. Meek WH. Selective adjustcmcnt of the speed of internal clock and memor3.' processes. J Exp Psychol Animal Behav. Processes 1983:9:171-201. Michon JA. The compleat time experiencer. In: Michon JA, Jackson JL. Time, Mind, and Behavior. 1085: 20-52.
Time Perception Measurements
203
Navon D. On a conceptial hierarchy of time, space and other dimensions. Cognition 1978"6" 223-228. Nichelli P, Alway D, Grafinan J. Perceptual and motor timing in cercbellar degeneration. Neurology 1994"44 (Supplement 2): A231-A232. Nichelli P, Venneri A, Molinari M, Tavani F, Grafman J. Precision and accuracy of subjective time estimation in different memory disorders. Cogn. Brain Res. 1993" 1" 87-93. Omstein RE. On the Experience of Time. Hannondsworth, UK: Penguin, 1969 Pentland A. Maximum likelhood estimation" The best PEST. Percept. Psychophys. 1980:28" 377-379. Poynter DG. Inferring time's passage. In" Levin I, Zakay D. Time and Human Cognition: A life-span perspective. Amsterdam" North-Holland, 1989" 305-331. Richards W. Time reproductions by H.M. Acta Psychol. 1973"37' 279-282. Shaw C, Aggleton JP. The ability of amnesic subjects to estimate time intervals. Neuropsychologia 1994:32: 857-873. Taylor M, Creelman C. PEST: Efficient estimates of probability functions. J Acoustic Soc. ban. 1967:41 782-787. Topping J. Errors of Observation and Their Treatment. London" Chapman and Hall. 1955" 119. Twersky A, Kahneman D. Judgment under uncertainty: heuristics and biases. Science 1974:185 1124-1137. Vroon PA. Effects of presented and processed information on duration experience. Acta Psychol. 197034' 115-121. Wcardcn J, Lejune H. Across the great divide: Animal psychology and time in humans. Time and Society 1993:2(I)" 87-106.
204
P. NicheUi
Wearden JH. Do humans possess an internal clock with scalar timing properties? Learn. Motiv. 1991a.22" 59-83. Wearden JH. Human performance on an analogue of an interval bisection task. Q. J. Exp. Psychol. 1991b:43B: 59-81. Wearden JH. Temporal generalization in humans. J. Exp. Psychol: Animal Behav. Processes 1992:18(2): 134-144. Wearden JH, McShane B. Interval procedure as an analogue of the peak procedure: Evidence for similarity of human and animal timing processes. Q J. Exp. Psychol. 1988:40B: 363-375. Wearden JH, Towse JN. Temporal generalization in humans" Three further studies. Behavioral Processes 1994:32: 247-264. Williams HG, Woollacott MH. Ivr3.' R. Timing and motor control in clumsy children. J. Motor Behav. 1992:24" 165-172. Williams JM, Medwedeff CH, Haban G. Memory disorders and subjective time estimation. J. Clin. Exp. Neuropsychol 1989" 11" 713-723. Wing AM, Kristofferson AB. Response delays and the timing of discrete motor responses. Percept. Psy,chophys. 1973a; 14: 5-12. Wing AM, Kristofferson AB. The timing of interresponse intervals. Percept. Psychophys. 1973b:13" 455-460. Zakay D. The evasive art of subjective time measurement. In: Block RA. Cognitive Models of Psychological Time. Hillsdale, NJ" Lawrence Erlbaum Associates, 1990: 59-84. Zakay D, Nitzan D, Glicksohn J. The influence of task difficulty and external tempo on subjective time estimation. Percept. Psychophys. 1983:34: 451-456. Zeiler MD, Scott GK, Hoyert MS. Optimal temporal differentiation. J. Exp. Anal. Behav. 1987;47" 191-200.
Time, Internal Clocks and Movement M.A. Pastor and J. Artieda (Editors) 9 1996 Elsevier Science B.V. All rights reserved.
205
THE DEVELOPMENT OF CENTRAL PATTERN GENERATORS FOR VERTEBRATE LOCOMOTION KEITH T. SILLAR
Gatty Marine Laboratory. ,School of Biological amt Medical Science. The University of St. Andrews. St. Andrews, Fife KY16 8LB. ,Scotland ABSTRACT. The networks of neurones in tile spinal cord which control vertebrate locomotion, commonly called central pattern generators or CPGs. are constn~cted very early in development, often before locomotion is even possible. These elementary networks are then modified during development to suit changing behavioural requirements. The CPGs must therefore serve the needs of an organism at its particular stage in development and at the same time lay Ihe foundations for the conlrol of movemen!s at later stages. How do the CPGs controlling locomotor behaviour change during maluralion to cope with this challenge? This article reviews recent data from several preparations, but especially the developing swimming syslem of amphibian embryos, in which lhe mechanisms underlying.~PG dcvelopment are beginning Io unfold. In Xenopus, the descending projections of brainslem neurones which release 5HT appear Io be instrumental in the poslc,nbry,onic development of the spinal CPG, via aclions on a range of cellular and synaplic mechanisms.
I. Introduction "You have to walk before you can run" conveys the idea that skilfi~l tasks are best accomplished via a series of intervening steps in which our abilities gradually improve with practice. Presumably, the origin of this adage is the observation that human infants are essentially immobile at the time of birth, are able to walk about 1 to 2 y,'ears later and can only run after a period of further development. In fact our locomotor abilities are tuned to perfection over a considerably longer period even than this. Most animals, both vertebrate and invertebrate, experience a similar time-dependent progression in their abilities to move through their environment. My purpose in this article is to review current knowledge on how the neural circuits which
206
K. T. Sillar
control vertebrate locomotion develop. The thread running through my review will be the maturation of a spinal circuit controlling swimming behaviour in postembryonic amphibians. This relatively simple and rapidly developing model system, the focus of my own research, offers an amenable experimental preparation in which to address many questions of general importance in the development of rh)ethm generating circuits. The rhythmic movements of the limbs or body which underlie the Iocomotory movements of vertebrates are generated by networks of neurones located within the CNS. These networks are organised in such a way that they can produce repeating cycles of impulse bursts in skclctal motomeurones and coordinate the relative timing of discharge in different motor pools. The pattern of discharge within each motor pool encodes the duration and intensity of contractions in its target muscle, while the frequency of the underlying rhbahm determines the rate at which locomotion can occur. In combination these two factors sum to dictate the speed of progression of the organism. The main function of the central circuitry. therefore, is to ensure the proper sequence, strength and repetition of contractions in the locomotor musculature. It is now well established that the basic rhythmic motor output for locomotion can occur in the absence of any timing cues from sensor), neurones. For this reason, the rhythm generating networks have been referred to as central pattern generators or CPGs and the fact that CPG activity can also be elicited in the absence of descending commands from the brain indicates that the essential elements of locomotor CPGs are located principally within the spinal cord.
2. Some features of vertebrate locomotor CPGs
The intrinsic CPG clock cycles in an almost metronomic fashion at a frequency which roughly matches that of the behaviour it controls. Since the frequency of locomotion in different organisms can vary enormously, this places demands on the nature of the underlying circuitry. For example, the CPG controlling s~vimming in embryos of the amphibian, Xenopus laevis (Kahn and Roberts, 1982: Roberts, 1990) has a clock frequency of between 10 and 20 Hz (Figure l a), while that controlling the same behaviour in the adult lamprey covers a different frequency range, 0.25 to 10Hz (Walldn and Williams, 1984). Mechanical constraints such as the size and shape of the animal will undoubtedly play a part in determining the frequency range over
The Development of Central Pattern Generators
207
which locomotion actually occurs. However, the differences that exist between the clock frequency, of CPGs in these two 'lower' vertebrates must also reflect differences in the structure and function of the networks involved because when the mechanical constraints are removed, when the spinal cord is isolated from the periphery, the intrinsic clock frequencies still resemble the behavioural frequencies. This matching of motor frequency to behavioural frequency is intriguing because the locomotor networks of the lamprey and the Xenopus embryo appear to be rather similar, at least in terms of the types of neurones involved and the transmitters they use (Roberts and Clarke, 1982" Roberts. 1990; Grillner et al., 1991). For example, in both species the excitator3.' drive for swimming is produced by the co-activation of NMDA and nonNMDA glutamate receptor subt319es (Dale and Roberts, 1985' Dale and Griilner, 1986). Furthermore, reciprocal inhibition, which is necessary for the alternation of activity between antagonists, involves the release of glycine from interneurones which cross the spinal cord (Buchanan, 1982: Dale, 1985). Similar transmitters and receptors are also involved in the nonaxially based locomotor systems of neonatal and adult mammals (Pratt and Jordan, 1982; Jordan, 1983: Fenaux et al., 1991" Cazalets et al., 1992). So, the basic networks responsible for vertebrate locomotor rh.x~hm generation appear to utilise similar synaptic mechanisms which transcend both phylogenetic and ontogenetic boundaries. They may therefore represent the ancestral vertebrate circuits from which both axial and non-axial motor systems evolved through a process of elaboration upon a basic wiring diagram. If this is the case. then the obvious differences which exist in the CPGs of adult vertebrates, compared to their immature precursors, may simply involve increases in the numbers of participating neurones and the influences of modulatory, control systems, rather than changes in the basic operation of the underlying circuitr3'. The main question to be addressed in this review then, is what developmental influences are responsible for the elaboration of the basic vertebrate CPG circuitry into its complex and flexible adult form'?
208
K. T. Sillar
3. The developmental CPG clock 3.1. CPGS ARE CONSTRUCTEDBF,FORE TttEY ARE NEEDEDFOR BEHAVIOIIR. In most animals the neural circuitry responsible for rhythmic locomotor movements is constructed very, early in development, even before birth or hatching, when locomotion itself is not even possible (reviewed in Sillar, 1994). This is true even in humans, where rhythmic locomotor-like movements have been observed in the womb after just 10 weeks of development (de Vries et al., 1985). The initial construction process is thus completed before the occasion at which the circuitry is first needed for locomotion. However, these early CPGs are generally lacking in the precision and flexibility of their adult counterparts. During maturation, CPGs incorporate, in parallel with the intrinsic clock producing the motor pattern, a second, much slower 'developmental' clock, in which changes in the motor output coincide with changes in the behavioural requirements of successive stages. This rather less tractable timepiece is important since it ensures that changes in the output of CPGs are coincident with the required alterations to the behaviour they control. In this sense the developmental CPG clock is predictive, anticipating the form it must take for the successfi~l execution of locomotor behaviours at later stages. 3.2. EARLYCPGS ARE RETAINEDAND MODIFIEDDURINGDEVELOPMENT. The same basic circuitry for rhythmic locomotor control appears to be retained and adapted during the development of adult behaviour. This makes sense for a mode of locomotion like swimming in which the requirements of a CPG change only qualitatively during development as the animal simply expands in size and retains essentially the same mode of locomotion throughout life. For example, in Xenopus embryos the swimming pattern at the time of hatching is highly stereotyped (Figure l a) and larval development witnesses the acquisition of flexibility in the motor pattern (Sillar et al., 1991" Figure l b). However, the basic mode of coordination upon which swimming behaviour relies (alternation across the body and a rostrocaudal phase delay) is retained during the maturation process indicating that the same or very similar circuitry, is responsible at all stages.
Ai
,5h,
iii R5
Bi
stage 3718 embryo
1
1
I
fitive swimming .I
4r--L-l4i
*L-+I 3
stane 42 larva
iii R4
+1
L9
Figure 1. Development of s\\.imn~ingin postcmbnonic .Yenopus. A. At the time of hatching (stage 3718). Xenopus embtyos (Ai), s\vinl b!. lateral body undulations at 10 to 20Hz. Aii. Tracings of swim q c l e from high speed video recording. 5ms interval between imagcs. Arrows at points of nlasinlum body curvature ~vhichpass from head to tail, alternately on each side. Aiii. Ventral root activity during fictive s\vimming in paralysed embno. Recordings made from right (R) and Le!? (L), rostra1 (5) and caudal (13) intermyotomal clefts. Note single biphasic impulse per cycle alternating across the body and passing rostrocaudally with brief delay, as in real s\vimrning. B. At stage 42* a day later, the l a n a (Bi) mims in a similar ashion (Bii). Fictive swimming now involves bursts of activih on each q c l e (Biii), but retains the same coordination. Recordings made by J. Wedderburn.
R Q
2
2Q
5
2
210
K. T. Sillar
However, it is not so immediately obvious that the same solution be adopted by vertebrates with jointed appendages. This is because the limbs may be used for specific behaviours early in development which are never likely to be deployed again. Thc hatching movements of chicken embryos is one example (reviewed in Bckoff. 1992) in which the left and right limbs are synchronised in episodes of rh.x~hmic thrusting movements. After hatching the motor programme is never usually required again and is substituted by one with an entirely different coordination which controls walking, where the left and right legs are now active in alternation. Walking begins just a few hours after hatching so there is an abrupt behavioural switch. This might imply that the neural circuitr3.' for hatching is simply turned off and lost after hatching to be replaced by different circuitry controlling walking. However, it seems more likely that the two behaviours utilise essentially the same neural circuitr).' and that sensor).' inputs associated with walking movements somehow trigger a switch in the linkage between separate CPGs controlling each limb. Evidence that circuitr)., generating the hatching motor programme is retained after hatching has accrued from experinaents where hatching movements can be elicited in post-hatched chicks when they are placed in surrogate glass eggs. 3.3. CPGS ()FTEN DEVEL()P ROSTR(K'AIIDALLYALONGTHE BODY AXIS. The maturation of many CPGs appears to follow a pre-set sequence along the rostrocaudal body axis, from head to tail. This is particularly true of axially based locomotor systems like swimming, where such a statement may seem obvious. However, it also occurs in mammals like the rat, where coordinated forelimb locomotion is possible at birth, but the hindlimbs become involved in walking a fortnight or so later in development (see also section 4.1, below). At~er hatching from the egg membranes the motor output for swimming in Xenopus larvae develops along the body axis (Sillar et al., 1991). At the time of hatching (developmental stage 37/8; Nieuwkoop and Faber, 1956) the swimming motor pattern is very simple and stereot333ed with brief motor bursts occurring on each cycle (Figure l a). This is because myotomal motomeurones fire only one impulse per cycle. During the first day or so of postembr3,onic development, however, this limitation is lost so that by stage 42. more robust bursts of discharge occur (Figure l b). The transition to the bursty larval motor pattern occurs rostrocaudally since at an intermediate stage (stage 40), about 12 hours after
The Development of Central Pattern Generators
211
hatching, the rostral segments generate bursts of discharge during swimming, but more caudal segments produce embryo-like swimming. These observations on spinal circuit development in Xenopus are suggestive of neuronal signals which progress rostrocaudally within the spinal cord during larval development to influence the activity of a pre-existing CPG (see section 5. below).
4. The components of spinal CPGs change during development 4.1 CtIANGESIN NI-IFRt)NALSYSTI-MS. Although the filndamental basis for spinal rhythm generation seems to be rather similar across a range of vertebrates and appears to be relatively independent of developmental stage (see section 2, above), the format of the mature CPG does change during behavioural development. This suggests that the neural components of the C PG are also altered. In some cases it has been possible to observe changes in the number and types of neurone present in the spinal cord as CPGs ,nature. In this regard the complement of neurones immuno-positive for the inhibitory amino acid neurotransmitter GABA, changes in a manner consistent with an important function for GABAergie neurones in CPG development. In the chicken embryo, for example, there is an inverse relationship between the appearance of neurones immunoreaetive for GABA and glycine (Antal et al., 1994), the two important inhibitory' transmitters i.~ the spinal cord. GABA-immunoreactive neurones are the first to appear (at embn,onic day 4 - E4), and their numbers peak at E8 before steadily declining. In contrast, glycine inununoreactive neurones appear rather later in development at E8 and the population then increases approximately linearly until they establish a distribution characteristic of the adult by E20. Interestingly. the sum of GABA-and glycine-immunoreactive cells is approximately constant from E8 onwards. These findings suggest that GABAergic inhibition is important early on in the production of embD'onic motor output but is then superseded by glycinergie mechanisms. Physiological evidence supports this idea since early embr),onic rhythm generation depends upon a mixture of GABAergic inhibition and glutamate receptor-mediated excitation, but as development proceeds the GABAergic component diminishes (O'Donovan et al., 1992). In terms of motor circuit filnction, the loss of GABAergic inhibition in the
212
K.T. Sillar
chicken embryo accompanies a developmental transition in the phasing of the motor output. In the neonatal rat spinal cord a reduction in GABAergie inhibition may be vital to the very expression of CPG output during development. The newborn rat pup can crawl using its forelimbs and the participation of the hindlimbs in quadrupedal locomotion becomes established only at~er a few weeks of further development. The isolated spinal cord of the neonate can generate fictive locomotion in the thoracic motor centres following local applications of NMDA (Cazalcts et al., 1994). Under these conditions the hindlimb nerves are silent but will themselves generate fictive locomotion when GABA receptor antagonists are applied to lumbar segments. The inference is that the caudal C PGs are present in the lumbar spinal cord at birth but are inhibited by GABAergic mechanisms. It is tempting to speculate from these data in the rat that a developmental reduction in GABAergic inhibition within the lumbar CPGs could be responsible, as in the chick, for the maturation of the locomotor system. 4.2 CIIANGES IN NEIJR()NAI. Iq~,()I'I'~R'I'~S
Embr3,onic vertebrate ncuroncs differ markedly from adult neurones performing the same or similar functions. They differ not only in their size and geometry, but also in thcir elcctrical properties and responses to various neurotransmitters. For example, the structure and function of transmitter receptors and ion channcls are knoxx,~ to change as development proceeds (e.g. Cherubini et al., 199 i" Hcstrin, 1992" Desarmenian et al., 1993; Walton et al., 1993). Changes in thcse molecules will undoubtedly influence the input-output propertics of ncurones within locomotor circuits. However, relatively little is known about how specific alterations in membrane receptors and ion channels might affect locomotor rhythm generation. Determining just how the dcvclopmental regulation of transmembrane proteins affects circuit fi~nction and motor behaviour will be a daunting but important issue to be addressed in future studies, In Xenopus embryos, myotomal motorneuroncs fire a single action potential on each cycle of swimming activity and this unusual behaviour has been ascribed to the presence of a slowly inactivating K + conductance which clamps the membrane potential below spike threshold after the neurones have discharged (Soffe, 1990). Howcvcr. during motor system development (Figure 1), the same motorncuroncs can now display multiple firing (Sillar et
The Development of Central Pattern Generators
213
al., 1992a). The firing properties of the motomeurones thus change in a manner which allows the expression of the more flexible, 'bursty' larval motor pattern. Presumably, for this to occur, spike accommodation must have been reduced, consistent with a reduction in the contribution of the slow K + conductance to neuronal firing.
5. Descending neuromodulation and CPG development Neurones whose processes descend into the spinal cord from higher centres, like the brainstem for example, continue to develop for a significant period after birth or hatching. It has been assumed that during this period these descending projections modt.latc the basic spinal CPG circuitr)., to adapt the developing motor output. While research into the role played by specific populations of such descending fibres is still in its infancy, one system stands out as being of potentially general importance. This system originates in the rostral, ventral medulla at early stages in development in a region knox~.n as the raphe. Raphc neurones can be characterised by their transmitter phenot3]ge since the majority contain the indoleamine, 5hydrox3.~ryptamine (5HT: serotonin). 5.1 A R()I,E F()R DESCEN1)IN(; 5Irl RELEASEIN SWIMMINGDEVEL()PMENT As described above, the spinal cord of the Xenopus emb~,o contains all the neural machinery necessar3.' to generate a basic locomotor rhytlma very early in development, and the same machinery is then modulated as development proceeds (Sillar et al., 1991" Figure 1). The rostrocaudal development of the swimming rh3~hm indicated that the spinal CPG could be progressively influenced by neurones whose projections descend into the spinal cord early in larval life and circumstantial evidence suggested that 5HT release from raphespinal projections could be responsible. Firstly, in adult vertebrates, 5HT is involved in the intrinsic modulation of locomotor activity (e.g. Viala and Buser, 1969" Harris-War, rick and Cohen, 1985" Wall6n et ai., 1989: Barbeau and Rossignol, 1990" Grillner et al., 1991). In each example, the amine facilitates the firing properties of spinal neurones to enhance the intensity and duration of locomotor bursts. This role of 5HT in adults is reminiscent of the acquisition of enhanced bursting during normal development in the Xenopus sxvimming system (Figure 1). Secondly, the
214
K. T. Sillar
descending raphespinal system innervates the spinal cord at approximately the same epoch in vertebrate development that motor systems mature. In Xenopus this correlation is particularly striking since the first descending serotonergic fibres reach the rostral spinal just before hatching, but then proliferate and extend to the caudal cord over the first few days of larval life (van Mier et al., 1986). Thirdly, 5HT receptors are thought to be important in developmental signalling. For example, the 5HTIa receptor has been proposed to be involved in neuronal differentiation fWhittaker-Azmitia, 1991). Furthemlore, the serotoncrgic system appears to play a role in s)~aptogenesis during development. For example, the pharmacological deletion of descending serotonergic projections in chicken embryos near the time of hatching significantly reduces the density of non-serotonergic synapses on spinal motomeurones (Okado et al., 1992). We hypothesised from these observations that the development of the swimming CPG in postembr),onic Xenopus could be due to the ingrowth of raphespinal projections and their release of 5HT. Initial evidence from the bath application of 5HT to different stages of development supported this hypothesis since the amine could mimic the normal progression of burst activity during swimming (Sillar ct al.. 1992b). Three additional lines of enquiry have now provided more direct evidence for a causal link between the development of the raphespinal system and the ontogeny and maintenance of the larval swimming pattern (Sillar et al., 1995; Figure 2). i) pharmacological blockade of the 5HTla-like receptors which modulate ongoing larval swimming (Wcdderbum and Sillar, 1994a) reduces the duration of larval motor bursts to the extent that they now resemble those occurring during embryonic activity, ii) surgical removal of descending influences, which include the raphespinal projections, similarly elicit embryo-like swimming from larval preparations (Figure 2C). iii) the pharmacological ablation of raphespinal axons using a monoamine neurotoxin (5,7 DHT), has the same effect in that the larval spinal cord, now naive to the effects of endogenous 5HT, produces motor output akin to that generated by the embr3.,onic CPG for swimming (Figure 2B). These new findings lead to the conclusion that the descending serotonergic system plays a very important, if not crucial role in the development of the locomotor CPG, by converting the stereotyped and inflexible embryonic CPG into a more adult-like and flexible form. Interestingly, the same serotonergic modulatory system which alters the timing and intensity of the intrinsic CPG clock in adults appears to be responsible, in Xenopus at least, for the
The Development of Central Pattern Generators
Ai
215
stage 42 - intact
spinal cord
iii
Bi
stage 42 - toxin-treated
iii
spinal cord
100~tM Ci
stage 42 - post-spinal
-=.=
.
.
10ms Fi~=re 2. Evidence for descending serotonergic i~ffluence on CPG development in Ak,nopus larvae. Ai-iii. A! slage 42 (i) CPG produces bursts of activity in each c2,'cle (ii" 16 consecutive cycles of roslral ventral root (vr) activi .ly superimposed). At this stage the spinal cord receives a descending serotonergic projection (iii: drawing of 5HT-immunoreaclive fibres in rostral spinal cord). Bi-iii. Neurotoxic ablation of descending seroloncrgic fibres prevents rh.xlhnl development, i. Drawing of larva raised in ImM 5.7 dihydrox31r?,.~plamine, a monoamine neurotoxin. Lan,a is a sibling of conlrol larva in Ai. ii. Toxin-treated larva produces embr).'o-like swimming wilh brief biphasic impulse in each cycle (c.f. Figure laiii). Spinal cord of Ibis larva showed no evidence of descending serotonergic projcclions. Ci.ii. Spinalized (at arrow) slage 42 larva (i) produces embryo-like activity (ii). suggcsting descending inputs maintain the burstv larval pattern. See text and Siilar el al (1995) for fi~rlher details.
216
K. 7". SiUar
developmental modulation of spinal locomotor circuitry into a form which is suitable for serotonergic modulation at later stages. It would be logical to propose from these studies that the mechanism through which 5HT mediates circuit development solely involves effects on the firing properties of spinal ncurones. Atter all, it is this mechanism which is largely responsible for the serotonergic modulation of adult locomotor circuit~' (Grillner et al., 1991). On the other hand it is already known that 5HT may play additional roles during development and indeed the modulation of the swimming CPG in Xenopus by 5HT is not restricted to effects on neuronal firing properties. For example, the amine can regulate the strength of synaptic connections between certain neurones within the CPG (Sillar and Wedderbum. 1994). In addition, the activation of 5HT receptors on motomeuroncs in both Rana and Xenopus tadpoles appears to modulate the actions of glutamate at the NMDA receptor, being responsible for the induction of intrinsic oscillator3' membrane properties (Sillar and Simmers, 1994" Weddcrbum and Sillar, 1994b). Moreover, this latter effect of 5HT is developmentally regulated in Xenopus such that the amine has no effect on NMDA receptors in cmbr3.'os, but can induce intrinsic oscillatory. behaviour in larvae (Weddcrbum and Sillar, 1994b). These two recent discoveries offer insight into the complex targets of the amine during the development of the locomotor C PG. Of considerable interest is the fact that the 5HT receptors involved in these diverse developmental effects appear to belong to a single subclass xvhich is pharmacologically similar to the mammalian ia-type receptor (Wcddcrbum and Sillar, 1994a).
6. Conclusions
The remarkable simplicity of the Xenopus embryo spinal cord, containing only 8 classes of differentiated neurone (Roberts and Clarke, 1982), together with the rapid development of the swimming system atter hatching, provides an amenable model system in xvhich the mechanisms which orchestrate CPG development can be readily explored. During the first day of larval life, the stereotyped motor output generating embryonic swimming is transformed into a more flexible larval form. Several lines of evidence support the hypothesis that the desce.~ding projections of serotonergic raphe neurones in the brainstem are causally linked to rhythm development. Raphespinal projections i~mervatc the spinal cord after hatching and, through an emerging
The Development of Central Pattern Generators
217
diversity of cellular and synaptic mechanisms, re-structure the output of the swimming CPG. It is perhaps too soon to speculate whether the serotonergic system might play similar roles in the development of other vertebrate CPGs. However, it is noteworthy that 5HT exerts vcr3.' similar effects on locomotion both during vertebrate development and during the intrinsic modulation of adult locomotion, that is, an increase in burst duration and intensity. Moreover, the receptor sub-type involved in Xenopus CPG development is already kno~.aa to be important in vertebrate brain development in general. Thus, the descent of serotonin-containing axons into the spinal cord at early stages in vertebrate development may bc a widespread developmental trigger for the alterations to pre-existing circuitry which herald the maturation of locomotor behaviour. Interest in how locomotor C PGs develop has increased recently with the advent of new preparations and techniques for exploration. The data to accrue from this work provides a tantalising glimpse of what may eventually turn out to be features of general importance in neural circuit development in vertebrates. However, the surface has only just been scratched.
7. References Antal M, Polgar E, Bcrki A. Bir3'ani A. Poskar Z. Development of specific populations of intcrneurons in the ventral horn of the embryonic chick lumbosacral spinal cord. Eur. J. Morphol. 1994: 32 201-06. Barbeau H, Rossignol S. The effects of serotonergic drugs on the locomotor pattern and on cutaneous reflexes of the adult chronic spinal cat. Brain Res. 1990:514: 55-67. Bekoff A. Neuroethological approaches to the study of motor development in chicks" aclaievemcnts and challenges. IReview]. J. Neurobiol. 1992" 23' 1486-505. Buchanan JT. Identification of intcrneurons with contralateral caudal axons in the lamprey spinal cord' synaptic interactions and morphology. J. Neurophysiol. 1982 47 961-75.
218
K.T. SiUar
Cazalets JR, Sqalli-Houssaini Y, Clarac F. Activation of the central pattern generators for locomotion b.v scrotonin and excitatory amino acids in neonatal rat. J. Physiol. 1992 455" 187-204. Cazalets JR, Sqalli-Houssaini Y, Clarac F. GABAergic inactivation of the central pattern generator for locomotion in isolated neonatal rat spinal cord. J. Physiol. 1994; 474" 173-8 !. Cherubini E, Gaiarsa JL, Bcn-Ari Y. GABA" an excitatory transmitter in early postnatal life. IRevie~vl. Trends Ncurosci. 1991" 14" 515-19. Dale N. Reciprocal inhibitory intcrneurones in the spinal cord of Xenopus laevis. J. Physiol. 1985" 363: 527-43. Dale N. Roberts A. Dual component antino acid-mediated synaptic potentials: excitator)' drive for swimming in Xenopus embryos. J. Physiol. 1985" 363" 35-59. Dale N, Grillner, S. Excitator3.' synaptic drive for swimming mediated by excitatory, amino acid receptors in the lamprey. J. Neurosci. 1986; 6: 266275. Desarmenian MG, Clendcning B, Spitzer NC. In vivo development of voltage-dependent ionic currents in embryonic Xenopus spinal neurons. J. Neurosci. 1993" 13" 2575-8 I. De Vries JIP, Visser GHA. Prechtl HFR. Fetal motility in the first half of pregnancy. In: Continuity of neural functions from prenatal to postnatal life. Clinics in Developmental Medicine. HFR Prechtl (ed). Spastics International Medical Publications. 1994 t)4" 46-64. Fenaux F, Corio M. Pallisses R, Viala D. Effects of an NMDA receptor antagonist, MK-80 I, on central locomotor programming in the rabbit. Exp Brain Res. 1991" 86: 393-401. Grillner S, Wall6n P. Brodin L. Lansner A. Neuronal network generating locomotor behavior in lamprey. IReviewl. Ann. Rev. Neurosci. 1991" 14" 169-99.
The Development of Central Pattern Generators
219
Harris-Warwick RM, Cohen AH. Serotonin modulates the central pattern generator for locomotion in the isolated spinal cord of the lamprey. J. Exp. Biol. 1985:116: 27-46. Hestrin S. Developmental regulation of NMDA receptor-mediated synaptic currents at a central synapse. Nature 1992" 357" 686-89. Jordan LM. Factors determining motoneuron rh.vthmicity during fictive locomotion. In" Neural Origin of rh3~hmic movements. (A. Roberts and B.L. Roberts, eds). Soc. Exp. Biol. Syrup. 1983" 37' 423-44. Kahn J A, Roberts A. The central nervous origin of the swimming motor pattern in embryos of Xenopus lacvis. J. exp. Biol. 1982" 99' 185-96. Nieuwkoop PD, Fabcr J. Normal tables for Xenopus laevis (Daudin). Amsterdam, North Holland. I t/56. O'Donovan MJ, Scmagor E, Sholomenko G, Ho S, Antal K, Yee W. Development of spinal motor nct~vorks in the chick embryo. J. exp. Zool. 1992:261' 261-73. Okado M, Cheng L. Tanatsugu Y. Hamada S, Hamaguchi K. Synaptic loss following removal of scrotonincrgic fibers in newly hatched and adult chickens. J. Neurobiol. 1992 24 687-98. Pratt CA, Jordan LM. Recurrent inhibition of motoneurons of decerebrate cats during controlled treadmill locomotion. J. Neurophysiol. 1982" 44: 489500. Roberts A. How does a nervous s.vstcm produce behaviour? A case study in neuroethology. [Review]. Sci. Prog. 1990: 74:31-51. Roberts A. Clarke JDW. Tile ncuroanatomy of an amphibian embo, o spinal cord. Phil. Trans. Roy. Soc. Scr. B 1982' 296' 195-212. o
Sillar KT. Synaptic specificity' development of locomotor rhythmicity. [Review]. Curr. Op. Neurobiol. I t~t~4: 4:101-07.
220
K. T. Sillar
Sillar KT, Simmers AJ. 5HT induces NMDA receptor-mediated intrinsic oscillations in embr3,onic anlphibian spinal neurones. Proc. Roy. Soc. Ser. B 1994; 255" 139-45. Sillar KT, Wedderbum JFS. Presynaptic modulation of glycine neurotransmission by 5HT in the spinal cord of Xenopus laevis embryos. Eur. J. Neurosci. (Supp) 1994: 7" 167. Sillar KT, Wedderbum JFS, Simmers AJ The development of swimming rhythmicity in post-embr3.,onic Xcnopus laevis. Proc. Roy. Soc. Ser. B 1991" 246: 147-53. Sillar KT, Silm~aers AJ. Wcddcrburn, JFS. The post-embryonic development of cell properties and synaptic drive underlying locomotor rhythm generation in Xenopus larvae. Proc. Roy. Soc. Ser. B 1992a: 249: 65-70. Sillar KT, Wedderburn JFS. Simmers AJ. Modulation of swimming rhythmicity by 5-hydroxytr3.'ptamine during post-embryonic development in Xenopus laevis. Proc. Roy. Soc. Ser. B 1992b; 250" 107-14. Sillar KT, Woolston A-M. Wcdderbum JFS. Involvement of brainstem serotonergic interneurons in the development of a vertebrate spinal locomotor circuit. Proc. Roy. Soc. Ser. B 1995; 259: 65-70. .
Soffe SR. Active and passive membrane properties of spinal cord neurones during fictive swimming in frog embr3'os. Eur. J. Neurosci. 1990; 2: 1-10. van Mier P, Joosten HJW, van Rheden R, ten Donkelaar HJ The development of serotonergic raphc spinal projections in Xenopus laevis, lnt J. Dev. Neurosc. 1986: 4: 465-76. Viala D, Buser P. The effects of DOPA and 5HTP on rhythmic efferent discharges in hindlimb nerves in the rabbit. Brain Res. 1969; 12: 437-43. Wall6n P, Williams TL. Fictivc locomotion in the lamprey spinal cord in vitro compared with swimming in the intact and spinal animal. J. Physiol. 1984: 347: 225-39.
The Development of Central Pattern Generators
221
Wall6n P, Buchanan JT, Grillncr S, Christenson J, H6kfeit T. The effects of 5-hydroxytryptamine on the afterhyperpolarization, spike frequency regulation and oscillator3.' membrane properties in lamprey spinal neurons. J. Neurophysiol. 1989:61" 759-68. Walton MK, Schaffner AE. Barker JL. Sodium channels, GABAa receptors and glutamate receptors develop sequentially on embryonic rat spinal cord cells. J. Neurosci. 1993" 13" 2068-84. Weddcrbum JFS, Sillar KT. Modulation of rhythmic swimming activity in post-embryonic Xenopus laevis tadpoles by 5-hydrox3~u'ptamine acting at 5HTla receptors. Proc. Roy. Soc. Ser. B 1994a: 255" 139-45. Wedderburn JFS, Sillar KT TTX-resistant, NMDA receptor-mediatcd membrane potential oscillations in spinal locomotor neurones of Xenopus iaevis larvae are 5HT-depende,lt. Neurosci. Abst. 1994b" 20' 763. Whittaker-Azmitia PM. Role of serotonin and other neurotransmitter receptors in brain development: basis for developmental phannacology. [Review]. Phannacol. Rev. 1991 43 553-61.
This Page Intentionally Left Blank
Time, Internal Clocks and Movement M.A. Pastor and J. Artieda (Editors) 9 1996 Elsevier Science B.V. All rights reserved.
223
AN HIERARCHICAL MODEL OF MOTOR TIMING BRIAN L. DAY
MRC Human Movement and Balance Unit. Institute of Neurology. Queen Square. London, WCIN 3BG. IlK ABSTRACT. This chapter questions tile view that the time of execution of a simple voluntary n~ovement is controlled solely by the person's intention to move. In a series of experiments, subjects were requested to make a predetermined movement in response to an external cue. Occasionally a transcranial shock to the motor cortex was given at a time between the cue and the usual time of movement onset. This had the effect of delaying execution of the movement without substantially changing its spatial characteristics. When subjects were trained to make simultaneous and bilateral movements and a cortical stimulus was applied to only one hemisphere, the limb movement contralateral to the stimulus was delayed whereas the ipsilaterai limb was little affected. Some classes of saccadic eye movements also could be delayed by a cortical shock. However, "express" saccades could not be delayed by the stimulus. The results suggest that the cortical stimulus delays movement without appreciably affecting the time of intention to move. This effect probably results from inhibition of the motor cortex which prevents it being engaged in the usual way during vohmtary n~ovement. To explain the results an hierarchical model with feedback elements is proposed for the timing of voluntar)' movement.
I. Introduction A simple concept related to timing of voluntary movements is that our limbs move when we tell them. According to this idea, once the internal instruction to execute a movement is issued, then the appropriate motor act inevitably follows after a relatively fixed delay. The delay is determined by the time taken for neural signals to pass between various central motor structures, culminating in the activation of muscles. In this chapter I shall challenge the view that the "when" of a voluntary motor act is controlled solely by the person's intention. The reason for doubting this simple model springs from a
224
B.L. Day
series of experiments in which subjects attempted to produce a single, predetermined movement in response to an external cue (simple reaction time paradigm) while a stimulus was delivered to the motor cortex in the interval between the cue and the movement (Day et al, 1989b). These experiments showed that the cortical stimulus could delav the onset of movement without appreciably affecting either the movement itself or the time of intention to move. I shall argue that the characteristics of this phenomenon suggest a more complex, hierarchical model for determining the "when" of a movement, which, if correct, has implications for the interpretation of motor timing data.
2. Cortical stimulation and motor timing The experimental procedure is briefly described below. The subject was seated with the shoulder abducted and elbow flexed to 90 degrees. The forearm was semipronated and clamped to a fixed horizontal surface. The hand was encased in a rigid splint which was free to rotate in the horizontal plane about the wrist joint axis. The position of the wrist was transduced by a potentiometer and displayed to the subject on an oscilloscope screen. One second after a visual warning signal (blanking of screen) a l kHz tone was delivered through headphones. This was the signal for the subject to move their hand as rapidly as possible through an angle of about 20 degrees. Electromyographic activity was recorded from pairs of surface disc electrodes overlying wrist flexor and extensor muscles. In one third of the trials, selected randomly, a cortical stimulus was given at a preset time relative to the auditor3; go signal. Two different methods were used to stimulate the cortex through the scalp. 1) An electrical method in which short duration, high voltage stimuli (501as time constant, 700V maximum) were applied between two disc electrodes glued to the scalp, and 2) a magnetic method in which a pulsed magnetic field (nlaximum of 2.IT at 230~ts) was applied via a 9cm diameter fiat circular coil held on the head. The main observation is illustrated in figure 1 which shows averaged movement and EMG traces recorded from one subject. His reaction time (mean + SEM interval between tone and onset of agonist EMG) was 136 + 3ms in control trials. A cortical stimulus delivered lOOms at~er the tone produced a short-latency muscle response just before the usual time of movement onset and the movement was delayed by around 50ms (185 +
An Hierarchical Model of Motor Timing
225
5ms). The duration of the delay was influenced by two main factors. The delay increased both with increasing stimulus intensity and the closer the stimulus was given to the expected time of movement onset (obtained in control trials). However, the triphasic pattern of muscle activity (Hallett et al, 1975) was preserved and showed nonnal form. Furthermore, the appropriate pattern of agonist/antagonist muscle activation was produced depending on whether the subject intended to make a wrist flexion or extension movement. This evidence suggests that spatial characteristics of the movement were largely unaffected by the stimulus.This phenomenon was present in all subjects studied and was similar for both methods of cortical stimulation despite subtle differences in the way these two techniques activate motor cortical cells (Day et al, 1989a). However, it could not be reproduced when a peripheral nerve stimulus was substituted for the cortical stimulus. A short train of supramaximal electrical stimuli to the median nerve at the elbow produces a series of events in tonically active wrist flexor muscles that are superficially similar to those produced by a stimulus to the motor cortex, i.e. a short-latency muscle twitch followed by a prolonged period of electrical silence in the muscle. When this peripheral nerve stimulus was applied during the simple reaction time task, in the same way as for a cortical stimulus, it failed to delay the movement (figure2). Instead, all that was seen was a large suppression of the burst of EMG in the agonist muscle with its timing apparently unaffected. The suppression is understandable if the silent period produced by the peripheral nerve stimulus reflects inhibitory, processes within the spinal cord. These then could exert a similar inhibitory, action on descending motor commands. What is more mysterious is why the silent period produced by the cortical stimulus does not have a similar effect and suppress or abolish parts of the incoming commands to the cortex. A crucial question is whether the cortical stimulus delayed the movement by interfering in some way with the subject's time of intention to move. Did the stimulus cause the subject to issue the internal instruction to "go" at a later than normal time? This is not unreasonable given that the rather crude scalp stimulus used in these experiments could well have activated areas other than the motor cortex. These could be areas involved either with detection of the 'go" signal or with the internal command to initiate movement. However, one set of experiments suggested that the time of intention to move was not substantially altered by the stimulus. For these
226
B.L. Day
/..,%
10"[
' j
too ~,v
[
/
/
/
; w.~:
,,o~,.o.
8
I ld JN
m
,:
'~176
/"%
Rexor
I ii
A~ ~. /
.-
~,',,,
9
Extensor EMG
i :i
"i
i!
,
e
e
Figure 1. Averaged position and rectified EMG traces from a single subject attempting to flex his wrist in response to an auditory 'go' signal given at the start of the sweep. The average control nlovement (solid lines and filled EMG traces) is characterized by alternating bursts of agonist/antagonist muscle activity (triphasic pattern). For the trials in which a magnetic conical stimulus was given lOOms after the 'go' signal (dotted lines), the movement and pattern of muscle activities are similar but delayed. Note that the conical shock (artefact) produced shortlatent' (around 15ms) responses in both flexor and extensor muscles leading to a small extension of the wrist prior to the voluntary flexion movement. (From Day et al, 1989b).
An Hierarchical Model of Motor Timing
227
experiments, subjects were trained to make simultaneous flexion movements of both wrists. Electrical cortical stimuli, intermixed with control trials as before, were randomly applied over either the right or the lett motor cortex. The reasoning behind this experiment was if the stimulus interferes exclusively with the subject's time of intention to move, then a unilateral stimulus should delay movements equally on both side of the body. If. on the other hand, it interferes with an executive process further downstream, then only the limb contralateral to the stimulated henaisphere should be delayed. As sho~aa in figure3a, the timing of movements ipsilateral to the stimulated hemisphere were slightly affected by the cortical stimulus but this effect was small compared to the delay seen in the contralateral limb. This discrepancy in the timing of the two limbs was apparent for a range of intervals between the go signal (auditor3, tone) and the cortical shock (figure 3b). Subjective experience as a subject reinforced the idea that the delay of movement was not explained by alterations of intent. With large shocks applied near to the usual time of movement onset it is possible to obtain long delays of around 150 ms. With long delay,s, the sensation is like a transient paralysis" one is aware of trying to move the hand but with the limb failing to respond. The intention was preserved but the execution blocked. Eye movements also can be delayed by a cortical stimulus in a similar way to limb movements (Priori et al, 1993). In these experiments, subjects were requested to make saccadic eye movements either in response to a jumping visual target or as a volunta~' movement in response to an auditor), tone. As for hand movements, an appropriately timed cortical stimulus delayed both voluntary saccades and visually triggered saccades (figure 4). However, there was one interesting point of departure from the hand when the procedure was changed so that subjects could make visually, triggered saccades with shorter than usual reaction times. These so-called "express" saccades were obtained when the central fixation light was extinguished for a short period before the appearance of the peripheral target light. Express saccades, which are thought to utilize collicular mechanisms without involvement of the cortex (Fischer, 1986). were not delayed by the cortical stimulus. This result supports the idea that the delay in movement occurs through disruption of conical motor executive areas. The motor cortex was strongly activated by the stimulus and is a prime candidate for the site at which disruption of timing occurred. Following the synchronous discharge of corticospinal neurones in response to a transcranial shock, the motor cortex undergoes a prolonged period of
B.L. Day
228
200
~v
'
:J.I
100 m s
o
Ie I $ Q
e
ii
200 ~ V
Figure 2. Comparison of the effect of conical stimulation (top traces) with that of peripheral nerve stimulation (bottom traces) on rapid flexion movements of the wrist. The traces are from rectified and averaged EMG records of wrist flexor muscles in a single subject. The signal to move was given at the start of the sweep. Control flexor muscle activity is shown by the filled traces. The average EMG activity, when an interposed stimulus was given, is shown by the dotted traces. Both the cortical (electrical stimulation) and median nerve stimuli (train of 3 supramaximal stimuli separated by 4ms) were timed to produce a direct muscle response just before the usual time of agonist vohmtary nmscle activation. The movement EMG pattern remains intact but delayed by the cortical stinmlus whereas the peripheral nerve stimulus inhibits the first agonist burst of the movement EMG pattern without affecting its timing. (From Day et al, 1989b).
An Hierarchical Model of Motor Timing
229
inhibition (lnghilleri et al, 1993). It is likely that this transient inhibition of the motor cortex prevents its normal engagement during the act of initiating a rapid movement; only when the excitability of the cortex returns to nonnal levels can the final stages of processing be completed successfully. Two main characteristics of the delay phenomenon can be explained on this basis. First, for a rapid voluntary movement, the motor cortex is brought into play towards the end of the preparatory period (Evans, 1972) and so a shock given at that time would have maximum delaying effect on the movement. With earlier shocks the delay would become progressively shorter. Second. the duration of motor cortical inhibition is longer for higher intensity stimuli (Inghilleri et al, 1993) and therefore greater delay would be expected to occur with larger shocks. This idea of inhibition preventing the nonnal engagement of the motor cortex for a fixed time, therefore, goes some way towards explaining the observed phenomena. However. it does not help to explain why the movement eventually 'popped-out' with nonnal form after the inhibition had abated. Why did the cortical inhibition not act to suppress incoming signals in an analogous way to the effect of the peripheral nerwe stimulus? One possibility is that upstream areas are able to store, at least for a short time, the information required by the motor cortex to produce the movement. These areas also wottld need to have some knowledge about the state of the motor cortex in order to influence the timing of information flow to it. Such knowledge could feasibly be gained from feedback infommtion that informs either on the state of readiness of the motor cortex to receive instructions or on the progress of the motor instructions through the cortex. Whatever the nature of this information, the important point is that the timing of delivery, of coded instructions to spinal motoneurones may depend upon at least two partially independent processes. A high-level process that instructs motor centres to complete the motor preparation and release the movement, and a subordinate middle-level process that exerts fine control over the precise timing of the movement depending upon infomlation received from the motor cortex. A simplified, schematic model of such a timing structure is illustrated in figure5. In this figure the black lines indicate flow of timing information and the white lines indicate flow of spatial infommtion. For a simple reaction task, in which the intended movement is known in advance of the 'go" signal. the subject is capable of a certain amount of early preparation of the movement which is represented by the pre-cue preparation box. The evidence for this is the fact that shorter reaction times are obtained in simple
B.L. Day
230
Slhm. L hemisphere
Stim. R hcmsspherr
J~i
z.so ~ v r
R flexor EMO
_~til I , .: t
~ ~f.
!\!\
'7"~'-'g2~ t
ib"l
L fl ....
L flexor E M G
EM~
120"
ira)
6o.
2o
~j"If"
Inr162 Ior~c--br~in ~tlm. (ms)
Figure 3. Effect of unilateral conical stimulation on attempted simultaneous. bilateral wrist flexion movenlents. Electrical stimuli were delivered either to the right or the left hemisphere. A rectified and averaged EMG traces from the right and left wrist flexor muscles of a single subject in response to a "go' signal given at the start of the sweep. Control movements (filled traces) are performed synchronously whereas movemelllS in which cortical stimuli are given (dolled traces) are performed asynchronously. The first agonist burst from the side contralateral to the stimulus is substantially delayed whereas the ipsilateral burst is little affected. B illustrates the mean (+/- SEM) delaying effect of a unilateral cortical stimulus on the contralateral and ipsilateral wrist flexion movements of 3 subjects. For contralateral movements, lhe cortical stimulus produced progressively greater delay the later it was given in the reaction period. (From Day et al, 1989b).
231
An Hierarchical Model of Motor Tinu'ng
I ~,Qms Tost
Figure 4. Delay of sonic t3.'pes of saccadic eye movements by magnetic stimulation of the cortex. A superimposed raw EOG traces from one subject making rapid eye inovements to fixate a peripheral LED which was illuminated 125ms before the start of the sweep. The saccades from trials in which a cortical stimulus is given (lower traces) are clearly later than Ihose of the control trials (upper traces). In test trials, the artefact is due to blocking the recording amplifiers during stimulation. In control trials, the artefacl is due to the firing of a magnetic stimulating coil, placed behind and off the head of the subject, to mimic the sound of the conical stimulating coil in test trials. B superimposed raw EOG traces of a subject making "'express" saccades. These were produced by blanking the central fixation point 200ms before the appearance of the target light. In test trials. conical stimuli were applied (double-headed arrow) 50ms after the target was illuminated (start of sweep). This produced an artefactual DC shift in the traces. The saccades begin some 60ms later (vertical line). In control trials, there is a similar artefact due to the discharge of a control magnetic coil placed behind the head. The cortical stimulus does no! delay these "'express" saccades. (From Priori et al, 1993).
versus choice reaction tasks in which the subject does not have advance knowledge of the required movement (Rosenbaum, 1980). However, even for the simple reaction task, the reaction time is usually greater than lOOms. Some of this time is consumed by processes concerned with detection of the 'go' signal and some by the time taken for the motor cortex to activate muscles (less than 20ms). I will assume that not all the reaction period is explained by these two processes and that some time is required for further processing of motor infonuation. This extra processing, which takes place after the 'go" signal has been detected and the decision to move has been made, is represented by the post-cue preparation box. As outlined above, the delay in movement onset caused by the cortical shock suggests the presence
B.L. Day
232
of internal feedback that modifies the time that the motor cortex is brought into play during voluntary movement. Because the delay was variable (depending upon stimulation conditions), and could be very short, it is unlikely that the cortical shock destroyed the information obtained during post-cue preparation. If this were the case, the process would need to be repeated resulting in relatively fixed and, presumably, longer delays. Instead, I suggest that the motor information obtained during this final preparatory stage is held in a short-tem~ memory store that allows rapid activation of motor cortex as soon as it is ready to receive and refine this information.
DECISION
~
~,~
" CUE " DETECTION
n SPINAL CORD ,
,
===iF Timing information ,-~- :~;; >
Spatial information
Figure 5. Schematic flow diagram of a possible hierarchical timing control system incorporating internal feedback elements. See lext for details.
An Hierarchical Model of Motor Timing
233
The proposed model, therefore, is essentially hierarchical with internal feedback elements. An important property of such a mechanism would be that our limbs would not necessarily move when we tell them. A complex timing model of this sort makes the interpretation of timing data from subjects with CNS lesions problematical. Abnormal motor rhytlma may be due to damage of the high-level timing circuitry, or to the proposed middlelevel timing circuitry or may even be due to abnormalities of motor cortical recover 3, time between movements.
3. References
Day BL, Dressier D, Maertens de Noordhout A. Marsden CD, Nakashima K, Rothwell JC. Thompson PD. Electric and magnetic stimulation of human motor cortex: Surface EMG and single motor unit responses. J Physiol. 1989a: 412: 449-73. Day BL, Rothwell JC, Thompson PD, Maertens de Noordhout A, Nakashima K, Shannon K, Marsden CD. Delav in the execution of voluntary movement by electrical or magnetic brain stimulation in intact man. Evidence for the storage of motor programs in the brain. Brain. 1989b: 112: 649-63. Evans EV. Pre- and post-central neuronal discharge in relation to learned movements. In" Frigyesi T, Rinvik E. Yahr MD, editors. Cortico-thalamic projections and sensorimotor activities. New York" Raven Press. 1972' 44958. Fischer B. Express saccades in man and monkey. Prog. Brain. Res. 1986" 64' 155-60. Hallett M, Shahani BT. Young RR. EMG analysis of stereotyped voluntary. movements in man. J. Neurol. Neurosurg. Psychiat. 1975' 38" 1154-62. Inghilleri M, Berardelli A. Cnlccu G, Manfredi M. Silent period evoked by transcranial stimulation of the human motor cortex and cervicomedullary junction. J. Physiol. 1993' 466' 52 i-34.
234
B.L. Day
Priori A, Bertolasi L, Rothwell JC, Day BL, Marsden CD. Some saccadic eye movements can be delayed by transcranial magnetic stimulation of the cerebral cortex in man. Brain 1993: 116: 355-67. Rosenbaum DA. Human movement initiation: specification of arm, direction, and extent. J. Exp. Psychol: General. 1980: 109: 444-74.
Time, Internal Clocks and Movement M.A. Pastor and J. Artieda (Editors) 9 1996 Elsevier Science B.V. All fights reserved.
I N V O L V E M E N T OF THE BASAL PERCEPTUAL AND MOTOR TASKS
GANGLIA
235
IN
TIMING
MARIA A. PASTOR and JULIO ARTIEDA
Neurophysiology ,Section. Department of Neurology and Neurosurger), Clinica Universitaria. Universidad de Navarra. 31080 Pamplona. ,Spain. ABSTRACT. Adequate spatial and temporal perception of tile external world and of one's body are required to achieve accurate motor performance. Deficits in tinting are evident in most abnormalities reported in Parkinson's disease such as: delayed reaction time, increased movement time, greater inter-onset latency in sequential movements and slowness in the performance of repetitive movements. There is some evidence which indicates that the tinting alteration of patients with PD may be secondary either to inadequate temporal perception or to the dysfunction of the internal generator responsible for triggering m o v e m e n t . Various aspects of timing have been examined in patients with PD. Compared to age-matched controls patients had higher temporal discrimination thresholds in the auditory, visual and somaesthelic sensory modalities. In time estimation and reproduction of time intervals tasks. PD patients overestimated the duration of time intervals reproduced by inlernal counting compared to normals. Patienls showed greater variability in the performance of repetitive alternating flexionexlension movements of the wrist al higher frequencies. The results suggest Iha! the tinting deficits of patienls wilh PD involve motor and perceptual tasks.
I. Introduction
Parkinson's disease (PD) is one of the best studied basal ganglia dysfunctions, due to the degeneration of the dopaminergic nigro-striatal pathway which produces denervation of the striatum. Traditionally, Parkinson's disease has been considered as a motor disorder, characterized by a symptomatic tetrad: resting tremor, rigidity, disturbances in postural control and akinesia~radykinesia, the latter being the most specific s}anptom. At present, akinesia or bradykinesia can be regarded as umbrella terms, describing several automatic and voluntary, movement dysfunctions. The origin of these clinical motor symptoms is poorly understood at present. However, several sensor3' and perceptual abnomlalities have been reported
236
M.A. Pastor and J. Artieda
recently indicating a broader sensorimotor integration.
physiopathology
which
may
involve
2. Automatic movements
The pioneers of motor control physiology, William James and Sir Charles Sherrington, considered movement as a chain of reflexes controlled by sensory afferents. This perspective was revised as a result of the report, by Graham Brown and Karl Lashley, that motor impulses could be produced in the absence of sensory afferents. As a result a new theory' based on the leading role of the central mechanisms of motor control was introduced. This approach was supported by showing that feeding, respiration, scratching, locomotion and other automatic movements involve rhythmical motoneuronal activity which may persist even when the rhythmically discharging neurons have been deprived of inputs from peripheral sense organs and other parts of the nervous system. Automatic movements are executed by the subject without the involvement of a voluntary drive and without the contribution of sensors' feedback. The neural networks which generate rhythmical outputs are commonly referred to as central pattern generators (CPGs). Several types of neurons are involved in the structure of a CPG but the main role is due to timekeeper neurons that maintain the sequence of discharge and rhythm of movement (see Sillar in this book). These CPGs might also constitute, in man, the basis of automatic movements such as breathing, locomotion and spontaneous blinking (Ponder and Kennedy, 1928). These automatic movements are altered in PD, their rate being reduced or even abolished. Frequency of spontaneous blinking is greatly reduced in PD. This abnormality is reverted with levodopa administration (Karson, 1983). Rest tremor is an abnormal automatic movement frequently present in PD. Intraneuronai recordings have shown rhythmic firing of the sensorimotor cortex and ventrolateral thalamus synchronized with parkinsonian tremor (Albe-Fessard et al., 1966: Jasper and Bertrand, 1966). These rhythmic discharges do not disappear following deafferentation of the tremoric limb or paralization with curare (Foerster, 1936; Joffroy and Lamarre, 1971" Lamarre and Joffroy, 1979). From these findings it is possible to infer that the timekeeper causing the tremor may be located in the cortex or thalamus. In fact, lesion of the ventralis intennedialis (Vim) results
Involvement of the Basal Ganglia in Timing
237
in the abolition of tremor (Bucy and Case, 1949" Narabayashi, 1982: Lenz et ai., 1994).
3. Disorders of voluntary movement in PD The genesis of a voluntary movement can be explained on the basis of a simple model from computer science. The first step in this process is the decision to move preceded by the idea of movement. For this purpose a motor plan must be determined, integrating temporal and spatial information from the environment and the body. The motor plan establishes which motor programs are to be used. their sequence, timing and spatial quantification. The motor program is considercd as the algorithm which establishes for a particular motor action the muscles involved (agonist. antagonist, synergists and posture fixers), their sequence of activation, amplitude, duration and timing between contractions. This is adjusted according to the force, speed, and amplitude of the required movement. When the movement has been formulated, the command is transmitted to the different interfaces (primar)., motor cortex, anterior horn motoneuron, neuro-muscular junction) for its execution. During the execution of movement there is a continuous flow of information regarding the traicctor3' which comes in from the sensory. afferents. This speculative approach, modified from Allen and Tsukahara (1974), may be useful in the analysis of sensorimotor integration in PD. Kinnier Wilson, in 1929, demonstrated, in a patient with PD affecting the left side of the body, a delay of 750 ms in the contraction of the left rectus femoris in comparison with thc contralateral muscle when the patient tried to sit up from a lying position. Ycars later he describcd a delay in the response to a simple reaction time paradigm to a visual stimulus in PD patients (K.Wilson, 1947). Several groups have confirmed these results using different stimulus modalities (Evarts ct al., 1981" Bloxham et al., 1984) but no correlation has been found between these abnormal responses and bradykinesia (Evarts et al., 1981' Jahanshahi et al,. 1992). These findings in reaction time studies do not. however, explain the slowness of movement. Bradykinesia is better related to the increased time employed in the execution of the movement (Flowers, 1975). Simple ballistic movements are good models for preprogrammed movement. These movements are executed as an "open loop", independent of sensory feed-back. They present a triphasic EMG pattern: the initial EMG burst of the agonist muscle which generates the force of the movement,
M.A. Pastor and J. Artieda
238
followed by the antagonist EMG burst stopping the segment displacement, and the third EMG burst of the agonist which helps to achieve the final position accurately. This stereot3,'ped agonist-antagonist-agonist EMG pattern is archetypal of this movement. The amplitude, duration and timing of EMG burst depends on the degree of the amplitude and the duration of the planned movement. In PD patients the amplitude of the first burst of the agonist is defective. This abnormality is present both in axial and distal muscles (Wiesendanger et al., 1969; Berardelli et al., 1984 1986), (Figure 1).
.CONllK)L
FIw.
-- ......
jkll-_
.at_
PAI~KINSON
__
,,
. . . .
•JJA jr
,.
....
..! lql
1~]
Figure 1. Recordings of ballistic movements of the wrist in two subjects, a healthy control subject and a Parkinson's disease patient. Flexor and extensor EMG surface electrode recordings and angular displacement of the wrist.
A hypothetical motoneuron hypoexcitability may lead to a hypometric ballistic movement by recn~iting (temporally and/or spatially) a decreased number of motor units in the agonist muscle. Although the firing pattern of motor units during a tonic contraction in PD patients follows the Hennemann principle, there is a decreased discharge frequency and a greater variability of the inter-discharge interval (Dietz et al., 1974; Milner-Brown et al., 1979). The alpha-gamma interaction is normal (Burke, 1977). There is some evidence of abnormal spinal and transcortical reflex mechanisms in PD (Obeso et al., 1985" Artieda et al.. 1986: Dick et al., 1984). However, the velocity of evoked movements by electrical stimulation of the motor cortex in patients with P D is normal in contrast to the increased time of performance of vohmtary movement (Dick et al., 1984). As in normal
Involvement of the Basal Ganglia in Timing
239
subjects, subthreshold cortical magnetic stimulation shortens reaction time in PD patients (PascuaI-Leone ct al., 1994) and normalizes the triphasic EMG pattern. In PD the initial premovement period of relative innexcitability found using this method, is short. What seems to influence the increase in a simple reaction time paradigm is the enlargement of the progressive excitability second period. Patients with PD are capable of modulating the amplitude and duration of the first agonist burst during a ballistic movement according to distance but movement remains abnormally slow and the agonist burst is scarcely "energized" (Berardelli et al., 1086). These abnormal findings again could be due to errors in temporal quantification of motor programs derived from either altered temporal and/or spatial perception or anomalous integration during the cerebral processing of movement.
4. Perception in PD Even though the original publication on Parkinson's disease does not describe sensory symptoms, in 1940 Kinnier Wilson wrote: "Though one or other sensory defect has figured now and again in the records of the disease, objective change is merely incidental. Reduction of cutaneous sensitivity to the electric current is doubtless largely accounted for by dry or cold skin. On the subjective side, however, pains, aches, feelings of stiffiless, "drawing", "dragging", "tightening", and so on can be ascribed in part at least to muscular rigidity; beside, the age of most parkinsonians is one at which rheumatic, arthritic, arteriopathic and other morbid factors may come into play". Forty three percent of PD patients complain of sensory alterations but it is infrequent to find abnomlalities in a conventional clinical examination (Snider et al., 1976 Koller, 1984). The pioneer studies on perception in PD were based on previous findings in frontal Iobectomized patients who each underwent an ample resection involving the basal ganglia (Teuber and Mishkin, 1954). Since then dysfunction of verticality judgment, body scheme, extrapersonal spatial orientation, and spatial perception have been reported (Proctor et al., 1964: Danta and Hilton, 1975" Bowen et al., 1972: Bowen, 1976). However most of these tasks included a motor component which may have masked the results. Errors in visuospatial tasks which only require memorization have been described (Mortimer et al., 1982; Lancey Home, 1973). Although Boiler et al. (1984) distinguishing the motor
240
M.A. Pastor and J. Artieda
component in spatial perceptive tasks found some pure perceptual abnormalities. Schneider et al.. (1986) studied somaesthetic perception in the trigeminal area and its relationship with orolingual movements and found an abnormal spatial proprioceptive perception from the temporo-mandibular joint in PD patients. These findings were not restricted to the oromandibular area, they subsequently demonstrated altered spatial perception in wrists and fingers (Schneider et al.. 1987). During the last two decades, several groups have been working on frontal lobe function in P D. Bowen et al. (1975) described impaired performance on the Wisconsin Card Sorting Test, involving "set shifting". Since then many authors have reported impaired performance of PD patients on neurophysiological tests sensitive to frontal dysfunction (Lees and Smith. 1983: Cools et al., 1984; Stem et al., 1984; Flowers and Robertson, 1985" Pillon et al., 1986- Taylor et al., 1986, 1987; Brown and Marsden, 1988" Calta-Girone et al., 1989; Levin et al., 1989). Brown and Marsden (1986, 1988. 1990) in particular, limited this inability to the use of internal clues to direct behaviour, to the so-called Supervisory attentional system (SAS). They found in PD patients using a simple visuospatial test, nommi right-left discrimination and utilization of information (Brown and Marsden, 1986). Using a processing negativity paradigm Stam et a1.(1993) tested selective attention in PD who had abnormal frontal tests, to find disturbed processing negativity directly correlated in magnitude with frontal impairment. A cognitive deficit involving the monitoring of the stimulus-response in a complex reaction time paradigm has been described recently (Cooper et al., 1994). Others afirm the abnormal learning of predictions of movement trajectories on tracking due to fronto-striatal dysfunction (Schnider et al., 1995). What is still a matter of discussion is the role of associated dementia in these findings, the result of the impairment of the caudate-thalamus-frontal cortex, "complex loop" (DeLong et al., 1983), and the invovlement of an impoverished mesocortical dopaminergic system.
5. Time perception and estimation in PD There are a limited number of studies in the literature regarding time perception in PD patients, fi~rthemaore they lack a precise clinical and neuropharmacological assessment of the subjects. Dinnerstein (1962, 1964) considered that the time needed for somaesthetic stimulus perception in PD patients to be greater. He compared the time used in the perception of a
Involvement of the Basal Ganglia in Timing
241
somaesthetic and a visual stimulus given simultaneously and in a second task the perception of an auditor3., and a simultaneous visual stimulus. He found visual perception preceding somaesthetic and auditory perception in PD patients, contrary to age-matched controls. These results matched the Venables and O'Connor (1959) and Talland (1964) findings: reaction time to auditory stimulus was significantly longer compared with reaction time to visual stimulus in schizophrenic patients treated chronically with neuroleptics. The delayed response to auditory stimuli was correlated to bradykinesia. Supported by these results Dinnerstein concluded that central mechanisms may be involved in the delay' of somaesthetic and auditory. perception. Riklan et al. (1970) studied, in 37 PD patients, the critical flicker fi~sion point (CFFP) that is the minimal frequency necessar), for a train of visual stimuli to be perceived as a continuous. The severity and clinical characteristics of the patients were not specified. They observed a significant correlation between the CFFP and the severity of the disease. In a second study by the same group (Maskin et al., 1974) the CFFP was studied in 3 groups of subjects: PD patients on Icvodopa therapy a short time before, PD patients on long term icvodopa therapy and age-matched nonnal subjects. CFFP was increased in PD patients in comparison with normals and within the PD group those patients who had been under levodopa therapy longest were the most severely affected. It is probable that these results were more directly related to the severity of the disease than to levodopa therapy per se. It is possible that the group of patients who had started levodopa treatment earlier were more severely affected at the time of the study. One approach to the study of temporal perception is the assessment of temporal discrimination. Temporal discrimination threshold is the minimum time interval required between two successive auditory, visual or somaesthetic stimuli for these to be perceived as separate. Temporal discrimination thresholds have been found to be significantly increased in Parkinson disease (Figure 2). The administration of a single levodopa/carbidopa oral dose significantly reduces the thresholds of the three sensory modalities, particularly the somaesthetic (Artieda et al., 1992). This abnormality is correlated with disease severity and cannot be related to cognitive impairment or dysfi~nction of sensory, receptors or pathways. The delay in visual evoked potentials found in PD patients (Bodis-Wollner and Yaar, 1978). due to the presence of dopaminergic cells in the retina, can hardly explain the impairment of visual temporal discrimination in PD which would require an interference betxvcen the perception of the first and second
242
M.A. Pastor and J. Artieda
stimulus of each pair, not just a delay in the arrival of the two stimuli to the visual cortex. Abnormal somaesthetic temporal discrimination thresholds have been found in PD, despite the presence of normal sensory nerve action potentials and somaesthetic evoked potentials recovery curves following double stimulation. Strikingly. the highest threshold was for somaesthetic stimuli and there was a significant correlation with this value and the clinical rating scale, simple reaction time, movement time and a tapping task. These findings and the fact that there is a significant improvement after levodopa, gives more than a suggestion of the implication of the dopaminergic system in controlling temporal discrimination.
I1U
140 120 100
rl
i " 4O 20 0
7
Somaesthetk
1
[] CONTROL [] PARKINSON T T
I Visual
Auditon/
Figure 2. Somaesthetic. auditon' and visual temporal discrimination thesholds in Parkinson's disease patients and age-matched controls. Significant differences between groups were obtained for the three modalities.
It was necessary to explore whether this abnormality was task dependent or if there was a wider implication of the basal ganglia in different types of temporal perception. The estimation of a short time interval and its subsequent reproduction was assessed in a group of Parkinson's disease
Involvement of the Basal Ganglia in Timing
243
patients and compared with age-matched controls. The method consisted of the presentation of time intervals delineated visually by a series of 15 numeric time markers. The timc markers were presented at three different rates: 5, 3.3, 1.6 Hz. This resulted in time intervals of three different durations; 3, 4.5, and 9 seconds, hrmaediately following the presentation of the last time marker (number 15) which signalled the end of the time interval, the subject was asked to reproduce the previous time interval by internal counting of 15 time markers at the same rate as that of the immediately preceding time sample. The end of the reproduced time interval was indicated by the subject by pressing a key. When required to reproduce a short time sample, however, patients with Parkinson's disease overestimated the duration of a time interval compared to nomml controls. For both the patients with Parkinson's disease and the controls, the overestimatation in the reproduction of time intervals increased in magnitude with the lengthening of the duration of the time sample. The increase in the rate of presentation of the time markers from 1.6 to 3.3 to 5 Hz significantly inflated the absolute error of temporal estinaation. These findings suggest that the patients with Parkinson's disease had difficulty adjusting their internal tempo to the rate of the external time markers especially when the latter were presented at the two fastest rates of 5 and 3.3 Hz. Thus, with the lay3~othetical internal clock nmning at a slower rate than that set by the external tempo this resulted in an over-production of the time intervals by the patients with Parkinson's disease. The severity of Parkinson's disease, had a significant effect on the accuracy of time estimation and reproduction tasks. More importantly, reaction and movement time, used to quantif).' slowness of movement initiation and execution, were significantl.v correlated with time estimation and reproduction accuracy. These findings suggest that the timing deficits in perceptual tasks such as those of the present study may have a common substrata with the slox~.lless invariable documented in the initiation and execution of movements by patients with Parkinson's disease (Artieda et al, 1992, Pastor et al, 1992a).
6. Time and movement in PD. Repetitive and sequential movements Complex movements may be generated either by a single complex motor programme or by assembling several simple motor programmes. Complex movements such as handwriting, typing, talking, walking, etc, might be
244
M.A. Pastor and J. Artieda
controlled by a single motor programme (Carter and Shapiro, 1984: Denier van tier Gon and Thuring, 1965, Terzuolo and Viviani, 1980; Tuller et al, 1982). In PD, these complex movements are the first to be impaired. The difficulties inherent in electrophysiological studies of complex movements have led scientists to limit analysis to isolated aspects of the simplest movement components. Schwab et al. (1954, 1958) were the first to recognize that PD patients have difficulty in performing sequential movements. Inter-onset latency of sequential movements employing two different motor programs is increased in Parkinson's disease and is better correlated with bradykinesia than simple movement motor time. These fi,~dings suggest that P D patients have both a dysfunction in quantifying simple motor programmcs and in the temporal assembling of motor programmcs in order to generate a motor sequence (Benecke et al, 1986, 1987). In PD patients, the perfonnance of repetitive movements has also been used as a grading parameter although its precise interpretation has not been known. In clinical practice common manoeuvres to examine akinesia are tapping with the hand or the foot. The first description of the decrement in amplitude of repetitive movements in PD patients was made by Souques in 1921, who examined pronation-supination movements of the wrist. A repetitive alternating wrist flexion-extension movement paradigm is a good model for the analysis of timed triggered movements. The performance of repetitive rhythmic movements in the presence of external auditory feedback and when such rh)~hnaic naovements are self-paced has been studied in patients with Parki,~son's disease compared to age-matched controls. Surface EMG have been recorded from the flexor and extensor carpi muscles of the forearm. The angular displacement of the wrist was measured by a potentiometer. The subject was asked to synchronize the flexion-extension movements with the auditory stimuli, (a series of 30 tones) and to continue the flexion-extension movements at the same rate after the tones had stopped. The block of trials was ended when 30 self-paced movements had been performed after cessation of the tones. The tones were presented at 5 different frequencies" I, 2, 3, 4 and 5 Hz. The time interval between the onset of 2 flexion movements were calculated. Both patients and controls were able to maintain the rate of movement after cessation of the auditory stimuli which provided an external pacing of the repetitive movements during the first 30 trials. PD patients were unable to reach normal frequencies of alternating rh.~hmic movements and they showed a lack of synchronization when following external auditor}, rhythmic stimuli
Involvement of the Basal Ganglia in Timing
245
particularly at rates higher than 3 Hz. The variability of IRI was greater than in normais and frequency related. The variation coefficient was significantly higher at frequencies of 1 and 2 Hz and at 4 and 5 Hz, these two groups of sequential movements appeared to be generated by different
40-1 84
SOOpv
o.~ s,,~.
Figure 3 Recordings of a sequence of ~lllernating movements of tile wrist. Upper traces: Control subject performing flexion-exlension movements at 5 Hz, is noleworthy the exaclitude of angul~lr displacement. Lower traces: Parkinson's disease patient, showing distorlion of movement consequence Io fragmentation of EMG bursts and lack of rh~ahm.
strategies, the slower one by a simple reaction time paradigm and the faster one by preprogranuning the sequence at the presented rate (Pastor et al. 1992b).
7. DA and time in animal studies Results of research using animals have led to the conclusion that dopamine antagonists such as haloperidol and drugs such as methamphetamine
246
M.A. Pastor and J. Artieda
decrease and increase respectively the speed of the internal clock. Rats treated with methamphetamine underestimated a time interval, and under the effect of haloperidoi overestimated the presented time interval (Maricq et al, ms ~X~
I
900IS(X)-
ON
700600t5008
&00-
I
300-
i
l
I
I
~.5
S
200100-
I
w
II
I
1.5
2
I
I
2.5 3
9
I
3.5 4
I
Hz
I]I
m s
1000!K)0eoo-
I1
700-
OFF
600s(x)t,o o 30021:1o100
!
1
Itlll
~.s 2
9
I
2.w $
e
9
9
9i
3.5 ~, t,.5 s
9 9
9
Hz
Figure 4. Effect of a single oral dose of levodopa on the performance of repetitive wrist movements. Mean and standard error of the inter-response interval at frequencies from 1 to 5 Hz. Lower diagram: patient tesled after withdrawal of medication. Upper diagram: The same patient after a single dose of levodopa.
1981; Maricq and Church. 1983). Other evidence suggests that the methamphetamine-induced impaimacnt of temporal discrimination may be due to changes in the response rate rather than be the result of any alterations in sensitivity to temporal cues (Segal, 1962; Robbins and Iversen, 1973). However, use of experimental procedures to isolate clock speed from factors such as changes in response rate. led Maricq (1981) to conclude that
Involvement of the Basal Ganglia in Timing
247
methamphetamine increased clock speed. Furthermore, in simple motor networks, such as the 14 neuron pyloric circuit of the lobster somato-gastric ganglion, the frequency and amplitude of pacemaker potentials have been shown to increase following exposure of the synaptically-isolated pyloric neurons to dopamine baths. The mechanism comprises excitation and phaseadvance of the piloric constrictor neurons in a rhythmic motor pattern at least in part by reducing the transient K+ current (Harris-Warrick and Flamm, 1986" Harris-Warrick ct al., 1995). As the speed of the internal clock appears to increase with dopamine agonists and decrease with dopamine antagonists, it has been proposed that dopamine might be a modulator of the internal clock (Maricq. 1981" Maricq and Church, 1983). In experiments on rats the grooming sequential actions, which follow characteristic patterns of serial order, were observed following lesions of the neocortex, cerebellum or striatum. Berridge and Whishaw (1992), discussed the role of the striatum as implementing the pattern by modulating the activation of completing sensorimotor and central pattern generating circuits elsewhere in the brain. In their opinion the striatum may promote the completion of a grooming chain through a phasic and selective facilitation of the access of the chain's brainstem syntax generator to motor output systems, while simultaneously channelling motor control access away, probably in a graded fashion, fiom completing programs and sensorimotor circuits. In sunmmr3' the role of the striatum is that of a hierarchic controller, which turns other circuits in the brainstem and elsewhere on and off as behaviour progresses. Some indirect evidence in support of this hypothesis can also be found in the clinical literature. Wahl and Sicg (1980) and Tysk (1983) have reported that chronic schizophrenics overestimate the duration of short time intervals, using a verbal estimation paradigm with the second as a measure, which can be corrected under treatment with neuroleptics (Wahl and Sieg, 1980). The effects of levodopa therapy in the present study also lends support to the role of dopamine in the modulation of the internal timekeeper. Following the administration of levodopa a significant improvement in time discrimination. estimation and reproduction by the patients with Parkinson's disease was observed, especially for those tests with faster rates of presenting the time markers (5 and 3.3 Hz). Compared to when tested off medication, administration of dopaminergic medication also resulted in significantly more accurate timing of repetitive sequential movements, especially at the higher frequencies. The mean interresponse intervals of the patients when tested on medication were not
248
M.A. Pastor and J. Artieda
different from those of the controls, even at higher rates of movement. Levodopa medication seems to improve the ability of patients with Parkinson's disease to speed up their internal clocks in order to adjust the internal tempo to that of externally presented stimuli. The participation of other neurotransmitters has to be defined, however, it is well known that serotoninergic, noradrenergic, neuropeptidergic systems, are affected in P D (Heim et al., 1986; Mann et al., 1983; Fann et al., 1971). A direct influence over the noradrenergic system through the transformation of levodopa into noradrenaline, could be possible. The degeneration of dopaminergic neurones in the ventrotegmental, mesolimbic and mesocortical projections might influence temporal perception.
8. Anatomophisyological basis in PD The abnormal findings in perceptual and motor timing in Parkinson's disease (Artieda et al., 1992; Pastor et al, 1992ab) suggest that the basal ganglia are involved in the internal timing processing. The striatum represents the main afferent structure of the basal ganglia. The most important afferents to the striatum are the dopaminergic nigrostriatal pathway and all cortical areas. The striatum projects somatotopically to both palidal segments (GPM and GPL) and through different cell populations with no overlapping. There is a direct pathway from putamen to GPM from a gabaergic-substantia P-dynorphin population. Another indirect pathway (Gabaergic-enkephalin subpopulation) arises from the putamen to the GPL. The GPL produces inhibition of the subthalamis nucleus, which sends excitatory afferents to the GPM. The GPM sends topographically organized projections to the thalamic ventral anterior and ventral lateral nuclei. These thalamic nuclei mainly project to the supplementary motor area. It is in these thalamo-cortical circuits that the final influence of the striatum on timing movement may be established. Lacruz et al. (1992) found a high incidence of abnormal tactile temporal discrimination thresholds, in patients with basal ganglia lesions. It is knox~aathat the main cortical output from the basal ganglia extends via the lateral thalamus to the supplementar3' motor area (SMA), which in turn projects heavily to the superior parietal lobe. In summary, the nigro-striatal dopaminergic deficit may induce timekeepers or CPG disfunction, underlying temporal perceptual and motor tasks and tremor generation, possibly as a result of the increased inhibitory pallido-thalamic output (Figure 5).
Involvement of the Basal Ganglia in Timing
249
I RI.~I~!I~"EDAIITOMATIC [ kltlA,~MENTS
TEMPtlJRAL'PERCEPTI(JN DlS,:,RDERS ~ ....
...,o
,.
9,
"~176176176176
I S'MPLEa':;v I
~
1
t
.~,,,~I~.~EN'rs
"
......
I TREM':'RI
~
,~R;ENTIAL MOV. I
t .,,,,,. 1 Figure 5. Representation of the hypothetical implications consequence to alteration of internal limekcepcrs
ACKNOWLEDGEMENTS. This work was supported in part by a grant from the Spanish governemen! DGICYT PB92-1)713. The authors wish to thank Mrs Isabel Sanchez and Maria Lopez for their help in editing this chapiter.
9. References
Albe-Fessard, D., Guiot, G., Lamarre,. Y. Activation of thalamo-cortical projections related to tremorogcnic proces. In Yahr M, Purpura D, (editors), The thalamus, Columbia Universityo Pres. New York: 1966: 237-54. Allen, GI., Tsukahara, N. Cerebrocerebeilar communications systems. Physiol. Rew. 1974:54:95 7-1006. Artieda, J., Quesada, P.. Obeso. J A.. Martinez-Lage, JM. Fisiopatologia de la rigidez en la enfermedad de Parkinson: Estudios electrofisiologicos. In"
250
M.A. Pastor and 3. Artieda
Obeso JA., Martinez-Lagc J M, editors. Enfermedad de Parkinson y movimientos anormales. Pamplona 9Eunsa. 1986 Artieda, J., Pastor, MA., Lacruz, F., Obeso, JA. Temporal discrimination is abnormal in Parkinson's disease. Brain 115: 1922: 199-210. Benecke, R., Rothwell, JC., Day, BL., Dick, JPR., Marsden, CD Motor strategies involved in the performance of sequential movements, Exp.Brain. Res. 1986: 63, 585-95. Benecke, R., Rothwell, JC., Dick. JPR., Day, BL., Marsden, CD. Disturbance of sequential movements in patients with parkinson's disease, Brain. 1987:110:361-79 Benecke, R., Rothwell, JC.. Dick. JPR., Day, BL., Marsden, CD. Performance of simultaneous movements in patients with Parkinson's disease, Brain. 1986: 109:739-57 Berardelli, A., Dick, JPR., Rothwell, JC., Day, BL.. Marsden, CD. Scaling of the size of the first agonist EMG burst during rapid wrist movements in patients with Parkinson's disease, J Ncuroi. Neurosurg. Psychiatr. 1986: 45: 1273-9. Berardelli, A., Accomero, N., Argenta, M., Meco, G., Manfredi, M. Fast complex ann movements in Parkinson's disease, J Neurol. Neurosurg. Psvchiatr. 1986. 49:1146-4t). Berridge, KC., Whishaw IQ. Cortex. striatum and cerebellum: control of serial order in a grooming sequence. Exp. Brain Res. 1992: 90: 275-90. Bloxham, CA., Mindcl. TA.. Frith, CD. Initiation and execution of predictable and unpredictable movements in Parkinson's disease, Brain. 1984: 107: 371-84. Boiler, F., Passafiume, D.. Keefc. N., Rogers, K., Morrow, L., Kim. Y. Visuospatial impainnent in Parkinson's disease, Role of perceptual and motor factors. Arch. Ncurol. 1984" 4 I: 485-90.
Involvement of the Basal Ganglia in Timing
251
Bowen, FP., Hoenh, MM.. Yahr. MD. Parkinsonism: alterations in spatial orientation as determined by' a route walking test, Neuropsychol. 1972; 10' 355-61. Bowen, FP. Behavioral alterations in patients with basal ganglia lesions, in: The basal ganglia Yahr MD, editors. Raven Press 9New York, 1976. Brown, RG., Marsden, CD. Visuospatial function in Parkinson's disease, Brain. 1986" 109' 987-1002. Bucy, PC., Case, TJ. Tremor: Physiologic mechanism and abolition by surgical means, Arch. Neuropsychol. 1949 41 741-46. Burke, D. Muscle spindle feedback in Parkinson's disease, in' Clinical Neurophysiology in Parkinsonism. Deiwaide. PJ.. Agnoli. A.. editor Amsterdam' Elservicr. 1977 I-8. Carter, MC., Shapiro, DC. Control of sequential movements evidence for generalized motor programs. J of Ncurophy'siol. 1984: 52 787-96. Danta, G., Hilton, RG. Judgement of the visual vertical and horizontal in patients with parkinsonism. Neurology. 1975: 25' 43-7. Deecke, L., Englitz, HG., Komhubcr, HH., Schimit, G. Cerebral potentials preceding voluntary movements in patients with bilateral or unilateral parkinsonian akinesia, in: Desmcdt J.E. editors Progress in Clinical Neurophysiology. Karger Bascl. 1977:151-63. De Lancey Home, DJ. Sensorimotor control in parkinsonism, J Neurol. Neurosurg. Psychiatr. 1973' 36 742-46. Denier Van Der Gon, JJ., Thuring. JP. 'The guiding of imman hand-writing movements' Kibemetik. 1965 2 145-48. Dick, JPR., Cowan, JMA., Day. BL., Bcrardelli. D., Kachi, T., Rothwell. J., Marsden, CD. The corticomotoncurone connection is normal in Parkinison's disease, Nature. 1984:310: 407-t).
252
M.A. Pastor and J. Artieda
Dick, JPR., Cantello, R, Bunlma, O., Gioux, M., Benecke, R., Day, BL., Rothwell, JC., Thompson, PD.. Marsden, CD. 'The bereitschafispotential, L-dopa and Parkinson's disease. Electroencephalogr. Clin. Neurophysiol. 1987; 66" 263-74. Dietz, V., Hillesheiner, W., Freund, JL. Correlation between tremor, voluntary contraction and firing pattern of motor units in Parkinson's disease, J Neurol. Neurosurg. Psychiatr. 1974" 37" 927-37. Dinnerstein, AJ., Frigyesi, T. Lowenthal, M. Delayed feedback as a posible mechanism in parkinsonism. Percept.Motor Skills. 1962: 15" 667-80. Dinnerstein, AJ., Lowenthal. M., Blake, G., Mallin, RE. Tactile delay in parkinsonism. J Nerv.Ment. Dis. 1964:139:916-26. Evarts, EV., Teravainen. H.. Calne. DB. Reaction time in Parkinson's disease, Brain. 1981" 104:167-86. Flowers, K. Ballistic and corrective movements on an aiming task', intention tremor and parkinsonian movement disorders compared, Neurology. 1975: 25" 413-21. Foerster, O. Symptomatologie dcr Erkrankugen des Ruckenmarks und seiner Wurzeln. In Bumke R, Foerster O. (eds), Hanbuch der Neurologie, SpringerVerlag, Berlin: 1936 5" 1-403. Harris-Warrick, RM., Flamm. RE. Chemical modulation of a small central pattern generation circuit. Trends Neurosci. 1986; 432-437. Ivry, RB., Keele. SW. Timing of functions of the cerebellum, J Cog. Neurosci. 1989: 1,134-50. Jasper, HH., Bertrand, G. Recording from micro-electrode in stereo-taxie surgeD' for Parkinson's disease. J. Neurosurg. 1966: 24:219-21. Joffroy, AJ., Lamarre, Y. Rhy~mic unit firing in the precentral cortex in relation with postural tremor in a deafferented limb. Brain Res. 1971; 27: 386-9.
Involvement of the Basal Ganglia in Timing
253
Karson, CN., Spontaneous eye-blink rates and dopaminergic systems. Brain, 1983" 106" 643-53. Koller, WC. Sensory symptoms in Parkinson's disease, Neurology. 1984" 957-59. Lacruz, F., Artieda, J., Pastor, MA., Obeso, JA. The anatomical basis of somaesthetic temporal discrimination in humans, J Neurol. Neurosurg. Psychiatr. 1992" 54" 1077-8 i. Lamarre, Y., Joffroy, AJ. Experimental tremor in monkey: activity of thalamic and precentral cortical neurons in the absence of feedback. Adv. Neurol. 1979:24 9109-22. Lenz, F.A., Kwan, H.C., Martin, R.L., Tasker. R.R., Doshorovskv. J.O.. Lenz, Y.E. Single unit analysis of the human ventral thalamic nuclear group. Tremor-related activity in filnctionally identified cells. Brain. 1994:117(!): 531-44. Maskin, MB., Riklan, M., Chabot. D. Effects of short-term versus Iong-tema L-dopa therapy in parkinsonism on Critical Flicker Frequency, Percept.Motor Skills. 1974: 38' 455-58. Milner-Broxsxl, HS., Fisher. MA.. Wr W. Electrical properties of motor units in Parkinsonism and a possible relationship with bradykinesia, J Neurol.Neurosurg.Psychiatr. 1979.42' 35-41. Mortimer, JA., Pirozzolo, FJ.. Hansch. EC., Webster, DD. Relationship of motor syntoms to intellectual deficits in Parkinson's disease, Neurology. 1982" 32" 133-37. Obeso, J A., Artieda, J., Qucsada. P.. Martinez-Lage. JM. The reciprocal inhibition in rigidity and dystonia, in: Restorative Neurology, Dclwaidc PJ., Agnoli A. editors, London. Elsevier. 1985. PascuaI-Leone A., Vails-Sole. J. Brasil-Neto, J.P., Cohen. L.G.. Hallct, M. Akinesia in Parkinson's disease. I Shortening of simple reaction time with focal, single-pulse transcraniai magnetic stimulation. Neurology. 1994" 44(5)" 884-91.
254
M.A. Pastor and J. Artieda
PascuaI-Leone A., Vails-Sole, J, BrasiI-Neto, J.P., Cammarota, A, Grafman, J., Hallet, M. Akincsia in Parkinson's disease. II Shortening of simple reaction time with focal, single-pulse transcranial magnetic stimulation. Neurology. 19t)4 944" 892-98. Pastor, MA., Artieda, J., Jahanshahi, M., Obeso, JA. Time estimation and reproduction is abnormal in Parkinson's disease, Brain. 1992; 115" 211-25. Pastor, MA., Jahanshahi, M., Articda. J., Obeso, JA. Performance of Repetitive wrist movements i.a Parkinso,l's disease, Brain. 1992" 115" 87591. Ponder, E., Kennedy, W. On the act of blinking, J Exp. Physiol. 1928" 18" 89-110. Proctor, F., Riklan, M.. Cooper. IS., Teuber, NL. Judgement of visual and postural vertical by parkinsonian patients, Neurology. 1964: 14" 287-93. Narabayashi, H. 'Surgical approach to tremor', in Marsden, CD., Falm, S. (editors), Movement disorders. London. Butterworths. 1982" 292-99. Riklan, M., Levita, E.. Misiak. E. Critical Flicker Frequency and integrative functions in parkinsonism, J Psychol. 1970.75" 45-51. Schneider, JS., Diamond. GG.. Markham. CH. Parkinson's disease: sensory and motor problems in areas and ha.ads, Neurology. 1985" 37' 951-56. Schneider, J.S., Diamond. SG.. Markham, CH. Deficits in orofacial sensorimotor function in Parkinson's disease, Ann. Neurol. 1986: 19" 27582. Schwab, RS., England. AC. Parkinson's disease. J. Chron. Dis. 1958: 8" 448. Snider, SR., Fahn, S., lsgreen. P., Cote, LJ. Primary sensory symptoms in Parkinsonism, Neurology. 1976: 26: 423-29. Souques, A. Rapport sur Neurol..(Paris). 1921" 37" 534.
les
syndromes
parkinsonienes,
Rev.
Involvement of the Basal Ganglia in Timing
255
Talland, GA., Schwab, RS. Performance with multiple sets in Parkinson's disease, Neuropsychologia. I t)64: 2" 45-53. Terzuolo, CA., Viviani, Y. Dctem~inants and characteristics of motor patterns used for typing, Ncurosci. 1980: 5" 1085-103. Teuber, HL., Mishkin, M. Judgment of visual and postural vertical after brain injury, J. Psychol. ! 954 38" 161-75. Tuller, B., Kelso, JAS.. Harris. KS. Intcrarticulator phasing as an index of temporal regularity in speech. J Exp. Psychol.' Hum.Percept. Perform. 1982: 8 460-72. Venables, PN., O'Connor. N. Reaction times to auditory and visual stimulation in schizophrenics and normal subjects, J. Exp. Psychol. 1959" 2: 175-79. Wiesendanger, M., Schneider, P., Viiloz. JP. Electromyographic analysis of a rapid volitional movement. Am. J. Psychol. 1969: 48: 17-24. Wilson, SAK. Modern problems in Neurology, New York, Winwood. 1020. Wilson, SAK. Neurology. Arn61d. Londrcs. 1947 793-95. Wing, AM., Kristofferson, A B. The timing of intcrresponse intervals. Percept. Psychophys. 1973: 13 455-60.
This Page Intentionally Left Blank
Time, Internal Clocks and Movement M.A. Pastor and J. Artieda (Editors) 9 1996 Elsevier Science B.V. All fights reserved.
257
EXPLORING THE DOMAIN OF THE CEREBELLAR TIMING SYSTEM SEAN CLARKE, RICHARD IVRY, ROBERTS and NAOMI SHIMIZU
JACK
GRINBAND,
SETH
Department of P.wcholo~v. llmvervity qf Cal!/brnia, Berkeley, CA 94720 ABSTRACT. Tile ability of all animal to process temporal information has adaptive significance across different temporal ranges. The ability to encode and utilize temporal information allows an animal to predict and anticipate events. However, the lime scales var3.' widely. The predictable event might be based on information that changes over relatively long periods such as a year or a day. or over periods comprising much shorter durations, events that change within a few minutes or milliseconds. Are there a single set of neural mechanisms thai are essential for representing temporal information over these different scales? Despite the facl thai numerous neural sln~clures have been linked to successfid performance on a varieD' of timing tasks, this question has received relalively little allen!ion. In this chapler, we will focus on the role of the cerebellum in a varie .ty of timing tasks. We will review the hypothesis that the cerebellum can be conceptualized as a relatively task-independent tinting mechanism. An important feature of this hypothesis is that the range of the cerebellar tinting s.vslem is assumed to be relatively reslriclcd. Specifically, we assume thai the cerebellum is capable of representing temporal information ranging from a few milliseconds to an upper bound of a few seconds. What remains unclear is whether the cerebellum is involved on tasks spanning longer durations. Cognitive processes such as attention and memor3.' become clearly imporlant here. and indeed, may dominate performance for longer inlervals. The animal literature points !o noncerebellar stn~ctures as playing a crilical role in these tasks and we will provide a brief review of this work. Finally. we will present the preliminary resulls from two experimenls designed to directly test the hypothesis that the cerebeilum's temporal capabilities are limilcd Io rclalivcly shorl durations.
1. Studies with patients with cerebellar lesions Ivry and Keele (1989) assessed the performance of a variety of neurological patients and age-matched control subjects on two tasks that were designed to
258
S. Clarke et aL
require the explicit reprcscntation of temporal information. For the time production task, the participants produced a series of simple keypresses, attempting to produce isochronous intervals between each pair of keypresses. For the time perception task, the participants judged whether a comparison interval was shorter or longer than a standard interval. There were three primary groups of patients" those with cerebellar lesions, those with Parkinson's disease which would indicate basal ganglia pathology, and those with cortical lesions cncompassing premotor regions. Anatomical models as well as consideration of the symptoms associated with cercbcllar lesions prompted the inclusion of the first group. The basal ganglia and cortical groups were included both for comparison purposes and because of earlier neuropsychological research implicating basal ganglia (Wing et al., 1984) or frontal/temporal regions (Milner, 1971) in time production or perception. In temas of variability on the repetitive tapping task, the patients with Parkinson's Disease perfonned comparably to age-matched control subjects. Surprisingly, these null results were obtained under both the on and off medication state. In contrast, patients with either cortical or cerebellar lesions were found to have increased variability on the repetitive tapping task. The total variability was decomposed into two components, that associated with central control processes and that associated with motor implementation (Wing and Kristofferson. 1973). From this analysis, the patients' deficits were attributcd to both sources. However, a second study focused on patients with unilateral cerebellar lesions, either in medial or lateral regions. Here, a double dissociation was obtained. Whereas medial lesions led to increased implementation variability, lateral lesions led to an increase in central variability (Ivr)., et al., 1988). This dissociation is in accord with neuroanatomical models which emphasize ascending projections from the lateral cerebellum and descending projections from the medial cerebellum. From the tapping results, it is not easy to determine whether the cerebellum is critical for regulating timing, or some other aspect of motor performance. While Wing and Kristoffcrson (1973) labeled the central component. "clock variability", this componcnt actually includes all sources of variability not included in the estimate of motor implementation variability (Ivry and Hazeltine, 1995). For this rcason, the perception task provides an opportunity to determine whether a particular structure was essential for internal timing. Correlational studies have suggested that a common mechanism is invoked in both motor and perceptual timing (Keele et al.,
Exploring the Domain of the Cerebellar
259
1985). From this, we might expect to find that lesions of a particular brain region will impair performance on both time production and perception tasks. Only the patients with cerebellar lesions showed this dual-deficit. They were significantly impaired on the time perception task, requiring a larger difference between the comparison and standard intervals in order to achieve a criterion level of pcrfommnce. The perceptual deficit was specific for time discrimination in that the cerebellar patients were unimpaired on an intensity discrimination task. Importantly,, the cortical group was normal on the time perception task, but impaired on the intensity task. Thus, the perception task provided a second double dissociation suggesting a special role for the cerebellum in both motor and perceptual tasks that require precise timing. As with the tapping results, the Parkinson patients perfomaed within nonnal bounds on the time perception task. These results required a reconccptualization of the domain of cerebellar function. This structure has generally been linked to motor functions, or sensorimotor learning. The fact that the patients were impaired on a purely perceptual task suggested that its domain should be specified in terms of a particular mental operation, namely the representation of the temporal relationships between events. We have hypothesized that this computational capability is invoked across a wide range of tasks that require this form of representation. For example. Ivr3' and Diener ( 1991) reported that cerebellar patients were impaired on a velocity perception task and that this perceptual problem could not be attributed to a problem in occulomotor control. Indeed, they proposed that some of the eye movement problems observed following cerebeilar lesions may reflect an inability to represent the metrical properties of a moving stimulus.
2. The cerebellum and sensorimotor learning
Impressive progress has been made over the past few decades towards identi~,ing the neural structures involved with different forms of learning and memoD'. Given the obvious advantages imposed by learning, it is not surprising that a large number of neural structures have been implicated in these processes. An important endeavor has been to specify the domain of these structures and develop computational models to explain their contributions.
260
S. Clarke et al.
One approach for understanding the computational requirements of different learning situations is to consider the temporal properties imposed by different tasks. For example, the learning process in classical conditioning is constrained by the temporal relationship between the CS and US (see Flaherty, 1985; Jenkins, 1984). A prerequisite for learning across a wide range of paradigms is that the onset of the CS precede the US. The most effective inter-stimulus interval (ISI), however, varies depending on the type of learning. Three general categories can be described. 1) Conditioning of simple skeletal reflexes such as the eyeblink reflex is limited to short ISls. being strongest when this interval is less than 1 see. 2) Conditioning of autonomic responses such as heart-rate conditioning can occur with these short ISis, but can also be robust when the ISI is extended to the minutes range. 3) Conditioning of avoidance behavior such as in food aversion experiments can be found when the CS and US are separated by durations up to many hours. Moreover. whereas the pairing of a CS and US may lead to multiple CRs, the timing of these learned responses can be quite different. Conditioning of the rabbit nictitating membrane response (NMR) has become a model paradigm for investigating the neural substrates of basic associative learning and memor3., processes associated with simple skeletal reflex responses. In NMR conditioning, a neutral CS such as a tone or light is paired with an aversive US (e.g.. an airpuff directed near the eye). After relatively few presentations, the animal begins to extend the membrane in response to the CS alone. The rate of NMR learning is highly dependent on the ISI. Smith (1968) reported that an ISI of 200 ms produced the highest percentage of CRs in comparison to ISis of 100, 400, and 800 ms (see also, Steinmetz, 1990). Few CRs were observed with ISis of 50 ms. Conditioned NMRs can be found with longer intervals, although the rate and efficacy of learning are reduced. In addition, the topography of the CR is highly constrained by the ISI. The maximum extension of the nictitating membrane occurs just prior to the presentation of the US. Indeed, it is this feature that makes this CR highly adaptive. It pem~its the organism to attenuate the aversive effects of the US. The importance of timing in NMR conditioning was made clear by the work of Kehoe et al. (1989). When rabbits were conditioned simultaneously with two ISls, they produced two CRs, each one timed to be maximal just prior to the onset of the upcoming US. Learning related changes during NMR conditioning have been observed in neurons in several brain regions, including the hippocampus (Berger and Thompson, 1978) and the cerebellum (McCormick and Thompson, 1984).
Exploring the Domain of the Cerebellar
261
However, lesion studies have provided compelling evidence that the cerebellum is essential for NMR conditioning. The exact site of plasticity within the cerebellum has been a source of controversy. Thompson (1986) has argued that the critical locus is the interpositus nucleus, whereas other studies have focused on the importance of the cerebellar cortex (e.g., Yeo et al., 1984). Given that the principal cerebeilar inputs, the mossy and climbing fibers, im~ervate both sites, it is reasonable to assume that learning-related changes may occur in both sites (Perret et al., 1993). If this is so, then we want to consider potential computational differences between nuclear and cortical learning. One possibility is that nuclear mechanisms might support a basic associative process for the fonnation of a CR, while changes in the cerebellar cortex are essential for shaping the topography of the CR. That is, the precise timing of the CR may result from changes in the cerebcllar cortex. This hypothesis is supported by the findings that lesions of the cerebellar cortex disnlpt the timing of the CR (Perrett et ai., 1993). It remains difficult to specif3.' the learning domain of the cerebellum (see Ivry, 1993). One possibility is that this structure is essential for forming sensorimotor associations that result in skeletal responses to avoid aversive stimuli (Thompson, 1990). This hypothesis emphasizes the task domain of the cerebellum and focuses on the fact that the climbing fiber pathway provides a salient error signal for shaping appropriate skeletal responses. An alternative h)~othesis is that the domain of cerebellar learning extends to those situations in which the animal must precisely represent temporal information. By this way of thinking, the cerebellum is associated with NMR conditioning because this type of learning is only adaptive if it is appropriately timed (Kecle and lvr3'. 1991). That is, learning an association and forming the temporal representation of that association can not be thought of as distinct. Of course, other types of associations may not have the same temporal requirements and. as such, would not be expected to be dependent on the cerebellum. For example, the timing of conditioned autonomic responses in the NMR paradigm seems to b e relatively independent of the ISI and this form of learning is unaffected by cerebcllar lesions, even when the NM response itself is abolished (Lavond et al., 1984). Buonomano and Mauk (1994) have presented a computational model of the cerebellum that produces the associations seen in NMR conditioning as well as the precise topography of the CR (see also, Bullock et al., 1904). This model does not depend on delay lines or arrays of oscillators, but rather emphasizes known anatomical and physiological properties of this neural
262
S. Clarke et al.
structure. For example, an input pattern across a set of mossy fibers will trigger dynamic changes in the response properties of many cortical ceils. Some of these changes will occur rapidly, e.g., within a few milliseconds. However, due to negative feedback loops and physiological processes such as slow IPSPs, other consequences of the mossy fiber activity may not be evident for hundreds of milliseconds. Thus, the eerebellar cortex can maintain a representation for an extended period of time after the onset of a CS. Coupled with an appropriate learning signal such as the complex burst of the climbing fiber, the system can learn to produce a response at a desired point in time.
3. Potential limitations of cerebellar timing In the Buonomano and Mauk (1994) model, the timing of a particular interval reflects the real-time properties of a set of neural elements. For example, activation over one set of units would correspond to a duration 250 ms and activation over a different set of units would correspond to a duration of 500 ms. There is no fundamental temporal unit from which larger scale durations are constructed via an oscillatory process. That is. the 500 ms duration is not created by two cycles through a 250 ms circuit (or multiple cycles of some fimdamental period). Oscillator), models have dominated the psychological and neural literature for a variety of reasons. Oscillation is ubiquitous in biological systems, both at the neural level and in behavior. An appealing feature of oscillator3, models is that a simple set of mechanisms can provide temporal processing over a wide range of durations since long temporal intervals can be created via multiple cycles of the fundamental period. Non-oscillator3' timing mechanisms do not have this feature. They would be expected to have an upper bound on their temporal range, a duration corresponding to the maximum duration that can be supported by the patterns of connectivity in the network. Longer durations might be represented by joining a real-time network with a memory process. While this has some resemblance to standard clock-counter models, it differs in that there is no basic oscillatory. We propose that the cerebellum is best characterized as a non-oscillatory timing system. At present, this hypothesis is best viewed as a conjecture, an idea intended to generate empirical tests. There are a few reasons to suspect an upper bound on a short range timing system. First, as noted above.
Exploring the Domain of the Cerebellar
263
conditioning of skeletal responses such as the NMR are difficult to obtain with ISis longer than I-2 scc. While the effects of cerebellar lesions on conditioning with long ISis has not been established, it does appear that the hippocampus becomes relatively more important under such conditions (Berger et al., 1986). One possibility is that when the ISI exceeds the cerebellar range, learning shifts to new neural sites. Alternatively, memory capabilities of the hippocampus may become combined with timing capabilities of the cerebellum, cspcciall.v in trace conditioning paradigms. Second, behavioral studies in humans suggest that there may be a qualitative change around 2-4 scc in our capacity to represent temporal infommtion. Below this duration, successive events are seen as belonging to a conunon temporally-defined group, regardless of whether this group has a rh}ethmic structure or lacks such organization. Above this duration, events are perceived as temporally isolatcd events, even if they occur periodically (Fraisse, 1963' Povel. 1981). Ps.vchophysical studies have also indicated an increase in the Weber fraction on duration discrimination tasks for intervals longer than 2 sec (Getty. 1975 but scc Allan and Gibbon. 1991). On the motor side, Mates et al. (1994) have shox~aa that when tapping with a periodic pacing signal, people shift from a predictive to an reactive mode as the target interval becomes longer. For intervals less than 2-3 sec. the subjects' responses tended to anticipate the tones. However, for longer intervals, the responses almost always followed the tones, and indeed, occurred with a latency suggestive of a simple reaction time.
4. Animal models of temporal discrimination
Several operant learning paradigms have been developed for exploring how animals process and discriminate temporal information. These tasks typically involve durations spanning from a few seconds (e.g., Allan and Gibbon, 1991) to many seconds (Roberts, 1981" Meck and Church, 1987). This work has led to rigorous theoretical models that succeed in accounting for a wide range of phcnonlcna. Moreover, physiological and phannacological manipulations have helped identify, some of the neural mechanisms associated with the component parts of the models. One popular method has been the peak procedure (e.g., Roberts, 1981" Meek, 1991). In this task, a signal is presented and an animal is provided a reinforcement for the first response emitted after a criterion period of time has elapsed. Under
264
S. Clarke et al.
such conditions, the animal will increase its response rate as the target duration is approached. By not providing a reinforcement of a certain percentage of trials, the researcher can observe how the animal's response rate decreases once the target duration has elapsed. Performance on this and related tasks have generally been evaluated within the context of an information processing model that includes timing, attentional, memory, and decision processes (e.g., Gibbon and Church, 1984). In this model the clock is conceptualized as a pacemaker, emitting pulses as a Poisson process. These pulses are gated by a switch into an accumulator, or counter process. The state of the gate may be under the influence of attention. The current value of the counter constitutes working memory and is compared to a target duration in reference memory, a trace formed from previously encoded time values. The final processing stage entails a decision process in which the two memory, Values are compared to determine whether a threshold value has been exceeded, thus triggering a response. Over the past decade, a number of studies have been designed to identify. the critical neural mechanisms associated with the psychological constructs. For example, Meek (1983) argued for a dissociation of clock and memory processes based on the differential effects of pharmacological agents. Dopaminergic agents caused a transient shiR in the peak time, a result consistent with the hy'pothcsis that these drugs affected the speed of the clock. In contrast, neuropeptide and cholinergic agents caused permanent changes in the psychometric functions, indicating that their effect was on reference memory. Based on a lesion study. Olton et al. (1988) concluded that attentional processes were localized to the frontal cortex areas and subcortical structures that proiect to this region. Lesions of either the frontal cortex or of the nucleus basalis magnocellularis disrupted performance when rats had to time two simultaneous stimuli without affecting the processing of either signal when presented individually. Lesions of the medial septal area and fimbria fomix did not disrupt performance on the simultaneous discrimination task, but did disrupt perfommnce when a gap was inserted during the stimulus presentation. This result is consistent with a deficit in working memory, implicating a hippocampal role in this process. In summary, these studies have emphasized that performance on temporal discrimination tasks involve a complex network that includes limbic, basal ganglia, and frontal structures. Implicit in this work has been the assumption that the same mechanisms will be invoked over a wide range of durations
Exploring the Domain of the Cerebellar
265
(e.g., Nichelli, 1993). For example, a conunon pacemaker will be triggered by stimuli of varying durations, with a critical difference being the number of counts that accumulate during the extent of a particular stimulus. Indeed, the speed of the pacemaker may vary, and its output is subject to attentional limitations. Thus, while there is a clock-like process, the representation of temporal information is influenced by a number of non-temporal, cognitive processes. To date, this animal research has ignored the cerebellum, perhaps because this structure has been assumed to be limited to the motor domain while these tasks focus on perceptual and memor), processes. Moreover, while some studies have used stimuli that are less than 1 sec (Allan and Gibbon, 1991), the maiority of this work has involved stimuli that are considerably longer, frequently ranging tip to 40 sec. As noted above, we have hypothesized that the ccrcbcllar timing system is limited to relatively short durations. Our working model is that this timing process is relatively immune to cognitive influences. The onset of a stimulus may automatically activate different sets of neurons, and mcmor), demands are minimal. Units active at the offset of the stimulus may become associated with certain responses, a model that does not have the working memor3, requirements of a counter or require that a current representation be compared to a reference memory. 4.1. EXPERIMENT I The first experiment was designed to test two hypotheses. First. we wanted to develop an animal model to cxplore the cerebellar timing hypothesis. Rats were trained to discriminate intervals that ranged from 200 to 850 ms. The effects of cerebcllar Icsions o,~ their pcrfommnce was assessed. Based on our human studies, it was predicted that these lesions would increase variability on this task without producing a change in bias. That is. the lesions would increase the noise in the cerebellar timing network without producing an overall change in clock speed (since there is no fundamental timing unit). Second, wc also trained the animals on a second duration discrimination task. but here the durations spanned 15-45 sec. Based on our conjecture that the ccrcbcllar timing system is limited to relatively short durations, we did not expect Icsio.~s of this structure to impair performance on this second task. Evidence of a selective deficit on the short range timing
266
S. Clarke et al.
task would provide initial evidence for a dissociation of the mechanisms involved in the representation of short and long range durations. Our human studies (Ivr3.' and Keele, 1989: Ivry and Diener, 1991) have indicated that the neocercbellum is critical for representing temporal information. The output of the neocerebellum is primarily projected via the lateral nuclei, the dentate and interpositus (composed of the globuse and empoliform in humans). Thus. lesions were targeted to center on the dentate nucleus, with the expectation that the damage would extend to the interpositus nuclei. 4.1.1. M e t h o d Subjects. Twelve naive rats (Fisher 344, Charles River Laboratories, Wilmington, MA) were tested. One rat became sick during the course of the experinaent and had to be euthanized. The animals were approximately 90 days old at the beginning of training and were maintained at 80% of their free feeding weight throughout the experiment. Rats were housed individually in plexiglas cages with water available ad libitum. Apparatus. Subjects were trained individually in standard operant boxes. each enclosed in a sound attenuating chamber. The front wall of the box contained a recessed food trav and two retractable levers, one on each side of the food tray. Water was available ad iibitum on the back wall of the box. Visual stimuli were displayed on circular light panels positioned above each lever and via a houselight situated at the upper left hand comer of the front wall. The presentation of the stimuli, response collection, and reinforcement delivery were controlled b.va PC computer. Timing was accurate to .01 milliseconds. Surgery. All surgical procedures were completed under aseptic conditions. After anesthesia with sodium pentobarbital (50 mg/kg), rats were placed in a stereotaxic apparatus. The skull surface was exposed and two small holes were made over the dentate nucleus on each side. An electrode was lowered into the holes on either side and a lesion was made by passing 50 mA at constant current for 20 seconds. Based on the Paxinos and Watson (1986) atlas, the targeted coordinates for the lesions were +/- 3.4 ML. 6.2 DV, and -11.3 AP from Bregma. Eight randomly selected rats received bilateral lesions in this manner. Three animals were given sham lesions by exposing the dura without lowering the electrodes or passing any current. Rats were allowed 1 week to recover, at which time test trials were begun. Procedure. Animals were tested on a daily schedule in two groups of six. Each session lasted four hours during the training phase and seven hours o
u
Exploring the Domain of the Cerebellar
267
during the test phase. On alternate days, rats were tested on the short range and long range timing tasks. For half of the rats, the overhead light was used for the short range (SR) task and both circular disks were used for the long range (LR) task. The two stimuli were reversed for the other group. For the short range task, the initial durations for the training phase were 300 and 1200 ms. For the long range task, the initial durations were 20 and 40 see. Filled intervals were used: the stimulus was present for its entire duration. On each trial, one of the two possible stimuli for that session (SR or LR durations) was presented. 100 ms after the offset of the stimuli, both levers were extended. Following a response or a six sec interval, the levers were retracted. Correct responses were rewarded with a 45 mg sucrose pellet. The mapping of stimuli to the response levers was counterbalanced, with one mapping used for three of the boxes and the reverse mapping used for the other three boxes. For all of the rats. the same mapping was used for both the SR and LR tasks (e.g., the 300 ms and 20 sec were both associated with the right lever). The inter-trial interval was 30 sec. A session consisted of approximately 750 trials for the SR task and 375 trials for the LR task. Perfommnce with the training values as3xnptoted at around 90% correct after 24 sessions (12/task). At this point, the stimulus durations were adjusted to 300 and 750 ms for the SR task and 25 and 40 sec for the LR task. After another 14 sessions, the animals' performance had returned close to the 90% asymptotic value. To obtain psychometric fi~nctions, the test phase included both the training durations and probe durations. For the SR task, there were nine probe durations ranging from 200 to 850 ms, with seven of these durations falling between the endpoint values. For the LR task, the nine probe durations ranged from 20 to 45 sec. On 50% of the trials, one of the two training durations was selected and correct responses were reinforced. On the other 50% of the trials, one of the nine probe durations was selected and no reward was possible. As in the end of the training phase, the tasks alternated by session. The rats completed 36 sessions of the test phase prior to surgery. After a one-week recover3., period, post-surgery testing commenced in the same manner and continued for an additional 23 sessions. 4.1.2. Results And l)iscnssion. Figure 1 presents the psychometric fiinctions for the two tasks. In each figure, the probability that the animals respond long is graphed as a fi~nction of duration. Only the trials on which a response was made were included. Panels a and c are the averaged data for
268
S. Clarke et al.
the lesion and sham groups on the SR task; panels b and d are for the LR task. The data are based on the last 15 presurgery sessions and the first 15 postsurgery sessions. The point of subjective equality (PSE) and standard deviation were estimated from these functions. The first measure is an indicator of bias. It corresponds to the duration at which the group will respond long on 50% of the trials. Note that an increase in PSE occurs when the probability of responding long decreases. The second measure is one 100 Sham: SR Timing
Sham: LR Timing
50 -
10(11=
i Lesion:
SR Tim,ng
Lesion:
LR Timing
50 -
0 -
o.',
0'8
l
I
I
I
I
l
20
25
30
35
40
45
DURATION (S)
Figure 1" P~'chometric filnclions for Expcrimen! 1. Unfilled symbols represent presurgery performance, filled s.vmbols represent postsurgery performance.
estimate of a difference threshold, or acuity. Steep functions yield low scores indicating that the animals consistently identify a particular stimulus as short or long. Shallow functions yield high scores, indicating more uncertainty in the response functions. The sham animals showed little change in performance on the SR task following surgery. Their PSEs were 533 ms and 539 ms for the pre- and post-surger3., sessions. They also showed a small increase in variability, with the standard deviation score rising to 201 ms from 183 ms (9.8%). In
Exploring the Domain of the Cerebellar
269
contrast, the group receiving ccrcbellar lesions showed significant changes in performance following surgery. In terms of the aggregate functions, these changes were seen in both the PSE and standard deviation measures. Their group PSE was 538 ms pre-surgery and 564 ms post-surgery. The standard deviation score for the lesioned group rose to 239 ms following surgery from a baseline score of 192 ms (24.4% increase). On the LR task, the PSE for the sham group changed from 33.4 sec to 33.4 secs while their standard deviation actually decreased (8.4 sec to 8.2 see). The lesion group showed a modest increase in standard deviation from 8.3 see to 8.6 sec (3.8% increase). As with the SR task, however, their PSE increased, going from 32.8 sr prc-surger 3, to 34.2 see post-surgery. Since there were not sufficient trials per stimulus duration in each session, we opted against using global measures and developed two trial-bytrial measures, one to reflect a change in bias and one to reflect a change in acuil~y. First, we established an average presurger3' psychometric function. The response on each trial was scored in terms of its difference from the predicted filnction. Long responses were assigned a value of 100 and short responses were assigned a value of 0. For our bias measure, we subtracted these numbers (100 or 0) from the mean percent long responses for that duration. For example, the mean percent that the animals responded long for the 850 ms stimulus was 90. When an animal responded long to this stimulus, their bias score for that trial was -10; if they responded short, their bias score was 90. This score was calculated for each trial. If there was no change in perfommnce, the average score would be 0 whereas an incrcascd likelihood to respond short would yield a positive score. A similar procedure was uscd to obtain a measure of consistency. The only modification was that the difference score was multiplied by -1 for the five longest durations. Again. this procedure was calculated for each trial individually and a mean score was obtained for each animal. As a whole, the higher the mean consistency score, the more consistent is performance (e.g., short durations are responded to as short and long durations are responded to as long). For example, responding long to the 850 ms stimulus would result in a score of 10 for that trial, whereas responding short would result in a score of-90. An appeali,lg feature of these measures is that thev result in scores of identical units for the two tasks. Pre- and post-surgery bias and consistency scores were calculated for each animal for both the S R and L R tasks. The mean values for each group are s h o ~ in Figure 2. Givcn that there were only three animals in the sham
270
S. Clarke et al.
group, the data were analyzed separately for the two groups (and we will not report these results for the shams since no effects reached significance). The bias and consistency scores were entered into two ANOVAs (one for each measure) with phase (pre vs. post) and task (SR and LR) as factors.
0
.
.
.
.
1
.
i
I
-!0
ii/ m1
--
SR
i, R
SR
LR
LESION ,
SHAM
Ql'resurgcrv llPosl.~urgerv .
I
.
J
,,.
>,
~
i
10 -T-.................................
- I 0
'
t
,
SR
-
1, I7,
!
SR
LESION
LR SHAM
i t'~i'resurgery IIPostsurgery ]
Figure 2: Bias and Consistency scores for Experiment 1.
On the bias measure, a significant effect was found for phase. F(I,7)=7.04.p<.05. The effect of task and the interaction were not significant. Thus. as was seen in the ps.vchometric functions, on both tasks, the lesioned animals were more likely to respond short following surger3'. In contrast, the consistency measure revealed a dissociation between the two tasks. On this measure, both main effects were not significant, but the phase X task interaction was. F(I.7)=8.40. p<.05. As can be seen in Figure 2, when the data are analyzed on a trial-by-trial basis, the animals with cerebellar lesions became less consistent on the SR task following surgery while they became more consistent on the LR task. This result is in accord o
o
Exploring the Domain of the Cerebellar
271
with the hypothesis that the temporal range of the cerebellum is limited to short intervals. The effect of the lesions was relatively transient. For the first two postsurgery sessions on the SR task, the consistency scores were -5.25 and 4.88. The scores remained greater than -1.5 negative for the next three postsurgery sessions, at~er which they hovered around 0. We can consider a number of reasons for the transient nature of the deficit. First, histological analyses indicated that our lesions rarely destroyed both lateral nuclci. Moreover, as noted above, it is likely that temporal information is represented in the cercbcllar cortex, although the output from cortical neurons would have to pass through the deep nuclei. Second, our animals received many more trials post-surger).,' than are commonly used in lcsion studies. For example, by the fifth post-surgery session on the SR task, the animals had participated in approximately 2,500 trials. This might have been sufficient for reorganization within intact regions of the cerebellum. 4.2. EXPERIMENT 2 The dissociation on the consistency mcasttre between the SR and LR tasks was the most positive result in support of the cerebellar timing hypothesis. This finding provides the first evidence of cerebellar involvement on what appears to be a perception of time discrimination in animals. However, there are other aspects of the results that suggest these findings be treated with caution. First, as noted above, the effects were relatively transient and not very. large. Second, there was a reliable change in bias for both tasks. While our tasks were designed to equate the response requirements for the two tasks, it is also important to note that the motor demands are not equivalent in the SR and LR tasks. Suppose the animals have to orient at~er stimulus onset and that ccrebcllar lesions produce a generalized slowing in this behavior. This deficit would be expected to be more disruptive on the SR task than the LR task because of the brevity of the stimuli in the former condition. If the rats were slow to orient, the brief SR signals might be near completion before the animals have begun to fully attend to the stimulus. In contrast, a slight increase in orienting time would have little effect on the LR task given that the minimuna duration was 20 sec. The orienting deficit might account for the fact that in both conditions, the lesioned animals were more likely to respond short.
272
S. Clarke et al.
We have completed a second experiment to test this hypothesis. Two tasks were used in this expcriment" a modified version of the SR task and a non-temporal perceptual task that also involved short signals. For this control task, the rats were required to judge the intensity of a 550 ms signal. If the cerebellar lesions are disrupting an internal timing system, then the rats should show a selective deficit on the duration discrimination task. If the lesions are producing a generalized slowing in orienting, then performance on both tasks should be disrupted. The selection of an intensity task was based on our human studies in which we found cerebellar patients to be unimpaired in judging the loudness of brief tones. 4.2.1. M e t h o d Subjects: Twelve new rats were tested in Experiment 2. The animals were housed and fed as in the first experiment. Apparatus: The same operant boxes were used with one modification. A speaker was mounted above the mesh ceiling in order to present an auditory stimulus consisting of white noise. Surgery. Surgical procedures were as in the first experiment. The only change was that two penetrations were made for the lesions on each side, with the second penetration being slightly displaced from the first. This change was adopted in an attempt to increase the size of the lesions. Bilateral lesions were perfomacd on nine animals with the remaining three undergoing the sham procedure. Procedure. There were a number of significant modifications in the procedure for Experiment 2. First. the SR task was modified so that the rats had to judge an empty interval. This modification was adopted to reduce the possibility that the animals were using a non-temporal cue such as energy' integration for the timing task. For each trial on this task, two 50 ms markers were presented, separated by an inter-stimulus onset time of the target duration. For the training phase, the initial intervals were 250 ms and 1000 ms. After perfommnce asymptoted, these values were changed to 300 and 800 ms. Because the lights in the operant chamber have slow decay rates, the short markers required that we use auditory, white noise stimuli for the 50 ms markers. In the test phase, nine probe durations were added ranging from 250 ms to 850 ms. Second, an intensity discrimination (ID) control task was substituted for the LR task. For this task, a continuous white noise was played for 550 ms on each trial. For the training phase, the sottest and loudest sounds were initially set to intensities of 60 dB and 80 dB, respectively. These values
Exploring the Domain of the Cerebellar
273
were adjusted to 65 dB and 79 dB, once performance had reached an asymptotic value. The nine probe stimuli ranged from a low of 62 dB to a high of 82 dB. Third, all sessions were four hours in duration and alternated by session between the two tasks. Wc did not include any mixed sessions since an auditory stimulus was used for both tasks. On each day, one group was tested on the SR task and the other group was tested on the ID task. All animals received extensive training on the two tasks prior to surgeD'. After a one-week recover3.' period, they were tested for an additional 26 sessions post-surgery. The following analyses are restricted to the last 16 sessions before the operation and the first 16 sessions after the operation. Thus, they include eight sessions per task, pre- and post-surgery. 4.2.2. Results And Discttssion. The psychometric filnctions for the two tasks are presented in Figure 3. For the S R task. the data points correspond to the probability that the animals would respond long for each duration. For the ID task, the data correspond to the probability that a particular intensity was judged loud. Sham operations produced minimal change in performance. On the SR task, PSE and standard deviation values went from 569 ms and 153 ms to 567 ms and 173 ms (a 13. i% increase in sd). On the ID task, the PSE was 72.2 dB presurger 3, and 72.4 dB postsurgery. The standard deviation also showed only a slight change increasing to 4.8 dB from 4.2 dB (14.3%). In contrast, the psychometric fitnction on the duration discrimination task became considerably flatter for those animals receiving bilateral cercbellar lesions. Correspondingly. the standard deviation for this group showed a large increase (190 ms to 264 ms. a 38.9% increase). This increase in variability occurred without any change in PSE" for the pre- and postsurge~, functions, the PSE estimates arc 569 ms and 567 ms, respectively. Thus, unlike Experiment !. the effect of the cerebellar lesions was only observed on our measure of variability without any change in bias. At present, we can not account for the discrcpent results on the bias measure. The cerebellar lesions did not affect performance on the intensity discrimination task. The PSE decreased slightly, going from 72.0 dB to 71.6 dB while the standard deviation scores were actually slightly lower post-surgery (5.1 dB and 4.9 dB for the pre- and post-surgery fimctions, respectively). Thus, ccrebcllar lesions do not produce a generalized impaimaent in the ability of rats to respond to brief signals. The rats in this
274
S. Clarke et aL
group were only impaired when they had to judge the duration of these stimuli. To statistically evaluate these conclusions, we computed bias and consistency scores as in Experiment 1 (Figure 4). These data were entered
100
tlJ
to Z
Sham: SR Timing
I
~
Sham: Amplitude
50-
W
6
q
,,,'
1011) -
Lesion: Amplitude
0
--
I
I
I
I
'
I
I
I
I
0.25 0.35 0.45 0.55 0.65 0.75 085 DURATION(S)
o
/
-28
I
-24
AMPLITUDE
I
-20
I
-16
(ATTENUATION
I
-12
I
-8
IN DO)
Figure 3" Psychometric fimclions for Expcrimem 2. Unfilled symbols represent presurgery performance, filled symbols represent postsurgery performance.
into ANOVAs with two factors: phase (pro- vs. post-surgery) and task (SR vs. ID). No main effects nor interactions were significant for the shams, a result tempered by the fact that there were only three animals in this group. Of greater interest, only the consistency ANOVA yielded significant effects for the cerebellar group. Significant main effects were found for both phase. F(1,8)=7.24, p<.05 and task, F(i.8)=8.87, p<.05. Furthermore, these main effects were constrained by a significant two-way, phase X task interaction, F(1,8)=13.06, p<.01. A substantial decrease in consistency on the SR task occurred after the animals were given ccrebellar lesions, with little change in the ID task.
Exploring the Domain of the Cerebellar
'~
i/fill
......
-10
SR
275
SR
A Inplitude LESION
A m pStude SHAM
IQ l'rt:su,'gcry lail Postsurgery
Z
1() -I ii
--
z
~-IO
~.~
I
~
~ SR
. . . . .
Atllplill, dc LESION
,
/ii/, j SR
AmpSlude SHAM
i ti
I
Figure 4: Bias and Consislenc.v dala for Expcrimen! 2.
In comparison to Experiment 1. the nuclear involvement and overall size of the lesions was larger in this experiment. This may have contributed to the fact that across all eight post-surgcr 3' sessions on the SR task. the consistency score was at least !.0 point lower than on the pre-surgery sessions. Nonetheless. the animals showed good recovery, with initial mean consistency, scores of-6.2 and -8.8 dropping for the first two post-surgery sessions improving to -.17 and -1.8 by the last two sessions. When considering this recover3', it is important to keep in mind that the animals received over 450 trials per session.
5. Conclusions We have proposed that the cerebellum plays a critical role in the representation of temporal infor,uation. The initial empirical support for this hypothesis came from research with the study of patients with cerebellar
276
S. Clarke et al.
lesions (Ivry and Keele, 1989). and included tests of both time production and time perception. To date, animal studies have continued to focus on how this timing capability ma.v be expressed in motor performance. The current studies provide initial evidence in the rat that the cerebellum may be essential for the perceptual representation of stimulus duration, even when that percept is only indirectly tied to motor responses. These findings emphasize the importance of considering the computational capacity of a particular neural structure when considering its domain. Rather than focus on particular tasks (e.g., reflex conditioning, motor control, perceptual processing), it can be useful to consider the operation performed by a structure and then explore how widely that operation is employed. In the case of the cerebellum, it appears that its timing capability can help account for why this structure is essential for coordinated movement, speech production, sensorimotor learning, as well as certain perceptual functions (see Ivry, 1993). Although the experimental record is at present quite scant, computational considerations also motivate our conjecture that the temporal extent of the cerebellum is limited to relatively brief durations. A timing system that is used for coordinating the actions of multiple joints or perceiving the velocity of a moving stimulus is unlikely to be useful for tasks that span many seconds or minutes. Researchers utilizing tasks such as the peak procedure have suggested vet3' different evolutionary pressures for why rats might benefit from being able to determine when 40 see has elapsed. Such a capability could help the animal determine when the expected return at a particular feeding area should begin to diminish (Roberts, 1983). It is unlikely that this representation would need to have the kind of temporal precision required in skilled movement. Indeed, performance on such tasks may not have the same real-time requirements, perhaps being sufficiently served by inferential mechanisms based on behavioral events (Killeen and Fetterman, 1988) or attentional processes.
6. References
Allan LG, Gibbon J. Human bisection at the geometric mean. Learn.. Motiv. 1991" 22 (1-2): 39-58.
Exploring the Domain of the Cerebellar
277
Berger TW, Thompson RF. Neuronal plasticity in the limbic system during classical conditioning of the rabbit nictitating membrane response. I, the hippocampus. Brain Res. 1978" 145 (2)" 323-46. Berger TW, Berry S D, Thompson RF. Role of the hippocampus in classical conditioning of aversive and appetitive behaviors. In: Isaacso RL, Pribram KH, editors. The hippocampus. New York' Plenum Pressl 1986: 203-39. Bullock D, Fiala J, Grossbcrg S. A neural model of timed response learning in the cerebellum. Neural Netxvorks. ! 994" 7' 1101-14. Buonomano DV, Mauk MD. Neural network model of the cerebellum: temporal discrimination and the timing of motor responses. Neural Comput. 1994: 6" 38-55. Flaherty CF. Animal learning and cognition. New York" Mcgraw-Hill. 1985. Fraisse, P. The psychology of time. New York: Harper. 1963. Getty DJ. Discrimination of short temporal intervals" a comparison of two models. Percep. Psychophys. 1975 18 ( I)" 1-8. Gibbon J, Church RM. Sources of variance in an infomlation processing theory of animal timing. In: Roitblat HL, Bever TG, Terrace HS. editors. Animal cognition. Hillsdalc, NJ' Erlbaum. 1984" 465-88. Ivry R. Cerebellar involvement in the explicit representation of temporal information. In: Tallal P, Galaburda AM, Llinas RR, von Euler C, editors. Temporal Information Processing in the Nervous System. New York" Ann. N. Y. Acad. Sci. 19t)3 9682' 214-30. lvry RB, Keele SW. Timing filnctions of the cerebellum. J. Cog. Neuroi. 1989;1(2)' 136-52. Ivry RB, Diener HC. Impaired velocity perception in patients with lesions of the cerebellum. J. Cog. Ncurol. 199 !' 3' 355-66.
278
S. Clarke et al.
Ivry RB, Hazeltine RE. Perception and production of temporal intervals across a range of durations- evidence for a common timing mechanism. J. Exp. Psych." Hum. Perc. Perf. 1995:21 (1)" 3-18. Ivry RB, Keele SW, Diener HC. Dissociation of the lateral and medial cerebellum in movement timing and movement execution. Exp. Brain Res. 1988; 73: 167-80. Jenkins HM. Time and contingenc.v in classical conditioning. In: Gibbon J, Allan L, editors. Timing and time perception. New York: Annals of New York Academy of Sciences. 1984- 423" 242-53. Keele SW, lvry R. Does the cerebellum provide a common computation for diverse tasks'? In" Diamond A, editor. The development and Neural Bases of Higher Cognitive Functions. New York" Annals of the New York Academy of Sciences. 199 I 608:179-2 i 1. Keele S, Pokorny R, Corcos D, lvr), R. Do perception and motor production share common timing mechanisms? Acta Psychol. 1985" 60" 173-93. Kehoe EJ, Graham-Clarke P, Schreurs BG. Temporal patterns of the rabbit's nictitating membrane response to compound and component stimuli under mixed CS-US intervals. Bchav. Neurol. 1989:103 (2)" 283-95. Killeen PR, Fettennan JG. A behavioral theo~, of timing. Psychiol. Rev. 1988; 95 (2): 274-95. Lavond DG, Lincoln JS. McConnick DA. Thompson RF. Effect of bilateral lesions of the dentate and inteq~ositus cerebellar nuclei on conditioning of the heart-rate and nictitating mcmbrane/c.vclid responses in the rabbit. Brain Res. 1984" 305 (2)" 323-30. Mates J, Radii T. Muller U. and Poppcl E. Temporal integration in sensorimotor sychronization. J. Cog. Ncurol. 1994:6 (4): 332-40. McCormick DA, Thompson RF. Neuronal responses of the rabbit cerebellum during acquisition and performance of a classically conditioned nictitating membrane-eyelid response. J. Neuroi. 1984:4 ( 11): 2811-22.
Exploring the Domain of the Cerebellar
279
Meek WH. Selective adjustment of the speed of internal clock and memou, processes. J. Exp. Psych." Animal Behav. Proc. 1983; 9 (2): 171-201. Meek WH. Modality-specific circadian rhythmicities influence mechanisms of attention and memory for interval timing. Learn. Motiv. 1991" 22: 15379. Meck WH, Church RM. Cholincrgic modulation of the content of temporal memory. Behav. Neurol. 1987' 101 (4)" 457-64. Milner B. lnterhemispheric differences in the localization of psychological processes in man. Brit. Med. Bull. 1971" 27: 272-77. Nichelli P. The neuropsychology of human temporal information processing. In" Boiler F, Grafinan J. editors. Handbook of Neuropsychology. Amsterdam: Elsevier Science Pub. 1993" 339-71. Olton DS, Wenk GL, Church RM, Meek WH. Attention and the frontal cortex as examined by simultaneous temporal processing. Neuropsychologia. 1988:26 (2) 307-18. Paxinos G, Watson C. The rat brain in stereotaxic coordinates. New York: Academic Press, Inc. 1986. Perrett SP, Ruiz BP, and Mauk MD. Cerebellar cortex lesions disnlpt learning-dependent timing of conditioned eyelid responses. J. Neurol. 1993' 13 (4): 1708-18. Povel D. Internal representation of simple temporal patterns. Psych." OP. Human Percep. Perform.. 1981" 7 (1)" 3-18.
J. Exp.
Roberts S. Isolation of an internal clock. J. Exp. Psych.' Animal Behav. Proc. 1981" 7: 242-268. Roberts S. Properties and filnctions of an internal clock. In" Mellgren RL. editor. Animal Cognition and Behavior. Amsterdam: Noah Holland Pub. Company. 1983" 345-85.
280
$. Clarke et al.
Smith, MC. CS-US Interval and US intensity in classical conditioning of the rabbit's nictitating membrane response. J. Comp. Physio. Psychiol. 1968; 66 (3): 679-87. Steinmentz JE. Classical nictitating membrane conditioning in rabbits with varying interstimulus intervals and direct activation of cerebellar mossy fibers as the CS. Behav. Brain Res. 1990; 38: 97-108. Thompson RF. The neurobiology of learning and memory. Science. 1986; 233:941-47. Wing AM, Keele S, Margolin DI. Motor disorder and the timing of repetitive movements. New York Academy of Sciences Conference on Timing and Time Perception: Timing of motor programs and temporal patterns. Ann. New York Acad. Sci.. 1984: 423: 183-92. Wing AM, Kristofferson, AB. Response delays and the timing of discrete motor responses. Percep. Psychophys. 1973; 4 (1): 5-12. Yeo CH. Cerebellum and classical conditioning of motor responses. In: Wolpaw JR, Schmidt JT, Vaughan TM, editors. Activity-driven CNS changes in learning and development. New York: Annals of the New York Academy of Sciences. 1991: 627: 292-304. Yeo CH, Hardiman MJ, Glickstein M. Discrete lesions of the cerebellar cortex abolish the classically conditioned nictitating membrane response of the rabbit. Behav. Brain. Res. 1984:13 (3): 261-66.
Time, Internal Clocks and Movement M.A. Pastor and J. Artieda (Editors) 9 1996 Elsevier Science B.V. All fights reserved.
281
TIMING IN P E R C E P T U A L AND M O T O R TASKS A F T E R D I S T U R B A N C E S OF THE BRAIN I
NICOLE VON STEINBI~ICHEL, MARC WITTMANN pOPPEL 1"2
I
and ERNST
Ilnstitut./~ir medizinische t:~'),chologie. Fakultat./iir Medizin. LudwigMaximilians-Universitat Mfinchen. (ioethestr. 31/'1. 80336 Mtinchen. Germany. 2Forschungszentrnm (K1;A). 52425 J#lich. Germany ABSTRACT. For an understanding of how functions are represented in the brain it appears usefi~l io dislinguish between content-related or complex fimctions, e.g. perception, learning, memory and action, and logistics-related or elementary fi~nctions, namely activation and temporal integration and organisation. Elementary fi~nctions thai temporally stnlcture the content-related fimctions are a necessary prerequisite of conscious experience and actions. With various experimental paradigms different levels of temporal processing and organisation of i~fformation and bchaviour are assessed, which can be differentially impaired depending on the nature and localization of damage to the brain. This will be shown for patients wilh focal brain lesions, for patients with symptomatic HIV infeclion and for paticnls with epilepsy who received different treatment.
I. Introduction Time perception and temporal control of perceptual and motor acts have been neglected topics for a long time. Only recently this area of research has gained new momentuna, because it has become clear that without including temporal mechanisms brain mechanisnas can hardly be understood. Such an argument is for instance derived from the persisting problem in the neurosciences how fimctions are represented in the brain. It is probably fair to say that the traditional view of a strict localization of function has been given up: a modified Iocalizational approach appears to be accepted by most neuroscientists, characterized by the idea that local neuronal modules are necessar)' but not sufficient for the implementation of functions. Complex filnctions like speech, vision or feeling pain are not
282
N. von Steinbachel et al.
represented in single circumscribed areas, but depend on the integrative action of distributed neuronal modules (v. Steinbiichel and P~Sppel, 1993). Such a view is for instance supported by studies measuring regional cerebral blood flow or local glucose consumption (e.g., Lassen et al., 1977; Talbot et al., 1991). The challenging question actually is, which specific temporo-spatial patterns represent defined subjective phenomena; it is hoped that by the complementar3." use of PET (positron emission tomography), fMRI (functional magnetic resonance imaging) and especially MEG (magneto-encephalography) this question can be brought closer to an answer. The spatial distribution of local neuronal modules that form complex functions introduces the "temporal question". What neuronal algorithms can one think of that link distributed activities in such a way that integrated percepts and acts are possible, as it obviously takes some time to send neuronal infommtion from one area to another? Another approach that stresses the importance of this question comes from a taxonomy of elementary functions (P6ppel, 1993). It has been suggested to strictly distinguish between content-related and logistics-related functions. Neuronal mechanisms that provide the content of conscious experience or actions appear to be implemented by distributed activities whereas neuronal naechanisms responsible for brain logistics appear to be non-local in their effects. According to this taxonomy as logistics-related functions activation and temporal organization have been identified; without a necessary activation mainl.v by brain stem mechanisms content-related functions run idle: without temporally stnmcturing the distributed activities mental content would be chaotic. It has been a long tradition in the neurosciences that pathological alterations allow a better understanding of nomml function. We present here evidence on timing in perception and action from various patient groups that support earlier findings on distinct temporal mechanisms in neuronal processing (v. Steinb~ichel and POppel. 1991). At first the question of system states in the temporal domain of tens of milliseconds is addressed; here we believe to have identified a neuronal mechanism allowing binding of distributed activities. Then we discuss some results on temporal integration which links successive elcmc.ltar).' events into units. It has been suggested (v. Steinbiichel et al., 1993) to interpret the "subjective present" as the phenomenal representation of temporal integration in the domain of a few seconds. Finally we come to an intermediate level where we describe neuronal mechanisms underl.ving simple voluntary motor acts like tapping or
Timing in Perceptual and Motor Tasks
283
reaction time. In all these cases evidence from patients with brain lesions provides better insight into the u,aderlying neuronal mechanisms.
2. System states: Effects after brain lesion 2.1. THEORETICAL CONSIDERATIONS There are scvcral reasons why the brain might require specific system states within which neuronal information is processed. It has already been mentioned that distributed activities pose a logistical problem. Another reason for introducing system states comes from physics and biophysics characterizing the different sense modalitics. By comparing the visual with the auditor), modality the transduction time of physical stimuli is considerably longer for vision then for audition: transduction of acoustic stimuli takes less then ! ms in the auditory and transduction of optic stimuli some tens of milliseconds in the visual modality (POppel et al., 1990). Additionally time is neccssar).' for the sound wave to reach a subject. Therefore a logistical problem of temporal availability of information of one object transducted by two different sensor)' systems arises. If the object is close, it will be processed earlier in the auditory domain - and then made centrally available - because of the rather short transduction time. At a distance of approximately 10 m visual transmission time (under optimal conditions) may correspond to the transmission time of sound" only under this rare condition one could obtain simultaneous central availability. (We have referred to this special distance as the "horizon of simultaneity"). For greater distances the visual information arrives in the brain earlier then auditory information. This problem together with the intra-ccrebrai logistical problem has led to the hypothesis that the brain itself creates system states in which all occuring stimuli are treated as "co-temporal". These system states are presumably based on neuronal oscillations, and experimental evidence suggests that their period length lies in the range of 30 to 40 milliseconds. It is believed that the temporal order of stimuli - and accordingly subjective perception of succession of events - can only be established, if their neuronal representations appear in t~vo different periods of these oscillations. Neuronal activity within one temporal window defined by these oscillations is classified as 'co-temporal' and integrated into one unit. even if located in modules distributed widely across the brain. The theoD' therefore implies
284
N. von Steinbachel et aL
discontinuous, temporally discrete information processing with units of 30 to 40 milliseconds lenght. 2.2. EVIDENCE 2.2.1. Experimental evidence. Evidence to support this theory is to be found
in several fields of the neurosciences; e.g. psychophysical, neuropsychological and ncuroph.vsiological studies including single cell as well as mass potential measurements provide essential data on this topic. Gardner and Costanzo (1980). for instance, showed temporal integrations of single cells in the somatoscnsor3, cortex in this temporal range. Studies on the effects of general anaesthesia on auditory evoked potentials showed a usually detectable 30 to 40 Hz oscillation to disappear under anaesthesia (Madler and P6ppcl, 1987). It has been concluded, that this oscillatory component is a prerequisite for integrated information processing, and that its suppression for instance by centrally acting anaesthetics renders information processing impossible. Thus. the building-blocks of conscious experience can no longer be fom~ed, as the primordial events cannot be derived from neuronal activities within such system states. Patients having gone through such states often report that "no time at all" has passed. Psychophysical evidence concerning visual and auditory perception as well as motor behaviour also supports the notion of temporally discrete infommtion processing in the domain of 30 to 40 milliseconds. For example in tasks that require the coordination of perception and action a multimodality with intemlodal distances of approximately 30 ms can be detected (Radii et ai., 1990). In such experiments on finger tapping subjects are asked to s3~chronize their motor response with a repetitive acoustic pattern. A regular synchronization error is found showing that subjects respond systematically ahead of the actual stimulus representation (negative phase-shift), although they had the subjective impression of being exactly in time. The analysis of swlchronization errors reveals multimodal distributions with intermodal distances of approximately 30 ms between the local maxima. Thus, these results point to temporally discrete psychomotor control. In studies on auditor3.', visual, and tactile temporal order thresholds [OT], where subjects are asked to decide which one of two consecutively presented stimuli appeared first, the minimal interstimulus interval [ISI] allowing reliable distinction is determined. In ease of auditory OT measurement pairs of laterally alternating clicks of I ms duration are presented via headphone randomly starting on the left or right side. The temporal distance between the
Timing in Perceptual and Motor Tasks
285
clicks is also randomly varied, and subjects have to decide, which click was first. Values converge generally at an ISI of 30 to 40 ms in young healthy subjects and of 40 to 60 ms in elderly subjects (v. Steinbiachel and POppel, 1991). This intemlodal convergence of order thresholds which doesn't vary, over the different sense modalities suggests a centralmechanism participating in temporal ordering of perceptual events. 2.2.2. Clinical Evidence. One question is whether a brain lesion might effect the processing of information on this temporal level. This is indeed the case as reported by Efron (1963) and Swisher and Hirsh (1972), who found an increased auditory OT in patients with brain injuries, especially in patients with lesions in regions responsible for speech perception and reception. If two successive phonemes are represented in the same pathologically extended period of a neuronal oscillation - instead of two consecutive periods as ill healthy subjects, the temporal order of the phonemes can no longer be recognized. Also the ability to recognize and distinguish stop-consonant-vowel syllables with different voice-onset-times (the time passing from the noise burst of a plosive to the onset of laryngeal pulsing) may be impaired in patients with aphasia (Tallal and Ncwcombe, 1978). In order to discriminate between voiced and unvoiced consonants (as in the syllables/da/vs. /ta/), temporal resolution has to bc in the range of some tens of milliseconds. These results lead to the formulation and subsequent testing of the hypothesis that an improvement in temporal discrimination abilities may also improve certain linguistic compctcnces in aphasic patients (v. Steinbiichcl, 1995a). It was found that auditor3.' order thresholds in patients with aphasia (left hemispheric lesions) were well above 100 ms (as compared to 32 ms in healthy controls), but reduced to nonuai values after OT training with positive verbal feedback. Two control groups of patients also trained with positive feedback in pitch discrimination (control group 1) or a visual searching task (control group 2). sho~vcd no such improvement (Figure 1). To test a possible transfer of this temporal improvement to speech recognition in the patients who had received OT training, the ability to discriminate stop-consonant-vowel syllables with different voice-onset-times was examined. Compared to normal controls, the three patient groups showed a marked deficit in separating the two phonetic categories. Only patients that had received OT training showed a significant improvement in differentiating between the syllables /da/ and /ta/ (Figure 2). Thus, this experiment shows, that only patients traincd in temporal processing with the perception but also in the context of language acquisition, emphasizing the
286
N. von Steinbachel et aL
200 gxperi3tental Group
160
-
1 6 0
-
Control Group I I ------
Control Group II Healthy V o l u n t e e r s
140 120 UI
lO0vl ~
806040-
0 bcs
Training
after 4h
aflcr 8h
Training [Is] Figure 1. Means of auditor3. order thresholds (OT [ms]) for three patient groups with aphasia" the experimental group and two control groups in comparison with healthy volunteers before, after 4 hours and after 8 hours of OT training. OT training paradigm showed an improvement in an elementary linguistic task. This result may bear significance nol only with rcspcc! to speech.
importance of high-frequency temporal information processing in the development of speech. Also. possible therapeutic implications must be considered. We propose, that in therapy of aphasia attention should also be directed to improving neuronal functions like temporal coding and decoding of stimuli. which may enhance the effects of a linguistically oriented therapy.
Timing in Perceptual and Motor Tasks
.:...-:\. ~:~:;,. ..............
:;:::.....
~p.~.,~t.~
--dr-
}............. ~ * " ~ ' 0 ~
G,o,,,p] ~ I
~'*~' Group
~!~:,.... C o n t r o l
80 ~.~:~ ....... .-..',~i X:::!!:.: ........
287
II
I
L,,m,d
~9
6o-
..... ~:.'.~:....... ".~.J......i~ii
0'I
o
40-
W)
u
A
"~"
"'A
20-
100 1 ,
I ........T
,,1,
:~iii!!i.-"-A _ _ ~, "".'::
,--,,,, 9 0 -
l
V----
"I"
"t-
m,
m
.~
-, "."
"~'a "I::I
. ::..:...."":Y 60
- "
~::~i!:. X x~~~
'"...
;':':':"-N..
:::.
,
:, ..'"
...:,
...........
,-iiii~:,!......... .,'. "r..,
";"
"~':i'-~!~;
... :%....."
I=
o
40-
9
::'":',...
... . ;:.':
"';::,,
':!i!ii:.:.
2O
""
~--&_.i__ ~k---A
0 -I......~........ ~..... T. . . . . 0 10 20 30
T. . . . .
40
V-----" ,
50
60
70
80
90
Voice-Onset-Time [ms] Figanre 2. Ida//ta/discrimination I%/da/I bcfore (upper figure) and after (lower figure) tile order threshold Iraininlg of tile experimental patient group with aphasia and tile two patient control groups wilh aphasia in comparison with heallhy volunteers.
288
N. yon Steinbachel et al.
On the basis of this result tile question was formulated whether a prolongation of OT is generally seen after brain injury or whether special focal lesions in areas responsible for the production or reception of speech are selectively associated with a deficit in temporal processing. Five groups of patients with focal brain injuries were studied together with a control group of patients who had orthopaedic injuries: 1. lef~ hemisphere pre-central (LH-pre) with nonfluent aphasia (Broca), 2. let~ hemisphere post-central (LH-post) with fluent aphasia (Wemicke- amnesic, transcortical-scnsor3. ). 3. left hemisphere without aphasia (LH non-Aphasia, including predominately patients with subcortical lesions) 4. right hemisphere pre-central (RH-prc) 5. right hemisphere post-central (RH-post). Groups 4 and 5 show lesions corresponding to the lesions in group 1 and group 2. The experiments show significantly prolonged auditory OTs only in the group of patients with left hemispheric post-central (LH-post) lesions leading to fluent aphasia (Figure 3). The performance of all other groups was not significantly different from that of healthy controls - including patients with left hemispheric lesions and nonfluent aphasia (LH-pre). At first sight the result concerning patients with nonfluent aphasia seems rather unexpected. However, already in the study of Swisher and Hirsh (1972) patients with left hemisphere lesions and Broca aphasia performed more accurately in the perception of temporal order than the patients with fluent aphasia. Thus, areas of the left hemisphere associated with speech recognition seem to be among the regions particularly vulnerable to disturbances in the temporal range of 30 to 40 ms (v. Steinbtichel et al., 1995). If neuronal oscillations with periods of approximately 30 to 40 ms are responsible for creating distinct successive system states this result suggests that a certain oscillatoo, process is selectively prolonged in the primarily or secondarily affected regions of the brain. From a neurophysiological point of view this suggests that oscillator3.' phenomena in neuronal nets are not one general property of the entire cortical mantle, but that certain local areas have their owaa neuronal machinery to deal with successive stimuli. It is worthwhile stressing in this context that one group of patients, those with left hemispheric injuries who do not suffer from aphasia and have predominately subcortical lesions show OTs that are similar to those of the controls, some of them even better. Timing then appears to be a selective .
289
Timing in Perceptual and Motor Tasks
140
-
-
120
100 m
-
80604020-
.L,
LH pr 9 Control Group
LH w i t h o u t Aph. LH post RH pro
RH
post
Figure 3. Means of auditory order Ihresholds lOT] for five patient groups wilh focal brain injurics and a neurologically healthy group: significant differences between groups arc shown.
property, and special algorithms must bc required to bind activities of different regions together.
3. Temporal integration after brain lesions Another level of temporal organization of neuronal information processing is observed in the integration of successive system states and sequences of events into one homogeneous 'gestalt' (POppel, 1978: 1994). Integration works up to approximately 3 seconds and the perception (or notion) of the so-called subjective presence seems to be associated with it. This mechanism appears to be automatic, i.e. prior to any semantic evaluation. It can be observed in perceptual and cognitive tasks and in motor behaviour. In ethological findings (Schlcidt ct al.. 1987: Kien et al., 1991), the exact measurement of durations of individual movement patterns in daily voluntar3, short-term activities -especially repetitive bouts of behaviour, e.g. while eating, working, scratching or speaking - revealed a marked peak in frequency distributions. Across different human cultures this peak is
290
N. von Steinbachel et al.
constantly produced close to 3 seconds. One can conclude therefore that intentional acts are temporally embedded in a "3-second-window". A similar segmentation can be seen in sensorimotor synchronization. This task requires that a subject synchronizes finger taps with a regular sequence of sensory stimuli. For short interstimulus intervals (ISI: for instance 600 ms) one observes a typical synchronization error. The subjects anticipate stimulus occurence by some tens of milliseconds. This anticipation is seen only up to approximately 3 seconds. In response to a further elongation of ISI the synchronization error drifts to increasing positive values (Mates et al., 1994). it appears, that up to a limit of 3 seconds coordination of motor programs with anticipation of sensory stimuli works well: above that limit temporal integration abilities of the brain - in comparing planning and outcome of motor actions - seem to break down. This experimental paradigm can easily be applied in patients with different brain lesions. Here xve report some observations on patients with epilepsy who were treated with either Phen.~1oin or Carbamazepine and a control group of healthy subjects (v. Stcinb~ichel et al., 1992: v. Steinb/Jchel, 1995b). Six different IS! with lengths ranging from 600 to 3600 ms were presented. Two different measures were taken, namely the key touching time (KTT), assessing the more motoric aspects of synchronization, and the synchronization error (SE) which indicates the efficiency of dynamic anticipator' timing aspects of s.vnchronization. It can be seen in Figure 4a that KTT is a function of ISI lengths. The longer the pause between successive stimuli becomes the longer the response key is pressed down. This appears to be a straight forward result, but poses some interesting questions, looking at the temporal dynamics. The entire movement is a complex ballistic movement, with KTT we measure the interval between the end of an agonistic and the beginning of an antagonistic movement. Such a movement is highly automatized and the result suggests that there is a transfer from the sensor3' domain to the programming of a movement that appears to be controlled rather peripherally. The second important result is that although the two patient groups and the healthy volunteers show the same general characteristics, they are differentially slowed down. in comparison to healthy controls the two patient groups are significantly slowed down. the Phenytoin group showing the longest values. This indicates that the disease (and/or the treatment of the disease) effects a specific aspect of motor programming in a systematic way leaving intact, however, the gcneral dynamics of organisation of motoric aspects of behaviour.
Timing in Perceptual and Motor Tasks
300
291
-
280 q l
El 2 0 0 ~
150-
100 -
--X-__s
50-
phenytos
Carbamazepine Healt:
_
. . . . . . . . T----
i
600
hu
Volunt:eers
I...... |1800 2400
1200
-
~ .... 3000 3600
lSI [ms]
Figure 4a. Means of key touching limes (KTT lms]) for six different ISis [ms] in a sensorimotor synchronisation t~sk of two patient groups with epilepsy treated with Phenytoin or Carbamazcpinc and of hcallhy volunteers.
so
i
|
o-t
.
9
I~
-50-
" -::
:.
"
,
.
.
.
.
.
.
.
. . .
....
.
/
"
J
-100
--
~
H e a l e l a y -
.
Phonytos
Carbama -150
i...-.
I
600
zepine V o l u n e e e r s
. . . . . . . . . .
,
-
2400
"
I
3600
ISI [ms]
Figure 4b. Means of tile synchronisation errors (SE Ires]) for three different ISis Imsl in a sensorimotor synchronis;rlio, task of two patient groups with epilepsy treated with Phen.~loin or Carb~lm;izcpine a.d of healthy vohmteers.
292
N. von Steinbachel et al.
A different result is observed in the measure of the efficiency of synchronization (Figure 4b). For the Phenytoin group we observe for all ISls a clear anticipation of stimuli (negative synchronization error) with a tendency to longer SEs with increasing ISI lengths. For the Carbamazepine group one sees a slight decrease of SEs with increasing ISl lengths. The same applies to the control group, only that with an ISI of 3600 ms they show a positive phase shift. With respect to s.~aachro,aisation, over all groups clear anticipations are observed only for the shortes~ ISi of 600 ms. In all other ISis the Phenytoin patients differ totally from the txvo other groups. This stands in contrast to the findings in the KTT measure. Although patients with Phenytoin or Carbamazepine treatment are slowed down compared to the control group. they show the same temporal pattcr,a as the healthy controls - with respect to the functional increase of KTT. These findings indicate that the processes for the anticipation of sensor3.' stimuli and the dependent motor commands are systematically different. Therefore we conclude that there are at least two different neuronal programs involved, one for motor anticipation of stimuli and one for the execution of movements. Temporal integration can also bc studied with the technique of nonverbal reproduction of temporal intervals" either optic or acoustic stimuli are presented to the subject, and the task is to reproduce as precisely as possible the stimulus duration, it has been observed that intervals up to approximately 3 seconds can be reproduced appropriately and with a small variance, reproduction being both accurate and precise. Intervals longer than 3 seconds are usuall.v underestimated with high intra-individual variance (P6ppel. 1978). It has been pointed out that two different mechanisms are responsible for temporal reproduction: one operating up to approximately three seconds, a second for longer intervals being characterized by a qualitatively different gain function. The aprupt transition from one mode of reproduction to another suggests two different neuronal mechanisms, which are described by fitting of two different regression lines. It has been suggested to relate this phenomenon of temporal reproduction to the well kno~a~ time-order error. If two stimuli have to be compared with respect to subjective intensit.v, the comparison is rather accurate if two stimuli are presented within a temporal window of only two or three seconds. If the interstimulus interval is longer, usually the first stimulus will be underestimated substantially, as if a central representation has faded with respect to its intensity. In experiments with the patient groups with focal brain lesions (mentioned under 2.2.2). the general observation of two
Timing in Perceptual and Motor Tasks
6000
293
+ A
FtH p r e
..... :MIV.....
..~ ~9-~.:!::-~..... !.~ LH
m
E SO00"
M
pre
l+lllll lttlll~
Control
lllll,ll
Group
Standard
..................
0
e~
m 40oo -
+mlo1411
.,, ...~-'~
....,....~. ,.:
o
3000 -
,l,,,I r
-
....
:+~+::.. 2000-
..~.
IX,
....
i- :~!!i"
......~i~','."
~
1000 -
I"~:
1+o00
~
....
-
,"-,
.....
li~
l
~
:V
~
i
-t .....
.....
1
. . . . . .
i-" .....,
~ 5000
~H
post
I,H
post
.,,o' ,.,'
Control
o
.,,,.'~
Group
..,,.'"
Standard
4.a
..... _.
.
.
.
-•
_
4000 i+lil~
'~S'":::;-.....
o soou....-. .:,,~~
O
....
2000 -
G Ioo0
0
-
=-
0
F-
i - + l - - - r -
'i'
i
:+
r
-
r
--
i
......
r
1000 1500 2000 2500 3000 3500 4000 4500 5000 5500
IS[ [ms] Figalre 5. Mean visual reproduclion limes [ms] for patients with acquired focal pre-cenlral lelt and righl lesions (uppcr figure) and acquired focal l~Sl-ce.tral left and right lesions (lower figa~rc) a.d for a group of neurologically healthy patients.
294
N. von Steinbachel et al.
mechanisms of temporal reproduction has only been observed for healthy volunteers and for patients with post-central lesions (v. Steinbiichel et al., 1993). As can be seen in Figure 5 (lower figure), intervals up to 2.5 seconds are reproduced very close to tile standard, and longer intervals are reproduced shorter. If one looks at reproduction of different time durations in patients with pre-eentrai lesions, a qualitatively different result is obtained (Figure 5, upper figure). Wheras the results from the patients with post-central lesions correspond to those of the healthy volunteers, there is a marked difference for patients with pre-central lesions. After right hemispheric lesions, a general tendency for a longer reproduction of intervals is observed; after left hemispheric lesions a tcndcnc.v for shorter reproduction of intervals can be seen. Possibly these data indicate a stabilizing mechanism of the frontal lobes with respect to temporal reproduction and thus temporal integration. After lesion of the right hemisphere one could assume that the eigen operations of the let~ hemisphere are set free. in this case one sees a longer reproduction of temporal intervals suggesting a natural tendency of temporal expansion of the left hemisphere. The opposite can be argued for the right hemisphere; after left hemisphere injuries, cigcn operations of the right hemisphere are set free which, according to the data. suggest a tendency of temporal compression. Thus, expansion of the left and compression of the right frontal lobes with respect to temporal reproduction operate against each other. If one assumes interhemisphcric i.ahibitor3' interactions these two opposing tendencies - like a l~ush-pull mechanism - control each other. Thus, under normal circumstance one sees a regular sequence of approximately three second integration intervals, and only after frontal injury the intrinsic operations of the left or right frontal lobe are made visible. These temporal dynamics allow fi~rther speculation, if the unilateral natural tendencies are modulated by other neuronal processes, the interhemispheric inhibition can be influenced and thus the intcgratio,1 intervals can be altered. This would create a certain flexibility of temporal integration in order to deal in an adaptive way with sensor3.' input or endogenously created information, for instance if attentional demands are changing. Temporal integration can be also studied with a paradigm of ambiguous figures (P6ppel, 1994) like the Necker cube or the Rubin figure (two faces versus one vase). In study'ing spontaneous alterations of ambiguous figures (v. Steinbflchel et al.. 1993) it could be demonstrated (Figure 6) that approximately three seconds arc nccdcd in healthy volunteers to keep one
295
Timing in Perceptual and Motor Tasks
,..,8ej u
0
U
U
3--
2"
0 .
.
.
.
.
.
.
.
!
. LH prc
I
! ......
r
u
.m7 O
~s
g, 2 t 0
.
.
.
Control Group
i LH without Apls. LH post ILH p r u
KH post
Figure 6. Means of perccplual durations Isec.] of the ambiguous figures. The Necker cube (upper figurc) and Ihc succcssion of syllabcls "soma-maso" (lower figure) for five palienl groups with acquired focal brain lesions and for a group with neurologically hcallh.v palic~ls: significant differences between groups are shown.
296
N. yon Steinbachel et al.
alternative in the focus of attention. Only patients with prefrontal lesions (either the left or the right ilcmisplaere) show a significantly longer alteration rate (Figure 6). In these patients not only visual but also auditory alteration rates were studied, using acoustic stimuli. In the experiments the patients had to indicate whether they heard the syllable sequence SO-MA or MA-SO. The task was to push a button when it was no longer possible to prevent hearing (seeing) the alternative syllable (figure)i As can be noted the temporal structure of auditor3.' alteration is the same as for visual alteration. Again healthy volunteers show a value of approximately 3 seconds. Patients with pre-central left and right hemisphere lesions show again considerably longer alteration rates. The fact that after brain injury the duration of the perception of the ambiguous alternatives, both for optic and acoustic stimuli, is only prolonged in patients with pro-central and not with post-central lesions suggests a general neuronal mechanism which is shared by both sensory modalities. One might think that temporal reproduction durations and spontaneous alteration rates of anabiguous figures both represent a common neuronal mechanism of temporal integration operating up to a limit of approximately 3 seconds. It has to be explained then. why there is a systematic difference for patients with pre-central lesions as observed here. In the experimental paradigm of temporal reproduction we see a clear dissociation for patients with lett or right hemisphere injuries: only' patients with lesions in the right hemisphere corresponded in their responses in the ambiguous figures task to their responses in the reproduction paradigm, showing a prolonged integration span. Patients with lesions in the lett hemisphere however performed with prolonged responses in the ambiguous figure task (as the patients with right pre-central lesions), but showed a significantly shorter duration time in the temporal reproduction task. We believe that the qualitative difference in these experiments is due to different attentional demands, although in both paradigms temporal integration is tapped. Attentional demand in the experiments on the spontaneous alteration of optic or acoustic stimuli is considerably higher because of the continuous dynamics of the task. Thus. this additional factor might be responsible for the different results. 4. Disturbances of motor control after brain damage We have suggested two operational levels in temporal control of perception and action, namely a high-frequency mechanism with periods in the range of
Timing in Perceptual and Motor Tasks
297
30 to 40 millseconds and a low-frequency mechanism in the domain of 3 seconds. These processcs sccm automatic in nature, i.e. endogenously implemented, and they provide a sequential dynamics of neuronal information processing. Embedded within these two temporal domains, there is a host of temporal phcnomc,m that relate neuronal information processing in a reactive way to sensor)' infonnation. Typical phenomena at this level of some hundreds of milliscconds are simple or choice reaction times and repetitive motor acts asscsscd by maximal tempo or personal tempo. We report here some observations from this i,~termediate level on patients with systemic changes of brain fi~nction and healthy controls in order to gain some better insight in central and peripheral timing of motor control. At first we want to report somc data on choice reaction time in health), volunteers and paticnts with HIV i,lfections. In this experimental arrangement two variables can bc determined, namely release time [RLT] which reflects the central cognitivc aspect of the reaction including choice, and movement timc IMTI which reflects a more peripheral motor aspect. The subject is askcd to hold a release key. and whenever a stimulus appears he or she has to move with the index finger to the correct response key chosen with respect to thc particular stimulus. It has been observed in earlier experiments that paticnts with HIV infection show a prolonged total reaction time (Perdiccs and Cooper. 1989). In our experiment it can be seen (Figure 7) that the two components assessed by the choice reaction time paradigm are prolonged in these patients (v. Steinbtichel, 1994). Both the MT and the RLT arc substantially lengthened in patients with HIV infection. This observation corresponds with the general hypothesis that the virus leading to AIDS affects the brain at an early stage and leads to a slowing d o ~ of ccntral processcs [RLTI. but as we see here also of peripheral motor responses [MTI. If one compares the mean coefficiant of variances of the patients both for MT and for RLT with healtlty volunteers, it becomes obvious that both components have bccome temporally less stable. Using the same reaction time paradigm in patients with epilepsy it was observed (v. Steinb~ichcl, 1995b) that such patients show also prolonged movement and release times, in particular patients treated with Phenytoin are characterized by a slo~vcd-down central process compared to patients treated with Carbamazcpinc or healthy controls. A slowing-down of movement time was sccn in both patient groups. A slowing-down of central and peripheral aspects of timing in patients with HIV infcction has also been observed in measuring the "personal tempo", hc subject is asked to press in a regular and personally
298
N. von Steinbachel et al.
RI.,T
MT 500
I
-
SD-MT Mean-MT
,-,
400-
$D-RLT
i
Mean-RLT !
i~,,q
O
300-
~
2oo-
I I I
I 100
-
'
I ! ._
HIV
H 9
Voluntc 9
HIV
Patients
Hcslthy
1
-
Pstl 9
Voluntccrs
Figure 7. Means and sl~lnd~lrd devil, ions of movement times (MT ims]) and release times (RLT Imsl) in a choice reaction task of patients with symptomatic HIV infection and for a group of hc;lllhy volunteers.
comfortable tempo a response kc.v. This simple task allows to determine at least three different variables: Key touching time [KTT] can be determined (the interval between the end of an agonistic and the beginning of an antagonistic movement)" the pause interval [PI| can be measured (the interval between successive key touching times" the end of the previous and the beginning of the next KTT). and additionally the sum KTT and PI which can be referred to as intcrrcsponsc interval [IRI] can also be analysed seperately. Although mathematically IRI is the sum of KTT and PI, IRI has to be treated independently because it is a question whether the addition is done with either the proceeding or succeeding interval. Without further going into a deeper analysis in this paper there is strong evidence using autocorrelational analysis that KTT has to be related to the succeeding PI, because strong correlation is observed. Thus, the sequence of KTT and PI are programmed as one unit. and the two elements are supervised by a general control mechanism.
299
Timing in Perceptual a n d M o t o r Tasks
KTT
i 9 ,i
1000 -
PI
|
~]
SD-KTT Mean-KTT
800
SD-PI
:
Mean-PI ,ii~
__
_
~~ ....~..~.y..?::.~... ~-~
600". "'"-."'"'.. ""'.. "".."~ ,."" -,,'"'".."'...'""'..""'...
I I
, .. . ....... ' ..................... !
i!!~!:ii~;~.i:: .. ~:ii!i~!:iL:::~::
,........'............"......"...... I ilili!!ili~iiiiiil/iiii~!iiiiiiii~ii~i ~ ,._""/C"-.,~"...;I".,.I .9" . . . '...... i ........... ~ !i................................ ~i:!~:~i~i~:~i!/~"i!i~i~i~?ii;!7
400-
.......
,
200
~<,<<<<
L .......
0 -~ HIV Healthy
Volunteers
Patients Healthy
HIM Patients Volunteers
Figure 8. Means and standard dcvialions of key touching times (KTT [ms]) and pause intervals (PI [msl) i, the pcrsotmi tapping tempo of patienls with symplomatic HIV infeclioll a,d for a group of healthy volunleers.
In Figure 8 again substantial differences in timing processes can be seen if healthy volunteers are compared to paticnts with HIV infection. The mean values for both variables are much longer in patients. As in the experiments on reaction time. both the motor timing component which is reflected in KTT, and the central timing component which is reflected in PI, are prolonged in this patient group. It is interesting to note that in both experiments, in a task which requires a reaction, and in a task that reflects dynamic timing, timing is significantly cffccted on two distinct levels (motor and central components) in these patients. As in the case of reaction time, a large intra-individual variance indicates an instability of timing in patients with HIV infection. The experiment on personal tempo has also been performed with patients having suffered circumscribed lesions in the brain. As can be seen in Figure 9, significant differences have been observed between healthy controls and the various patient groups. The patients with left hemispheric pre-central lesions are significantly prolonged in central timing aspects (PI), compared to healthy controls or patients with left hcmispheric lesions not leading to
300
N. von Steinbachel et al.
aphasia. Patients with left laemispheric post-central lesions are also slowed in this aspect, but to a smaller degree than patients with left hemispheric precentral lesions. Patients with right hemispheric pre-central lesions resemble the control group (patients with right hemispheric post-central lesions were not included in this survey because not enough patients (only 4) were available in the group for statistics). In the several above mentioned tasks one group of patients did not fit in the general picture, namely the patients with left hemisphere injuries (predominantly subcorticai lesions) not leading to aphasia. These patients show for instance a rather low order threshold (Figure 3) and much shorter key touching times and pause intervals in their personal tempo compared to all the other groups, it looks as if. because of these injuries central timing mechanisms as well as peripheral sensory and motor mechanisms are accelerated. This could mean that under normal circumstances structures in the left hemisphere which arc not related to speech processing might play an important role in slowing-down central and peripheral sensory and motor mechanisms.
5. Conclusion
Looking at different levels of temporal processing it becomes clear that cerebral timing mechanisms are not simple phenomena. There are different neuronal processes which arc responsible for high-frequency and lowfrequency processes: filrthcrmorr there are different mechanisms observed for the intermediate level. Motor and sensory timing have to be distinguished. Different areas of the brain seem to be differentially involved in different timing processes. In particular, pre-central mechanisms may play an important role in stabilizing temporal integration. Systemic cerebral diseases can effect both central and peripheral timing. One of the remaining questions is how the temporal mechanisms that can be identified in different timing domains and which arc associated with different brain regions are interlinked with each other so that homogeneity of experience and continuity of time is possible. We believe that in order to address this question, the different timing mechanisms have to be identified and dissected before an integrative model pattern can be developed.
Timing in Perceptual and Motor Tasks
Key
Touching
301
Time
--
300
-
,..~200
I , ,
, , , . .
,~:.~
~,
,
,----.,
100
+. . . . . .
i+ :,~ +?~i
-
0, Interval
Pause -t
1000
II00
---, 9
--
I~O0-
B ~
41oo -
200
-
0 -
N
//+
_m Interrcspons 9 I n t e r v a l
1200
-
iloo-
40O
II
_ramira..
0
Control
LH Group
pre
LH LH
post
without
Aph. R.H
pre
Figure 9. Means of key Iouching lilncs (KTT [IllS]), pause intervals (PI [ms]) and interresponse intervals (IRI Imsl) in the personal tapping task for four patient groups with acquired focal brain lesions and for a group with neurologically healthy patients: significant diffcrcnccs bclwccn groups are shown.
302
N. yon Steinbachel et al.
6. References
Efron R. Temporal perception, aphasia and d6jb, vue. Brain. 1963: 86: 40324. Gardner EP, Costanzo RM. Temporal integration of multiple-point stimuli in primary somatosensor3., cortical receptive fields in alert monkeys. J. Neurophysiol. 1980: 43" 444-t~8. Madler Ch, POppcl E. Auditor3.' evoked potentials indicate the loss of neuronal oscillations during general anaesthesia. Naturwissenschatten. 1987; 74: 42-3. Kien J, Schleidt M, Sch6ttncr B. Temporal segmentation in hand movements of chimpanzees (pan troglodytes) and comparison with humans. Ethol. 1991" 89: 297-304. Lassen NA, lngvar DH. Skinhoj E. Brain filnction and blood flow. Scientific American. 1977" 239 (4)" 50-t~. Mates J, Mf~iler U, Radii T. P6ppel E. Temporal integration in sensorimotor s3.aachronization. J. Cogn. Neurosc. 19t/4" 6" 332-40. Perdices M. Cooper DA. Simple and choice reaction time in patients with immunodefiencv vinis infection. Aim. Ncurol. 1989" 25" 460-67. P6ppel E. Time Perception. In" Held R. Leibowitz HW, Teuber HL, editors. Handbook of Sensor3.' Physiology Vol. 8" Perception. Berlin: Springer. 1978" 713-29. POppel E. Taxononay of subjective phcnonlena: A neuropsychological basis of functional assessment of ischcmic ot traumatic brain lesions. Acta Neurochir. Suppl. 1993" 57:123-2t/. P6ppel E. Temporal mechanisms in perception. Int. Rev. Neurobiol. 1994; 37: 123-29. P6ppel E, Schill K. v. Steinb~ichel N. Sensory integration within temporally neutral system states: a hypothesis. Natu.wissenschafien. 1990; 77: 89-91.
Timing in Perceptual and Motor Tasks
303
Radii T, Mates J, Ilmbcrgcr J. P6ppci E. Stimulus anticipation in following rhythmic acoustical patterns by' tapping. Experientia. 1990: 46: 762-63. Schleidt M, EibI-Eibesfr !. POppcl E. A universal constant in temporal segmentation of human short-term bchaviour. Naturwissenschaften. 1987; 74" 289-90. v. Steinbi.ichel N. Gcsundhcitsbczogcnc Lebensqualit/it bei HIV-positiven Menschen. In: J~gcr H. editor. HIV-Mcdizin: Mfglichkeiten der individualisierten Therapic. Wisscnschaftlichc Ergcbnisse in der Mitte der 90er Jahre Landsberg/Lech: ecomcd. I t)t~4:337-40. v. Steinbiichel N. Temporal systcm states in speech processing. In: Hernnann HJ, Wolf DE. P6ppcl E. editors. Workshop on supercomputing in brain research: from tomograph.x' to neural networks. Singapore: World Scientific Publishing Companx. I t~t)5a: 75-82. v. SteinbiJchei N. Gcsundhcitsbczogenc Lebensqualitht ais Beurteilungskriterium fiir Behandlungscffcktc bci Patienten mit Epilepsie. Z f Reha Pr~iven. 1995b: 7(3)' 139-46. v. Stcinbi.ichei N. P6ppcl E. Temporal order threshold and language perception. In: Bhatkar VP & Rcge KM. editors. Frontiers in knowledge-based computing. Ncxv Delhi: Narosa Publishing House 1991" 81-90.
v. Steinbflchcl N. P6ppcl E. Domains of rehabilitation: a theoretical perspective. Behav. Brain. Rcs. l t~t~3 956" 1-10. v. Steinbtichcl N, P6ppcl E. Hiltbrunncr B. Independent temporal factors underlying cognitive processes. Abstract. 22nd Annual Meeting of the Society for Neuroscienccs. A.mhcim. I t~t)2. v. Steinbiichel N, Rciscr M. P0ppcl E. Temporal constraints of cognition: fi~rther evidence of automatic temporal integration. Abstract. The 23rd Annual Meeting of the Society fbr Ncuroscience, Washington. 1993.
304
N. yon Steinbflchel et al.
v. Steinbiichel N, Reiser M, Wittmann M, Szelag E. Temporal constraints of cognition: temporal information processing in different patient groups with acquired brain lesions and in controls. Abstract. The Meeting of European Neuroscience, Amsterdam. 1995. Swisher L, Hirsh IJ. Brain damage mid the ordering of two temporally successive stimuli. Neurops.vchoi. 1972; 10: 137-52. Talbot JD, Marrett S, Evans AC. Mc.ver E, Bushnell MC, Duncan GH. Science. 1991, 251, 1355-57. Tallal P, Newcombe F. Impairment of auditory perception and language comprehension in dysphasia. Brain Lang. 1978" 5" 13-24.
305
AUTHOR INDEX African mormyriform fish, 29 Alvira, R., 127 Allan and Gibbon, 12-3" 147; 195-6; 263; 265 Amagai, S., 5; 27; 40; 42" 45 Ar6chiga, H., 95" 107:109- I 0 Artieda, J., 1; 5: 7" 10: 14: 18: 21; 235,238" 241" 243: 248-50: 253-4 Block, R.A., 18" 104, I l !-2 1436; 148" 150" 152-4: 157-8, 160: 163; 197" 200:204 Bressler, S.L., 53-4.59: 60' 62: 64-5 Cart, C.E., 3-5, 18-9' 27" 29' 33: 35" 37-8" 40-2; 44: 46: 76:89-90 Campbell, J., 115 144" 159 Clarke, S., 207; 214, 217' 278 Day, B.L., 223-6, 228" 230, 2334; 250; 252 Grinband, J., 257 Ivry, R., 148; 160; 192" 196" 202" 204: 252; 257-9; 26 I" 266" 276-8 Miall, R.C., 69; 76-81 87: 90: 92-3 Nichelli, P., 12-3" 22" 141' 187: 198 203; 279 Pastor, M.A., 1 10; 14' 18 2 I" 235; 243 245: 248; 250:253-4 P/Sppel, E., 23 141 165: 167-77: 180; 182-5; 181" 282-5 289: 292: 294' 302-3 Roberts, S., 146; 162' 202 2067; 214; 216-7: 263" 276:279 Shimizu, N., 257
Sillar, K.T., 208-14; 205; 208; 217-20 Steinbiichel-Rheinwall, N.V., 184; 281-2; 303-4 Wittmann, M., 184:281 Zakay, D., 143-6; 150; 152; 154; 157; 163-4; 192; 204
This Page Intentionally Left Blank
307
SUBJECT INDEX 5HT receptors, 213-14 Attention to time, 149; 152-54 Auditory fusion threshold, 167; 181 Auditory nerve, 33-5; 45-8 Auditory., visual and tactile temporal order thresholds (OT), 284 Automatic movements. 236 Being conscious, 134:165 179 Biosonar signals, 28" 31 Bisection point, 147" 193; 195 Category rating. 189-90:193-95 Causal relations, 120-23:125 Central mechanisms of motor control, 236 Central pattern generators, 205-6" 216; 236 Cerebellar cortex, 261-2' 271' 280 Cerebellar lesions, 257-9; 261" 263; 265; 269; 270-4 Cerebellar patient, 196" 259; 272 Circadian pacemakers. 99" 101-2" 105-6; 109-10:112 Circadian rh,,ethms. 96-100:1078; 111" 113-4: 144" 148 Cochlear hair cells, 34 Cochlear nucleus magnocellularis. 34" 36 Coding of temporal information. 27-8 Cognitive counter, 150; 154-5 Coincidence detection. 39; 42 Comparison of duration, 12
Conditioning of the rabbit nictitating membrane response (NMR). 260 Continuity of subjective time. 165 Convergence, 9; 29; 33; 37: 41" 62; 102; 285 Cortical hierarchy, 55 60 Cortical integration, 54" 67 Critical flicker fusion point (CFFP), 241 Cyclical conception of time, 118 Decision mechanism, 149 Delay lines, 3" 4' 9: 24; 36" 39" 42-4" 46"51' 74; 84; 86-7; 89 Descending raphespinal system. 213 Detection of phase differences. 40 Differential-phase-sensitive cells. 41-2 Diseases of time, 127 ageing and death, 137 madness, 127:137 weakness. 127:133" ! 3 7 Dopaminergic agents, 264 Duration discrimination, 91' 189; 196" 263" 265" 272-3 Duration estimation, 144: 150-2" 154-5 163 Duration scaling tasks, 189 Eigenmannia. 30; 40-42; 47:50 Electric organs. 28.45 Electrorcceptors. 28: 32; 41" 50 52 Electrosensor3, system, 28. 30. 32, 34 Empty time. 196 Endbulb of Held. 34 Entrained oscillator. 122
308
Subject Index
Episodic memory, 124; 125" 188 Estimation by analogical comparison, 192 Execution of movement, 237; 243; 292 Experienced duration, 188; 197 Feedforward networks, 87 Filled time interval, 196 Gabaergic neurones in CPG. 210 Globus Pallidus Lateralis, 248 Globus Pallidus Medialis. 248 Gymnarchus JAR, 41 Gynmotoid electrosensory lateral line lobes, 38 Horizon of simultaneity. 172:283 Infradian rhythms, 97; 103 Intensity discrimination, 259 272-3 Interaural phase differences, 30-1: 33-6; 42 Internal clock, 9: 13; 24: 74: 80: 148-9; 153" 157" 159; 162: 192; 198: 202; 204; 243; 246-8; 279 Interpositus nucleus, 261 Large-scale networks, 55-6, 60 Latent inhibition (LI) method, 151 LFP waveforms, 59 Linear time. 118: 120:123-5 Local field potential. 57:67 Magnitude estimation, 189; 192; 202 Mammalian bushv cells. 35 Medial superior olive, 25: 35: 44: 52 Memory stores. 149 Methamphetamine-induced impairment of temporal discrimation, 246
Mismatch negativity (MMN). 16, 179 Models, 1-2; 6; 9; 12; 33; 37. 69: 72-4: 79-80; 85-6; 88-9; 102; 104; 143-4; 148-50; 152-6: 158. 163; 180; 200; 237; 258; 262-3; 277 attention model, 150 attentional-gate model of duration estimation, 144, 154 digital models, 86 Jeffress model, 3-4; 9; 36: 3940: 42; 44 Nichelli model, 12 the oscillation model, 78 timing-with a timer models. 148 timing-without-a-timer models, 148 Molecular mechanisms of rh,~xhm generation. 102. 105 Monnyrid nELL, 38 Mormyrid nucleus of the electrosensory, lateral line lobe, 38 Motor and perceptual timing. 258 Motor programs, 55; 233: 237: 239: 244:251; 280; 290 Narrative structure, 123 Neural network models, 69:85 Neuronal oscillations in the 40 Hz, 173 Neurons of the nucleus laminaris. 44 Non-oscillatory timing sy,stem. 262 Nucleus laminaris. 35-6; 39; 436; 48; 51
Subject hldex Nucleus magnocellularis in birds, 38 Offset potential, 14 Onset Potential, 14 Order of succession, 2; 9 Original time, 129 Oscillators, 73; 75-6: 78-9; 82: 95; 113~ 121 Parkinson's disease. 235: 242: 245 Per protein rhylhm, 106 Perception of simultaneity, 1-2. 9,19 Perception of duration. 1" 2" 12" 188 Perceptual representation of stimulus duration. 276 Perceptual successiveness. 170 Perfommnce of repetitive movements, 235; 244 Photic entrainment, 108-9 Place coding of interaural phase difference, 42 Precision. 46:71-2: 88:191" ! 934: 198; 208:276 Prospective estimation. 145 putamen, 248 Raphe neurones, 214 Recovery. curves of the sensory nerve potentials. 7 Recurrent networks, 87:93 Reference memory. 148: 149: 155; 162: 195:264-5 Remembered duration, 188-9 Representation of temporal information, 258: 265: 275:277 Reproduction of temporal intervals, 174: 292:294
309
Retrospective estimation, 145 Rhythmic activities at 40 Hz Rhythmic movement, 206:217" 244 Sensorimotor associations, 26 I Silent period produced by the cortical stimulus, 225 Silent period produced by the peripheral nerve, 225 Simple reaction time task, 225 Single pacemaker neurons, 74 Social-psychological relations. 125 Somatosensor3' evoked potential. 7-8 South American gymnotifoml fish, 29 Spatial pattern analysis, 57 Spinal rhythm generation, 210 States of being consciuous (STOBCON). 179 Stationarity, 169 Supplementary motor area (SMA). 248 Sustained potential, 14-6" 22-3 Synchronization, 19; 52; 59: 63" 90: 97; 110: 177-8' 181-2: 185" 189; 191" 284; 290; 292" 302 Svnthesize time, 134 Temporal and spatial infomlation. 237 Temporal coding schemes, 84 Temporal discrimination, 1-2" 59; 12; 21" 90; 235; 241-2; 246" 248" 253" 263-4:285 temporal discrimination auditory, 241-2
Subject Index
310
temporal discrimination somaesthetic, 241-2 temporal discrimination visual, 241-2 temporal discrimination threshold, 5; 7-9: 235: 242: 248 Temporal hyperacuity, 29.37:39 Temporal infommtion. 18. 22. 27-8; 35; 37-8: 44: 46: 69. 73. 141; 149; 151-3; 156: 161: 197: 202; 257-8: 261: 263: 265: 266: 271: 275; 277; 279: 286. 304 Temporal jitter, 37:72 Temporal order threshold. 166-7: 169; 284 Temporal overestimation. 190 Temporal production. 12:189: 191 Temporal reproduction. 12: 174: 189: 193; 292; 294:296 Temporal underestimation. 190 Temporally discrete information processing, 284 Thalamo-cortical circuits The "binding" problem. 1 I The "go" signal, 229 The Barn owl, Tx~o alba. 31 The decision to move. 231:237 Threshold of discrimination of the duration, 12 Threshold of simultaneity. 2 Time and matter. 127 Time and measure, 128-9 Time and space, 127 Time as end or aim, 128 Time as mediation. 128 Time as origin, 128 .
Time bisection, 193-6 Time coding pathways, 38-9 Time of intention to move, 223-5" 227 Timing of voluntary movements. 223 Transient inhibition of the motor cortex, 229 Ultradian d~3r 102-3 Verbal estimation, 12, 145:174: 190; 193; 198:247 Weakly electric fish, 28 Weber ratio, 193 Working memory, 13, 64" 149: 177: 195: 199:201 Xenopus iaevis, 206