Cognition, Vol. 2, No. 4

1

Psychological reactivity to discrepant events: Support for the curvilinear hypothesis*

PHILIP R. ZELAZO J. ROY HOPKINS SANDRA JACOBSON JEROME KAGAN Harvard University Abstract 80 male and 60 female infants were observed under 5 a priori judged levels of discrepancy to assess whether sustained attention was linearly or curvilineariy related to degree of discrepancyfrom an experimentalstandard. Following habituation, 7&month-oldinfants were exposed to either the repetition of the standard, a minimal, first- or second-level moderate discrepancy, or a novel stimulus having no relation to the standard. The related stimuli, varying in elongation, were sphere, pear, club, and cylinder-shaped objects; the novelstimulus was a di_ffrent colored, tooth-like object. 80 infants observed the sphere as the standard and the cylinder as the second-level moderate discrepancy; 60 infants were exposed to the reverse order with the cylinder as the standard. The use of dcxerent stimuli at each discrepancy level controlled for spec$c stimulus effects. Habituation and recovery of responding were observed in an operant paradigm. Lever pressing, jixation, and vocalization increased most to the second-level moderate stimuli anddecreasedmost to thefamiliar andnovel objects: fretting was highest to the redundant stimuli and lowest to the moderate objects. There were no stimulus main eflects or The results support the hypothesis of a curvilinear relation between interactions. stimulus discrepancy and sustained attention, excitement, andpreference.

As attentional processes become more central in psychological theory the issue of environmental determinants of attention becomes a major theme for empirical inquiry. The traditional view of the relation between external event and recruitment of attention in infants and young children was phrased in absolute terms, and it was postulated that the attention-getting power of a specific event could be stated in terms of fixed * This research was supported by grant HD4299 from National Institute of Child Health and Human Development, USPHS; grant OCD-DB-174 from the Office of Child Development, and a grant from the Spencer Foundation, Chicago, Illinois. We thank C.

Super and J. Reiser for their suggestions and help during the early phases of this research. Requests for reprints should be sent to Philip R. Zelazo, Department of Psychology and Social Relations, Harvard University, Cambridge, Massachusetts 02138, U.S.A. Cognition 2(4),

pp. 385-393

386

Philip R. Zelazo, J. Roy Hopkins, Sandra Jacobson and Jerome Kagan

qualities-degree of complexity or amount of contour, for example (Dember and Earl, 1957). That view has been amplified by a series of experiments indicating that the familiarity of an event is an important determinant of sustained attention (Fantz, 1964; Fantz and Nevis, 1967; McCall and Kagan, 1967; Lewis, 1969; McCall and Melson 1969; Kagan, 1970, 1971). However, this relativistic position - called the discrepancy hypothesis - has at least two alternative forms. The first states that the relation between degree of discrepancy from the organism’s internal abstract representation of the event and sustained attention to the event itself is linear (Saayman, Ames and Moffett, 1964; McCall and Kagan, 1970; Jeffrey and Cohen, 1971). The more different the stimulus is from the organisms’ prior experience, the more attention it will recruit. An alternative view posits a curvilinear function with an optimal degree of discrepancy recruiting maximal attention and with very familiar and very discrepant (or novel) events recruiting relatively less interest (Hebb, 1949; McCall and Kagan, 1967 ; McCall and Melson, 1969 ; Kagan, 1970, 1971; Super, Kagan, Morrison, Haith and Weiffenbach, 1972). Cohen and his associates (Cohen, 1973; Cohen, Gelber and Lazar, 1971) have wisely differentiated between events that produce an initial orientation and events that elicit sustained attention following the initial response. Sustained attention has not been adequately examined, in part because assimilation occurs too rapidly with the simple stimuli that have typically been used. Paradoxically, it is desirable to use simple stimuli in order to establish a single dimension of change and to estimate the distances between stimuli (c$ Thomas, 1971). Thus, a procedure was developed that allowed for the gradual exposure of relatively simple, a priori ordered stimuli, in effect, amplifying the transformation period. The operant habituation-dishabituation procedure used by Siqueland (1969) and the suggestion that operant responding for a visual reward may function in a manner analogous to other attentional measures (Zelazo, 1972) lead to the development of an operant procedure that permitted brief stimulus exposures. Consistent results for fixation, vocalization and additional measures such as operant responding and fretting can also strengthen the argument that the curvilinear discrepancy function observed in earlier studies reflects an underlying cognitive process and not simply individual response characteristics. Moreover, a curvilinear operant pattern would imply that a visual reward may derive its reinforcing effectiveness from the child’s level of schemata growth for that stimulus. The purpose of this paper, therefore, is to determine if multiple indices of attention, including an operant measure, are consistent with the linear or curvilinear forms of the discrepancy hypothesis when the transformation period is prolonged and a priori ordered stimuli are used.

Psychological reactivity to discrepant events

387

Method Subjects and design One hundred and forty normal Caucasian infants, 80 boys and 60 girls, were recruited from advertisements in a local newspaper and from birth records. Eighty-six children were excluded because of mechanical problems, failure to reach criterion or excessive fretting. All children were 73 months old (plus or minus 2 days) and 84 % were from college educated families; 45% of all parents had some post-graduate training. There were five discrepancy groups: No change, minimal change, two different levels of moderate change and a major change (or novel) group. A series variable was formed by reversing the four groups from no change to minimal to moderate change in order to distinguish the effect of discrepancy from the effects of specific stimuli. Thus, discrepancy groups, series and sex were independent variables with the standard and transformation periods (i.e., the means for the last two minutes of the standard and the first two minutes of the discrepant stimuli) as a repeated measure in the analysis of variance. Stimuli The five stimuli were approximately equal in volume and made of papier-mache. Figure 1 illustrates the sphere, pear-shaped form, club-shaped form, cylinder and irregular tooth-shaped objects. The first four objects were all red with 3 inch black stripes; the tooth-shaped object was green with no stripes. For the 80 infants tested in Series 1, the sphere was the standard to which they were habituated and the discrepant object that was presented following habituation was either the pear-shaped form (minimal discrepancy), the club or cylinder (first- and second-level moderate discrepancies) or the tooth (novel or major discrepancy). For the 60 infants seen in Series 2 the order of the red and black objects was reversed to control for specific stimulus effects. Hence the cylinder was the standard to which they were habituated and the discrepant object was either the club (minimal discrepancy), the pear or sphere (first- and second-level moderate discrepancies) or the tooth (novelty). The control subjects who received no change in object continued to see the sphere in Series 1 or the cylinder in Series 2. The stimuli were prepared with the aid of eleven adult judges who were shown the five objects simultaneously in random formation. They were asked to rank the objects according to degree of elongation using the sphere as the reference and to estimate the intervals between adjacent objects. The four related objects were restricted to the

388

Philip R. Zelazo, J. Roy Hopkins, Sandra Jacobson and Jerome Kagan

Figure 1. The sphere, pear-shaped object, club-shaped object, cylinder and tooth-like object used as stimuli in the experiment.

range between 0 and 10; the novel object was allowed to vary from 0 and 100. Eight judges ordered the pear, club and cylinder as progressively more elongated from the sphere; the median scores were 3, 6, 10 and 0 respectively. The tooth-like object was judged as unrelated and assigned a median value of 20. Three of the judges placed the irregular object among the related objects because of its height. Procedure

Each baby was seated on his mother’s lap facing a 14-inch square window in front of a two-compartment box resembling a two-channel tachistoscope. Ambient illumination in the room was kept low and constant throughout testing. A bright yellow padded lever extended from beneath the window of the apparatus. A lever press lighted one compartment and provided a 2+ second view of the three dimensional object. All objects revolved on a turntable at 16 revolutions per minute in order to maximize the attention-eliciting properties of the stimuli. The standard stimulus was changed to the discrepant stimulus by closing a switch in the control room which caused the discrepancy chamber to light on the next lever press. The infant’s instrumental responses were automatically recorded on a polygraph.

Psychological reactivity to discrepant events

389

An observer watched the subject through a one-by-four inch window behind one of the chambers and recorded duration of lixation, vocalization and fretting on a button box linked to a polygraph. Responses were easily coded because the procedure permitted a clear, face-forward view of the infant. Fixation was scored only if the infant looked directly forward. All positive sounds were scored as vocalizations while ambiguous and negative sounds including crying were scored as frets. Interobserver reliabilities were calculated for five infants for each of these variables. Scores for two coders were compared for each 20-second period throughout each of these records. Observers were always blind with respect to experimental treatment and each other. Pearson product moment correlations for the four measures and the five pairs of infants were all above 0.90. Intercoder reliabilities for fixation and vocalization from a recent experiment in this laboratory were comparable : rs = 0.96 and 0.97 respectively (Zelazo, Kagan, and Hartmann, in press). Every 20 seconds the accumulated number of instrumental responses to occur accompanied by fixation of the stimulus during any portion of the 23 seconds of illumination (called reinforced instrumental responses) was tallied on-line from the automatic stimulus marking channel of a Grass polygraph. These calculations were immediately graphed and used to determine when each S had displayed criteria1 habituation and was ready for introduction of the discrepant stimulus. The following criteria for habituation were used: (a) Reinforced responses increased over base level, (b) reinforced responses following peak responding in two consecutive periods decreased below 50% of the mean of the previous two 20-second periods, (c) the standard period must have lasted at least three minutes from the child’s first reinforced response, and (d) the S must have had at least ten reinforced responses. When these criteria were reached the discrepant stimulus was illuminated. A final requirement was that the transformation period must have lasted a total of at least three minutes.

Results

The results revealed a main effect for the standard-transformation factor1 and a significant standard-transformation by group interaction for all four dependent variables. There were no series or sex main effects and only one third-order interaction for fixation time (group by sex by standard-transformation; F= 2.49, df = 4/120, pc.05) among the 60 statistical comparisons made for the four measures. The 1. The standard-transformation variable was used in the analyses rather than transformation scores alone because reliable differences oc-

curred among groups during the last minute of the standard.

390

Philip R. Zelazo, J. Roy Hopkins, Sandra Jacobson and Jerome Kagan

higher-order interaction revealed a clearer curvilinear pattern of recovery among groups for boys. The repeated finding of no difference for the series variable indicates the minor contribution of specific stimuli. Change scores were derived from the standard-transformation by groups interactions by subtracting the means for the standard from the means for the transformation for each discrepancy group and each dependent variable. Trend analyses were calculated for the change scores in the five discrepancy groups (combined over the nonsignificant series and sex variables) using coefficients computed for unequal intervals based on the adult judgments (Gaito, 1965). Table 1 shows the means for the standard-transformation by groups interaction and the derived mean change scores. Results for the trend analyses for each variable are also presented in Table 1. Table 1. Behavioral changes following presentations of discrepant objects Degree of discrepancy

Reinforced responses (minutes) Fixation time (seconds) Vocalizations (seconds) Fretting (seconds)

n

No change

Minimal

STd

STd

Trend analysis

First-level moderate

Secondlevel moderate

Novel

STd

STd

STd

Source

df

F

p<

linear 1
linear 1 2.13 .15 1 4.22 .05 29.45/24.00/ 33.31/24.00/ 33.00/29.91/ 31.68/34.85/ 41.50/38.51/ quadratic +3.18 -2.99 residual 2<1ns -5.46 -6.05 -3.09 error 135 -3.3813.371 -0.02

2.6613.471 +0.81

2.90/4.03/ +1.13

1.81/5.25/ +3.44

3.0412.271 -0.77

linear l
ns .Ol ns -

1.15/10.52/ +9.38

1.66/5.75/ +4.09

1.78/4.16/ f2.39

0.42/1.69/ f1.27

0.92/4.02/ +3.09

linear 1 4.04 quadratic 1 7.43 residual 2<1 error 135 -

.05 .Ol ns -

28

28

28

28

28

S represents the mean score for the last two minutes of the standard, T represents the mean for the first two minutes of the transformation and d indicates the difference between these values.

Reinforced responses increased most to the second-level moderately discrepant stimuli and decreased most to the novel and minimally discrepant objects. The trend analysis produced a nonsignificant quadratic component (F= 2.30, df = 1/I 35,~ = .16). The significant residual and the pattern for the distribution of means implies a cubic

Psychological reactivity to discrepant events

391

trend that is no doubt the result of increased irritability and restless activity in the control group. The no-change control subjects may have shown less of a decrement in instrumental responding than the minimal change group because of excessive restlessness created by the repetition of the same stimulus. The no-change infants showed over twice as much fretting as any other group (see Table l), and this distress was accompanied by increased activity and pressing of the lever. Figure 2 illustrates the change scores for the five groups and the four measures: Reinforced responses, fixation time, vocalization and fretting. Reinforced responses, fixation time and vocalization increased most to the second-level. moderate discrepancies; the quadratic Figure 2.

Change scores for mean differences in instrumentalresponses, fixation time, vocalization and fretting between the Iast two minutes of the standard and the$rst two minutes of the transformation for aN five discrepancy groups lo.5 r

VOCALIZATION

REINFORCED

-9.01

I HO CHANGE

I

I

MINIMAL FIRST DISCREPANCY LEVEL MODERATE

INSTRUMENTAL

I

SECOND LEVEL MODERATE

I

UNAELAIEO DISCREPANCY

392

Philip R. Zelazo, J. Roy Hopkins, Sandra Jacobson and Jerome Kagan

component was significant for fixation time and vocalization. Fretting also yielded a quadratic function and was inversely related to the other measures. Fretting, which implies boredom, was lowest to the second-level moderate stimuli, greatest to the no-change groups and revealed a tendency to increase to the novel stimulus.

Discussion On the whole, these results support the curvilinear form of the discrepancy function. The second-level moderate discrepancies (both cylinder and sphere) not only recruited longer epochs of sustained attention but also generated more positive excitement and instrumental behavior. Thus, these data support the suggestion made by Piaget (1952), Hebb (1949), Kagan (1970, 1971) McCall (1967, 1969) and others that the human infant has a cognitive preference for information moderately different from his existing knowledge. The use of brief stimulus exposures permitted a gradual unfolding of sustained attention to the discrepant stimuli. This important procedural change, in effect, slowed down the transformation period sufficiently to reveal the curvilinearity consistently across groups for all four measures. This combination of reactions also supports the suggestion (Zelazo, 1972) that the discrepancy function may predict reinforcer effectiveness for some stimuli. That is, moderate stimulus-schema discrepancy may form the basis for the intrinsic reward inherent in some visual and auditory stimuli. The reduction in operant responding to a repetitive visual stimulus, also reported by Siqueland (1969), may be a function of level of schemata growth for those stimuli. The results of this experiment indicate that degree of recovery of operant responding following habituation to a standard, like fixation time, is in part a function of degree of discrepancy from that standard. REFERENCES Cohen, L. B. (1973) A two process model of infant visual attention. Merrill-Palmer Q., 19, 157-180. -Gelber, E. R., and Lazar, M. A. (1971) Infant habituation and generalization to repeated visual stimulation. J. exp. child Psychol., 11, 379-389. Dember, W.N., and Ear&R. W. (1957)Analysis of exploratory, manipulative, and curiosity behavior. Psychol. Rev., 64,91-96. Fantz, R. L. (1964) Visualexperienceininfants: Decreased attention to familiar patterns relative to novel ones. Science, 146, 668670.

-

and Nevis, S. (1967) Pattern preferences and perceptual-cognitive development in early infancy. Merrill-Palmer Q., 13, 77108. Gaito, J. (1965) Unequal intervals and unequal n in trend analyses. Psychol. Bull., 63, 125-127. Hebb, D. 0. (1949) The organization of behavior. New York, Wiley. Jeffrey, W., and Cohen, L. (1971) Habituation in the human infant. In H. Reese (Ed.), Advances in child development and behavior, Vol. 6. New York, Academic Press. Pp. 63-97.

Psychological

Kagan, J. (1970) Attention and psychological change in the young child. Science, 170, 826-832. (1971) Change and conrinuity in infancy. New York, Wiley. Lewis, M. (1969) A developmental study of information processing within the first three years of life: Response decrement to a redundant signal. Mono. Sot. Res. child Devel., 34, (133). McCall, R. B., and Kagan, J. (1967) Stimulusschema discrepancy and attention in the infant. J. exper. child Psychol., 5, 381-390. and Kagan, J. (1970) Individual differences in the infant’s distribution of attention to stimulus discrepancy. Devel. Psychol., 2, 90-98. and Melson, W. H. (1969) Attention in infants as a function of magnitude of discrepancy and habituation rate. Psychon. Sci., 17, 317-319. Piaget, J. (1952) The origins of intelligence in children. New York, International Universities Press.

reactivity

to discrepant events

393

Saayman, G., Ames, E. W., and Moffet, A. (1964) Response to novelty as an indicator of visual discrimination in the human infant. J. exp. child Psycho& 1, 189-198. Siqueland, E. R. (1969) The development of instrumental-exploratory behavior during the first year of human life. Paper presented at the Meeting of the Society for Research in Child Development, Santa Monica, Calif. Super, C. M., Kagan, J., Morrison, F. J., Haith, M., and Weiffenbach, J. (1972) Discrepancy and attention in the fivemonth infant. Gen. Psychol. Mono., 85, 305-331. Thomas, H. (1971) Discrepancy hypotheses: Methodological and theoretical considerations. Psychol. Rev., 78, 249-259. Zelazo, P. R. (1972) Smiling and vocalizing: A cognitive emphasis. Merrill-Palmer Q., 18, 349-365. -Kagan, J., and Hartmann, R. (in press) Excitement and boredom as determinants of vocalization in infants. J. Gen. Psychol.

R&urn.& Les auteurs ont present6 ii 140 bCbCs des deux sexes, lgCs de 7 mois l/2, 5 niveaux, choisis a priori, de modification dun objet afin de voir si les reactions suscitees avaient un caractere lineaire ou curvilineaire. Apres les avoir habitues a dormer une reponse operante a un stimulus &talon, on presentait aux sujets soit un stimulus sans modification, soit un stimulus comportant une modification minime ou un de deux degres dune modification mod&&, soit un stimulus tout a fait nouveau. Les quatre stimulus parents Btaient parfois inverses afin de controler des effets dventuels du stimulus. On constate que les pressions sur un levier, les

fixations et les vocalisations presentent le degr6 maximal d’intensite pour les stimulus comportant une modification mod&e de degre 2 et diminuent d’intensite pour les objets familiets et les objets nouveaux. Les stimulus redondants donnent lieu a l’agitation la plus marquee et les objets comportant une modification moder&e de niveau 2 a l’agitation la moins marquee. On ne trouve pas d’effet du stimulus. Les resultats corroborent l’hypothese d’une relation curvilineaire entre les modifications du stimulus dune part, l’attention, I’excitation et la preference d’autre part.

2

Sensory-motor

intelligence

and semantic relations in early child grammar*

DEREK EDWARDS** Universityof Sussex

Abstract Based on case grammar and on the jindings of studies of early child language, a system of semantic clause-types is elaboratedfor the description of the relational meanings that are apparently expressed universally in the two-word speech of young children. These semantic relations are compared to the concepts invoked in Piaget’s descriptions of sensory-motor intelligence, in particular to the concepts of permanent objects and their spatial relations, to the dual concepts ofpersons as physical objects and as active beings, and to the role of persons as causers of changes in the locativity of objects. A close correspondence is found, and it is claimed that the nature of sensory-motor intelligence severely constrains the range of reIationa1 meanings expressed, including even the child’s notions of possessive relations between persons and objects, of attributes of objects and his use of apparently ‘experiential’ verbs.

1. Recent developments in the study of child language The ‘pivot grammars’ of children’s two-word utterances that were formulated in the early 1960s are now generally considered to be descriptively inadequate (Bloom 1970b; Brown, 1973). The basic notions that pivot words never occur alone, never in combination and always in fixed position, have all been questioned. However, quite apart from such doubts about the validity of pivot grammars on their own worddistributional criteria, we may consider them inadequate on other points; the impression they give is that when children talk they are not so much attempting to communicate messages and express meanings about the world as combining words from distributional classes in an effort to be syntactical. But syntax surely serves + I wish to acknowledge the debt I owe to Roger Goodwin for guidance, encouragement

and many useful discussions. ** Now at Loughborough University.

Cognition2(4), pp. 395-434

396

Derek Edwards

pragmatic and semantic purposes - it is the organization of words in sentences for the expression of meanings. A fuller description of child language would look at child syntax in terms of the meanings it serves to express. Bloom (1970a) has shown that if meanings (derived from the situational contexts of utterances) are taken into account, the simple ‘pivot-open’ dichotomy is too simple and too superficial. Thus the noun+noun combinations in Bloom’s data are described as expressing a number of different relational meanings: Conjunction [boot umbrella), attribution (party hat), possession (Daddy hat), subject-locative (sweater chair) and subject-object (Mommy book). One combination was assigned two different meanings because it occurred in two quite different situational contexts: (1) Mommy sock Context: M putting K’s sock on K (subject-object) (2) Mommy sock Context: K picking up M’s sock (possessor-possessed) (K = the child Kathryn, M = her mother) It must be stressed that the ‘meanings’ involved here are not referential meanings that hold between objects and the words which label them but rather relational meanings that hold between words in sentences, the province of ‘deep syntax’, paralleling the roles and relations between the persons and things in the situation or event described. Recent attempts to classify children’s early utterances in terms of their relational meanings (Schlesinger, 1971a, 1971b; Brown, 1970; Slobin, 1970, 1972)have tended to classify two-word utterances as expressing semantic relations between pairs of elements such as Agent + Object or Possessor + Possessed and to list a few examples of each two-term relation. There has been no attempt to integrate these relations within a general linguistic system, nor within a cognitive theory which would seek to link the semantic concepts to the kinds of cognitions the child has about the world described. This paper is directed towards these two tasks. Influenced by the work of Bloom and of Schlesinger, Brown (1970) has drawn up the classification reproduced as Table 1. It is a list of the relational meanings expressed in two-word utterances, omitting the ‘Modality’ categories of Negative, Imperative and Interrogative (see Fillmore, 1968, for the Proposition-Modality distinction and the compatible notions in Halliday, 1967-1968, 1970). Brown claims that his categories describe 75 % of the many and various utterances he has examined. Brown’s classification is, however, in three ways still oriented to the surface structure of the utterances rather than to their meanings: (1) The ‘Relations’ are crossclassified as ordered pairs of distributional classes from adult grammar - noun, verb and adjective. And the ‘Operations of Reference’ are basically types of pivot structures semantically classified, since on a purely semantic basis they involve some ‘relational meanings (e.g., ‘that hot’ would involve attribution). (2) There is no indication of any underlying semantic elements that may be unexpressed in an utterance, but the

Sensory-motor intelligence and semantic relations in early child grammar

397

postulation of which may be essential to Brown’s interpretation of what elements are expressed. (3) Related to (2) is the fact that there is no account taken of how the separate Table (1)

two-term

1.

relations

are semantically

related to each other.

Brown’s (11970) semantic classzjkation of two-word utterances

Operations of Reference a) Nominations ‘that’ (or ‘it’ or ‘there’) + book, cat, clown, hot, big, etc. b) Notice ‘hi’ + Mommy, cat, belt, etc. c) Recurrence ‘more’ (or ‘nother’) + milk, cereal, nut, read, swing, green, etc. d) Nonexistence ‘allgone’ (or ‘no-more’) x rattle, dog, juice, green, etc.

(2) Relations a) b) c) d) e) f) g)

Attributive Possessive Locative Locative Agent-Action Agent-Object Action-Object

Adj + N N+N N+N N+V N+V N+N

V+N

(big train, red book, etc.) (Adam checker, Mommy lunch, etc.) (sweater chair, book table, etc.) (walk street, go store, etc.) (Adam put, Eve read, etc.) (Mommy sock, Mommy lunch, etc.) (put book, hit ball, etc.)

For example, what does it mean to classify an utterance as expressing the relation Agent + Object? Any such description if at all appropriate must involve somewhere the notion of an event or action. Agents are defined as performers of actions and instigators of events. If a child says ‘Mommy sock’ and from the context we interpret the semantic relation as Agent + Object, then we are thereby saying that the child was not talking merely about his mother and a sock, but about his mother doing something to a sock. Similarly, what is the relation between, say, the utterances ‘Mommy put’ ‘put sock’ and ‘Mommy sock’ when these occur, as is possible, in identical situational contexts? And does not the action ‘put’ involve a further ‘locative’ element as part of the underlying meaning? Semantic interpretation demands for each of these two-term utterances the postulation of underlying, but unexpressed, semantic-relational elements that are conceptually required by the interpretation and which may actually be realised in other contextually related utterances (cJ Braine’s ‘replacement sequences’, 1971). Slobin (1970, 1972) has presented strikingly similar lists to Brown’s of the semantic relations common to child language across a variety of very different cultures, stressing the universality of these relations in early child language and postulating their basis inp utative ‘universals of structured experience’ (1970, p. 175). More specifically, Brown (1970, p. 223) suggests that ‘The meanings expressed . . . seem to be extensions of the kind of intelligence that has been called “sensory-motor” by the great developof the first sentences are an mental psychologist Jean Piaget . . . If the meanings extension of sensory-motor intelligence then they are probably universal in mankind’.

398

Derek Edwards

The concern of this paper is with the interface between universal semantic relations and universal sensory-motor cognitions, which meet in the content and process of language acquisition at the beginning of syntax in two-word utterances. In the remainder of the paper a semantic-relational framework is elaborated which seeks to be consistent with child language data, with current linguistic theory and with what Piagetians tell us is the structure of the child’s knowledge of the world at the end of the sensory-motor period. The acquisition of meanings is characterised as a process in which the meanings of input speech are assimilated to the child’s perception of the world and events referred to. Perhaps here it should be made clear that sensory-motor intelligence is not being put forward as an explanation of language acquisition. Speech production and the organization and use of a symbol system involve much more, or else chimpanzees would talk. What the cognitive structures provide is the structures of knowledge and meaning via which language refers to the world and makes sense; it does not provide for the realization of those meanings in speech.

2. The system of semantic roles and relations Slobin’s and Brown’s comments on the universality of particular semantic concepts in child language are excitingly reflected in the fact that some theoretical linguists are also trying to arrive at universal categories in adult language. For example, Fillmore (1968) suggests that ‘the case notions comprise a set of universal, presumably innate, concepts which identify certain types of judgments human beings are capable of making about the events that are going on around them, judgments about such matters as who did it, who it happened to and what got changed’ (p. 24). If we can show that sensory-motor intelligence involves parallel nonlinguistic organizations we may ultimately be investigating some universal aspects of human thinking. The semantic relations we shall be concerned with are those that are involved in expressing basic propositional meanings within clauses - with what Halliday (1970) calls the ‘Ideational’ component of language. We shall not be concerned with Modality (questions, commands, etc. ; Halliday’s ‘Interpersonal function’). We are concerned here with the semantic categories and relations with which language describes situations and events in the world and particularly with those categories that, according to Brown’s and Slobin’s lists, occur universally in early child language.

2.1

Case relations

The semantic relations used by Schlesinger, utterances are derived mainly from Fillmore’s

Brown and Slobin to describe early (1968) ‘case grammar’. In Fillmore’s

Sensory-motor intelligence and semantic relations in early child grammar

399

later writing (1970, 1971a, 1971b) and in the independent work of other linguists whose descriptions involve semantic relations similar to cases (Chafe, 1970; Anderson, 1968,197l; Halliday, 1967-1968, 1970) we find a variety of overlapping categories and terminology that has to be standardized before we can discuss its relevance to children’s cognition and language. The list of cases presented below is inevitably provisional and is likely to be modified and supplemented by further developments in linguistic theory. Anderson (1971, p. 10) defines cases as ‘the grammatical relations contracted by nouns which express the nature of their “participation” in the “process” or “state” represented in the sentence’. Cases represent generally the types of roles played by persons and things (the archetypical notional ‘nouns’ - Lyons, 1966) in the relations, actions and states identified by the ‘verb’. Most surface structure verbs and adjectives and the locative prepositions, are considered here to belong to a class of ‘deep verbs’, these being the semantic elements that describe the kind of state, action or relation that holds between the various nominal ‘case’ elements. The ways in which cases combine in ‘deep clauses’ around a central verbal-relational element correspond to the definition of the various verb-types and clause-types discussed in Section 2.2. Agent. Fillmore (1968, p, 24) defines ‘Agentive’ (later called ‘Agent’) as ‘the case of the typically animate perceived instigator of the action identified by the verb’. It is the case of ‘John’ in la-lc. (1)

a. John opened the door b. The door was opened by John c. John kicked the door

‘Agent’ is similar to what Anderson terms ‘Ergative’, Halliday ‘Causer’ and Chafe also ‘Agent’. Following Halliday (1968), Anderson (1968) and Lyons (1968), the case Agent is discussed below in close relation to the notions of transitivity and causativity in verbs. Instrument. Chafe (1970, p. 152) defines Instrument as ‘something which the Agent uses’, while Fillmore’s definition includes Chafe’s and more besides: ‘The case of the inanimate force or object causally involved in the action or state identified by the verb’ (1968, p. 24). Thus Fillmore’s definition includes not only ‘key’ in 2a-2c but also ‘wind’ and ‘ship’ in 2d and 2e. (These sentences are examples used by Fillmore and Chafe). (2)

a. b. c. d. e.

The key opened the door John opened the door with the!key John used the key to open the door The wind opened the door The ship destroyed the pier

400

Derek Edwards

In Chafe’s system some ‘noun elements’ such as ‘ship’ and ‘wind’ are considered to have the semantic feature ‘potent’, which enables them to act as Agents in their own right, as the instigators of actions: ‘The ability of a noun to occur as Agent depends on its semantic specification as a thing which has the power to do something, a thing which has a force of its own, which is self-motivated’ (c$ Halliday, 1970, pp. 148-149 on ‘natural force’). The crux of the issue is not so much animacy versus inanimacy as whether the thing (or noun) in question is understood to be the causal source of the action. Thus, sentence 3 is ambiguous between an agentive and an instrumental interpretation of John’s role in the action: He may have slipped and fallen or been thrown against the wall (instrumental), or else he may have deliberately struck the wall (agentive). (3)

John hit the wall

In the agentive interpretation John plays both roles: Both causer-initiator of the event as well as the physical thing that actually makes contact with the wall. Object. Fillmore (1968, p. 25) defines ‘Objective’ (later ‘Object’) as follows: ‘The semantically most neutral case, the case of anything representable by a noun whose role in the action or state identified by the verb is identified by the semantic interpretation of the verb itself; conceivably the concept should be limited to things which are afected by the action or state identified by the verb’. This case corresponds more or less to Chafe’s ‘patient’, to Anderson’s ‘nominative’ and to Halliday’s (1968) ‘affected’. It is the role of the entity said to be in a particular state or condition, or to undergo a change of state or position, or to be affected by an Agent and/or Instrument. The Object is italicized in sentences 4a-4g. (4)

a. b. c. d. e. f. g.

The table is red The glass broke John broke the glass (with a hammer) The glass was broken by John The rock broke the window The chair is in the bedroom John moved the chair into the kitchen

Experiencer. This is the case of the animate being who is said to be having some kind of mental experience. The type of ‘experience’ involved is identified by the experiential verb. For example, ‘John’ is Experiencer of a ‘perception in 5a-5c (5)

a. John heard a noise b. John saw the book c. Bill showed John the book

Sensory-motor intelligence and semantic relations in early child grammar

of a ‘cognition’

(6)

a. b. c. d.

in 6a-6d

John knew the answer John believed that he would win We persuaded John that he would win It was apparent to John that he would win

and of a ‘reaction’ (7)

a. b. c. d.

401

in 7a-7d

John liked the play The play pleased John John was sad (about missing the opera) Mary interested John in opera

‘John’ is Halliday’s (1967) ‘processer’ and the experiential verb types are three of Halliday’s types of ‘mental process’ verbs. Sentences 6b-6d are Fillmore’s (1968) which is examples of the case then called ‘Dative’, renamed (197 1a) ‘Experience?, also Chafe’s term for it. ‘Experience? is preferred here because the tetm ‘Dative’ is likely to be confused with what is called in this paper ‘Beneficiary’ (see below). Experiencer is a special version of Object; it is the role of the ‘affected’ entity, in or undergoing change in a particular state, when that state or process is ‘experiential’ (cf: Halliday, 1968, p. 195). Phenomenon. This is Halliday’s (1967) term for the thing or fact which is known, perceived or reacted to by the Experiencer. It is exemplified in 5a-7d by ‘the answer’, ‘the book’, ‘that he would win’, etc. Chafe (1970, p. 144) classifies the Phenomenon as ‘patient’, that is, as Object. However, we prefer to have Experiencer as a kind of Object, with Phenomenon a special case introduced by experiential verbs. The Phenomenon is not said to be in a state, nor to undergo change, nor to be ‘affected’ in any sense at all ; it is the Experiencer which undergoes the process identified by the verb.

Location. This may be defined as the spatial position or orientation of the Object. The term is Chafe’s, and it resembles Fillmore’s case ‘Locative’, except that the latter is defined as the location of the state or action rather than of the Object. Location is the case of the items italicized in 8a-Sc, the other NPs being Object. (8)

a. Henry is in Liverpool b. The book is on the table c. The envelope contains five pounds

The term ‘Location’ will be restricted to those spatial positions with respect to which the Object (or clause-event embedded to the case Object) is static and unchanging. Source and Goal. Where a locative verb specifies movement of the Object towards,

402

Derek Edwards

into, out of, away from, etc., the named position, then that position is termed Source or Goal as appropriate. Fillmore (1971a) defines Source as ‘the place from which something moves’, and Goal as ‘the place to which something moves’ (the ‘something’ being the Object). (9)

a. b. c. d.

The book fell off the table John left London Henry threw the book onto the table John walked home

(Source) (Source) (Goal) (Goal)

In 9b and 9d ‘John’ is Agent as well as Object. In 9c, ‘Henry’ is Agent and ‘the book’ Object; Henry is not himself described as changing position onto the table. Possessor. This is the case of the typically human possessor of the Object, where the possessive relation is static. It is the possessive version of the locative case Location and is italicized in sentences lOa-1Oc. (10)

a. It is John’s book b. John owns the book c. The book belongs to John

Bene$ciary. Where the verb specifies that the Object changes possession, possessor is called Beneficiary (corresponding to the locative case Goal): (11)

the new

a. Henry got a bicycle for Christmas b. John gave Henry a bicycle c. Henry sold the bicycle to George

The similarities and differences between Possessor, Beneficiary and Goal will be examined below when we look at possessive and locative verbs. Note that the Beneficiary we are concerned with here is the possessive Beneficiary, the ‘beneficiary of an object’ (Halliday, 1970). There is another case sometimes called ‘beneficiary’ or ‘benefactive’ that is not part of the grammar of possession and which does not occur in early child grammar; this is the ‘beneficiary of a service’: (12)

a. John cut the grass for Mary b. John sold the car for Mary

Comparison of 1 lc and 12b shows the essential difference. Result. Fillmore (1971a) defines this as ‘the entity that comes into existence result of the action’. Examples are: (13)

a. Millers grind flour b. They built a house near the stream c. God created the world

as a

Sensory-motor intelligence and semantic relations in early child grammar

403

The verbs in 13a-13c are what Lyons (1968, p. 439) calls ‘existential causative’; they specify an action that brings an object into existence. ‘Result’ is another sub-type of Object, this time being the thing ‘effected’ by the action identified by the verb. Fillmore’s (1968) case ‘factitive’ and Chafe’s (1970) ‘complement’ both include what we call here ‘Result’, but they also include a rather different notion which Halliday (1967) terms ‘range’: (14)

a. George played tennis b. Henry did the housework c. Mary danced the Charleston

These ‘range’ elements are not objects brought into existence and continuing to exist as a result of the action. Rather, they are specifications of the nature of the action itself. ‘Range’ is therefore notionally verbal rather than nominal and is not a true ‘case’ in the sense of being a participant role in the action. Halliday (1967) calls it a ‘pseudo-participant’. The case Result does not figure in Brown’s and Slobin’s lists of universal case relations in early child language, and will not figure in the rest of this paper. 2.2

Clause types

The definitions of the cases make it plain that they are elements in relations; cases are defined by the manner of their participation in a state of affairs or in a process or interaction of some kind. Thus the definition of cases involves their potentialities for combining with each other, and these potentialities are made explicit when we subclassify verbs according to which cases they select and formulate deep semantic clause-types which consist of a single verb with its associated cases. Although the central role of the verb in the semantic relations of a clause is implicit in the work of several linguists (Anderson, 1971; Halliday, 1967-1968) it is Chafe (1970) who has argued the matter most explicitly. Using ‘verb’ and ‘noun’ in their notational senses, Chafe states that ‘semantic structure [is] built around a central verb, which [is] then accompanied by nouns related to it in several ways’ (p. 11). It is the ‘central role of the verb’ that determines which cases are selected. In 15a and 15b the semantic relation of noun to verb is different (Agent in 15a, Object in 15b), but it is the verbs which determine these relations. The nouns themselves are of course identical. (15)

a. John fought b. John died

While we can directly classify the verb own as possessive and red as attributive, we cannot do so for nouns; John may play just about any case role at all, depending on the manner of his participation in the verb-process.

404

Derek Edwards

The basic verb-types 1. States

2. Processes

3. Actions

4. Action-processes

that Chafe describes are these (1790, p. 95) : The wood is dry The dish is broken The elephant is dead The wood dried The dish broke The elephant died Michael ran Harriet sang The tiger pounced Michael dried the wood Harriet broke the dish The tiger killed the elephant

‘States’ are verbs which attribute ‘a certain state or condition’ to the object (or ‘patient noun’). ‘Processes’ are changing states, describing the object as undergoing a change of state. ‘Actions’, however, are quite different: ‘They have nothing to do with either a state or a change of state; instead they express an activity or action, something which someone does’ (p. 100). And the accompanying case is Agent. ‘Action-processes’ are simply a combination of actions and processes; they select both Agent and Object and describe the Agent’s role in bringing about a change of state in the Object. Action, process and state are ‘selectional features’ by which verbs are sub-classified; other such features are ‘benefactive’ and ‘locative’. One thing that these features select is a particular set of cases, so that each verb-type has its own case structure (c$ Fillmore, passim). Although Chafe’s basic approach is exciting, there are some points at which his treatment is unsatisfactory. The main problem arises in his treatment of ‘action’ verbs. Despite defining ‘patient’ (i.e., Object) as something in or undergoing change in ‘a certain state or condition’, Chafe goes on to define the deep objects of verbs like hit and kick as ‘patients’ and the verbs themselves as ‘action-processes’ even though they imply no necessary change of state in the object. Hit and kick are types of surface contact actions and ought not to be described as ‘processes’ (cJ Fillmore’s treatment, 1970). Also, it is not clear that Chafe’s so-called ‘action-processes’ actually specify any ‘action’ performed by the Agent at all; it might be argued that both the transitive and the intransitive use of break and dry actually describe only the process undergone by the Object, with the Agent’s activity not described at all - the Agent’s role is merely to bring about the named Object-process. Another inadequacy in Chafe’s system is that he simply lists verb features like ‘action’, ‘process’, ‘benefactive’ and ‘locative’ without giving credit to the fact that

Sensory-motor intelligence and semantic relations in early child grammar

405

while ‘benefactive’ is specific to possessive relations, introducing the possessive case Beneficiary, ‘action’, ‘process’ and ‘state’ are features of very wide use and importance, involving the pervasive cases Agent and Object. Indeed, Chafe recognizes the need for further work to ‘clarify the apparent priority of the patient and agent relations’ (p, 347). Perhaps the importance of these relations is more clearly expressed in the work of Lyons (1968), Anderson (1968, 1971) and Halliday (1967-68), who discuss roles similar to Agent and Object within the context of transitivity and causativity in verbs and clauses (see Ingram, 1971, for some relevant discussion of early child language). To such notions we now turn. Table 2.

Static, dynamic and causative sentences (after Lyons, 1968) Possessive

static dynamic causative

16a John has a book 16b John received a book 16c Bill gave John a book

Locative 17a The book is on the floor 17b The book fell on the floor 17~ Bill put the book on the floor

Table 2 is adapted from Lyons (1968). In each of the possessive sentences (16a-16c) there is a constant relation of possession between the book (Object) and John (Possessor/Beneficiary). Similarly, each of the locative sentences (17a-17~) concerns a locative relation that obtains between the book (Object) and the floor (Location/ Goal). In the ‘static’ clauses the possessive or locative relation is unchanging. In the ‘dynamic’ clauses these relations undergo change, this change being specified by the verb. In the ‘causative’ clauses there is again a changing relation between the same two participants, but now the change is described as caused by an Agent - ‘Bill’. The three conditions in which the locative and possessive relations are presented - static, dynamic and causative - have elements in common with three of Chafe’s ‘selectional features’: States, processes and the compound ‘action-processes’. Whereas Chafe’s states and processes were attributes of single Objects, Table 2 happens to show relations between an Object and another entity (Possessor/Goal/etc.). However, a more important difference for the present discussion is that instead of listing ‘possessive’ and ‘locative’ along with ‘state’ and ‘process’ as equal-status features of verbs, Table 2 represents the system as involving two orthogonal dimensions, which are (1) the type of state or relation (locative, possessive, attributive) and (2) whether the given state or relation is an unchanging state of affairs, one that involves change or one that is caused to change or happen by an Agent and/or Instrument. The introduction of an Agent to render a change or movement ‘causative’ is a common paradigm in many languages for making intransitive verbs transitive (Lyons, 1968; Anderson, 1968; Halliday, 1967-68). Halliday and Anderson argue that the notion of causativity should also be extended to transitive verbs that have no intransitive counterparts.

406

Derek Edwards

Table 3.

Some basic clause types

Notation: (x) = x is optional (x)(y) = either x or y or both

x y= [I

either x or y but not both 1 DIRECTED

STATIC ACTION

DYNAMIC

CAUSATIVE

Ir+o+

lCAg)+I+O+

V.DIR.ACT.

V.DIR.ACT.

(the

ACTIONAL

I

2

MOVEMENT

3

LOCATIVE

4

POSSESSIVE

(RELATIONAL

STATDIE. 5 EXPERIENTIAL

brick hit the (John hit Bill floor) with a brick) 0 + Vdyn.Mov (Ag)(I) + 0 + (the wheel turned) Vcaus.Mov. (Bill turned the wheel with his elbow) 0 + Vstat.roc + 0 + Vdyn.Loc + (Ag)(I) + 0 + L Vcaus.Loc + (the book is on S (theGbook fell off the chair) [I the table/on (Jo: took the the floor) book off the shelf/put it on the shelf) 0 + Vstat.Poss + 0 + Vdynross + (Ag)(I) + 0 + B P vcaus.Poss + B (the book belongs (Bill got a to Bill) (John gave Bill present) a present) E + Vstatzxr + E + Vdyn.rxr + (A&I) + E + Vcaus.ExP + (Ph) (Ph)

s [I

(John likes opera)

1 6

ATTRIBUTIVE

(John got interested

in

(Ph) (Mary interested John in opera)

opera) 0 + Vstat.ATm

0 C Vdyn.Am

(the jar is broken)

(the jar broke)

(Ag)(I) + 0 + VCSLUS.ATTR

(Bill broke the jar with his

I

I

0 = Object I = Instrument L = Location Cases : Ag = Agent S = Source P = Possessor B = Beneficiary E = Experiencer

]

fist)

G = Goal Ph = Phenomenon

Table 3 shows a way of organizing verbs and cases into a set of deep semantic clause-types. The clause-types are divided into two major groups, Actional and Stative, and they have the same names as the verb-types which are their central relational element. Each clause contains a single verb and the cases it selects, and all verbs are either Actional or Stative.

Sensory-motor intelligence and semantic relations in early child grammar

407

(A) Actional verbs are distinguished by their lack of inherent specification of any necessary change of state or of spatial position of the Object affected. (a) DIRECTEDACTION verbs describe a type of physical contact between Instrument and Object which may be instigated by an Agent, who may be using his own body as the Instrument. Examples are hit, stroke, kick, punch. The latter two, besides being Instrument-specific (as regards the body-part used), are probably always ‘causative’, since they refer to actions typically performed intentionally by Agents. (b) MOVEMENTverbs describe the type of motion or activity gone through by the Object and include roll, turn, walk and spin. Walk, march and run are causative MOVEMENTverbs (V.caus. MOV in Table 3) that typically have reflexive Agents, where the person running or walking is both Agent (the instigator) and Object (the thing actually undergoing the movement). Although MOVEMENTverbs in themselves merely describe the kind of activity undergone by the Object, they are often combined with LOCATIVEverbs (see 2.2.1 below) in what may be called ‘compounded clauses’ which describe both the motion undergone by the Object and also a change in its spatial position. Thus, ‘Bill walked to work’ is a sentence in which two deep clauses are compounded, a ‘causative MOVEMENT’and a ‘causative LOCATIVE’clause, which share the same reflexive Agent + Object (‘Bill’). The Goal element ‘work’ is introduced by the LOCATIVEclause, centered on the verb ‘to’. (B) Stative verbs are those which describe a state of affairs or a changing state of affairs where the establishment of a new state is implied. For example, although the verb buy necessarily involves the role Agent and has no ‘static’ variant, it nevertheless specifies a change in a state of affairs, this being the possessive relation that holds between Beneficiary and Object. Stative verbs are all classified according to the kind of state of affairs they describe, or in which they describe a change. Thus the adult English surface verbs buy, own, acquire and give are all POSSESSIVE;(be) on, enter and evacuate are all LOCATIVE;~~~, break and shattered are all ATTRIBUTIVE.Each of these general types of state or relation has a distinctive ‘case frame’, or selection of cases, associated with it. The ‘Relational’ clauses are centered on relations between two elements, the Object (or Experiencer in EXPERIENTIALclauses) and the special relational case introduced by the particular verb-type: Possessor/Beneficiary in POSSESSIVErelations, Location/Source/Goal in LOCATIVE relations and Phenomenon in EXPERIENTIAL clauses. The ‘Characterizing’ clauses predicate a single characteristic to the Object, this being the state itself, described by the verb. ATTRIBUTIVEverbs include most English adjectives and also many surface verbs (break, $atten, shrink, etc.). EXPERIENTIAL verbs may be either Characterizing (John is sad) or both Characterizing and Relational (John saw the snake; John is sad about X), depending on the semantic presence or

408

Derek Edwards

absence of the case Phenomenon, which is particularly likely to be absent (semantically, not merely deleted in surface structure) in the ‘reactional’ types (see section 2.1 above). Stative verbs, especially ATTRIBUTIVE and LOCATIVEtypes, include among them certain verbs which express only the state or relation itself and not the ‘static-dynamic causative’ dimension of the clause. Examples are the ATTRIBUTIVES ugly, smooth, yellow, heavy and the LOCATIVE relations in, on, under, across. In the word classes of surface structure English these are ‘adjectives’ and ‘prepositions’ though in other languages they may appear in surface structure as tully inflected verbs (Chafe, 1970). Thus in English these words are not able to carry verbal inflexions such as for tense, number and aspect. They are therefore accompanied in surface structures by ‘dummy verbs’ (Lyons, 1968) that carry the inflexional categories for them. Typical dummies are be, do, have, go and become. All clauses have deep verbs and are built around them, even though the ‘verbal’ word in the surface structure may be merely a carrier of tense and aspect, with the actual state or relation itself being expressed in an adjective, a preposition, or even unexpressed except in the inflection ‘s which is suffixed to the Possessor in a static POSSESSIVEclause. 2.2.1 LOCATIVES LOCATIVEclauses describe an Object’s static or changing spatial location. Contrary to Anderson’s (1971, p. 81) treatment of locative prepositions as ‘superficial case markers’ which happen to be ‘not semantically equivalent’, we have agreed with Chafe’s (1970, p. 159) treatment of them as deep verbs which express the actual locative relation between Object and p1ace.l In static LOCATIVE clauses with prepositions, the surface verb ‘to be’ invariably carries the ‘static’ aspect. Figures 1-3 show how the semantic relations of some LOCATIVE clauses are described in terms of the verb-types and cases of Table 3, with cases shown as dependent on verbs. (Tense is lexicalised merely for convenience.) 2.2.2

POSSESSIVES POSSESSIVE clauses and verbs have already been discussed passim and are sufficiently similar to LOCATIVES that we need not dwell on how they fit into Table 3. Some differences between POSSESSIVES and LOCATIVES are considered in 2.2.3. However, we must f&t distinguish three types of possession that are relevant to the description of both child and adult language.

1. It is interesting to note that my daughter Helen and also another subject ‘Alice’ (unpublished data), spontaneously inflected both our and up as surface verbs; e.g., ‘lorry outing

a window’ (is out of); ‘Helen upping steps’ (is climbing up). Some more data from Helen is included below.

Sensory-motor intelligence and semantic relations in early child grammar

Figure 1. ‘The book is on the table’ (static

LOCATIVE)

Clause I

Vstat. LOC

oyyt

/ 1\ / \ I,----_ the book

Figure 2.

Location /\ ‘1 /’ -/--_-_ \ the table

(is) on

'George ftiIl of the table’ (dynamic LOCATIVE)

Clause I

Vdyn. LOC Source Ifi\ / L --‘1 the table

fell + off

‘George entered the room’ (causative LOCATIVE)

Figure 3.

Clause

I

Vcaus. LOC

.

\

\\

\ ./ ’

Gedt-ge

/

/

Goal /“\

/

/ I

entered

\

L’____L.

the room

409

410

Derek Edwards

Possession may be alienable or inalienable. ‘John’s face’, ‘John’s head’ and other body-part expressions involve inalienable possession (c$ Fillmore, 1968, p.61 ff.). Alienable possession is of two kinds, ‘transitory’ and ‘permanent’ (cf. Chafe, 1970, p. 148; Anderson, 1971, p. 113). Permanent possession implies ownership; the Possessor has proprietary rights over, or privileged access to, the Object possessed, as in ‘It is John’s car’. Transitory possession concerns things that are temporarily ‘in somebody’s possession’ - usually on his person or in his hands. (18)

a. b. c. d.

The thief was caught in possession Is that my book you have there? George lent his car to Fred John handed me his watch

of Henry’s waZlet

The italicized phrases in 18a-18d are relations of permanent possession. The relationship between the Object possessed and the other person in each sentence (the thief, you, Fred, me) is one of transitory possession. Without this distinction, to call all the relations in 18a-l8d ‘possessive’ would be utterly confusing. A person has to be able to ‘give’ an object to someone else without relinquishing proprietary rights over it, and to ‘have’ or ‘be in possession of’ an object that is actually the lawful property of somebody else. We may note here that ‘inalienable’ possession is probably always ‘static’ ; indeed, this is what ‘inalienable’ means. Although it may be possible for a person’s limbs to be cut off and owned by somebody else, such ownership would then be alienable‘permanent’. (The solution to the semantic puzzle posed by transplant surgery is not yet clear.) The two types of alienable possession more freely become ‘dynamic’ and ‘causative’. Sell and donate are causative permanent possessive verbs. Give is ambiguous : (19) George gave Henry a book In sentence 19, George may merely have handed Henry a book which actually belongs to either of them, to some third party, or even to nobody at all; otherwise he may be conferring rights of ownership of the book on Henry, in which event neither person needs ever to lay hands on the book. The former interpretation is transitory, the latter permanent possession. 2.2.3 Transitory POSSESSIVES and LOCATIVES Given that transitory POSSESSIVE relations involve no proprietary rights of Possessor or Beneficiary over the Object, no actual ‘ownership’ in the legalistic sense, why do we call these relations POSSESSIVE rather than ACTIONAL or LOCATIVE? An obvious reason is that transitory possession involves the general vocabulary and morphology of possession: The suffix ‘s’ (e.g., ‘Henry’s seat’ in the cinema) and the words give,

Sensory-motor intelligence and semantic relations in early child grammar

411

o&t, have, got,possession and so on. But let us consider a rationale based on differences in semantic interpretation which are also reflected in the surface structure of English. It is this: The Possessor or Beneficiary in transitory POSSESSIVE clauses is understood to be ‘animately involved’ in holding on to, or in actively receiving, the Object. On the other hand, in LOCATIVE clauses the Location or Goal may be animate or inanimate but is not understood to be ‘animately involved’ in the same way. Consider sentences 20a-20f which are compounded Actional + Stative clauses with the DIRECTED ACTION throw constant throughout and the Stative relation varying between LOCATIVE and POSSESSIVE. Clause-type (20)

a. b. c. d. e. f.

John John John John John John

threw threw threw threw threw threw

the ball to BiII the ball at Bill Bill the ball the paint to Bill the paint onto Bill Bill the paint

(POSSESSIVE) LOCATIVE POSSESSIVE (POSSESSIVE) LOCATIVE POSSESSIVE

Case of ‘Bill’ (Beneficiary) Goal Beneficiary (Beneficiary) Goal Beneficiary

In 2Oc and 20f Bill is understood to be actually receiving the Object thrown - he is ‘animately involved’. In 20b and 20e, however, and with any other permissible LOCATIVE preposition apart from to, Bill plays the role merely of a physical object in space and not of an active participant in an interaction. From 2Oc and 20f one has the impression that Bill actually catches the Object in his hands and indeed, in 2Of, almost that the paint is in a tin for this very reason. In 20b and 20e, however, Bill need not even be aware of what John is doing and could even be dead. Sentences 20a and 20d are ambiguous between POSSESSIVE and LOCATIVE interpretations, though perhaps one’s first interpretation of them would be as POSSESSIVES similar in meaning to 2Oc and 20f. Notice that if we were to paraphrase 2Oc and 20f by inserting a preposition it would have to be to; it could not possibly be at or over. There seem therefore to be both formal and semantic-interpretational criteria which distinguish between LOCATIVES and (non-static) transitory POSSESSIVES, even when the locative Goal is animate. Two things deserve emphasis: (1) Transitory POSSESSIVE clauses may be compounded with ACTIONAL clauses jUSt as LOCATIVES may. It is the relation between Object and Beneficiary, in which Beneficiary is ‘animately involved’ as receiver of the Object, that defines the clause as POSSESSIVE. (2) The semantic distinction between Beneficiary and Goal is not based merely on the feature [& ANIMATE], but on whether or not the animate being is understood to be behaving animately, or else merely as a physical object in space. This distinction will later be related to the sensory-motor child’s dual concepts of persons both as active beings and as physical objects. It is

412

Derek Edwards

the same distinction as was drawn earlier with regard to the roles of persons as Agents and of persons’ bodies as Objects and Instruments. 2.2.4 EXPERIENTIALverbs EXPERBNTIALverbs were discussed in Section 2.1 as being of three main types - reaction, perception, cognition - and as describing the mental experience of an Experiencer: The entity or fact ‘experienced’ (known, reacted to, seen, etc.) being the Phenomenon. In the causative clause of Table 3 (‘Mary interested John in opera’), Mary is Agent. She is the instigator and not the object of interest. In the clause ‘Mary interested John’, Mary would be Phenomenon, as would opera in ‘opera interested John’. An example of Instrument rather than Agent as causer would be say, ‘the TV programme interested John in opera’. (Again, ‘the TV programme interested John’ would be static EXPERIENTIALwith the TV programme as Phenomenon). The prepositions in, about (sad about X), over (upset over X) and so on often seem with EXPERIENTIALverbs merely to introduce the Phenomenon. Although often vaguely semantically appropriate these prepositions do not have the same systematic semanticrelational status as they have as LOCATIVEverbs, despite Anderson’s (1971) treatment of the clauses as ‘abstract location’. The in of ‘interested in X’ is a lot more arbitrary than that of ‘lying in bed’. It is as if the description of mental processes arrived later in the evolution of languages than the description of the locativity of objects and borrowed the latter’s vocabulary and syntax more or less idiomatically (cjY Chafe, 1970, on the role of idioms in language evolution). Some EXPERIENTIALverbs are much more explicitly locative, however and combine with locative prepositions in a fully meaningful way; for example, ‘look at/under/over/ through the door’. The importance of this for the description of certain verbs as they occur in two-word child language and the relation of it all to child cognition is discussed below in Section 3.2.3. 2.2.5 ATTRIBUTIVEdaUSt?S ATTRIBUTIVE clauses describe the Object as having, or as undergoing change with respect to, a particular attribute. Attributes include characteristics and dispositions of persons (fat, clever, etc.) insofar as these are not mental states experienced by persons. The distinction between experiences and attributes of animate beings underlies the ambiguity of sentence 21. (21)

John felt cold

Either John is experiencing the cold himself, or else his body is described as a physical object cold to somebody else’s touch and possibly even dead. There are of course various sub-types of ATTRIBUTIVEverbs in English. Some label

Sensory-motor intelligence and semantic relations in early child grammar

413

attributes that can generally be said to be true or false of a given object in itself, such as square,flat and red. Others label attributes that are dimensional and relative to a norm or criterion, such as wide, heavy, long (see Chafe, 1970, pp. 119ff, on ‘relative states’). Some early ATTRIBUTIVES in child language are examined in 3.2.1. 2.3

Conclusion of Section 2: The case Object

Arising from the definitions of the cases in 2.1 and the clause-types in 2.2 is the central importance among the cases of the case Object (of which Experiencer is a special variant). Object is the case in relation to which all of the others are defined. Moreover it is the only case to appear in all of the clause-types. It is the ‘pivotal participant’ around whose relation to the verb the whole system operates (cx Halliday’s, 1968 ‘Affected’ participant). This can be seen most clearly in the progression of Stative verbs through the static-dynamic-causative paradigm : (22)

a. The door is open b. The door opened c. John opened the door

The case Object is introduced neither by this latter dimension (as Agent is specifically causative) nor by the Actional-Stative dimension (as Location, Source and Goal are specificially LOCATIVE), but is common and basic to both. If we compare our Table 3 to Brown’s list of semantic relations expressed in early child grammar (Table l), it is clear at once that Table 3 is far more complex. This is because Table 3 is not a description of actual utterances of young children, but rather a general framework of semantic relations in terms of which child language may be described. It is a matter for empirical research (in which the present author is engaged) how these relations are realized in children’s utterances, which cases are realized from a given clause and in what order, and what is the frequency and order of appearance and why. Table 3 only presents the underlying relations that are necessary to any description of a child’s utterance that involves such notions as Agent or Locative. Semanticrelational interpretations of children’s utterances are implicitly classifications of them into semantic clause-types and this is what necessitates the postulation of unexpressed cases that are nevertheless part of the meaning of the utterance. However, two basic reservations must be made about such interpretations. Firstly, how do we know that we are not attributing to the child a too sophisticated, or perhaps a language-specific model of relations between entities in the real world? Secondly, even if the child possesses such concepts and behaves in terms of them, what kind of evidence tells us what his actual utterances mean when spoken? Does the

414

Derek Edwards

child use, say, want (if he uses it at all) as an EXPERIENTIAL verb, and if not, what is its meaning and how do we find out? These general issues are now taken up in Section 3.

3. Sensory-motor

intelligence

and language

The evidence presented by Slobin (1970, 1972) that certain semantic relations are probably universally present even in the earliest stages of child language suggests that those relations may correspond to functional roles and relations in the 13 to 2-year-old child’s cognitive model of his world. Moreover it is likely that such conceptions are ontogenetically and phylogenetically prelinguistic not only because they develop during the sensory-motor period in children, but also because they appear to be expressed (albeit not by word order) in the meanings of the sign-language utterances of the chimpanzee ‘Washoe’ (Gardner and Gardner, 1971; McNeill, 1972). Chimpanzees’ manipulative and problem-solving abilities involve a well developed sensory-motor intelligence. Piaget’s (1952, 1954) work on sensory-motor intelligence in children is still the most thorough and theoretically most integrated work available and much of the more recent work on infant cognition has been concerned with issues arising directly from it. Moreover Piaget deals with concepts both in relation to the child’s own spontaneous behavior and in relation to the child’s world-picture as a whole, attempting to arrive at a general understanding of how the child conceives of the world about him by the age of 18 to 20 months which is generally when the earliest two-term utterances have appeared. We shall therefore use Piaget’s framework as the basis of our understanding of how 14 to 2-year-olds conceive of the world.

3.1

The child’s construction of reality

In the book whose title we have borrowed for this section Piaget (1954, henceforth CR) separates out for detailed study four interrelated aspects of sensory-motor intelligence: The concepts of Objects, Space, Causality and Time. These aspects are separated only for the purposes of exposition; in fact they are inextricably interdependent and mutually defining. We shall consider them two at a time (cJ CR, pp. 364-380) in order to avoid undue repetition and also to emphasize their interdependence. Since we are interested mainly in what the child’s model of events in the world is like when he starts talking about them, we shall be concentrating on sensory-motor stages 4 to 6, omitting discussion of development before the age of about 8 months.

Sensory-motor intelligence and semantic relations in early child grammar

3.1.1

415

Objects and space

‘Could it not be said that the object exists in substance from the very beginning, only its localisation in space being subject to difficulties? . . . such a distinction is in fact meaningless; to exist as an object is to be ordered in space, for the elaboration of space is precisely the objectification of perceived images’ (CR, pp. 43-44). An object is conceptually ‘permanent’ and discrete only insofar as it can be abstracted from its different movements, orientations and locations vis-d-vis other objects, from its responses to the child’s actions upon it and from the child’s own body. The concept of ‘object permanence’ occupies a central position in Piaget’s discussion, being the cornerstone of sensory-motor intelligence and the concept in terms of which he unites all four aspects (objects, space, causality, time) into a general picture of the child’s conception of the world : ‘The elaboration of the object is bound up with that of the universe as a whole’ (CR, p. 94). Piaget’s most important evidence for his description of the development of the object concept comes from experiments involving a series of ‘displacements’ of objects, where objects are hidden behind or beneath a succession of screens. The criterion responses are whether and where the child looks for the hidden object. So clearly Piaget’s operational definition of the concept of object permanence actually involves an understanding of spatial relations between objects and screens and of the ways in which object-screen relations are changed by the action of the child or of the experimenter. Conversely, the concept of space is the concept of objects-inlocations, these relations being learned via the child’s own actions on objects. Paiget distinguishes three periods (sub-stages 4 to 6 of sensory-motor intelligence) in which the child’s behavior in the displacement task develops. One basic procedure is that the experimenter hides an object under screen A where the child is allowed to find it and then removes it to screen B and perhaps again to under screen C, all in full view of the child. In stage 4 (before 12 months) the child generally continues to look for the object under screen A despite having seen it moved and hidden elsewhere. In stage 5 (12-16/18 months) the child is able to follow the sequence of displacements and find the object where it was last hidden. However, if the experimenter conceals the object in his hand or inside a box and then removes and leaves the object behind screen B or C, the stage 5 child will search only beneath a screen where the object has been visibly placed (screen A). He is unable at this stage to infer the possibility that the object has been removed from the container while out of sight behind B or C. Stage 6 (after 16/18 months) brings the full maturity of the object concept with the ability to infer such invisible displacements. The conceptual difference between stages 4 and 6 is not merely a matter of memory capacity. Indeed the stage 4 response of looking under screen A after a series of displacements involving other screens consists in choosing the most time-distant

416

Derek Edwards

location in preference to more recent ones. Rather, we must infer that the stage 4 child has no coherent and integrated model of the world in terms of which to organise his memory of events. What Piaget calls the ‘spatial group’ is such an organisation ‘if the child does not remember the order of displacements, it is because in such cases he does not construct a coherent spatial group’ (CR, p. 78). The notion of spatial groups is discussed below in 3.1.2.3. Piaget (CR, p. 175) also observed in his own children’s spontaneous behavior that at 10 months of age they neither looked for nor showed any awareness of alternative routes that might avoid obstacles in their search tar lost or hidden objects. By 18 months, however, it is a different story: ‘. . . Jacqueline throws a ball under a sofa. But instead of bending down and searching for it on the floor she looks at the place, realizes that the ball must have crossed under the sofa and sets out to go behind it. But there is a table on her right and the sofa is backed against a bed on the left; therefore she begins by turning her back on the place where the ball disappeared, goes around the table and finally arrives behind the sofa at the right place. She has thus closed the circle by an itinerary different from that of the object . . . through representation of the invisible displacement of the ball and of the detour to be made in order to find it again’ (CR, p. 205). Jacqueline is clearly able now to imagine and take alternative routes to particular goals, demonstrating an implicit organization of spatial relations between objects that include her own body as an object whose displacements can be planned in advance and then smoothly executed. The child’s conception of itself and of other persons as physical objects like tables and chairs, with permanence and solidity and occupying positions in space, is of importance for our general thesis concerning cognition and case grammar. Persons are physical objects in exactly the same way as other objects - they are substantial objects that can be pushed and slapped, that cannot be walked through, that can fall down and so on. The child’s concept of himself as such a physical object is inferred by Piaget from detour behavior like Jacqueline’s and from observations such as this: ‘At 1; 3 (10) Jacqueline, in her playpen, discovers the possibility of letting herself fall down in a sitting position; she holds the bar and lowers herself gently to within a few centimetres of the floor, and then lets go of her support . . . foreseeing the trajectory her movement of falling will follow independently of any activity on her part’ (CR, p. 291). Piaget interprets this as showing that ‘the child henceforth considers himself dependent on laws external to himself or as submitting to causes independent of himself’ (p. 291) and again, ‘the child henceforth conceives of his own body as being . . . subject to the actions of things as well as a source of actions which operate on them’ (p. 292). The concept of objects and of persons as physical objects that occupy spatial positions and that are affected by the actions of other persons and objects seems to parallel closely the case Object - the entity described in a clause as in a particular

Sensory-motor intelligence and semantic relations in ear& child grammar

417

position or state or to be affected by change or movement or by the directed action of another entity. And like the object concept in sensory-motor intelligence the case Object occupies a central position in defining the roles and relations among the rest of the conceptual system, subsuming passive roles played by both animate and inanimate objects. Both the case Object and the concept of physical objects are detlned and conceptualized in the same kind of actional and relational terms. Permanent objects are not simply things to which a language can attach labels; they are also things that are understood and known in terms of the relations they contract with the context in which they are perceived. There is also, however, a direct empirical relation between children’s object labelling and their performance on object permanence tasks (Roberts and Black, 1972) and this can be linked to the findings of LeCompte and Gratch (1972) in which infants’ object concepts were tested by a sensitive ‘trick’ method in which the experimenter would on some trials trick the child by surreptitiously substituting a different object from the one hidden for the child to discover. Children aged 18 months showed ‘high puzzlement’ and continued searching for the missing toy. They clearly had a concept of a particular missing toy and therefore a nameable one, rather than merely an object or merely a place in which to look. The capacity to attach verbal labels to objects is a symbolic and linguistic capacity not ‘given’ by sensory-motor knowledge alone. However, nameable objects do have to be perceived, recognized and distinguished from their surroundings. In what Brown (1956) called ‘The Original Word Game’, played by mother and child, there are two complementary routines. In object-naming the child supplies a name for a gesturally located object; in pointing at named objects he indicates their locations. What is given in one situation is supplied in the other. It could be argued that while the latter situation is parallel to ‘static LOCATIVE' clauses, the former, Brown’s ‘Nominations’ (see Table 1 above), is a precursor of those clauses in adult grammar in which a particular object is identified or named, as in ‘the vicar is Mr. Johnson’, or ‘Henry is a hooligan’. The first-named element (‘the vicar’ or ‘Henry’) is linguistically introduced as the thing to be identified or given a category-label (c$ Halliday, 1967-1968); in object-naming by children, the ‘thing to be identified’ is simply pointed at. The identifying labels, whether labels of definite entities (Mr. Johnson) or of indefinite categories (a hooligan) correspond to the child’s label. The fact that the ‘thing to be identified’ is usually definite (‘the N’, or a proper noun or pronoun) in the adult naming-clauses is paralleled in the fact that the child’s object-to-be-named is deictically located by the looking and pointing gestures (cJ, Thorne, 1972, on the locative basis of the notion ‘definite’). Another ‘Operation of Reference’ in Brown’s list is the category ‘Nonexistence’, expressed by such well-known pivot structures as ‘allgone -’ and ‘- gone’. However,

418

Derek Edwards

such structures more typically occur in the context of the absence or disappearance rather than the ‘nonexistence’ of the named object. Indeed, Brown (1970, p. 222) defines Nonexistence as referring to something ‘which has been in the referent field or was expected to appear in it [but] is now not to be seen’. These are clearly objectlocation structures which depend for their meaning on the fact that objects when out of sight do continue to exist - that is, they depend on object permanence and are occasioned by the disappearance or absence of an object from an expected location. 3.1.2 Causality and time In Piaget’s system the development of the concepts of Causality and of Time is ‘completely analogous’ to that of Objects and Space: ‘The causal and temporal series . . . constitute merely the other face of the objective and spatial series envisaged hitherto’ (CR, p. 221). In order for the child to understand ‘Causality’, that is, how he and other persons act on objects and cause changes in their positions and states, it is obvious that he has to conceive of persons and objects as discrete entities, related systematically in a common space. And conversely, the concept of objects and their spatial displacements depends on the child’s appreciation of how his actions and those of other persons affect the world. Piaget distinguishes two types of Causality in sensory-motor intelligence : ‘Physical Causality’ which evolves from an earlier ‘phenomenalism’ and ‘Psychological Causality’ which evolves from the child’s sense of his own ‘efficacy’. These two types of Causality involve causal relations and physical contact, both object-object and person-object. They bear a close comparison to Fillmore’s distinction between the case roles Instrument and Agent. Equally interesting is the fact that these cases are also intimately connected with ‘causativity’ in the system of verb and clause types. 3.1.2.1 Physical and psychological causality. Piaget defines the concept of Physical Causality as ‘an objectified and spatialized causality affecting the interrelations of things’. It is the causal action that one object exerts on another through direct physical contact. Piaget suggests that ‘the child begins by discovering fortuitously that by hitting an object the stick can put it in motion’ (CR, p. 282). The role of the stick in the child’s subsequent planning of such events compares with Fillmore’s definition of Instrument: ‘The case of the inanimate force or object causally involved in the action . ..’ and with Chafe’s ‘something which the Agent uses.’ Sensory-motor children share with Kohler’s chimpanzees the ability to solve problems by taking advantage of the ways in which objects can be used to change the spatial positions of other objects through physical contact and so enable goal-objects to be reached. While younger children may discover such solutions by trial and error, by the age of 15 to 20 months they tend to do so without hesitation, apparently by

Sensory-motor intelligence and semantic relations irl early child grammar

419

‘insight’ (Lezine, Stambak and Casati, 1969). The older children show in advance of solving the problem a conceptual awareness of the ways in which objects can be forseeably rearranged - what Woodward (1971, p. 127) calls ‘the sequencing of component schemata before the performance of the actions’ - and of how such changes can be effected through the use of objects wielded as instruments. ‘Psychological’ Causality is defined as ‘a causality . . . uniting intentions with acts’. It is essentially the child’s concept of persons (and probably animals) as being capable of self-initiated movement, of causing events. So as well as conceiving of himself and other people asphysicaI objects, we may add that the prelinguistic child also conceives of himself and of others as agentive beings (c$ Fillmore’s Agent: ‘The typically animate perceived instigator of the action . ..‘). How do we know the child conceives of himself as an agentive being? Pre-verbal self-concepts are infuriatingly inaccessible, but Piaget offers us this argument: ‘To the extent that the child renounces considering external phenomena as the mere extension of his own action and confers upon them, along with objectivity and spatiality, a causal structure which is truly physical and independent of the self, it is very probable that he becomes aware of his own activity as direct power exerted by his intentions upon his organism’ (CR, p. 287). One of the first responses children make to parental speech is to obey commands. Ingram (1971, p. 904) argues that the I+-year-old ‘has felt the sense of agent in himself. He can obey simple commands calling for him to act as an agent’. Other behavioral evidence for this putative agentive self-concept is the child’s imitations of the agentive actions of other persons, implying that the child has a concept of ‘actions’ abstracted from both himself and the other as the persons who perform them.2 The child’s concept of other persons as agentive beings is somewhat more accessible than his self-concept; it can be inferred from his overt behavior towards them. Before the age of 8 to 12 months children seem to treat objects and persons in much the same way - as e&ities whose actions and movements depend on the child’s own ‘magical procedures’ which range from crying in the cot to make Mummy come, to Jacqueline’s routine of arching her back in an effort to get inanimate objects to repeat interesting movements, or else to get her father to push his tongue out again. By the arrival of their first birthday, however, children are well on their way towards appreciating that events in the world are systematically and physically rather than magically related to their own actions and that some events occur quite independently of the child’s 2. This was suggested to me by Roger Goodwin. It might be noted that imitation of auditoria input need involve no such process and may be done without perception of the emitter. However, imitating bodily the actions

performed by persons does seem to require the perceptual abstraction of the person from the action and the translation of it into a similar action performed by the self.

420

Derek Edwards

actions and might be instigated by other people. By this age a child ‘looks for a causal agent on seeing the movement of an object that he did not himself act upon’ (Woodward, 1971, p. 83). Piaget (CR, p. 276) reports that at 15 months Lucienne ‘like Jacqueline . . . holds out to her parents boxes she cannot open or toys she cannot operate, or else she points with her finger to objects that are too far away, so that they will be brought to her’. These actions in relation to other persons are prelinguistic acts of communication in which the addressee, as in all imperatives, is required to act as an agent. Piaget states that ‘from this time on, the child considers the person of another as an entirely autonomous source of actions’. That is, the sensory-motor child has a nonlinguistic concept of other persons as instigators of actions that is very similar to the role Agent in case grammar. Ingram (1971) refers to such nonlinguistic agentivity as ‘semantic transitivity’, which links these notions very nicely to the discussion in 2.2 on transitivity and causativity in the clause system. The child’s acquisition of ‘semantic transitivity’ begins not merely when he can utter a transitive verb or clause but much earlier when he first learns about the nature of persons and objects and their interactions in the world about him. Woodward (1971, p. 85) suggests that ‘getting beyond “magical” behavior is thus a matter of the child’s first learning to distinguish between the outcomes of his own actions on objects, and those of other people or of inanimate forces’. These cognitive categories and relations parallel in a remarkable way, considering their completely independent formulation by cognitive-developmental psychologists, the causativity relations that hold between Agent, Instrument and Object in case grammar. 3.1.2.2 Persons and their bodies in cognition and grammar. We have seen in the previous sections that sensory-motor children conceive of persons in two ways, as physical objects and as agentive beings, according to the manner of involvement of persons in action-relations (as objects affected or located in space versus as instigators of actions and events). It has to be stressed that this is not simply a matter of recognising physical differences between persons and other entities - that they have faces, legs, etc. - nor even simply a matter of their ‘animacy’. Rather, it is a matter of roles played in situations and interactions. The same animate person can ‘behave’ differently either as physical object or as agent. In 2.2.3 we saw that the cases Beneficiary and Goal behave differently in English syntax and that this could be related to their semantic involvement in the action described by the verb. Similarly persons play the case roles of Agent or Instrument also according to whether they are ‘animately involved’ or involved merely as a physical object, depending on whether we understand the action to be deliberate or not. We can now relate these distinctions to the sensory-motor child’s dual concept of persons both as active beings and as members of the class of general physical objects. If we

Sensory-motor intelligence and semantic relations in early child grammar

include the notion that adult grammar’s case role ‘Experiencer’ involved’ kind of Object affected, we can construct Table 4. Table 4.

421

is a special ‘animately

Case roles played by persons and objects Affected roles

Relational roles

Causer roles

Objects and persons inanimately involved

0 bject

Phenomenon, Goal, Source, Location

Instrument

Persons animately involved

Experiencer

Beneficiary, Possessor

Agent

The sensory-motor child’s concept of persons is essentially dualistic; the agentive ‘animately involved’ aspect of persons is related to their intentionality, while the ‘physical object’ aspect is a function of persons as bodies. Piaget (CR, pp. 288-289) speaks of the development of the child’s awareness of ‘the particular powers that his intentions, desires and efforts have over the perpetual body which is his own body’. for prehension and consequently Similarly, ‘the hand becomes . . . an instrument subject to conscious intentions on the child’s part’ (p. 230). The structure of semantic relations seems to be correspondingly dualistic, as Table 4 implies. Parts of a person’s body may play the semantic role of Instrument wielded by the person-as-agent: (23)

a. John broke the vase with his fist b. Henry pushed the door open with his foot c. Bill used his elbow to turn the light off

Similarly, the ‘inalienable POSSESSIVE’ relation between person-as-Possessor and his own body parts is dualistic, as are ‘reflexive Agents’ where the Agent and Instrument, or Agent and Object of a clause are realised as the same lexical entity. The notion of a person initiating a movement of his own body is clearly dualistic. 3.1.2.3 Causality in groups of spatial displacements. The concept of ‘groups’ o spatial displacements presupposes the concept of object permanence and consists in an understanding of the fact that if an object in location A is moved to location B then the object is now in location B and the child in location X can bypass A and go straight to B to find it. The spatial relations between objects are organized so that movements of objects are ‘reversible’ - an inverse movement will return an object to its previous location. The ‘group’ characteristics of object-place relations involved in the object displacement paradigm depend in fact not only on object-place relations but also on

422

Derek Edwards

Causality. The changes of position between the actions of persons; and the reversibility

object and screens are changes caused by of the object-place relations is in fact the

capacity for the object to be moved back to its original location by an inverse action. In his solution of the ‘invisible displacements’ problem the child was able to infer the operation and element which would complete a spatial group - that is, to infer that the experimenter must have removed the object while it was out of sight behind screen B or C and left it there. Note that the child is not inferring merely an objectlocation relation but an agent-action-object-place relation. Observations of spontaneous behavior similarly show that the l&month-old child can infer missing elements from caused events, such as when he looks for an agent outside the immediate visual field in order to account for the movement of an inanimate object. Piaget (CR, p. 295) quotes this example: ‘Jacqueline is examining the arm of an old armchair, unfamiliar to her, with an extension leaf used for trays, which I operate from behind. Jacqueline has not seen me do this and does not see my arm when I push the leaf. Nevertheless when it stops Jacqueline immediately turns to me, looks at my hand, and definitely shows by her behavior that she considers me the cause of the object’s movements. Hence this involves mental reconstruction of the causes of a perceived effect’. Concerning the ability to infer, or ‘mentally reconstruct’ unseen events, Woodward (1971, p. 117) concludes that ‘the following appear to be “represented” or recalled: static objects not in the perceptual field; the results of the movement of an object by the action of the child himself and the actions of other people and the outcomes of them’. What Woodward lists here are in fact the component elements of a spatial group and these can be represented quite neatly in terms of semantic locativity relations as in Figure 4. The semantic relations in Figure 4 express the following aspects Figure

4.

Semantic State

structures derivedfrom 1

Operation State 2

spatial groups

Object + V.stat.Loc

+ Location,

Agent + Object + V.caus.Loc

Object + V.stat.Loc

f

Source [ Goal

1

+ Location,

of a simple object displacement procedure. ‘State 1’ is the initial static relation between object and location, such as an object being beneath screen A. The ‘operation’ is the action of an agent (the experimenter : Instrument is omitted for convenience) in moving (V.caus.Loc) the object from its Source position (Location,) to a Goal position (Location,). ‘State 2’ is the new static locative relation between the object and another screen. If the ‘operation’ is reversed, Location, becomes Source and Location,

Sensory-motor intelligence and semantic relations in early child grammar

423

Goal, the object is returned to Location, and State 1 is restored. Given State 1 and subsequently State 2, the operation can be inferred. Given State 2 and the operation, State 1 can be inferred. Thus, the spatial groups concept is quite readily describable in terms of the locativity relations which were outlined earlier in Table 3. It is this isomorphism between the cognition and the semantics of person-object-space relations that enables the child to set about discovering how such relations are realised in the language spoken to him. 3.1.2.4 Time. The sensory-motor child’s concept of Time is merely what is implied in what has already been said about objects, space and causality. The concept of a spatial group of agent-action-object-space relations demands that the actions performed and the relations that are changed are conceptual units or ‘elements’ in the group. Events across time are conceptually ‘parsed’ into an ordered succession of states and actions. The conceptual reversibility of the operations depends on their conceptual separation from the static relations which they change. Thus the sensory-motor concept of time is simply the concept of states and events occurring in sequence. It is in terms of an ordered sequence of states and events that the child’s memory is organized - ‘the child remembers the sequential displacements of the object and sets them in the proper order’ (CR, p. 342). The solution of the object displacement problems of stages 5 and 6 involve remembering and even (in stage 6) reconstructing a temporally ordered sequence of displacements. A re-examination of Figure 4 shows that it is this concept of ‘Time’, of an ordered series of unitary states and events, that corresponds more or less to the concept of a deep semantic clause. Our notional definitions of ‘verb’ and ‘clause’ thus correspond to the way in which language encodes people’s perceptions of the flow of activity in the world about them in terms of a succession of states and events. The concept of sequentiality of discrete events, though insufficient for the immediate acquisition of much of temporal reference (Cromer, 1968), does provide a cognitive basis for the child’s very early use of quantifiers such as ‘more’ and ‘nother’ in such utterances as ‘more cereal’, ‘nother toy’ and so on. Bloom (1970a) characterises these as comments on or requests for ‘another instance’ of the object or event named. They therefore refer to the recurrence of a state or event, such as the mother giving the child another helping of cereal, and thus refer directly to unitary actions and events occurring in sequence. In Brown’s system (see Table 1 above) these are classified as ‘Operations of Reference : Recurrence’.

3.2

ATTRIBUTIVES,POSSESSIVES

Our discussions

and

EXPERIENTIALS

of the sensory-motor

concepts

underlying

some semantic

relations

424

Derek Edwards

has dealt with concepts of actions, agents, instruments, the movement and locativity of objects including persons as objects and the notion of a temporal succession of events. The relatively straightforward child language data expressing these notions will not be presented here, so that we may deal more fully with three basic types of verbs and their associated cases outlined in Section 2, which are still cognitively unaccounted for. The reason for this is partly the absence from Piagetian theory and data of the relevant concepts quite so nicely ready-packaged for our use. Also, some of these early meanings are found to be regulated by parental use of language during socialization processes, particularly rewarding, punishing and controlling the child’s behavior and such processes are beyond the scope of Piaget’s cognitive theory. Nevertheless, we shall tid that these meanings also are constrained by the nature of sensorymotor intelligence. 3.2.1 ATTRIBUTIVES The ATTRIBUTIVES that occur in early child language tend to label those features of objects that are physically and perceptually salient, which describe particularly those perceptual discriminanda which are important for the child’s differential actions upon the objects concerned. The ATTRIBUTIVE adjectives listed below, extracted from Bloom’s (1970a, p. 245-246) first corpus of data from a little girl (‘Kathryn’) aged 21 months, include several that describe the tactile qualities of objects as well as their visual characteristics : big fuzzy sharp open (verb) cold heavy sticky stuck dirty little tiny careful funny pink tire(d) hot There was considerable overlap between these ATTRIBUTIVES and those listed in the lexicons for the other two children studied by Bloom. In my own data on a little girl (‘Helen’), ATTRIBUTIVES were much scarcer than in the Kathryn data. The ones that had occurred by the age of 28 months were, in order of appearance, baddy, poor, hot, red, bad, wet, naughty, fluffy, clever, shut, open, sore. Taking account of situational contexts, these ATTRIBUTIVES,together with Bloom’s, can be put into three preliminary semantic categories. 1) physical properties, e.g., hot, wet, pink, fluffy, little, sharp, heavy, dirty, red 2) reward-punishment, e.g., baddy, clever, poor, naughty 3) actional and reversible, e.g., open, shut The first category contains many that are probably learned from parents in the context of constraints on the child’ actions (hot, wet, sharp, heavy, dirty, etc.). These latter are physical properties which have a special importance in determining the

Sensory-motor intelligence and semantic relations in early child grammar

425

manner and extent of the child’s actions on the objects concerned. Indeed they are probably acquired often with idiosyncratic meanings related to their appropriateness of use in the context of such constraints, so that sharp may refer not so much to sharpness as to some kind of ‘not to be touched’ property that adults seem to attribute to kitchen knives. Similarly, tiredand busy may be acquired in the context of a mother who is in various ways unavailable for play and conversation. Kathryn said ‘Mommy busy’ when her mother was not even present (Bloom, 1970a, pp. 47, 53). The second category contains the words Helen applied to herself, to parents and even to offending objects in the context of reward and punishment. They were always vehemently expressed and are clearly bound up with the process of socialization and again are context-specific with regard to a particular type of social relations where a person’s behavior is being moulded and regulated by another who assumes authority. Thus the so-called ‘Interpersonal function’ (Halliday, 1970) of parental language is important to early child language not only in its own right but also in providing the context in which various types of Ideational meanings are learned, defining by interpersonal criteria the appropriateness of use of particular lexical items which may not at first take on their proper adult referential meaning. We shall find a similar constraint imposed by regulatory socialization on Ideational meanings when we look at the child’s notion of ‘permanent possession’. The third category can perhaps be reclassified as non-ATTRIBUTIVE.Open and shut are generally used by young children (e.g., Bloom’s ‘Kathryn’ and my own subjects Helen and Alice) in the context of actions on objects (doors, flaps, lids, boxes, etc.) rather than in descriptions of ‘states of affairs’; that is, their usage corresponds to their adult usage as surface verbs rather than as adjectives. The verbs open and shut are indeed ATTRIBUTIVE in adult grammar only insofar as they describe the doorway, not the door itself (consider the ambiguity of ‘he walked through the door’). Young children probably refer to the object ‘door’ rather than to the state of the doorway, as in ‘open the lid’ rather than in ‘open the box’. Thus open and shut in early child language are probably best classified as LOCATIVE, since they refer to the locative orientation of the physical object ‘door’ and describe no state or change of state of the door itself. Open and shut are Relational rather than Characterising in child speech. The true ATTRIBUTIVES of categories 1 and 2 tend to be static only. This may relate to the fact that (in contrast to open and shut even in adult usage) changes in such states are both infrequent as well as non-reversible in the Piagetian sense. Compared to the extent to which persons and objects perform actions, are involved in changes of locativity and so on, it is clear that objects will be witnessed in the process of becoming sharp, fluffy, heavy or red a lot less often. ‘Non-reversibility’ means here that if one breaks a window, say, one cannot perform the inverse operation to restore it - the logic of actions in spatial groups does not apply to changes of ATTRIBUTIVE states.

426

Derek Edwards

Thus ATTRIBUTIVES in early child grammar tend to be static only and to label visually salient discriminanda of objects, a major subset of which includes those attributes of persons and objects that are criteria of constraints on the child’s behavior towards them. This subset may include ‘dimensional’ ATTRIBUTIVES such as Helen’s word heavy which even at 24 years was not used as a value on a perceptual dimension on which all objects are ordered. Rather, objects were simply heavy or they were not and the criterion was whether or not Helen had any difficulty in lifting them. Objects like cars, buses and cows were not described as heavy; the word was acquired in the context of difficulty in lifting or moving objects in the home, as was Alice’s very similar word ‘stuck’ and it was to these experiences that the meaning of the adult’s word was assimilated. 3.2.2 POSSESSIVES In his discussions of sensory-motor intelligence Piaget has not dealt with any notions of possession of objects by persons. Perhaps such notions were considered to be culturally relative and unlikely to be a fundamental part of the structure of human cognition as are object-space relations. However, the concepts and relations involved in the various meanings of ‘possession’ that occur in early child language are those that have already been seen to be fundamental in sensory-motor cognition. All three types of POSSESSIVErelations - inalienable, permanent and transitory - occurred in Helen’s early two-word speech. Let us examine what kind of person-object relations they were that Helen was talking about. 3.2.2.1 Inalienable POSSESSIVES. Naturally enough, these occurred only in situations where the relation was ‘static’. Here are some examples: AGE

1) 2413 2) 25:o

UTTERANCE

H: daddy neck (M: where’s daddy’s face?) H : daddy face daddy face

(M: no, that’s mummy’s face) H : mummy face dolly face

3) 25:o

(M: good girl, that’s the dolly’s face) H: thumb daddy thumb

(D: daddy’s thumb, yes)

CONTEXT

H puts a hand on D’s neck H pointing at D’s face H looking at M ; not pointing (M probably misinterprets H) H pointing at M’s face H holding doll in one hand with forefinger of other hand on doll’s face H holding D’s thumb

Sensory-motor inteliigence and semantic relations in early child grammar

H: thumb Helen thumb

427

H gets hold of H’s left thumb with right hand

(D: yes, that’s Helen’s thumb) These are straightforward static inalienable POSSESSIVES. They involve two basic concepts: 1) The part-whole concept. Parts of objects that are physically, perceptually or functionally discrete will be perceived as such by a child who regularizes his perception of the world into states and movements of objects. Parts of objects tend to be described in adult English as attached to the whole object via the grammar of possession. Thus, we say ‘this car has three doors’, ‘the teapot’s lid’, etc. 2) The person-body dualism. The notion that body parts are physical objects possessed, moved around and used as instruments by the persons who ‘own’ them seems to be common to both child and adult language and to sensory-motor cognition (see section 3.1.2.2 and others above). All of Helen’s inalienable POSSESSIVESup to the age of 26 months were person-body-part structures. 3 There were no kinship inalienables such as ‘Helen’s daddy’. Early use of the words ‘daddy’ and ‘mummy’ is as the names of actual persons rather than of kinship relations between persons. And note that kinship relations are not encompassed by either the part-whole or the person-body relations which fall within the scope of sensory-motor intelligence. 3.2.2.2 Transitory POSSESSIVES. These occurred in Helen’s early utterances only as ‘causatives’; that is, referring only to situations in which persons were passing objects to each other. AGE 1)

25:0

UTTERANCE

CONTEXT

H: dr jelly

H looking at D (There is a packet of jelly sweets on a shelf out of H’s reach) (dr is a special category of sounds used by H in requests) H handing doll to D (gimme is closer to adult ‘give’ than to ‘give me’) H holding out to D a cup from which H has been drinking (drink refers to cup) D accepts the cup

dr jeliy daddy

2) 25:l

H: gimme dolly

3) 25:l

H: gimme drink

3. We have to include Helen’s dolls and teddy bears as ‘persons’ since they functioned as such in make-believe play. However, Helen did not really conceive of them as animate and

agentive beings; she would never make requests of them or wait for them to act spontaneously. All of their human-like actions were acted out on them quite deliberately by Helen herself.

428

Derek Edwards

I-I: gimme daddy gimme mummy (M: thank you)

H D H M

offering a magazine to D accepts the magazine picks up a cup; offering to M accepting the cup

Helen’s transitory POSSESSIVES, of which these examples are typical, described the relations involved when people hand objects to each other. Invariably there was hand to hand exchange. Moreover, as example 3 demonstrates, where Helen herself was Agent she would generally not simply drop or place the object in the receiver’s lap but would stand there offering it, waiting for the receiver (Beneficiary) to put his hand out and grasp it. Helen’s concept of Beneficiary in these transactions is clearly that of an ‘animately involved receiver’ of an object (see sections 2.2.3 and 3.1.2.2). In some partially analyzed data I have descriptions of children of 17 to 18 months who show the same behavioral pattern, standing within arm’s reach of the receiver, offering some object and looking from object to receiver’s hand to receiver’s eyes until the receiver reaches out and grasps it. Helen’s verb gimme encoded this kind of transitory POSSESSIVE transaction; its usage was restricted to hand-to-hand exchanges and was never used in the context of non-possessive changes of location. Helen never said “gimme table when putting things on the table. The verb was clearly POSSESSIVE and not LOCATIVE, so that Helen distinguished these relations semantically as well as cognitively. 3.2.2.3 Permanent POSSESSIVES. Helen’s permanent POSSESSIVES were still all static at the age of three years. She had no concept yet of the transference of proprietary rights whether involving gifts, donations, buying or selling. The reason for this is that her concept of static permanent possession was not in fact quite the abstract, legalistic conception of ‘ownership’ that adults in our culture have. Here are some typical examples of the contexts in which they occurred : AGE

UTTERANCE

CONTEXT

1) 24~3

(M: off. Leave them alone Helen) H : mummy glasses (M: where’s daddy’s glasses?) H : daddy glasses

H reaching for M’s glasses on top of TV set H pulls back, points at glasses

2) 25:l H: daddy daddy book (D: yes. That’s daddy’s book)

H points at glasses D is wearing H picks up one of D’s books, looks at D, puts it down (H is generally forbidden to play with D’s and M’s books and has some of her own)

Sensory-motor intelligence and semantic relations in early child grammar

3) 25:l

H: Helen button (M: Helen’s button)

4) 32~3

5) 32~3

(D: Hey, come out of there) H: Is/ daddy’s television (D : put the brush down, Helen) H: mummy’s brush

429

H holding button on H’s cardigan, looking at M H creeping behind TV set which is backed against corner of room H dangerously swinging longhandled broom about (M absent) H puts broom back in cupboard

These examples cover a wide age range and show Helen’s concept of permanent possession to be bound up with the same kind of interpersonally imposed restrictions on her actions with particular objects that characterised some of her ATTRIBUTIVES. The notion expressed here, however, is a Relational one - of privileged access of particular persons to particular objects. This type of possession is clearly what Brown (1970, p. 222) is referring to: ‘The children’s possessive typically divides spaces and objects in the house among family members: it expresses a kind of primitive notion of territoriality and property’. Examples 4 and 5 are especially interesting. Helen’s parents lay no special proprietorial claim to the TV set nor to the broom. These utterances are diagnostic of Helen’s general concept of permanent possession, a concept not only of the association of persons with the things they habitually wear and use, but also of constraints that are imposed on Helen’s actions on objects to which her parents have privileged access. And this is a definition of ‘ownership’ that is within the scope of prelinguistic cognition. Indeed, I have unpublished descriptions of the behavior of prelinguistic children which suggest the presence of such concepts, as, for example, when a child, reaching towards some forbidden object such as a book or a telephone receiver, looks up to see that he is being watched and suddenly draws back with a submissive grin. Moreover it will be appreciated that the child’s restriction to static permanent possession is in keeping with these notions; possessed objects are so defined by consistent person-object relations rather than by a readily exchangeable contract or right of ownership. 3.2.3 EXPERIENTIALS Brown’s table of semantic relations in child language (see Table 1 above) and Slobin’s (1970, 1972) similar lists of possible universals do not include the category of EXPERIENTIAL verbs nor their associated cases. This seems correct, because it would be improbable that children really talk about mental states and experiences at the twoword stage given our previous suggestion that they are cognitively limited to the description of ‘overt perceivable events’ (whether real, inferred or imaginary). Table 5, abstracted from Bloom’s data on the speech of three children, lists the verbs that might

430

Derek Edwards

if they occurred in adult speech be classified as EXPERIENTIAL. (Letters stand for the children; numbers refer to successive speech samples for each child. Words italicized occurred in syntactic contexts, the others as single words) Table.

5.

EXPERIENTIAL

verbs in Bloom’s (1970a) data

AGE MLU Kl Gl G2 El E2 E3

21.0 19.1 20.2 19.1 20.2 22.0

1.32 1.12 1.34 1.10 1.19 1.42

look

read

look look

read

read

see

want

see

watch watch watch

see see see

look

show

show

want want

watch

Given that read is a version of ‘look at’ (a special verb for looking through books perhaps) all of these verbs except want are visual-perceptual. None describe ‘reactions’ (sad, happy, like, etc.) and none describe ‘cognitions’ (know, think, believe, etc.). Want usually functions in child language in the context of requests for people to do things or give things to the child and not as a description of a mental state (CA Gruber, 1967b). Helen’s use of want was at first exclusively in requests for objects to be given to her - as an imperative causative transitory POSSESSIVE;other children may use the word in a less restricted sense, to encode any kind of request. The remaining EXPERIENTIAL verbs we have to discuss are the visual-perceptual ones: look, read, see, show and watch. Before 27 months Helen used only two EXPERIENTIALverbs, and these also were look and read. Let us examine this small but persistent group of verbs and determine what their meanings might be in child speech and perhaps explain why young children seem so restricted in their range of ostensible mental process verbs. Apart from being visual perception verbs they are all, with the exception of see, Agentive in adult grammar. That is, they do not refer merely to the passive process of an Experiencer (as do the adult verbs hear, know and like) but also to the action of an Agent. With look, Agent and Experiencer are the same person, while with show they are different persons (c$ Gruber, 1967a; Fillmore, 1968, p. 30; also Chafe, 1970, p. 146). In fact see is sometimes used Agentively too, as in the imperative ‘see page 94’, to direct a person’s attention to something. And this seems to be precisely the meaning that all these verbs have in child grammar - roughly the meaning of look. The special characteristic of look is that in contrast to most EXPERIENTIALS,the looker can be seen to be looking; it is the name of an overt action. To look at something is to direct one’s head and eyes to it, which is a LOCATIVE relation. Indeed, it was pointed out earlier that Zook freely combines in adult grammar with most locative prepositions and hardly ever occurs without one. Similarly watch and show, even their adult usages,

Sensory-motor intelligence and semantic relations in earIy child grammar

431

involve overt locativity of the head and eyes of the seer towards the seen. The important point for their acquisition in early child language is not so much that these words are Agentive in adult grammar as that they are used in the context of overt and perceivable relations between persons and things in the physical world. Slobin’s (1972) classification of ‘see doggie’ as Identification rather than, say, Experiential (and cf. Gruber, 1967b) justifiably ignores the ostensible mental-process meaning implied by see. The child’s attention is directed towards the named object itself, rather than to a process of visual experience. It is surely no coincidence that ostensible EXPERIENTIALS in early child language are limited to this recurrent small set of visual verbs which situational context descriptions show to be invariably used in the contexts of agents, actions, objects and locations. Since there is no evidence, either cognitive or linguistic, that these children are thinking about mental experiences per se, we may suggest that verbs like look and see are first acquired as LOCATIVES, or else as what Brown calls ‘Nominations’ (cJ: above). In terms of the classification ot Table 3, children’s first use of these verbs is restricted to the Relational part of their meaning and excludes the Characterizing part which would be concerned with the nature of the visual experience. There are probably no true EXPERIENTIAL verbs in two-word child language. The meanings that are and are not expressed in the first sentences of children are, therefore, constrained not merely by some kind of deep semantic ‘complexity’. ‘John likes Mary’ is not obviously syntactically or semantically more complex than ‘John hits Mary’, but the latter is an overt action and its meaning is therefore assimilable by the knowledge structures of sensory-motor intelligence. It must be emphasized that what cognition constrains is not directly the actual words acquired, since these are of themselves mere bundles of sounds, but rather the meanings and usages of those words. We have seen some examples of assimilated meanings already: See, look, read, open, shut, more, want. 3.3

Summary and conclusion

The semantic relations expressed in two-word speech are essentially similar to the conceptual relations used in the adult contextualized speech addressed to the child and which are comprehensible to him via his knowledge structures as statements or requests relating to the world as he understands it to work. These concepts and relations are analysed here from two complementary perspectives: 1) The types of states of affairs, relations and interactions that obtain between persons and objects; 2) the roles played by those persons and objects in the states and interactions. These two perspectives correspond respectively to a basic set of 1) clause-types and deep verbs and 2) cases in linguistic theory.

432

Derek Edwards

Summary of links between sensory-motor cognition and language i SEMANTIC RELATIONS COGNITIONS

Table 6.

ROLES IN CAUSATIVITY AND ACTIONS

.

.

Psychological causer (role of persons as initiators and causers of events)

;

Agent

,

. + causative actions Physical causer (role of object or person’s body in affecting another object)

.. .. .. .

.

i

Instrument

that is in a particular state or 1OBJECT relation or undergoing change )

j

movement or the effects of action

ROLES IN STATES AND RELATIONS

1) spatial locativity of obiects 1 i 2) privileged a&s of’ : persons to objects I 3) persons receiving objects in i hand-to-hand exchanges I : i 4) persons and their body parts I : j 5) perceptual discriminanda of behavioral importance

V.LOCATIVE + Location/Source/Goal V.perm.PossEssIvE + Possessor V.caus.trans.PossEsswE

+ Beneficiary

V.inal.PossEssrvE + Possessor V.AT~RIBUTIVE

In defining both the clause-types and the cognitions, the notion of an ‘object affected is central. Table 6 summarizes the basic points of intersection between the first semantic relations children express and their cognitions about the world they are talking about. The notion ‘Object’ includes both the case Object and the sensory-motor concept of physical objects and is shown as the hub of the cognition-language interrelations. Relationships in Table 6 are to be read from left to right via the Object box. MacNamara (1972) has described the language-learning process as a semantically based paradigm in which children ‘initially take the main lexical items in the sentences they hear, determine referents for these items and then use their knowledge of the referents to decide what the semantic structures intended by the speaker must be’. ‘Knowledge of the referents’ is clearly a crucial component of such a process and would involve an understanding not only of the things that words refer to but also of how situations and events are structured according to perceived relationships between objects, between persons and between persons and objects. This paper has tried to outline the kinds of links between cognition and language that would enable the child to acquire language in a comprehension-based process in which the child interprets the meanings of adult speech by mapping meanings onto the situations described, given some knowledge of the referentia1 meanings of one or two words.

Sensory-motor intelligence and semantic relations in early child grammar

433

REFERENCES Anderson, J. (1968) Ergative and nominative in English. J. Ling., 4, l-32. (1971) The grammar of case. Cambridge University Press. Bloom, L. (1970a) Language development: Form and function in emerging grammars. Cambridge, Mass., M.I.T. Press. (1970b) Why not pivot grammar? J. Speech Hearing Dis. Braine, M. D. S. (1971) The acquisition of language in infant and child. In C. Reed (Ed.), The learning of language, New York Appleton-Century-Crofts. Brown, R. W. (1956) Language and categories. In J. S. Bruner, J. J. Goodnow and G. A. Austin (Eds.) A study of thinking. New York, Science Editions Inc. (1970) The first sentences of child and chimpanzee. In Psycholinguistics: Selected readings. New York, The Free Press. ---(1973) A first language: The early stages. Cambridge, Harvard University Press. Chafe, W. L. (1970) Meaning and the structure of language. Chicago, University of Chicago Press. Cromer, R. F. (1968) The development of temporal reference during the acquisition of language. Unpublished doctoral dissertation, Harvard University. Fillmore, C. J. (1968) The case for case. In E. Bach and R. T. Harms (Eds.) Universals in linguistic theory. New York, Holt, Rinehart and Winston. (1970) The grammar of hitting and breaking. In R. A. Jacobs and P. S. Rosenbaum (Eds.) Readings in English transformational grammar. Waltham Mass., Ginn and Co. (1971a) Types of lexical information. In D. D. Steinberg and L. A. Jakobovits (Eds.) Semantics: An interdisciglinary reader in philosophy, linguistics and psychology. Cambridge University Press. (1971 b) Some problems for case grammar. Georgetown Univ. Mono. Ser. Lung. Ling., 24, 35-56. Gardner, B. T., and Gardner, R. A. (1971) Two-way communication with an infant chimpanzee. In A. M. Shrier and F. Stollnitz (Eds.) Behavior of non-human primates. New York, Academic Press

Gruber, J. S. (1967a) Look and sec. Language, 43 (4), 937-947. (1967b) Correlations between the syntactic constructions of the child and of the adult. Presented at Biennial Meeting of the Society for Research in Child Development. Halliday, M. A. K. (1967, 1968) Notes on transitivity and theme in English. In 3 parts: J. Ling. 3 (1967) 37-81; 199-244; 4 (1968) 178-215. (1970) Language structure and language function. In J. Lyons (Ed.) New horizons in linguistics. Harmondsworth, Penguin. Ingram, D. (1971) Transitivity in child language. Language, 47 (4), 888-910. Lecompte, G. K., and Gratch, G. (1972) Violation of a rule as a method of diagnosing infants’ levels of object concept. Child Devel., 43, 385-396. Lezine, I., Stambak, M., and Casati, I. (1969) Les Etapes de l’intelligence sensorymotrice. Centre de Psychologie Appliqute. Lyons, J. (1966) Towards a ‘notional’ definition of the ‘parts of speech’. J. Ling., 2, 209-236. __ (1968) An introduction to theoretical linguistics. Cambridge University Press. Macnamara, J. (1971) Cognitive basis of language learning in infants. Psychol. Rev., 79 (l), l-13. McNeil& D. (1972) Some signs of language. Unpublished paper, Psycholinguistics Laboratory, University of Chicago. Piaget, J. (1952) The origins of intelligence in children. New York, International Universities Press. (1954) The child’s construction of reality. New York, Basic Books. Roberts, G. C., and Black, K. N. (1972) The effect of naming and object permanence on toy preferences. Child Devel., 43, 858-868. Schlesinger, I. M. (1971a) Production of utterances and language acquisition. In D. I. Slobin (Ed.) The ontogenesis of grammar. New York, Academic Press. (1971b) Learning grammar: From pivot to realisation rule. In R. Huxley and E. Ingram (Eds.), Language acquisition:

434

Derek Edwards

Models andmethods. New York, Academic Press. Slobin, D. I. (1970) Universals of grammatical development in children. In G. B. Flores d’Arcais and W. M. Levelt (Eds.) Advances in psycholinguistics. Amsterdam, NorthHolland.

-

(1972) Seven questions about language development. In P. C. Dodwell (Ed.) New horizons in psychology, Vol. 2. Harmondsworth, Penguin. Thorne, J. P. (1972) On the notion ‘definite’. Found. Lang., 8,562~568. W. M. (1971) The development of behaviour. Harmondsworth, Penguin.

Woodward,

RdsumP

En s’appuyant sur la grammaire des cas et sur les resultats d’dtudes portant sur les premiers langages des enfants, l’auteur a Blabore un systeme de classification semantique des propositions pour decrire les relations exprim&es de facon apparemment universelle dans les enonces de deux mots des jeunes enfants. Ces relations semantiques sont comparees aux concepts invoquees par Piaget pour d&ire l’intelligence sensori-motrice. En particulier aux concepts d’objet permanent et de ses relaions spatiales, aux concepts en dualite de

persomres en tant qu’objets physiques et d’etres agissants et au role des personnes en tant que cause de changement dans la localisationdes objets. On trouve une correspondance nette qui permet d’avancer que la nature de l’intelligence sensori-motrice determine Ctroitement le champ des relations exprimees par l’enfant, jusqu’a et y compris les notions de relations de possession entre personnes et objets, celles d’attribut des objets et l’utilisation par l’enfant de verbes qualifies d’ ‘experientiels’.

3

Acquiring

Concepts and words in the IS-month-old: concept names under controlled conditions*

KEITH

E. NELSON

JOHN

D. BONVILLIAN

Stanford University

Abstract The young child’s use of available evidence in acquiring words was examined Iongitudinally. In comparison with previous work on early word meaning, two essential departures in design were employed - word use by the children was exhaustively detailed in an experimentaI setting, and this setting permitted systematic control and limitation in patterns of referent object encounter and of adult naming for the objects (toys). Although the children (n =lO) were presented relatively few examples from which to Iearn word and concept meanings, they succeeded in learning to use many of the words appropriately. Variations in the course of learning for dyerent children are discussed, with special attention to the ways in which nonverbal action or mother-child interaction could have injIuenced the child’s progress. The power of words to evoke meaning lies at the heart of human communication through language. As adults we take this semantic power for granted, but we recognize the achievement marked by the child’s first meaningful words. Yet how these first words or later words acquire meaning for the child is a question which has attracted little research to date, notwithstanding the recent upsurge in research on syntactic aspects of language acquisition. The limited data which are available on the meanings of the child’s early words come primarily from a few subjects, all studied within an observational, naturalistic approach, but by a diverse set of researchers with diverse methodological and theoretical predilections. Neither the observations nor interpretations are easily wedded to form generalizations, but on one point most investigators * Preparation of this report was supported in part by NIH Grant 1 R03 MH19826-01Al. Thanks are due Jeffrey Allyn, Donald Steele, Terry Taylor and the mothers of the ten children. The second author was supported, when

this study was conducted, by a National Science Foundation Graduate Fellowship. Reprint requests should be sent to Keith E. Nelson, New School for Social Research, 65 Fifth Avenue, New York, New York 10003, U.S.A.

Cognition 2(4), pp. 435-450

436 Keith E. Nelson and John D. Bonvillian

have agreed (e.g., Moore, 1896; Leopold, 1939, 1949; Lewis, 1957; Werner and Kaplan, 1963; McNeill, 1970): Many words do not initially mean for the child what they mean for an adult. Conformity to adult usage is only gradually achieved. The most impressive documentation of this point is provided by Leopold’s accounting of word use by his daughter, Hildegard, in her second year (Leopold, 1939, 1949). An example is Hildegard’s use of ‘wau-watt’ to refer to dogs, a stone lion and horses, as well as to request her own baby pictures. If we consider word meaning in the sense of the referential category underlying naming, then we can use the young child’s changing patterns of naming to trace the changing meanings of words. When horses and lions as well as dogs are named ‘dogs’ by the child, we know that the meaning of the word differs for the child and adults ; the child’s use of the word appears to be based upon a systematic, but erroneous, hypothesis about the word’s meaning in the language. As this hypothesis is modified, the meaning of the word for the child as well as the child’s pattern of word use will change. But what evidence leads a child to adopt an initial hypothesis about a word’s meaning and what further evidence causes the child to revise this hypothesis? The observational accounts are virtually silent on these questions, as they provide only very scattered information on the verbal or nonverbal evidence available to the child. In the case of examples of word use available to the child, we don’t know how often the child heard individual words used by others or what the reterents were for such naming. Similarly, in the case of the child’s experience with objects which may underlie formation of a concept, few accounts have been given of which referents are encountered or how often they are encountered. The child faces problems of concept formation and concept naming, but how the child solves these problems may well be obscure until it is determined how the child responds to relevant evidence. In the present study we explored this question by examining early word acquisition under conditions where evidence available to children was controlled. We limited the set of evidence quite sharply, controlling how often the child encountered objects and names for the objects. The child’s progress in learning to apply the words to appropriate referents was charted.

1. Subjects The subjects children and neighboring 17.5 months (boys=16.5

were ten white, English-speaking children (five boys and five girls). The their mothers resided in the predominantly middle-class communities Stanford University. The subjects ranged in age from 16.2 months to at the commencement of the study, with a mean age of 16.5 months months, girls =16.4 months). The distributions of family size and birth

Concepts and words in the l&month-old

437

order within the sample of five girls were identical to the corresponding distributions for the five boys.

2. Method The children were observed regularly in a six-month longitudinal study of their acquisition of words tar different experimental objects. The objects presented to the children were selected as appropriate referents for sixteen words, with six objects serving as examples for each word. The children’s production of the sixteen words was examined in relation to systematic variations in their mothers’ patterns of object presentation and naming. First, in order to determine whether the children would extend their use of the words to examples that they had never heard named, for each word half of the referents were named and half were not named by the mothers. We also studied whether the acquisition of an experimental word would be facilitated by having all the referents for a particular experimental word appear consecutively, or, conversely, whether alternation of objects with different names would enhance discriminability, in turn leading to more rapid concept formation and word acquisition. In addition, measures of the children’s object manipulations and measures of the mothers’ behaviors were correlated with the children’s progress in learning to appropriately apply the names to the experimental objects.

3. Experimental words and objects Sixteen nouns were selected as the experimental words: Barrel, bobber, caboose, canteen, compass, eyebolt, handcuffs, hedgehog, nozzle, oiler, pulley, sifter, silo, sinker, snorkel and whetstone. The experimental words were chosen to meet the following criteria: (1) No child in the sample had used any of the words at the start of the experimental sessions; (2) referents for the words either did not appear in the homes (typically) or the parents agreed not to speak the words or use a referent of the words in the child’s presence; and, (3) to aid scoring of word use, each of the experimental words was two syllables long, and there was no duplication of syllables in the set of sixteen words. In choosing the six particular objects for each experimental word, we considered the degree to which objects differed. Sets of objects were chosen so that a ‘variability’ score of twelve - based on one point for each distinct difference (rated by two observers) within a set in colors, sizes and shapes - was achieved for each set of six examples. Eight experimental words were randomly assigned to each of two groups which

438

Keith E. Nelson and John D. Bonvillian

differed in methods of word and object presentation. In a low-contrast group all six objects serving as referents for a particular word were presented consecutively (for example, bobber,, bobber,, bobber,, bobber,, bobber,, bobber6; oilerI, oiler,, oiler,, oiler4, oiler,, oiler,). In the high-contrast group, objects serving as referents for different experimental words were alternated in pairs (for example, sifter,, sifter,, pulley 1, pulley2, sifters, sifter,, pulley,, pulley,, sifter,, sifter,, pulley,, pulley,). Three objects from each set of six objects per word were randomly assigned to a ‘named’ category, and the remaining three objects were assigned to an ‘unnamed category. The named objects were named twice by the mother each time the object was presented. The unnamed objects were never named in the experimental sessions. Half of the experimental words and objects (eight different words, 48 corresponding objects) were presented in each of the experimental sessions; the remaining half were presented in the immediately following session. Thus, two experimental sessions were required before the entire set of sixteen words and a complete ‘series’ of 96 objects were presented. The order of presentation of the objects and words in each series was randomized within minimal constraints.

4. Procedure The children were observed about once every other week. All ten subjects completed ten biweekly sessions, and nine subjects completed an additional four sessions. Two weeks prior to the commencement of the experimental sessions the mothers and children visited the comfortably furnished experimental room, so that the children became familiar with both the room and the experimenters. Each mother was instructed to employ whatever method she judged the most natural to draw her child’s attention to the objects and teach him the appropriate words for the ‘named’ experimental objects. One constraint applied to the mothers’ behavior concerned ‘named’ and ‘unnamed’ objects; mothers named only every other object in each session - for example, the first, third and fifth objects in each session were ‘named’ objects and each was labelled twice, but the second, fourth and sixth objects were not named. The only other major constraint imposed on the mother’s behavior was that she provide no corrective feedback. The experimenter handed one of the objects directly to the mother about every 45 seconds and so an experimental session lasted approximately 40 minutes. The experimenter recorded each child’s use of a name for a referent object along with the context in which the naming occurred. Measures of the child’s appropriate and inappropriate uses of the sixteen experimental words were calculated only for two-syllable utterances in which each syllable had a phonological resemblance to

Concepts and words in the I%-month-old

439

correct adult pronunciation. Besides the experimental words, other words which the child applied to the sixteen experimental objects were also recorded. The audio tapes of sessions 9 and 10 (recorded on a Uher 7000 recorder) were transcribed in order to obtain measures of the mothers’ language use. The mean length (MLU) in words and total number of utterances provided estimates of the complexity and total amount of the mothers’ speech. Two measures of the mothers’ attempts to structure the interaction were scored - requests that the children look at the experimental objects and requests that the children produce the objects’ names. Finally, the tapes were analyzed for signs of the mothers’ approval (e.g., ‘That’s right’, ‘Good’) and disapproval (e.g., ‘No’, ‘ That’s wrong’) of the childrens’ behavior. This latter analysis resulted in three variables: Positive reinforcement for use of the experimental words, positive reinforcement for speech exclusive of reinforcement for correct naming, and positive reinforcement for object manipulation. (Signs of disapproval, as we intended, were very infrequently given by the mothers.) The stability of these behaviors was estimated by correlating them with the same measures obtained from transcription of session five tapes. The correlations were as tollows: Mean length of utterance, r = .772; total number of utterances, r = .853; requests for looking, r = .430; requests for name, r = .981; positive reinforcement for naming, r = .305 ; positive reinforcement for speech, r = .634; and, positive reinforcement for object manipulation, r = .773. These correlations indicate consistent verbal interaction over a period of several months by the mother, with the exception of requests for looking and reinforcement for naming - and these exceptions might be a result of changes in the childrens’ behavior rather than the product of changes initiated by the mother. In each of sessions 9 and 10 an observer, watching through a one-way mirror, made detailed records of the child’s object manipulations. Object manipulations in the following categories were scored: Mouths, shakes, bangs, throws, drops, rolls, squeezes, moves moveable parts, dangles, blows into, looks through, drinks from, holds collection of objects, bangs objects together, inserts one object into another, stacks and groups objects spatially. Reliability of object manipulation scoring was calculated for four independent scorers who observed session 8. The average percentage agreement for all categories combined was 90.5 %, with a range across categories from 76.9 % to 100.0%. Interscorer agreement averaged 91.2x, with a range from 85.7% to 98.1%.

5. Results and discussion Several overall findings take priority over details of the children’s progress in word

440

Keith E. Nelson and John D. Bonvillian

acquisition. The first overall finding is that the children did, indeed, learn to use many of the experimental words appropriately. Naturally this finding makes possible some detailed analyses, but more importantly it shows that the evidence presented to the children was sufficient to allow them to recognize objects as concept examples and to label concept examples with appropriate words (‘hedgehog’, ‘sifter’, etc.). It is important to note that the children’s learning was based upon relatively limited data - for each class of objects (e.g., sifters) the children encountered referent objects about 30 times (six examples, once in each of five series) and heard adults label referent objects 30 times (three named examples two times in each series) by the end of the fifth presentation series. The children, on the average, used the experimental words 35.6 times and correctly labelled 17.8 of the 96 objects by the end of the fifth session. Progress was more rapid for the 48 objects named by adults than for the 48 ‘unnamed’ objects. By the end of series 2, 3, 4 and 5, respectively, an average of 2.1, 5.3, 8.8 and 12.1 ‘named’ objects had been correctly labelled by the children. But corresponding figures were only 0.6,2.5,4.3 and 5.7 for the unnamed toys. These means mask some interesting variation across children and across object-classes, topics we will explore below. Nevertheless, the means indicate that rapid mastery of some new words on the basis of limited evidence was achieved by at least some of these young children. A more differentiated analysis can begin at the end of the second presentation series. By this point in time sufficient word use by the children was observed to permit variance analyses of the children’s naming. Accordingly, analyses of variance were performed for the data available at the end of sessions 2,3,4 and 5. For each point in time, two dependent variables were employed - up to the point of analysis, the number of different objects appropriately named by each subject and the total number of times the experimental words had been used appropriately by each subject. Consider first the number of objects named. At the end of the second series the performance by the girls for named referents was above the performance by boys for named and unnamed referents and by girls for unnamed referents. This pattern held for the children’s

Table

I.

Mean number of objects named by end of series

Series Series Series Series

2 3 4 5

Objects named by adults

Objects not named by adults

Girls

Boys

Girls

Boys

3.8 9.0 14.0 19.6

0.4 1.6 3.0 4.6

1.0 2.8 6.0 8.4

0.2 2.2 2.6 3.0

Concepts and words in the l&month-old

441

naming not only at series 2 but also at the end of presentation series 3, 4 and 5, as Table 1 shows. At each point in time after series 2 the girls had much more frequently labelled those objects named by adults than those objects unnamed by adults, in each case F (I$) >9.03, p < .02. At the fifth series only, boys (as well as girls) had labelled adult-named referents significantly more frequently than adult-unnamed referents F (1,8) = 17.72, p < .Ol. Girls showed significantly more naming than boys for adult-named objects at the end of each series (F (1,8) >8.41, p < .02) and for adult-unnamed referents at series 5 (F (1,8)=201, p< .OOl). The interaction of children’s sex by adult naming was significant in the analyses of performance at the end ot series 2, 3, 4 and 5, (in each case 5.70
Mean instances of word (appropriate name) use by end of series

Series Series Series Series

2 3 4 5

Objects named by adults

Objects not named by adults

Girls

Boys

Girls

Boys

4.4 12.2 24.8 45.0

0.4 2.0 3.8 6.2

1.2 4.4 10.4 16.6

0.2 2.4 2.6 3.4

6. Correlations with mothers’ behavior Mothers who relatively often reinforced their children’s naming tended to be mothers

442

Keith E. Nelson and John D. Bonvillian

whose children were relatively far along in use of the names for the experimental objects. This correlation, r = .70 (df= 8), was the only statistically significant association observed between the mothers’ behavior and the children’s naming or amount of object manipulation. However, within the set of correlations reported in Table 3 there is a common pattern of interest. Mothers’ reinforcement for their child’s naming, mothers’ requests that the child name an object and mothers’ total number of utterances share the following pattern - positive correlation (I = .45 to .70) with both measures of the children’s progress in naming and negative correlation (r=-.40 to -.61) with the amount of object manipulation by children. One interpretation of Table 3.

Correlations between measures of mothers’ verbal behavior and children’s naming and manipulations of objects Measures of mothers’ verbal behavior All utterances Mean Total length number

Frequency of name use by child .19 Number of objects named by child .18 Frequency of object manipulations by child .03

47 .47

-.40

Requests that child: Examine Name toy toy

Positive reinforcement for child’s: Naming Toy manipSpeech ulations

.02

45

.70

.19

-.12

.47

.54

.ll

-.09

-.Ol

-.40

-.40

-.61

-.30

.Ol

this set of findings is that children who manipulated objects a lot rather than a little tended to name the objects infrequently rather than frequently and also tended to elicit from their mothers relatively few utterances. Furthermore, these same children elicited fewer requests for object names and fewer reinforcements of such naming occurred. On the other hand, the findings may largely reflect a chain of influence in which mothers who spoke frequently and specifically requested and rewarded objectnaming by the child tended to engage the child’s attention relatively more toward speech and object-naming and relatively less toward object manipulation. Only studies in which adult styles of interaction with children are experimentally varied can confirm either of these lines of interpretation. When such studies are contemplated, it would be wise to note three maternal variables which in this study had negligible association with the children’s levels of naming or object manipulation - mothers’ reinforcement for object manipulation, mothers’ reinforcement for verbal behavior other than object-naming and the mean length of the mothers’ utterances.

Concepts and words in the 18-month-old

443

7. Manipulating objects and naming objects Two more analyses, in addition to the correlational analyses of maternal behavior, relate to the question of what factors influenced how quickly words were learned. These two analyses concern how well objects were discriminated in action, the basic premise being that the more successfully any class of objects is discriminated from other classes of objects, then the more quickly the child might learn the concept and the name fitting the class. The definition of how well object-classes were discriminated rested upon how many of the objects were acted upon in a class-specific way: For each subject it was determined how many of the 96 objects in the sixteen object-classes (silos, hedgehogs, etc.) were acted upon in a way which was uniquely restricted to the objects within a class. This information, derived from the observation sheets for the fifth series, entered two analyses. First we asked if subjects who showed relatively high discrimination in action between object-classes were relatively advanced in word use at the same point in time that action-discrimination was assessed, i.e., the end of the fifth series. This relationship held, but only quite weakly - correlations (r’s) of object discrimination scores for the subjects were .27 with number of different objects named and .31 with total frequency of experimental words use. The second question we asked was whether object-classes averaging high action discrimination scores across children were the classes which received (by the end of series 5) the most naming by the children. Surprisingly, just the reverse was true - object-classes with high objectdiscrimination scores were generally the classes which had elicited relatively low frequencies of naming (r= -.54, p< .05). One possible basis for this outcome is that the children, rather than using action upon the objects as a means of learning about object characteristics and thus about classes underlying name use, instead may have tended to either act or to name the objects. In this sense, a ‘bobber’ might at first be treated as something to name or as something to throw, but not as both. For 17- to 24-month-olds, spoken words sometimes may serve as actions having few special priorities or characteristics relative to other actions, and actions of all kinds may serve in the process of ‘sorting’ objects into classes based upon characteristics (they roll, they are round, they receive ‘bobber’ as a name). Similar arguments for the role of early words as action have been made elsewhere (Sinclair, 1971; Werner and Kaplan, 1963). In the present study, the view that naming may been an alternative to other forms of action fits the finding that children with low frequencies of object manipulation tended to be children with high frequencies of appropriate naming, r = -.85, p < .002. Similarly, as suggested above, this view fits the finding that object-classes poorly discriminated in action tended to be the classes most frequently named. The children’s actions toward canteens and sifters help to

444 Keith E. Nelson and John D. Bonvillian

illustrate these patterns. The distinctive act most often applied to canteens was to unscrew the cap (lid), while sifters (as for sifting flour) were mechanically operated by squeezing or turning a handle. Each of these distinctive actions was more frequent than the distinctive actions for any other object-classes, but the names ‘sifter’ and ‘canteen’ were among those most rarely used by the children. It should be noted also that the sex differences observed in this study are consistent with the idea that naming and other actions often may have served as alternatives for the young child - in general, the boys, relative to the girls, manipulated the objects more frequently but named them less frequently.

8. Concepts and over-generalizations

in individual children

In trying to characterize the process by which children learned to use the words which appropriately labelled the sixteen experimental concepts, much can be learned by moving group data to examine all instances of concept acquisition by the children. In the course of such examination, across all presentation series available for each child, we come to some additional information not yet considered - the extent and nature of the children’s generalization of the words beyond the ‘appropriate’ conceptual boundaries employed by adults. We take as our minimum criterion for concept acquisition that a child appropriately labelled at least two of the six objects which exemplified the word. By this criterion, the mean number of concepts acquired was 8.3, out of a possible sixteen, and the range was from two to sixteen. Thus, all ten children and all sixteen concepts are involved in the subsequent descriptions of various paths to concept acquisition. Adult-unnamed as well as adult-named concept examples typically were included, as in most instances (80%) of concept acquisition the child named at least one adult-unnamed concept example. The following figures more fully document the children’s extension of the concept names to some examples adults never named - on the average, for 7.7 of the sixteen experimental concepts the children appropriately applied the concept label to at least one unnamed example, for 2.8 concepts at least two unnamed examples were appropriately labelled and for 1.2 concepts the children appropriately named all three unnamed examples. The child’s labelling of these adult-unnamed examples is especially strong evidence of concept acquisition, but we have chosen the broader criterion just stated because the order in which the child learned to apply a word across both adult-named and adult-unnamed examples is intormative about the course of concept acquisition. In acquiring the conceptual meaning adults attach to an experimental word, the child may take a direct and error-free route. That is, the child may appear to adopt

Concepts and words in the I8-month-old

445

the ‘correct’ (adult) hypothesis from the start and therefore avoid any errors of overgeneralization. For example, ‘compass’ would be applied to compasses only and no other experimental words would be applied to compasses. Such error-free histories of word use accounted for 69% of all instances of concept acquisition. As would be expected, the child typically began using a word by correctly naming one of the toys the adult had labelled with the word. But in about 20% of the cases of error-free learning, the first of the toys named by the child was a toy never named by adults. This finding establishes that children often formed an hypothesis about the conceptual meaning of a word before directly or indirectly imitating adult labelling of particular objects. Further evidence to this effect is provided by errors in naming. These errors, considered next, certainly must have been based upon some idiosyncratic hypothesis by the child rather than upon imitation of what adults said in reference to the particular objects. Most children extended few experimental words beyond the conceptual boundaries employed by adults. As examples, ‘compass’ in general was restricted to compasses and ‘hedgehogs’ to hedgehogs. When over-generalizations did occur, they were usually narrowly based and idiosyncratic. The idiosyncratic nature is readily shown - within all inappropriate pairings of names and object-classes, only one such overgeneralization was shared by more than one child, this one instance being the labelling by two children of a toy barrel (plastic, orange) as a ‘caboose’ (a word previously applied by these children to plastic cabooses of yellow or orange hue). That the children’s over-generalizations typically had rather narrow bases is indicated by the following observations : 1) When a child over-generalized a word, the word was most commonly overgeneralized to just one object-class and to just one of the objects within that class; 2) when an object-class was inappropriately named by a child, only one inappropriate name usually was applied to the class; 3) when individual objects are considered, it is clear that the child in most cases (66 %) applied over-generalizations to objects which he had not yet named appropriately. With regard to the last observation above, an important related observation must be discussed. This is the fact that even though a particular mislabelled object (say, a red barrel) typically had not been labelled appropriately (‘barrel’) before, nevertheles, other objects (say, yellow or brown barrels) within the same class usually had been labelled appropriately (‘barrel’) prior to the mislabelling. Taken together with the above observations, the most common form of over-generalization can be imagined easily if any two classes of objects are considered, such as pulleys and bobbers. At the time an over-generalization occurs, each of the words has been used to refer to some but not all of the six appropriate examples for the word. When one of the objects, say

446

Keith E. Nelson and John D. Bonvillian

a green bobber, is mislabelled with the label for the second class (‘pulley’), then an error in drawing the adult boundary between the classes occurs. Actually this error is two errors - the overextension of one class boundary and the underextension of the other class boundary. Further support for this characterization of the usual form of ‘conceptual boundary dispute’ in this experiment is provided by data indicating that in most cases (90 %) the overextension of a label to a class of objects was mutually exclusive (as noted above) of that class receiving mislabelling with any of the other experimental words. When a child learned to name all examples for a word, several series over a period of weeks nearly always were involved. If we take as starting point the first series in which a child appropriately named one of the objects in a cIass (e.g., hedgehogs), the median number of series before the child had named the last of the six objects within that class was two. The idiosyncratic nature of concept acquisition has been discussed above in terms of over-generalizations of the concept labels. The idiosyncratic nature of the children’s learning can be seen also when the successful instances of concept IabeIling are considered in detail If we look at all the instances in which two or more children acquired the same concept, we can ask for each child which of the six objects in a class were named and in what order, and, given this information, we can ask how frequently two or more children showed identical patterns of learning to apply the class name across the class members. There were three instances in which two (never more) children followed precisely the same pattern of learning to name members of a concept. In all other cases, each of the children who learned a concept followed a unique pattern of learning.

9. Bases for over-generalizations If a child mislabelled an object, what features or uses of the objects led to such oveigeneralization? We shall consider that question both for over-generalizations of the sixteen experimental words and for words learned outside the lab but extended inappropriately to the experimental objects. Some of the mislabellings suggest several possible bases. As examples, mislabellings of barrels as ‘bobbers’ or ‘cabooses’ may have been based upon hollow plastic construction as well as size and color similarities, and ‘clock’ applied to compasses may have involved both shape and a mistaken idea of function. Other over-generalizations would seem interpretable in terms of one rather than many features - the ‘roll-able’ wheels on a caboose called ‘pulley’ (of course with a wheel) or ‘compass’ (wheel-shaped and rollable), the long-tubed snorkel called ‘oiler’ (also long-tubed), the round sinker (despite color and texture and weight

Concepts and words in the I8-month-old

447

differences) mislabelled ‘bobbers’ (all round). Size, in a special sense, was involved when the smallest of the barrels or the smallest of the hedgehogs was labelled ‘baby’. The wide variation in bases for over-generalization which these examples suggest could be documented by many additional examples, but we will consider just two. First, there is evidence that shared function was a factor in over-generalization. Shared function could be assessed from session 5 records of object manipulation, and these records make clear that if a child labelled two or more classes of objects in common with one of the experimental words, then the child typically also applied one or more actions in common - such as rolling the ‘wheels’ on both pulleys and cabooses. A second example is ‘ball’, a word widely applied by many children, which may have often referred to the handiness of most of the toys for throwing. But another, and broader, possible basis may have been that ‘ball’ was being used as a superordinate term meaning ‘toy’. Finally, there were some idiosyncratic over-generalizations - such as ‘birdie’ to sifter and ‘compass’ to a very large silo - which to adult experimenters made no sense in terms of function, shape, color, texture, density or any other easily identified characteristic. Since the experimental sessions were but a small part of the child’s linguistic experience, it was by no means surprising to find that experimental objects were often mislabelled (as mentioned above) with words the children learned elsewhere. Such over-generalization did not slow progress toward mastery of the experimental words; on the contrary, children who most frequently used the experimental words appropriately also most trequently labelled the objects with non-experimental words (r = .85, p < .002). This latter finding may reflect the varying tendency of different children to actively seek matches between nonverbal concepts and words (experimental or non-experimental). The conditions of the present experiment may have provided more systematic exposure to relevant examples than young children usually experience when learning a new word. Both the extent of over-generalizations and the bases for over-generalization could have been affected by this factor. More particularly, the child’s attention to relevant teatures underlying a concept might well have been heightened under the structured conditions of the present study as compared with conditions in a child’s home, and, as a result, relatively narrow bases for over-generalizations and also relatively few over-generalizations may have occurred in this study. However, the validity of this interpretation cannot be determined until clear accounts are given of the conditions typically prevailing when children learn new words and their underlying concepts.

448 Keith E. Nelson and John D. Bonvillian

10. On diverse routes to concept mastery The present findings together with related previous observations indicate considerable diversity in the young child’s concept formation and concept labelling (cJ Katherine Nelson, 1973b). We can discern four general routes a child may take in acquiring the meaning adults attach to a word. The child may adopt the ‘correct’ hypothesis from the start, applying the word appropriately to new examples and showing no errors in the use of the word. This error-free route has been given minimal attention in previous work on early word meanings, but it could well be the most common; under the present conditions, with relatively tew examples from which to learn, it accounted for over two-thirds of all cases of concept acquisition for the experimental words. Another route in acquiring word meaning is initially to shift hypotheses unsystematically, focusing first on one feature and then another, basing each hypothesis on a feature or features shared by only a limited subset of the referents appropriately named by the word. Such drifting of hypotheses seems to characterize well many of the examples given in naturalistic observations (cJ Leopold, 1939) of word acquisition and some of the overgeneralization of the experimental words in the present study may also fit this pattern. A third alternative is that the child’s hypotheses may shift systematically from initial naming, based only upon the most general features of a word’s referents - as when ‘dog’ is applied to all animals similar in overall shape features - to naming progressively based upon more and more specific features of a referent, including finally the specific features uniquely characteristic of the referent (e.g., a dog’s bark). General-to-specific shifts of this sort indeed appear to fit some cases of word acquisition in young children (Clark, 1973 ; Donaldson and Wales, 1970). A final path preceding mastery of conventional word meaning is for the child to initially use the word in a consistent but overrestricted manner, basing naming upon very specific features or information. Thus ‘dog’ might be applied only to dogs of a certain size and color, or, as Katherine Nelson (1973a) has shown,‘ball may be applied to objects that function like balls even if they are not shaped very much like balls. Why a child chooses one of these routes to word acquisition over another route for any particular word cannot be answered from current findings. We know that the error-free route was common under the conditions of the present experiment, but the success of the children in avoiding over-generalization of experimental words left little evidence to trace for routes to mastery that begin with error. And previous research, despite its emphasis on the child’s errors, has not related the evidence available to the child to the child’s progress in word acquisition. Thus, the roots of the child’s non-adult hypotheses about word meaning will havt: to be tracked down in future work. It is our speculation that patterns of evidence - referents encountered and adult examples of naming encountered - will be crucial in influencing

Concepts and words in the 18-month-old

449

the child’s choices 01 hypotheses. At least we can see no compelling leasons that small differences in available evidence might not lead the child to different hypotheses, since in the present study - in which all children received the same evidence - a small amount or evidence was a sufficient foundation for error-free learning of concepts and concept names.

11. Conclusions Three general conclusions emerge from this study. First, 1%month-olds can learn new concepts and the words which label the concepts on the basis of quite limited evidence - very few encounters with objects fitting the concepts, very few adult demonstrations of how the objects should be named and virtually no corrective feedback. A second conclusion is that the particular routes to word acquisition vary markedly across children. That is, different children used the same available evidence in different ways, a conclusion supported both by the ways in which over-generalizations varied for different children and by the various paths to concept mastery when no errors of over-generalization occurred. A final conclusion is also a question; some children learned to label many more concepts than other children, and the bases for such sharp individual differences is a question wide open to future inquiry. The answer could well lie in how nonverbal actions and words (verbal acts) are related in early thinking, a topic touched upon only briefly in this or previous research. REFERENCES Brown, R. W. (1973) A first language. Cambridge, Harvard University Press. Clark, E. V. (1973) What’s in a word? On the child’s acquisition of semantics in his first language. In T. E. Moore (Ed.) Cognitive development and the acquisition of language. New York, Academic Press.

Donaldson, M., and Wales, J. (1970) On the acquisition of some relational terms. In J. R. Hayes (Ed.) Cognition and the development of language. New York, Wiley. Leopold, W. F. (1939, 1949) Speech development of a bilingual record. Northwestern

child:

A

University

linguist’s

Press, Evanston, Illinois. Lewis, M. M. (1957) How children learn to speak. London, Harrap. McNeil, D. (1970) The acquisition of language.

New York, Harper and Row. Moore, K. C. (1896) The mental development of a child. Psychol. Rev. Mono. SuppI., 1 (3).

Nelson, K. (1973a) Some evidence for the cognitive primacy of categorization and its functional basis. Merrill-Palmer Q., 19, 21-39.

(1973b) Structure and strategy in learning to talk. Mono. Sot. Res. child. Devel., 38 (l-2, Serial No. 149). Sinclair, H. (1971) Sensorimotor action patterns as a condition for the acquisition of syntax. In R. Huxley and E. Ingram (Eds.) Language acquisition: Models and methods. New York, Academic Press. Werner, H. and Kaplan, B. (1963) Symbol formation. New York, Wiley. -

450 Keith E. Nelson and John D. Bonvillian

RdsumP Dam une etude longitudinale, on a cherche a voir comment le jeune enfant utilise, dans l’acquisition des mots, les don&es dont il dispose. Dans son organisation, cette etude &carte sur deux points essentiels des travaux anterieurs sur la signification des premiers mots: dune part l’utilisation des mots par l’enfant depend exclusivement du cadre experimental, d’autre part le cadre permet de controler et de limiter systematiquement l’apparition des objets de reference et la denomination par I’adulte de ces objets (jouets). Bien

que les enfants (n = 10) n’aient et6 confront& qu’a un nombre relativement limit6 d’exemples a partir desquels apprendre la signification des mots et des concepts, ils sont parvenus a acqutrir l’utilisation adequate dun grand nombre de ces mots. Les auteurs analysent les differences entre enfants en tours de l’apprentissage et, plus particulibrement, en quoi l’action non-verbale ou l’interaction mbreenfant peuvent avoir une incidence sur les progres de l’enfant.

4

Dative

questions:

A study

to grammaticality

in the relation

of acceptability

of an English

D. TERENCE

sentence

type*

LANGENDOEN

City University of New York

NANCY

KALISH-LANDON

L 0 yola University

JOHN

DORE

City University of Ne w York

Abstract Generative grammarians have contended that English sentences of the type Who(m) did you give the book? (what are here called ‘dative questions’) are ungrammatical. The incorporation of the necessary restrictions in the grammar of English to account for this, however, requires a weakening of linguistic theory. It would be desirable, therefore, to account for the restriction within performance theory, as has been proposed by Jackendofland Culicover (1971). Theirparticular account isshown here to be inadequate. In the course of trying to devise a better account, wefound, by two diflerent questionnairetype experiments, that some English speakers, all from metropolitan New York City, accept dative questions. On the basis of this finding, we theorize that the observed variation in acceptability of dative questions is best accountedfor by d$erences in the perceptual strategies for determining the grammatical relations in perceived clauses that different populations of English listeners use. There are thus no dialect d@erences, strictly speaking, for dative questions: they are all grammaticalfor all English speakers.

1. Grammaticality

and acceptability

It is not always easy to determine language, for at least three reasons. are unacceptable do not guarantee

the grammatical status of expressions in a given First, speakers’ judgments that certain expressions that they are ungrammatical, since there may be

* This work was supported in part by a faculty research award of The City University of New York. 1. We use the term ‘language-independent’ throughout this paper specifically to mean ‘independent of the rules that generate the

sentences of any given language’. The principles are certainly not to be considered ‘independent of language itself’, since they are intended to be the rules according to which linguistic structures are perceived.

Cognition 2(4), pp. 4SI-478

452

D. Terence Langendoen, Nancy Kalish-Landon and John Dore

language-independent’ perceptual principles that make certain perfectly grammatical sentences seem ungrammatical (Chomsky, 1965; Bever, 1970; Kimball, 1973). Second, the converse is also true: Judgments that expressions are acceptable do not guarantee that they are grammatical (Otero, 1972; Langendoen and Bever, 1973). Finally, there are many expressions for which speakers do not give consistent acceptability judgments; such inconsistency has misled many linguists into postulating the existence of ‘dialects’ for which there is often no geographic, socioeconomic or other languageindependent basis (Labov, 1972). In this paper, we consider a type of English sentence which has, until very recently, been considered ungrammatical because the various linguists who have studied it have all found it unacceptable. We question the decision to label this type of sentence ungrammatical; first because it appears to lead to an ad-hoc complication of the rules of English syntax, second because there are many English speakers who find the sentence-type in question to be acceptable and third because there seems to be a language-independent perceptual principle that accounts for both the unacceptability of the sentence type and the variability found in those acceptability judgments.a

2. The interaction of the Dative transformation formations in English

with various other movement trans-

There is a well-known syntactic transformation in English that puts the underlying indirect object into the position of the direct object; that is, immediately following the verb. The characteristic preposition of the indirect object (to or for in most cases) is then deleted (whether by the same rule that performs the movement of the indirect object or by a later rule of Preposition Deletion is of no concern to us here). We call this movement rule the Dative transformation (Fillmore, 1965; Ross, 1968; Kuroda, 1968; Klima, 1970; Fischer, 1971; Jackendoff and Culicover, 1971; Edmonds, 1972 all treat the Dative transformation in some detail). According to this transforwe may derive the sentences mation in some detail). According to this transformation, in (2) from the structures that also underlie those in (1). (1) a. The traveler gave documents to the clerk b. My grandmother bought a radio for my sister (2) a. The traveler gave the clerk documents b. My grandmother bought my sister a radio We now consider sentences in which the underlying indirect object is moved to a new position in the sentence; first those in which Dative has not applied, and second 2. We say ‘seems to be’ rather than ‘is’ because we have not carried out empirical investiga-

tions on the principle in question. See below, Section 6.

Dative questions

453

those in which Dative has applied (the asterisk ‘*’ indicates those sentences which are unacceptable to the linguists that have studied sentences of this type). (3) a. i. *The clerk was given documents to by the traveler ii. *My sister was bought a radio for by my grandmother b. i. The clerk was given documents by the traveler ii. (*) My sister was bought a radio by my grandmother3 (Passive) (4) a. i. Harriet is tough to write letters to ii. Harriet is tough to buy clothes for b. i. *Harriet is tough to write letters ii. *Harriet is tough to buy clothes (Tough Movement) (5) a. i. It’s Harriet (that) I gave the watch to/ It’s Harriet to whom I gave the watch/ It’s to Harriet (that) I gave the watch ii. It’s Harriet (that I bought the watch for/ It’s Harriet for whom I bought the watch b. i. *It’s Harriet (that) I gave the watch ii. *It’s Harriet (that) I bought the watch (Clefting) (6) a. i. These people, I wouldn’t send a penny to ii. Elsie, I wouldn’t buy anything for b. i. *These people, I wouldn’t send a penny ii. *Elsie, I wouldn’t buy anything (Topicalization) (7) a. i. This is the person (that) Selma sold the car to/ This is the person to whom Selma sold the car ii. Do you know the person (that) I made this dress for?/ Do you know the person for whom I made this dress? b. i. *This is the person (that) Selma sold the car ii. *Do you know the person (that) I made this dress? (Relativization) (8) a. i. Who(m) did you give this book to?/ To whom did you give this book? ii. Who(m) did you make this dress for?/ For whom did you make this dress? b. i. *Who(m) did you give this book? 3. Fillmore (1965) judges all sentences like (3bii) unacceptable. Jackendoff and Culicover (1971, p. 400) hold that such sentences vary in acceptability, depending on whether or not the indirect object comes to ‘have’ the direct object. Thus for them (3bii) would be accept-

able, but not (i). (i) *My sister was played a tune by my grandmother The issues raised by this difference of opinion on acceptability are interesting but not within the scope of this paper.

454

D. Terence Langendoen, Nancy Kalish-Landon and John Dore

ii. *Who(m) did you make this dress? (Question Formation) According to the judgments just given, the interaction of Dative and Passive is the opposite of that of Dative and the other movement transformations considered. Dative must apply if the underlying indirect object is to undergo Passive; it must not apply if the indirect object is to undergo Tough Movement, Clefting, Topicalization, Relativization or Question Formation .4 It is easy to see why failure to apply Dative leads to unacceptability when the indirect object is made subject by Passive: Such sentences are ungrammatical because Passive can only move the noun phrase of a

preposition phrase into subject position if the preposition phrase immediately follows the verb (Jackendoff and Culicover, 1971, p. 398). What is not so easy to see is why the examples in (3b) are acceptable whereas the (b)-examples in (4)-(8) are not. One possible explanation makes use of a fundamental formal difference between Passive and the other movement transformations under consideration. Passive is not an unbounded movement rule; the others are. 5 One could then claim that what makes the (b)-examples of (4)-(8) unacceptable is their ungrammaticality, and that they are ungrammatical because their derivations violate the principle that no indirect object that has undergone Dative may be moved by any unbounded movement transformation.

3. An attempt to explain the interaction from universal grammar, which fails It would be highly desirable if we could formulate a principle of universal grammar from which the English-specific principle just formulated would follow as a direct result. The reason for this is that the most adequate formalization of the Englishspecific principle is in terms of a derivational constraint, the undesirability of which notion has, in our judgment, been amply demonstrated. (Chomsky, 1972; Baker and Brame, 1972; Langendoen and Bever, 1973).6 A language-universal principle that has 4. It is immaterial to us whether Relativization and Question Formation are two rules in English or two manifestations of the same rule (@%-Fronting). Evidence that they are two rules is presented in Langendoen (1973). 5. On this distinction, see Ross (1968). We say that a movement transformation is unbounded if it can move a constituent across an unlimited number of unmatched left brackets labeled S; otherwise a movement transformation is bounded. Thus, for example, Question Formation is unbounded because

sentences like (i) are grammatical. (i) Who(m) did you say that Georgette found out that Marian was known to have been seen with? But Passive is bounded because sentences like (ii) are ungrammatical. (ii) *Frieda was believed that they had heard from by Irene (i.e., as Passive of: Irene believed that they had heard from Frieda) 6. The derivational constraint would be formulated along the following lines. In a derivation, a noun phrase that has been moved

Dative questions

455

been suggested recently by Klima (1970) and Ruwet (1973) looks promising in this connection. In its crudest form, the principle states that syntactic transformations may not create structures that permit the existence of syntactic ambiguity that depends soZeZyon the grammatical relations of two constituents in a sentence; we may call this the constraint on relational ambiguity principle (CRAP).’ Among other things, CRAP can be used to explain the observation made by Chomsky (1965, p.128) that although it is generally possible to topicalize direct objects in German, sentences in which the direct object cannot be distinguished inflectionally from the subject cannot undergo direct-object topicalization. That is, Chomsky claims that although there is a general process in German involving Topicalization and Subject Postposing that permits the derivation of sentences like (9) from structures like those underlying (IO), (11) cannot be obtained from (12) because, if the derivation of (11) were allowed, (11) would have the same surface structure as (13), in which the two noun phrases bear the opposite grammatical relations. (9) a. Heute kommt die Frau. ‘Today, the woman comes’ b. Den Mann sieht die Frau. ‘The man, the woman sees’ (10) a. Die Frau kommt heute. ‘The woman comes today’ b. Die Frau sieht den Mann. ‘The woman sees the man’ (11) *Das Madchen sieht die Frau. ‘The girl, the woman sees’ (12) Die Frau sieht das MSidchen. ‘The woman sees the girl’ (13) Das MBdchen sieht die Frau. ‘The girl sees the woman’ As Chomsky observes, the ordinary mechanisms of transformational grammar would

by Dative may not be moved again by any unbounded movement rule. The best altemative to the derivational constraint within an Aspects-type theory would be to mark all noun phrases that undergo Dative with some arbitrary feature and add to the structural conditions on all unbounded movement transformations in English the stipulation that they are inapplicable to noun phrases carrying that feature. 7. Klima’s version of CRAP is somewhat different. It reads: ‘When there are multiple occurrences of the same category in one construction, without lexical or morphological differentiation, then a simple algorithm exists for distinguishing their function and no transformation will have such an effect as to interfere with the effectiveness of the algorithm’ (quoted in Ruwet, 1973, p. 426). This formula-

tion, however, is defective in at least two critical respects. First, for ‘transformation’, Klima should have something like ‘transformational derivation’, since presumably he would want to allow the possibility of a derivation in which a transformation applies so as to interfere with the algorithm only to have a second transformation undo its effect. Second, his formulation is not couched in universal terms, since the algorithm Klima refers to will differ from language to language. In English, for example, the algorithm Klima posits for distinguishing the function of direct object from the function of indirect object is that the indirect object is the noun phrase that immediately follows the verb (in the absence of morphological evidence). Moreover, this algorithm, if true, is not a grammatical principle, but rather a principle of language

456

D. Terence Langendoen, Nancy Kalish-Landon and John Dore

not be able to account

for the cases in which the application of Topicalization and Subject-Postposing in German would result in an unacceptable sentence, since the cases involve ‘accidents’ of morphology to which the transformations in question could not possibly be sensitive. CRAP, however, provides what seems to be the most direct and intuitively satisfying account.8 Returning to the problem of the interaction of Dative with the various movement transformations in English, we see that CRAP also predicts that the (b)-examples in (4)-(8) are ungrammatical because there is nothing in the structure (including morphology) of those examples that permits us to determine which of the two object noun phrases in each is the direct object and which the indirect object. That is, if those examples were grammatical, they would be relationally ambiguous. This fact becomes clearer if we consider an example in which the direct object is animate. (14) *Who(m) did Selma send the doctor? If (14) were grammatical, there would be no dispute about its relational ambiguity. It would mean the same thing as either (Isa) or (15b).s perception (that is, a language-independent principle in the sense given in footnote 1). Hence it can have no bearing on the question of whether a given sentence is grammatical. For further discussion of Klima’s version of CRAP, see footnotes 9 and 16. 8. However, although the existence of CRAP as a linguistic universal would remove the need for a high-powered constraint (presumably a transderivational constraint) from the grammar of German, CRAP is itself a device as powerful as a transderivational constraint. In effect, it is an instruction to block a derivation given the existence of another derivation which results in the same surface structure as the first but with the constituents in different grammatical relations (or, both derivations may be blocked; see footnote 9). Thus the explanatory power of CRAP is strongly limited. It does not prevent the existence of grammars with rules that could create relational ambiguity but only prevents the derivation of relationally ambiguous sentences when those rules are used. It would be more interesting if there were a principle that really limited the class of grammars that could be acquired by stipulating that certain rules could not be a part of a grammar that had certain other rules because of the problem of relational ambiguity. But, apparently, there is no such principle. 9. As formulated in footnote 6, the English

specific derivational constraint would actually block only the derivation of (14) from the structure that underlies (15a). The derivation of (14) from (15b) would be permitted, since Questicn Formation is moving the direct object, not the indirect object. From the version of CRAP given in the text, on the other hand, it would follow that both derivations would be blocked, since in both cases syntactic transformations that create an ambiguous structure are being applied. This contrasts with the German situation involving Topicalization and Subject Postposing, in which one derivation is not blocked because it does not involve the application of relational-ambiguity-creating transformations. Since none of the linguists we have cited who have investigated the problem under consideration took into account sentences like (14), we have no way of knowing whether they would judge (14) as unacceptable on both readings, or unacceptable only on the reading of (15a) and acceptable on the reading of (15b). We suspect that opinion would be divided on this matter, some finding that (14) is unacceptable on both readings, in conformity with our version of CRAP, and others that (14) is acceptable on the reading of (15b) only, in conformity with Klima’s version of CRAP and the derivational constraint formulated in footnote 6. We take up this matter of varying acceptability judgments below in Section 6.

Dative questions

457

(15)

a. Who(m) did Selma send the doctor to?/ To whom did Selma send the doctor? b. Who(m) did Selma send to the doctor? Similarly, convincingly ambiguous examples could be constructed using any of the other unbounded movement rules together with Dative. However, CRAP falsely predicts that the examples in (3b), which illustrate the interaction of Dative and Passive, are also ungrammatical, since structurally the surface-subject noun phrase could be taken to be either the underlying direct or indirect object. This fact, again, emerges most clearly when we consider an example that contains an underlying animate direct object. (16) The rich client was offered the young lawyer by the senior partner That is, (16) is the passive version of either (17a) or (17b). (17) a. The senior partner offered the rich client the young lawyer (i.e., The senior partner offered the young lawyer to the rich client) b. The senior partner offered the young lawyer the rich client (i.e., The senior partner offered the rich client to the young lawyer) To save CRAP, we must therefore specify that it blocks a derivation of a relationally ambiguous sentence only if at least one unbounded movement transformation is applied in it.l” 10. Alternatively, one could eliminate reference to boundedness in the statement of CRAP if one could substantiate the claim that (16) caot be derived from (17b) and that hence (16) is not relationally ambiguous. This could be done by strictly enforcing the requirement that the noun phrase made subject by Passive must either be immediately postverbal or contained in a preposition phrase that is immediately postverbal; in (17b), the noun phrase undergoing Passive is separated from the verb by another noun phrase. By so restricting Passive, however, we would also be predicting that both sentences in (i) are ungrammatical. (i) a. A book was given Mary by Nancy b. *A dress was bought Mary by Nancy According to both Fillmore (1965) and Jackendoff and Culicover (1971, pp. 398, 400), only sentences in which a for-dative is made subject by Passive are unacceptable. Nevertheless, Jackendoff and Culicover accept the limitation on the structural description of Passive discussed above and derive sentences like (ia) from the structure that underlies (ii) by a later, optional rule they call, simply enough,

To Deletion (1971, p. 404). (ii) A book was given to Mary by Nancy. But if Jackendoff and Culicover are right, then sentences like (16) remain relationally ambiguous, and CRAP continues to make the wrong prediction. To save CRAP without imposing the limitation on boundedness, one would have to insist that the proposed rule of To Deletion is not part of the grammar of English and that (ia), while acceptable, is nonetheless ungrammatical. Now, as we observed in the opening paragraph of this paper, it is possible that a class of sentences can be considered acceptable but ungrammatical. But to show that such a class [for example, the class of sentences like (ia)] must be viewed in that way, one must be able to demonstratealanguage-independent behavioral principle according to which those sentences are acceptable, despite their ungrammaticality. In particular, it will not suffice to say simply that such sentences are acceptable ‘by analogy’, for then the questions arise, by analogy to what, and why this analogy and not some other, In the case of (ia), we know of no language-independent principle that could be

458

D. Terence Langendoen, Nancy Kalish-Landon and John Dore

But even this limitation is inadequate. Ruwet (1973) points out that the application of Question Formation (an unbounded rule) and a rule he calls Stylistic Inversion creates sentences with relational ambiguity in French; one example he cites is (lS).ll (18) Quels soldats commandent ces officiers ? ‘Which soldiers command these officers?’ or ‘Which soldiers do these officers command?’ Exactly the same sort of ambiguity appears in German and English (in English it is limited to sentences containing the main verb have and no auxiliary verbs), as in the examples in (19). (19) a. Was hat deine Katze in den Pfoten? b. What has your cat in its paws? In light of such examples, it is clear that there can be no syntactic mechanism for ruling out relational ambiguity in all cases, whether language-universal or languagespecific. This raises the possibility that there is some alternative explanation for the unacceptability of the (b)-examples in (4)-(8), which does not classify those examples as ungrammatical. One such explanation for the cases in (8b), in which Dative interacts with the application of Question Formation to the indirect object resulting in sentences which we henceforth shall call dative questions (DQs), has been proposed by Jackendoff and Culicover (1971). We consider now their proposal in detail.

4. Critique of Jackendoff unacceptability

and Culicover’s of dative questions

perceptual-strategy

explanation

for the

Although we limit our attention in this section and in what follows to DQs, what we have to say will largely carry over to sentences like the (b)-examples in (4)-(7).12 Jackendoff and Culicover’s motivation for proposing a perceptually based account of the unacceptability of DQs, like ours, is based on the realization of the difficulty of accounting for the unacceptability of that sentence type in the grammar of English appealedto; indeed it is the very acceptability of (ia) that gives rise to the ambiguity of (16)! We conclude that (ia) is acceptable because it is grammatical, that (16) is relationally ambiguous, and therefore that CRAP, if it is to be saved ,from this particular objection, must make reference to the boundedness of the transformational rules involved. 11. Although Ruwet (1973) devotes the bulk of his article to a defence of CRAP from French syntax, he is prepared to abandon it if the two readings would not both be likely in a given context. But this amounts to an admission that

the principle is not part of linguistic theory but at most a rhetorical principle: Relationally ambiguous sentences are avoided except in the case where only one interpretation is likely given the context. 12. We have some informally collected evidence that there is some variation in the acceptability of the (b)-examples in (4)-(7); in particular that the examples in (3b) are considerably less acceptable than the others. We are not prepared at the moment, however, to expIain this variation.

Dative questions

459

(see especially p. 411 of their article). I3 According to their account, DQs are unacceptable because a listener cannot determine from their perceived structure what grammatical relation to assign to the interrogative constituent. To show why this is so, Jackendoff and Culicover propose the following two-stage account of how listeners assign grammatical relations to noun phrases. First, grammatical relations for noun phrases that occur in their underlying positions by undoing a bounded-movement rule (such as Passive, Dative or Subject-Verb Inversion) are established. Second, grammatical relations for noun phrases that have been moved by unbounded rules14 are established by fitting them into any remaining perceived gaps. The crucial point is that although an unbounded-movement rule will always leave a gap (namely, the position formerly occupied by the moved constituent), the gap will not always be readily perceivable. Let us call this second stage of Jackendoff and Culicover’s proposal for the perceptual determination of grammatical relations the Gap-Filling Principle (GFP). To see how GFP works, consider its application to DQs and their variants in which Dative is not applied. (20) a. What did John give to Mary? b. What did John give Mary? c. Who(m) did John give a book to? d. *To whom did John give a book? To explain the acceptability judgments in (20), Jackendoff and Culicover (1971) reason as follows (we change their example numbers to conform to ours). In each sentence, the presence of the w/z-word signals that the interpreter of the sentence must look for a gap into which the w&word can fit. In (2Oa), to follows give, which can never happen in a declarative sentence. One can thus conclude that what must have been fronted from between these two words. In (2Ob), 13. However, Jackendoff and Culicover, in another passage, express the view that the mechanisms that account for perceptual difficulty are to be accounted for in the grammatical description of a language. In connection with the transformation called Extraposition from NP, they say the following: ‘ .. . the constraint on Extraposition from NP, which is very awkward to state in terms of conditions on application of transformations, becomes much clearer in terms of the difficulty of correctly interpreting the resulting strings. By permitting problems of string interpretation as possible sources of ungrammaticality [emphasis ours], we can eliminate this otherwise unexplained constraint. However, we must leave open for the present the question of how

to incorporate this innovation into the theory of grammar’ (1971, pp. 406-407). Given what they say on p. 411, we charitably interpret this passage as containing a lapse; Jackendoff and Culicover, we believe, are not really proposing that a theory of grammar should contain within it a theory of how sentences are perceptually processed. 14. Jackendoff and Culicover’s exact wording (1971, p. 410) is that ‘NPs which have been moved away from arbitrary positions (such as NPs fronted by w/z-fronting) are then fitted into remaining gaps’. As far as we can determine, rules which move constituents from arbitrary positions are coextensive with rules which are unbounded.

460

D. Terence Langendoen, Nancy Kalish-Landon and John Dore

give Mary is a permissible

sequence in a declarative sentence, so what need not have come from between them. In fact, if it had, the impossible string *give something Mary would have to be the corresponding declarative VP form. However, nothing follows Mary, and the verb give requires two objects. Give Mary something is a possible declarative VP form, so one can conclude that what has been fronted from the end of the sentence. In (~OC), give a book is a possible string in a declarative VP, and the bare preposition at the end shows that whom must have come from the end of the sentence. In (2Od), again give a book is a possible string, and so no gap is noticed at the stage where Whom did John give a book has been perceived. At this stage, the listener’s hypothesis is that whom has been fronted from the end; hence the preposition to is expected to follow book, as in (20~). Imagine the hearer’s consternation when the expected to does not arrive. The sentence is therefore judged unacceptable, since it is expected to be (20~) and then fails to conform to that expectation (p. 409). To summarize, sentences like (20d) are said to be unacceptable because the gap left by the fronting of the questioned constituent is not noticeable; sentences like (2Oa-c) are acceptable because the gap is noticeable. There are several respects in which this explanation for the judgments in (20) is faulty. First, consider the treatment of (20a). While it is certainly unlikely that the string give to Mary would appear in a declarative sentence,15 it is certainly not impossible, since examples like (21) are certainly acceptable. (21) Whenever he is feeling charitable, John gives to Mary More seriously, consider a sentence like (22), which has exactly the same structure as (20a). (22) What did John write to Mary? Since to can follow write in a declarative sentence, and probably does so quite frequently in ordinary usage, it would seem that listeners many times, if not invariably, would fail to notice a gap between write and to in (22); nevertheless, no one would find (22) unacceptable. Now consider (20d). According to GFP, this sentence is unacceptable because no gap can be perceived within its VP. Now consider (23), which is structurally parallel to (20d) except that where a nominal direct object appears in (20d) a sentential one appears in (23). (23) *Who(m) did John tell that Mary was staying? Example (23) contains the surface VP tell that Mary was staying, which is impossible 15. Jackendoff and Culicover do not intend for us to conclude that interrogative sentences are transformed back into declarative sentences

in perception; it is just that in declarative sentences constituents generally appear where they also happen to appear in deep structure.

Dative questions

46 1

in a sentence not transformed by some movement rule. Moreover, the element that has been moved out of that VP must appear between the verb tell and the sentential object, and hence a gap must be noticeable in that position. According to GFP, therefore, (23) should be acceptable, but it is not. Still another problem with GFP as it applies to (20d)-type sentences is that it predicts that sentences like (14), repeated here for convenience, should be acceptable on the interpretation that the postverbal noun phrase is the indirect object. (14) *Who(m) did Selma send the doctor? This is so because who(m) can be fitted into the gap following the postverbal noun phrase (but see footnote 9). Thus, GFP, despite its intuitive plausibility, cannot as such provide a perceptually based explanation for the unacceptability of DQs. We must conclude either that there is some other language-independent account of their unacceptability or that, indeed, DQs are ungrammatical. One piece of evidence that would tip the scales against the latter conclusion would be the existence of a population of otherwise ordinary English speakers who find DQs acceptable. For such speakers it would be impossible to maintain that DQs are ungrammatical, since an acceptable sentence type can only be considered ungrammatical if there is a language-independent explanation for its acceptability (see Langendoen and Bever, 1973, for discussion). But for the sentence type under consideration, the problem up to now has been just the opposite: To find a language-independent basis for their unacceptability. Furthermore, there is no problem in accounting for the acceptability of DQs within the grammar of English, since they arise upon application of generally accepted transformations to well-formed base structures. Furthermore, the difference between the population that accepts DQs and the one that does not need not be viewed as a dialect difference in the strict sense that the two populations possess two slightly different internalized grammars. Rather, the difference could be just as plausibly explained on the basis of a difference between the perceptual mechanisms by which the two populations attempt to understand DQs. In the next section, we present the evidence that the relevant population exists; in the section following that, we provide an alternative to Jackendoff and Culicover’s explanation of the unacceptability of DQs that accounts for the different acceptability judgments regarding DQs.

5. Demonstration

that a population that accepts DQs exists

5.1 Result of a pilot study on DQs In a pilot

study

conducted

with undergraduates

at various

campuses

of The City

462

D. Terence Langendoen, Nancy Kalish-Landon and John Dore

University of New York and at Rutgers University in New Brunswick, New Jersey, we found that our subjects had no objections at all to DQs of any sort. In particular, we found that many of our subjects spontaneously interpreted genuinely ambiguous DQs like (14) in exactly the way opposite to that predicted by both GFP and Klima’s version of CRAP (see footnote 9). That is, such a sentence would often be interpreted as (15a) rather than as (15b). On the basis of this pilot study, we conducted in the spring of 1971 more careful studies to substantiate the claim that many people from metropolitan New York City find DQs to be acceptable. We were particularly interested in subjects’ responses to genuinely ambiguous sentences like (14) because rather than ask for acceptability judgments directly (a methodologically unsound technique, since one has no way of knowing whether the responses are bona fide acceptability judgments), we wanted to infer those judgments by indirect techniques. The fact that subjects can respond to DQs does not necessarily mean that they accept such sentences, but if they consistently respond to genuinely ambiguous DQs in a way that indicates that they interpret them in the manner of (15a) (in which the postverbal noun phrase is taken to be the direct object), then for them DQs are acceptable because such interpretations are only possible if DQs are acceptable.16 5.2 Two experiments on DQs In the first experiment, we constructed a questionnaire consisting of fifteen sentences, in which Ss were instructed to ‘add the word “to” once to each of the.. .sentences so as not to change the meaning’. This questionnaire is given in Appendix 1; for convenience we refer to it as the To-Insertion Form (TIF). The fifteen sentences were written in full capital letters with equal spacing between the words. They were of three types. (1) Five sentences were of the type (24). (24) What did you show the landlord? [Example (1) on TIF] These were control sentences, since they are undisputedly grammatical, and only the postverbal noun phrase can be reasonably construed as the indirect object. (2) Five sentences were DQs of the type (25). (25) Whom did you give the ball? [Example (3) on TIF] These were also control sentences, since, although they are of the type that we have 16. That is, if an ambiguous sentence like (14) is interpreted as (15b), the S may be responding to it on the pattern of acceptable non-DQs, such as (i). (i) What did Sehna send the doctor? If, however, the S’s response indicates that he or she construes the postverbal noun phrase in an ambiguous DQ as a direct object, there is no non-DQ model that he or she could

possibly be using as a basis for that response; hence the acceptability of DQs for that S. For Ss that do not accept DQs, the ability of interpreting (14) as (15b) is not to be taken as evidence that GFP or Klima’s version of CRAP (see footnote 9) is correct for those Ss, since again, those Ss may be using (i) as a basis for that interpretation in a task situation like those describe below in section 5.2.

Dative questions

463

seen have been considered ungrammatical, the postverbal noun phrase can be reasonably construed only as a direct object, and perhaps because of the coercive effect of the experimental situation it was almost always so judged. (3) Five sentences were ambiguous DQs of the type (26). (26) Whom did you send the woman? [Example (12) on TIF] These were the experimental sentences that we believed could be interpreted in two ways. Besides (26), we used the following sentences. (27) Whom did you offer the man? [Example (2) on TIF] (28) Whom did you lend the team? [Example (4) on TIF] (29) Whom did you show the woman? [Example (9) on TPF] (30) Whom did you refer the person? [Example (14) on TIF] The fifteen sentences were arranged pseudo-randomly as to type. TIF was administered to 48 subjects consisting of undergraduates at Brooklyn College and at Rutgers University, and of professional persons at Bell Laboratories in Piscataway, New Jersey. The S’s were instructed that ‘there are no right or wrong answers: We are interested only in where you think the word [to] can be added’. The results of this experiment suggested several changes in the form. First, the task did not in any straightforward way require Ss to interpret sentences as they would if they were to encounter them spontaneously in speech or writing. The form was structured much like a typical Scholastic Aptitude Test question and so may have involved the Ss’ conscious knowledge about English grammar. Second, we felt that the ordering of the first two sentences on TIF [examples (24) and (27) above] might be introducing perceptual-set bias, since the initial what of example (24) might start a pattern of response in which the postverbal noun phrase is the indirect object throughout. Third, we felt that examples (28) and (30) [examples (4) and (14) on TIF] were not entirely satisfactory. Example (28) seemed to cause may Ss trouble, and (30) was being uniformly interpreted only in one way (unlike the others), namely with the postverbal noun phrase as the direct object. Fourth, we felt that the use of whom rather than who throughout may have promoted the interpretation of that word as the indirect object (for many persons, whom is only used spontaneously when a preposition precedes). Accordingly, a new form was devised, a form in which Ss were asked to ‘answer the . . . questions with a full sentence using the same verb’. This form is shown in Appendix 2; we refer to it as the Answer Form (AF). In this form, we replaced whom by who(m), with the instruction to ‘read “Who(m)” as either “Who” or “Whom”, depending on how you would say it in ordinary conversation’. We interchanged the positions of examples (24) and (27), so that (27) became example (1) on AF and (24) became example (2) on AF. We replaced examples (28) and (30) by the following sentences.

464

D. Terence Langendoen, Nancy Kalish-Landon and John Dore

(31)

Who(m) did you recommend the man? [Example (4) on AF] (32) Who(m) did you direct the person? [Example (14) on AF] We also instructed Ss that if they felt that more than one kind of answer was appropriate, to write additional answers, but not to cross out or erase their first answer, as we were particularly interested in their first response. AF was administered to 79 Ss, who were undergraduates at Baruch and Hunter Colleges of C.U.N.Y. Shortly afterwards, AF was administered to part of the same population that originally received TIF (30 in all), and TIF was administered to part of the same population that originally received AF (44 in all). This was done to insure against a confound of population differences with form differences. Finally, we decided to try out TIF, with who replacing whom throughout, as shown in Appendix 3. This form was administered to 68 undergraduates at Brooklyn and Hunter Colleges and Rutgers University, none of whom had been previously used as Ss in this study, and to 44 undergraduates at S.U.N.Y. at Buffalo (we thank David Hays for his help in administering the form there). This was done specifically to test our hypothesis that whom is more likely to attract the word to than who, and to run the test on Ss tram outside metropolitan New York City. Table

1.

Forms vs. populations for the three sentences in common to the To-Insertion (Whom) and Answer Forms

Sentence

Population*

Postverbal noun-phrase is: Direct Ambig- No N object UOUS Interpretation

Form**

Indirect object

AF AF TIF TIF

21 73 17 14

4 6 23 23

3 0 6 1

2 0 2 6

30 79 48 44

A B A B

AF AF TIF TIF

14 55 12 10

12 22 28 29

4 1 8 5

0 1 0 0

30 79 48 44

A B A B

AF AF TIF TIF

15 45 13 13

12 30 28 30

2 3 6 1

1 1 1 0

30 79 48 44

(27) Who(m) did you A offer the man? B [Ex. (1) on AF, Ex. (2) on TIF] A B

(29) Who(m) did you show the woman?

[Ex. (9) on AF and TIFl (26) Who(m) did you send the woman?

[Ex. (12) on AF and TIF]

* Population A: Brooklyn College, Rutgers University, Bell Laboratories Population B: Baruch College, Hunter College. ** AF: Answer Form; TIF: To-Insertion Form

Dative questions

465

5.3 Results In Table 1, the grammatical

relations assigned to the postverbal noun phrases in the first TIF (using whom) and AF for the three experimental sentences that they have in common are tabulated. Comparing the same forms against the different populations on which they were administered, we find that there is no significant population difference, using the Chi-Square method. For example (27), p < .5; for example (29) p < .5; for example (26), p < .5. If we now add the populations together by form we obtain the results given in Table 2 (the results of TIF using who on metropolitan New York City populations are also included in this table). The differences between AF and TIF using whom are significant, using the Chi-Square method. For example (27), *p < .OOl ; for example (29), *p < .OOl ; for example (26), *p < .OOl. The differences between AF and TIF using who are significant for two of the three sentences; for (27), *p < .OOl; for (29), p <.3; for (26), *p < .OOl.

Table 2. Answer Form vs. To-Insertion Forms (both Whom and Who) for metropolitan

New York City populations* Sentence

(27) Whom(m) did you offer the man?

(29) Who(m) did you show the woman?

Postverbal noun-phrase is: Direct AmbigNo object uous Interpretation

Form

Indirect object

N

AF TIF (Whom) TIF (Who)

94 31 30

10 46 30

3 7 6

2 8 2

109 92 68

AF TIF (Whom) TIF (Who)

69 22 33

34 57 30

5 13 5

1 0 0

109 92 68

AF TIF (Whom) TIF (Who)

60 26 20

42 58 42

5 7 4

2 1 2

109 92 68

(26) Who(m) did you send the woman?

* For AF and TIF (Whom), the populations are Baruch, Brooklyn and Hunter Colleges, Rutgers University and Bell Laboratories. For TIF (Who) the populations are Brooklyn and Hunter Colleges and Rutgers University.

In Table 3, the grammatical relations assigned to the postverbal noun phrases for the experimental sentences not in common to AF and the two TIFs are tabulated

466

D. Terence Langendoen, Nancy Kalish-Landon and John Dore

for the metropolitan New York City populations (again, since internal population differences are nonsignificant, the figures are added together). For the five experimental sentences, the differences between TIF using whom and TIF using who are nonsignificant, except for example (29). For (27), p < .2; for (29), *p < .Ol; for (26), p < .7 ; for (28), p < .7; for (30), p < .3. Table 3.

Answer Form and To-Insertion Forms (Whom and Who) for the sentences not common to those forms for metropolitan New York City populations*

Sentence

(31) Who(m) did you recommend the man? [Ex. (4) on AFI (32) Who(m) did you direct the person? [Ex. (14) on AFI

Form

Indirect object

Postverbal noun phrase is : Direct AmbigNo object uous Interpretation

AF

34

63

4

8

109

AF

2

94

2

11

109

N

(28) Who(m) did you lend the team? [Ex. (4) on TIFI

TIF (Whom) TIF (Who)

30 26

51 39

10 2

1 1

92 68

(30) Who(m) did you refer the person? CEx. (14) on TIFI

TIF (Whom) TIF (Who)

7 8

77 56

5 1

3 3

92 68

* See Table 2 for the population

breakdown by forms.

Finally, in Table 5, we give a detailed analysis of the AF responses to examples (29) and (26) [examples (9) and (12), respectively, on AF]. In this table, we correlate the responses to (29) and (26), which are the only minimal pair among our experimental sentences (they differ only in the verb). In addition, we distinguish between responses which are syntactically and semantically appropriate and those which are inappropriate. For example, (33) is an appropriate response to (29), whereas (34) is inappropriate. (33) I showed the woman a friend (34) I showed the woman a hat Both responses indicate that the postverbal noun phrase the woman in (29) was

Dative questions

Table

4.

467

To-Insertion Form (Who) vs. population (metropolitan New York City and BuflaIo, New York)

Sentence

Population*

Indirect object

Postverbal noun-phrase is: Direct AmbigNo object uous Interpretation

(27) Who did you offer the man?

C D

30 38

30 2

6 2

2 2

68 44

(29) Who did you show the woman?

C D

33 27

30 9

5 6

0 2

68 44

C D

20 27

42 11

4 4

2 2

68 44

Who did you lend the team?

C D

26 13

39 24

2 5

1 2

68 44

(30) Who did you refer the person?

C D

8 3

56 38

1 2

3 1

68 44

N

(26) Who did you send the woman?

(28)

* Population Population

Table 5.

C: Brooklyn and Hunter Colleges and Rutgers University D: SUNY at Buffalo

AnaIysis of responses to examples (29) and (26) in Answer Form

Postverbal noun phrase is:

Both responses appropriate

One or both responses inappropriate

N

1. 1.0. in both (29) and (26) 2. 1.0. in either (29) or (26); D.O. in the other 3. 1.0. in either (29) or (26); ambiguous in the other 4. D.O. in either (29) or (26); ambiguous in the other 5. Ambiguous in both (29) and (26). 6. D.O. in both (29) and (26) 7. No interpretation for one or both (29) and (26) Total of 4, 5, 6 Grand total

42 12

9 12

51 24

2

0

2

2

0

2

3 19 0

0

3 25 2

(24) 80

6 2

(6) 29

(30) 109

468

D. Terence Langendoen,

Nancy Kalish-Landon

and John Dore

interpreted as indirect object; however the latter response is inappropriate because the noun phrase that answers the question who(m)? is inanimate. A syntactically inappropriate answer would be (35). (35) I showed the woman In (35), there is no second object; in fact the question is not really answered at all. For purposes of classification, the woman is assumed to be direct object, since that is how the sentence would be analyzed out of context.

5.4 Discussion We consider here the following matters: (1) The differences in the responses to the two tasks, (2) the differences between TIF with whom and with who and the difference between the Buffalo and New York City populations, (3) the intrasentential differences, particularly on AF, and (4) the establishment of a New York City population that accepts DQs. Concerning task differences, we find that in general AF favored the interpretation of the postverbal noun phrase as indirect object, whereas TIF favored its interpretation as direct object (but less so when the interrogative word was who rather than whom). Thus TIF provided a task that enhanced the appearance of the ability to interpret DQs in a way contrary to the predictions made by both GFP and CRAP. However, we are willing to discount that evidence on the basis of the relative artificiality of that task as compared to the question-answering task in AF and to rest our case for the existence of a population that accepts DQs on the AF responses. Concerning the difference between TIF with whom and TIF with who, we note that although the difference is nonsignificant except for example (29), the difference is in the predicted direction: Whom is more likely to be taken as indirect object than who. The reason significance was achieved for (29) may be that there were relatively many ambiguous responses to it in TIF with whom. These responses were not included in the calculation for significance. However, we have no explanation why just this example deviated from the general pattern. On the other hand, the difference between the New York City and Buffalo populations on TIF with wlzo convincingly shows that a population selected from outside New York City will treat DQs in conformity with the linguistic and psycholinguistic descriptions considered in Sections 2 through 4. The most striking intrasentential variations are to be found between the three sentences in common to the two forms (26), (27) and (29) and two of the four sentences not in common to them, namely (30) and (32). For AF, the contrast is most striking between (27), in which 94 Ss considered the postverbal noun phrase to be the indirect object and only 10 Ss considered it the direct object, and (32), in which only 2 Ss considered the postverbal noun phrase to be the indirect object and 94 Ss considered

Dative questions

469

it the direct object. For TIF (both who and whom varieties, New York City populations only), compare (27), in which 61 Ss took the postverbal noun phrase to be an indirect object and 76 took it to be a direct object with (30), in which 15 Ss considered the postverbal noun phase to be an indirect object and 133 5’s considered it a direct object. The reason for this set of differences has to do with a grammatical oddity inherent in examples (30) and (32). If we put them back into declarative form [replacing who(m) by someone], considering who(m) to be the indirect object, as most of our Ss did, we obtain ungrammatical sentences. (36) *You referred someone the person (37) *You directed someone the person The results are similarly ungrammatical if who(m) is considered the direct object. Thus we must assume that examples (30) and (32) are also ungrammatical, and that our Ss were forced in the experimental situation to interpret them in terms of the closest salient grammatical sentences (Katz 1964). For (30) and (32), they are the following.17 (38) Who(m) did you refer the person to? (39) Who(m) did you direct the person to? This accords with the experimental facts, since in both (38) and (39), the postverbal noun phrase is the direct object. We take up now the matter of most importance to us: The establishment of the existence of a population that accepts DQs, and for whom therefore DQs are grammatical. This, we claim, is shown in Table 5, in which we examine the responses to two non-problematic DQs in AF, examples (29) and (26). In section 5.1 above, we argued that if Ss are consistently able to interpret the postverbal noun phrases of DQs as direct objects, then DQs for them must be acceptable, since there are no analogical grammatical sentences on the basis of which they could interpret DQs in that way. The number of Ss that treated the postverbal noun phrases as direct objects in both (29) and (26) without error is 24, or 22 percent of the total number of Ss that filled out AF. This means that at least one-fifth of our Ss from metropolitan New York City find DQs acceptable. We suspect that the actual population, expressed as a percentage of fluent metropolitan New York City English speakers, is considerably larger than this figure would indicate, but we are content with having shown that it is at least this large. 17. This is so, since listeners would conclude from the fact that neither refer nor direct undergo Dative that the postverbal noun phrase must be the direct object. The minimum syntactic change, therefore, is to add the appropriate preposition (tg) to the interrogative

word. Our purpose in including these patently ungrammatical sentences in the questionnaires was to see whether they would be treated differently from DQs whose grammaticality status was the subject of investigation. And indeed they were.

410

D. Terence Langendoen, Nancy Kalish-Landon and John Dore

6. An explanation based on the perceptual determination of grammatical relations for the variation in the acceptability of DQs

In this section, we show how the variation in the acceptability of DQs could be accounted for as a consequence of the different perceptual rules used by Englishspeaking listeners for assigning grammatical relations to noun phrases. We assume that such relations are assigned while the sentence itself is being actively processed, and that the end of each perceived constituent is a potentia1 decision point for determining the grammatical relations that obtain within that sentence.l* In (40), we give a schematization of a DQ, with the potential decision points (DPs) indicated underneath. (40)

Who(m) 0

did

you

1

V 3

2

NP 4

5

In (41), we state the perceptual rules that are applicable in the processing of DQs and indicate for each rule the decision point in (40) at which it is applicable first.19 (41) a. Subject and object in inverted sentences (DP 3)

[Err.

]

1 [Obj. lf V.] 1

-

[Lx.]

ii

18. These assumptions can be strongly motivated simply from consideration of the rapidity and precision of speech comprehension. Although they have not received direct experimental support, the assumptions have also been used by other psychoIinguists in the construction of more elaborate hypotheses that have received experimental support. 19. We assume that the system of perceptual rules is designed to recover essentially the surface structures of sentences (as in Kimball, 1973) and ‘annotations’ of those surface structures that indicate the deep grammatical relations and perhaps certain other aspects of deep structure, where those are different from surface structure. The rules in (41) are a subset of the annotating rules. We cast them in the form

-

NP

[Subj.“,,]

-

x

’

*

of transformational rules because that is a convenient and revealing framework. The subscript ‘i’ on the category symbol ‘V’ (for ‘verb’) is an abstract inidicator of the clause in which that verb is the main verb; for simplicity, we assume that all the constituents mentioned in the rules of (41) are members of the same clause. The feature specification [t Dative] indicates that the verb permits application of Dative within its clause. The specifications [Dir. Obj. of Vi1 and [Ind. Obj. of VJ are assumed to substitute for the specification [Obj. of Vi]. Thus rule (41a) may mark certain noun phrases as objects of a particular verb; rules (42b,c,d) may then replace that specification with the more precise specifications as direct or indirect objects.

Dative questions

b. Interrogative

indirect

1 Ind. Obj. of V, c. Interrogative

object (DP 4) --x v, [ +Dative

1

1

Y

2

direct object and postverbal

EEfv]

-

471

x [TDative]

indirect -

(NP)

object (DP 5) -

y

i 1 1 Dir. Obj. of Vi d. Postverbal

1

2 2

3 3 Ind. Obj. of V,

1

direct and indirect objects (DP 5) xv, (NP) NP 1 2 3 1 2 3 Ind. Obj. of Vi Dir. Obj. of V,

I[

1

4 4

Y 4 4

=r

*

Rules (41a) and (41d) are independently needed for the perceptual processing of a wide variety of English sentence types and must be considered part of the set of listening strategies of everyone who is fluent in English. Rule (41~) embodies a specific case of Klima’s ‘simple algorithm’ for picking out direct and indirect objects in the event an object has been moved out of postverbal position. It, too, may be presumed to be shared by everyone who is fluent in English.20 Rule (41b), however, is limited to those speakers who accept DQs; it accounts in fact for their ability to interpret those sentences in the manner in which the grammar of English dictates that they must be interpreted. To see this, consider first how rules (4la,b,c,d) assign grammatical relations to sentences of the type (40), and second how rules (4la,c,d) do so. At DP 3, rule (41a) marks who(m) to be an object of the main verb and you the subject. At DP 4, rule (41b) specifies that who(m) is in fact the indirect object of the 20. Hankamer (1973, p. 52) reports that he does not accept sentences such as (i) in which the direct object in a sentence that has under-

gone Dative is fronted. (i) What did Harry sell Jerome? We can explain his inability to accept sentences

412

D. Terence Langendoen, Nancy Kalish-Landon and John Dore

verb. At DP 5, rule (41~) is inapplicable (the interrogative noun phrase is already specified as the indirect object), but rule (41d) is applicable; it marks the postverbal noun phrase to be the direct object of the verb. If rule (41b) is left out, the following perceptual derivation is obtained. At DP 3, as before, rule (41a) marks who(m) as an object of the main verb and you as subject. At DP 4, none of the rules of (41) are applicable. At DP 5, rule (41~) applies, marking the interrogative noun phrase to be the direct object and the postverbal noun phrase the indirect object. In case the postverbal noun phrase is animate, the result is not unacceptable, but if it is inanimate, the result is unacceptable, since it is not possible for an indirect object to be inanimate while the direct object is animate.21 Thus DQs are acceptable in case listeners have internalized rule (41b) and unacceptable in case they have not. At this point, we must emphasize that we have not demonstrated experimentally that English speakers who accept dative questions have internalized a perceptual rule like (41b), but only that the postulation of such a rule does provide an explanation for their ability to accept such sentences, without having to posit that they speak and understand a different dialect of English from everyone else. However, it should be possible to test experimentally whether a perceptual rule like (41b) exists. First, one must isolate two populations, one containing Ss that accept dative questions and the other Ss that do not. If Ss that accept dative questions do so on the basis of their having a rule that marks interrogative animate noun phrases as indirect objects as soon as they have heard a main verb that undergoes Dative, then such subjects should have demonstrably greater difficulty (as measured, for example, by latency to respond) in dealing with sentences like (42) than Ss that do not accept DQs. (42) Who(m) did you send home? The reason is that who(m) must be the direct object of send in (42), but Ss who employ (41b) upon hearing send will mark who(m) as indirect object and will have to correct that assignment once they discover that there is no other direct object. On the other hand, the presence of a perceptual rule like (41b) should enhance Ss’ ability to understand sentences like (43). (43) Who(m) did you write yesterday? like (i), an inability which he alone among the numerous linguists that have investigated the dative construction has testified to, by saying that he lacks rule (41~). Accordingly, the interrogative noun phrase will be marked as an object, but not any particular kind of object, such as direct or indirect. 21. Thus the rules in (41), like GFP and Klima’s version of CRAP, fail to account for

the unacceptability of ambiguous DQs, if indeed they are unacceptable (see footnote 9). One possible source for their unacceptability within the framework of (41) would be an interaction of rules (41b,c) in which the interrogative noun phrase becomes simultaneously specified as indirect and direct object. This must remain, however, a conjecture for the time being.

Dative questions

473

In (43), who(m) is the indirect object despite the fact that no direct object is present in surface structure. Thus the claim that the acceptability of DQs depends on the presence of a particular perceptual rule that not every English speaker possesses is empirically testable in a straightforward way. Regardless of the outcome of experiments testing the hypotheses advanced in this section, however, we believe that a solid basis for considering that dative questions are grammatical in English has been established.22

22. We regret that we read too late for inclusion in our theoretical discussion the interesting and provocative paper by Hankamer (1973), in which he asserts, among other things, that he can see no basis for distinguishing between rules of grammar and so-called perceptual strategies. We hope that this paper will be viewed as providing at least one such basis. Concerning Hankamer’s proposed ‘no-ambi-

guity condition’, which bears considerable resemblance to the principles discussed and dismissed in Sections 3 and 4 and which appears superior to them in a number of respects, we note simply that it is supported by remarkably idiosyncratic and otherwise unjustified acceptability judgments, such as the one pointed out in footnote 20. If those judgments cannot be supported, neither can the condition.

474

D. Terence Langendoen, Nancy Kalish-Landon and John Dore

Appendix 1 To-Insertion Form Using WHOM ADD THE WORD

‘TO’ ONCE TO EACH OF THE FOLLOWING

THE MEANING.

THERE

IN WHERE

1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

11. 12. 13. 14. 15.

YOU THINK

ARE NO RIGHT THE WORD

OR WRONG

DID

YOU

SHOW

THE

LANDLORD

WHOM

DID

YOU

OFFER

THE

MAN

WHOM

DID

YOU

GIVE

THE

BALL

WHOM

DID

YOU

LEND

THE

TEAM

WHOM

DID

YOU

OFFER

THE

CANDY

WHAT

DID

YOU

LEND

THE

CAPTAIN

WHOM

DID

YOU

SEND

THE

PACKAGES

WHAT

DID

YOU

GIVE

THE

TEACHER

WHOM

DID

YOU

SHOW

THE

WOMAN

WHOM

DID

YOU

SHOW

THE

DRESS

WHAT

DID

YOU

OFFER

THE

STUDENTS

WHOM

DID

YOU

SEND

THE

WOMAN

WHOM

DID

YOU

LEND

THE

MONEY

WHOM

DID

YOU

REFER

THE

PERSON

WHAT

DID

YOU

SEND

THE

BOY

LANGUAGES

OTHER

LIST IN ORDER

THAN

ONLY

? ?

NUMBER ENGLISH

OF DECREASING

HAVE YOU TAUGHT

SO AS NOT TO CHANGE

WE ARE INTERESTED

CAN BE ADDED.

WHAT

NAME OR SOCIAL SECURITY

SENTENCES

ANSWERS:

ENGLISH?

YOU

SPEAK

OR HAVE STUDIED.

IF MORE THAN

ONE,

FLUENCY

YES

-NO

Appendix 2 Answer Form

Please answer the following questions with a full sentence using the same verb. For example : Q. Who(m) did you see? A. I saw the man.

Dative questions

475

Read ‘Who(m)’ as either ‘Who’ or ‘Whom’, depending on how you would say it in ordinary conversation. If you feel that more than one kind of answer is appropriate, write additional answers. Please do not change your first answer to each question. There are no right or wrong answers. Your first answer is the best one. 1.

2.

Q. Who(m) did you offer the man? A. Q.

What did you show the landlord?

A. 3.

Q. A.

Who(m) did you give the ball?

4.

Q. A.

Who(m) did you recommend the man?

5.

Q. A.

Who(m) did you offer the candy?

6.

Q. A.

What did you lend the captain?

7.

Q. A.

Who(m) did you send the packages?

8.

Q. A.

What did you give the teacher?

9.

Q. A.

Who(m) did you show the woman?

10.

Q. A.

Who(m) did you show the dress?

11.

Q. A.

What did you offer the students?

416

12.

D. Terence Langendoen, Nancy Kalish-Landon and John Dore

Q.

Who(m) did you send the woman?

A. 13.

Q.

Who(m) did you lend the money?

A. 14.

Q.

Who(m) did you direct the person?

A. 15. Q. A.

What did you send the boy?

Name or Social Security Number Are you a native speaker of English? _Yes

_No

How long have you been a resident of Greater New York City Have you ever taught English _Yes

_No

Appendix 3 To-Insertion Form Using WHO ADD THE WORD

‘TO’ ONCE TO EACH OF THE FOLLOWING

THE MEANING.

THERE

IN WHERE

YOU THINK

ARE NO RIGHT THE WORD

OR WRONG

SENTENCES

ANSWERS:

CAN BE ADDED.

1.

WHAT

DID

YOU

SHOW

THE

LANDLORD

2. 3. 4. 5. 6. 7. 8. 9.

WHO

DID

YOU

OFFER

THE

MAN

WHO

DID

YOU

GIVE

THE

BALL

WHO

DID

YOU

LEND

THE

TEAM

WHO

DID

YOU

OFFER

THE

CANDY

WHAT

DID

YOU

LEND

THE

CAPTAIN

WHO

DID

YOU

SEND

THE

PACKAGES

WHAT

DID

YOU

GIVE

THE

TEACHER

WHO

DID

YOU

SHOW

THE

WOMAN

10.

WHO

DID

YOU

SHOW

THE

DRESS

?

SO AS NOT TO CHANGE

WE ARE INTERESTED

ONLY

11.

WHAT

DID

YOU

OFFER

THE

STUDENTS

12. 13. 14. 15.

WHO

DID

YOU

SEND

THE

WOMAN

?

WHO

DID

YOU

LEND

THE

MONEY

?

WHO

DID

YOU

REFER

THE

PERSON

?

WHAT

DID

YOU

SEND

THE

BOY

?

NAME OR SOCIAL LANGUAGES

SECURITY

OTHER

LIST IN ORDER

THAN

Dative questions

471

IF MORE THAN

ONE,

?

NUMBER ENGLISH

OF DECREASING

YOU

SPEAK

OR HAVE

STUDIED.

FLUENCY

HAVE YOU EVER TAUGHT ENGLISH?.

_YES

-NO

REFERENCES Baker, C. L., and Brame, M. (1972) Global rules: A rejoinder. Language, 48, 51-75. Bever, T. G. (1970) The cognitive basis of linguistic structures. In J. R. Hayes (Ed.) Cognition and the development of language. New York, Wiley. Pp. 279-362. Chomsky, N. (1965) Aspects of the theory of syntax. Cambridge, Mass., M.I.T. Press.

-

(1972) Some empirical issues in the theory of transformational grammar. In S. Peters (Ed.) Goals of Linguistic theory. Englewood Cliffs, N.J., Prentice-Hall. Pp. 63-130. Emonds, J. (1972) Evidence that indirectobject movement is a structure-preserving rule. Found. Lang., 8, 546-561. Fischer, S. (1971) The acquisition of verbparticle and dative constructions. Unpublished Ph.D. dissertation, M.I.T. Fillmore, C. J. (1965) Indirect object constructions in English and the ordering of transformations. The Hague, Mouton.

Hankamer,

J. (1973) Unacceptable

ambiguity.

Ling. Znq., 4, 17-68.

Jackendoff, R., and Culicover, P. (1971) A reconsideration of dative movements. Found. Lang., 7, 397-412. Katz, J. J. (1964) Semi-sentences. In J. A. Fodor and J. J. Katz (Eds.) The structure of language: Readings in the philosophy of language. Englewood Cliffs, N.J., Prentice-

Hall. Pp. 400-416. Kimball, J. (1973) Seven principles of surface

structure parsing in natural language, Cog. 2 (I), 15-27. Klima, E. S. (1970) Regulatory devices against functional ambiguity. Unpublished paper. Kuroda, S.-Y. (1968) Review of Fillmore, Lang.,

44, 374-378.

Labov, W. (1972) For an end to the uncontrolled use of linguistic intuitions. Unpublished paper. Langendoen, D. T. (1973) The problem of grammatical relations in surface structure. In K. Jankowsky (Ed.) Monograph series on languages and linguistics, 26, Washington, Georgetown University Press. and Bever T. G. (1973) Can a not unhappy person be called a not sad one? In S. Anderson and P. Kiparsky (Eds.) A festschrift for Morris Halle. New York, Holt, Rinehart & Winston. Pp. 392-409. Otero, C. (1972) Acceptable ungrammatical sentences in Spanish. Ling. Znq., 3, 233242.

Ross, J. R. (1968) Constraints on variables in syntax. Bloomington, Indiana University Linguistics Club. Mimeo. Ruwet, N. (1973) How to deal with syntactic irregularities: Conditions on transformations or perceptual strategies? In F. Kiefer and N. Ruwet (Eds.) Generative grammar in Europe. Dordrecht, The Netherlands, D. Reidel. Pp. 419-444.

478

D. Terence Langendoen, Nancy Kalish-Landon and John Dore

RtsumC

Les grammariens de la Grammaire Generative ont soutenu que les phrases anglaises du type who (m) did you give the book? (que nous appellerons interrogations au datif) ne sont pas grammaticales. Cependant, l’incorporation, dans la grammaire de l’anglais, des restrictions necessaires pour rendre compte de ce fait necessite un affaiblissement de la theorie linguistique. 11serait done souhaitable de pouvoir rendre compte de cette restriction dans le cadre de la theorie de la performance. Cela a BtCpropose par Jackendoff et Culicover (1971). On montre ici que cette position nest pas recevable. En essayant de trouver une meilleure proposition on a remarque, avec deux ex-

periences de type questionnaire, que quelques locuteurs de langue anglaise - tous originaires de la ville de New York - acceptaient les interrogations au datif. A partir de la on a dmis l’idt5e que l’on rendait mieux compte de la variation observte dans l’acceptabilite des interrogations au datif par des differences dans les strategies perceptives utilisees par les differentes populations de locuteurs pour determiner les relations grammaticales des propositions percues. Ainsi il n’y a pas de differences de dialecte, ii strictement parler, pour les interrogations au datif; elles sont toutes grammaticales pour tous les locuteurs anglais.

Discussion

‘Do not adjust

your

mind,

there is a fault in reality’Ideology in neurobiology*

STEVEN

P. R. ROSE

The Open University

HILARY

ROSE

London School of Economics

Abstract The paper considers the relationship between scient$c knowledge and the social context within which science is done. Because of the relationship of science and scientists to the state, scientiJic paradigms may be ideological, and much that is done in the name of science may itself be ideologically saturated. The dominant ideoIogy in neurobiology is that of reductionism, and indeed reductionism is argued by its proponents to constitute the scientiJic method, with ethical overtones, whilst opponents of science cite reductionism as exemplifving the inevitably oppressive nature of science. Within neurobiology, several types of reductionism are considered. Molecular reductionism seeks biological causes for socially observed events, for instance, schizophrenia and depression. Its consequences, in relationship to ‘minimal brain dysfunction’ and psychosurgery, are considered. Genetic determinism in relation, e.g., to intelligence (IQ), is another form of this type of reductionism. Evolutionary reductionism attempts to explain human behavior in terms of that of primates and other non-human animals, hence seeking to just$y the existing social order as dependent on a biological base. Category reductionism includes behaviorism and machine reductionism; in this latter, attempts are made to reduce the brain to the interactions of simply modellable logical neuronal networks. Against reductionism, autonomism as a separate type of paradigm is briefly discussed and rejected. A non-ideological, and hence scientific and non-oppressive paradigm, would be a version of interactionism, diaIectica1 materialism; such a science cannot be fully realized except in a transformed society.

* An earlier version of this paper was given at the International Sociological Association

Research Committee in the Sociology of Science, London 1972. See Whitley (1974). Cognition 2(4),

pp. 479-502

480

Steven P. R. Rose and Hilary Rose

1. Introduction Within the last few years there has been a re-emergence of a questioning of one of the basic convictions which underly the apparent certainty with which most scientists go about their business - that of the divorce between fact and value, science and ideology. Amongst scientists, this questioning has taken the form of debate about the question of the ‘neutrality’ of science in the face of its evident use for socially oppressive purposes (for an analysis of this debate, see Rose and Rose, 1969, 1972, 1973). Amongst sociologists of science, it has taken the form of a renewed interest in the sociology of scientific knowledge, as against the sociology of scientific institutions or of scientists (Whitley, 1974). Amongst philosophers of science there has been a wave of discontent with the classical Popperian view of the nature of scientific hypotheses, which has taken sustenance from the concepts of scientific revolutions and of paradigms introduced by Kuhn (e.g., Musgrave and Lakatos, 1971). We have pointed out elsewhere (Rose and Rose, 1972) that the re-emergence of this discussion of the relationship between the science that is done and the values and ideology of the society within which it is done represents a return to concerns which began to be developed many years ago as part of a now-classical Marxist analysis ot the role of science (e.g., Hessen, 1931). They were truncated in the West as a result of the profound historical changes in the relationship of science to the state which began with the 1939-45 war and in the Soviet Union with the administrative and ideological repression which developed with the Stalin period (Graham, 1971). Much of the discussion on these issues, especially amongst scientists, has focussed on the obvious socio-political constraints which, in American or European society, serve to direct science in the interests of oppression. However, there has been a strong current of argument to maintain that such a use of science is intrinsic to the nature of science itself; it forms part of the ideology of science as an institution. This has been the position, for example, of Feyerabend, Habermas and Marcuse. ‘The principles of modern science were a priori structured in such a way that they could serve as conceptual instruments for a universe of self-propelling, productive control; theoretical operationalism came to correspond to practical operationalism. The scientific method which led to the ever-more-effective domination of nature.. . Today, domination perpetuates and extends itself not only through technology but as technology, and the latter provides the great legitimation of the expanding political power, which absorbs all spheres of culture’ (Habermas, 1971). In this view, it is the scientific method which is at fault. Scientific method with its emphasis on rationality is seen as inevitably oppressive and itself providing the ideological legitimacy for rationally oppressive society. Such an attack on science and hence a stressing of the values of non-rational forms of knowledge typifies

‘Do not adjust your mind, there is a fault in reality’

481

writers like Roszak (1969). But is there such a thing as ‘the scientific method’, or is what is commonly perceived as the method of science itself an ideological product, reflecting the society in which it is practiced? In this paper, we wish first to discuss the possible sources of ideological input into science in general, and second, to examine a particular group of sciences, neurobiology, for evidence of the role of ideology.

2. Ideology in natural science Let us see how ideology and science become confounded. We take as given a materialist position: There exists an objectively real world of which humans are part. Our knowledge of this external objectively real world is obtained through our senses and interpreted by our brain in relationship to past experience. The very nature of the brain itself imposes a transformation upon incoming sensory information; the act of comparing the transformed information against stored memories of other transformed sensory information imposes another transformation. As a result, between the objectively real world and human knowledge thereof, there exists a series of mediations. The epistemological framework developed by the natural sciences begins as a series of procedures and rules for interpreting the evidence of our senses so as to provide reliable information (that is, predictive powers) with respect to the real world. As a science develops, the rules and interpretations become assembled in particular domains, the logical coherence of which is maintained by theoretical constructs which are attempts to unify these interpretations; the past experience against which new data is being compared is selected and organized in particular ways. Thus there emerges the artificial world of human knowledge - a world of mediated sensory information, a store of organized past experience and unifying theoretical constructs. All knowledge is in the artificial world; what distinguishes scientific knowledge from other types of knowledge is the degree of correspondence between the artificial world and the external real world, for it is in the nature of the artificial world of scientific knowledge that it generates predictions concerning the behavior of the external real world; these predictions can be tested in the real world and the results of such tests interpreted once again in the artificial world. In creating the artificial world, two types of inputs are involved : The actuality of the external real world and the constructs within the artificial world with which events in the real world are compared. For any scientist there are two classes of such influences in the formation of these constructs, the internal history of the subject, the domain of knowledge with its own particular organizing paradigms into which data is to be

482

Steven P. R. Rose and Hilary Rose

fitted, and the social history of the individual scientist himself/herself including his/her apprenticeship within the subject. The two histories arise, of course, from other inputs from the external real world, which may themselves be made the subject of a different scientific enquiry. They are distinguished from the set which directly form the subject matter of the particular science as being outside its domain and are traditionally excluded from account, but they mold the types of prediction the scientist makes concerning the real world and the way in which the sensory information arising in response to the tests of such a prediction are mediated and assimilated into the artificial knowledge world. What forces generate the paradigm? Clearly the many past inputs from the external real world into the domain of knowledge in the artificial world, and the social histories of the individual scientists who have mapped out the particular domain. But whereas for any individual scientist the personal idiosyncracies of his/her past history may affect his/her artificial world in a variety of disparate ways, insofar as the paradigm forms a shared part of the artificial world of many individuals, the individual idiosyncracies become of diminishing importance, whilst the common social histories of those who share the paradigm become of increasing importance. Insofar as these common social histories represent universal features of human existence, such as fundamental aspects of human perceptual processes, brain limitations, etc., they represent a permanent aspect of the artificial world. However, many aspects of the external social world are not universal but represent particular historical circumstances and conjunctures; they are thus not permanent but variable. While science was a relatively autonomous activity within society, with little articulation into the class and state structure, the role of the variable social inputs was perhaps less important. This situation is no longer the case. It is a commonplace that over the last 100 years, but particularly over the last few decades, science and the state have moved closer together. Under these circumstances the major variable input into scientific work becomes closely related to the needs of the state itself and the interests of the dominant class group within society. There are a number of routes by which these inputs enter the activity of scientific work. The most obvious is that scientific work requires a substantial budget. Unlike the situation in the seventeenth, eighteenth, and to a lesser extent, nineteenth centuries, scientific work can not be done by the disinterested amateur; it is a professional activity involving a significant proportion of the work force of industrialized nations, and it is paid for almost exclusively by the state and by industry; hence the problems which are set for scientific work are those which are of direct interest to state and industry and whose solution is to be sought with the existing social framework. Research funding thus both determines research directions and is an expression of class interests within society.

‘Do not adjust your mind, there is a fault in reality’

483

Intertwined with, but analytically distinct from, this social determinant is that of ideology. The paradigms of a science are its philosophical framework; they emerge from the interplay of data from the external real world, models within the artificial world and ideas drawn from outside both. It is with these ideas that ideology enters. Ideology represents both a set of approaches to the external, objectively real world and a set of epistemological tools. These either derive directly from an individual’s position in the dominant class within society or may reflect a false consciousness concerning the individual’s relationship with that dominant class. ‘The ideas of the ruling class are in every epoch the ruling ideas; i.e., the class which is the ruling material force of society, is, at the same time, its ruling intellectual force. The class which has the means of material production at its disposal, has control at the same time over the means of mental production, so that thereby, generally speaking, the ideas of those who lack the means of mental production are subject to it. The ruling ideas are nothing more than the ideal expression of the dominant material relationships . . . .’ (Marx and Engels, 1846, p. 39). Thus, because of the relationship of science to the state, and of the class position of scientists within society, the ruling ideas which enter into the shaping of scientific paradigms are ideological. When we talk of ideology in scientific work, therefore, we refer first to ways in which the social context, internalised by individual scientists, finds expression within their experimentation and their attempted interpretation of the external real world within the artificial knowledge world, and second, to ways in which ideas and models from the artificial knowledge world are reflected back into the social context again, to provide those ‘ruling ideas’ which help justify the existing social structures or those technologies which help preserve it physically. In the traditional sense in which the term ideology is used, it is contrasted with science. If a particular idea is ideological, it cannot be scientific. This is the distinction made, for instance, by Mannheim. However, more recent developments in the sociology of knowledge have tended to question this distinction (e.g., Whitley, 1974). Part of the problem arises because the way the word science is commonly used has become very broad. In the senses in which we are using it here, we imply not merely ‘the body of scientific knowledge embedded in the journals’ but also ‘the pursuit of scientific knowledge by experimentation and analysis’ and ‘the institutional framework within which scientific work is done’. In this broad sense of science, there is no incompatability in claiming the existence of ideology within science, and whilst it would be desirable to reassert the old distinction, to do so would, as we show below, require us to re-label a great deal of neurobiology as not science at all, but only ideology. Such a re-labeling would only cause confusion at the present time. Another distinction which is sometimes made is between ideology in thepursuit of science and the resultant body of scientific knowledge itself. Thus it is argued that,

484

Steven P. R. Rose and Hilary Rose

for instance, whilst Lore&s sympathy for the Nazis may have provided reasons for his development of ethology, the resultant knowledge is pure and untainted. Similarly, whatever Eysenck or Jensen’s paradigmatic framework, or personal motivations, ‘the facts about IQ’ remain as part of scientific knowledge and must be respected. This, however, reflects muddled thinking; if a paradigm is ideological, the research conducted within its framework must, itself, be ideological. Insofar as the paradigm and the research it sponsors are ideological, they must be in error, for ideology introduces, to a greater or lesser degree, erroneous ideas into the paradigm. To avoid confusion, we must stress that in saying this, we are neither questioning the overt motivations of the individual scientists concerned, which we can grant to be sincere, nor are we challenging the objectivity of their measuring instruments. The issue is not that of Piltdown Man-type hoaxes, nor is it, in general, the deliberate bending of data, the ‘stretching’ of evidence or of measuring instruments to fit a hypothesis (though this may sometimes apparently occur, as Kamin, 1973, has argued in relation to Burt). Within a paradigm, scientific research may be impeccably objective; it may even work in terms of the generation of results which can be fitted to the model; it certainly can produce powerful technologies, as we show below. The point is, however, that the data which has been collected will be fitted into a scheme within the artificial world which does not correspond to objective reality but distorts it; the measure of this distortion is the measure of the extent of ideological penetration into the paradigm. In this context we may consider another commonly used phrase, at least amongst Anglo-Saxon scientists: ‘Bad science’. ‘Bad’ is not used in an ethical sense but to indicate poorly designed experiments, inadequate methodology or faulty paradigms. In this sense, we must maintain that an ideological paradigm inevitably generates bad science, although the reciprocal does not of course follow. A non-ideological (scientific in the narrow sense) paradigm may still not automatically produce ‘good’ science, and the paradigm itself may be faulty for internal and hence non-ideological reasons. Is it possible to have non-ideological paradigms ? Within the context of present society and our present understanding, we cannot answer such a question unequivocally. First, it implies that we can stand outside our own society, and whilst we would reject the pessimistic relativism which would maintain that we cannot achieve a classless, and hence ideology-free, society, our own perspectives as to what sort of science such a society would produce must be limited. Second, we have so far analysed the mediations between the external real world and the artificial world of scientific knowledge as though the relative importance of direct inputs and ideological inputs is the same for all sciences and at all times. However, there is no particular reason to assume that this is, of its nature, the case. It may well be that it is only at such times as the present, of social and intellectual crisis, that we can

‘Do not adjust your mind, there is a fault in reality’

485

glimpse the interconnections between science and the social order. Particular domains of knowledge may or may not be particularly susceptible to the interplay of direct and ideological inputs. There may indeed be substance in the often-heard argument that, as one progresses from the social through the biological and towards the physical sciences, so the contribution of the ideological inputs diminishes by comparison with the contribution of the direct inputs. Whilst there is no proof of this commonly held belief, it does seem to us that the nature of ideology in the biological sciences, and particularly neurobiology, is easier to see and analyse than in, say, physics or mathematics. However, this may equally well reflect our greater familiarity with the subject matter of neurobiology.

3. Reductionism as ideology The history of any science is characterised by a changing series of paradigms within which the principal problems of the discipline are set and solutions sought; often such paradigms - and their ideological inputs - are strikingly revealed by the language in which they are described or the metaphors which their models employ; so we find that in neurobiology, there has been a changing pattern of metaphors for human brain and behavior from the hydraulic and clockwork models of the seventeenth and eighteenth century to today’s computer, monkey and rat analogies (Rose, 1973). However, for orthodox Western neurobiology in the 197Os, there is no doubt that the dominant paradigms are set in an overall reductionist mould. Reductionism as a doctrine states that: (a) Sciences are arranged in a hierarchical order, varying from high-level disciplines such as economics and sociology to lower-level ones, such as biology, chemistry, and, at the base, particle physics; and (b) that events in higher-level sciences can be reduced on the basis of a one-for-one correspondence to events, and hence laws, appropriate to the lower-level science; ultimately, therefore, that physical laws can be derived which will subsume and explain sociology (as some sort of ‘special case’, naturally). The alternatives to reductionism are of three main types. The first argues the autonomy of the higher-level sciences and is content to provide explanations for a phenomenon within economics, sociology or whatever, without reference to lowerlevel sciences. The second includes varieties of vitalism which introduce special features, such as god, life forces or free will, when approaching the questions of the relationship of mind and brain. The third may be described as interaction&, which in its most philosophically well defined form becomes dialectical materialism. This argues that, whilst events at any given hierarchical level represented by the different

486

Steven P. R. Rose and Hilary Rose

sciences must correspond to events at higher or lower levels, they cannot be reduced, by the application of causal laws or relations, to lower ones; biology cannot be invoked to explain away sociology; instead there will be a dialectical interaction between them. To take a crude example of the differences between the reductionist, autonomist and interaction% types of paradigm, consider the question of the relationship between alcohol and drunkeness. For the reductionist, the ingestion of alcohol and its consequent biochemical interactions in the body are the cause of the behavior pattern described as drunkeness; for the autonomist, drunkeness results from the social situation; in the short-term a party, in the long-term, an individual’s unhappy past history or the entire structure of a society in which smashing the mind, at least temporarily, is a way out of social contradictions. For an interactionist, the biochemical/pharmacological interactions of C2H,0H with enzymes at the molecular level and changed patterns of cell firing at the physiological level are not seen as causes of the behavioral events but as alternative and equivalent statements about the system at different levels of analysis; not causal but identity statements are invoked, and the interactionist also recognises that the interaction works both ways, that the cellular events associated with C,H,OH ingestion themselves vary from individual to individual and time to time, depending upon a host of environmental circumstances; one way, vertical causality is rejected. In general terms, we are arguing that the characterisation of the scientific method as oppressive, as in the quotation from Habermas, should properly not be applied to science in general but to reductionist science in particular, and that this reflects the relationship between reductionism and the dominant ideology (see also Putnam, 1973). Thus, when we examine reductionist neurobiology, we find within it just those ideological elements which the previous section postulated. The position has become more complicated in that the modes of thinking of reductionism have become so dominant in recent years that they have come to constitute what may almost be described as an ideology of science itself, which not only claims that reductionism has universalistic importance, superseding all other forms of knowledge, but also denies the significance of any social input into the artificial world of scientific knowledge, which becomes in this conception nothing less than an increasingly accurate copy, in the artificial world, of the external real world. The ideology of reductionism is thus positivistic. But it also has ethical overtones, claiming to provide rules for the proper conduct of human society. The only true goal for mankind becomes, in this view, the systematic extension of the artificial world of science and the matching within it, point-for-point, of the external, objectively real world; the rationality and objectivity of reductionism replace all else; they provide their own guide to human progress. Science, a social product, becomes both the goal and the method for all

‘Do not a&st your mind, there is a fault in reality’

487

society. The clearest exponent of this viewpoint has been the molecular biologist and archetypal reductionist Jacques Monod; he has argued this case extensively in Chance and Necessity (1973). In addition, the reductionist ideology of the institutions of science, hierarchical, authoritarian, sexist and (in their assumption of the superiority of western modes of thought, racist) cannot be ignored.

4. Reductionist ideology in neurobiology In the case of neurobiology, we can isolate for description several different subparadigms, which we can class as molecular reductionism (including a sub-category we may refer to as genetic determinism), evolutionary reductionism and two classes of category reductionism - behaviorism and machine reductionism. The existence of these different forms of reductionism is at least partly due to the existence of differing forms of biological explanation, which may be defined (following Lewontin, 1969) as (1) explanation of biological phenomena in terms of the chemical and physical properties of the system under study, in response to the question ‘what is this system made of?’ and (2) explanation of biological phenomena in an evolutionary sense, in response to the question ‘how did this system arise?’ Reductionist tools are those which, either by simplifying the system under study (as, for example, by choosing a worm or sea slug, both favored laboratory animals), or by limiting the aspects of the system chosen for examination (as, for example, by considering only certain forms of ‘emitted’ behavior) make the experimental problem more approachable and subsumable with the general methods and theories of science developed for less-complex systems, living or non-living. Reductionist tools are well established in biology and have proved, especially over the last three decades, so powerful that to elevate the tool into a philosophical principle has been a ready step. Molecular biologists such as Monod (1973) or Crick (1973) have argued that in the long run all of biology is to be derived from a study of the properties of the macromolecules of which the cell is composed (such as DNA) and their interactions and may be best understood by studying the chemistry and organisation of the intestinal bacteria Escherichia coli or - even more reduced - a bacteriophage. Crick’s proposal that a major combined effort of the world’s biologists be made to ‘solve’ E. coli, the so-called ‘Project K’, is an example of the sort of Los Alamos thinking that this reductionism generates; it is noteworthy that, over the last decade, there has been a steady invasion of neurobiology by some of the world’s leading molecular biologists (e.g., Crick, Brenner, Nirenberg and even Monod). It must be emphasised that the utility for certain purposes of biochemical descriptions of brain phenomena or of the techniques of experimental psychology and

488

Steven P. R. Rose and Hilary Rose

ethology in analysing behavior is unquestioned. It is the philosophical stance adopted by many of the proponents of these viewpoints which concerns us here. In the following sections we will discuss some examples of the workings out of the reductionist paradigms. Because several of these, especially those drawn from the areas of evolutionary and category reductionism, are likely to cover familiar territory, we will concentrate here on discussing molecular reductionism, and refer only briefly to the others. Molecular reductionism Molecular reductionism can be found at its sharpest in the field of mental disorder. What is the cause of schizophrenia? Is it to be seen - as the school of ‘orthomolecular psychiatry’ would argue - in the absence of certain key chemicals in the brain, or in the presence of abnormal metabolites due to genetic disorders? If so, treatment is to be found by dietary modification or the development of drugs which antagonise in some way the abnormal metabolites. Following the lead of such individuals as Osmond and Smythies (1952) and (in recent years) Pauling, this school argues strongly that there is an organic, brain-located cause for the behavioral manifestation. This belief has a long history, for at all stages in the development of biochemistry, the fashionable molecule of the moment has tended to be implicated as the cause of schizophrenia, from glutamate in the 1950s through an abnormality of ATP metabolism in the 1960s to today’s attention to the problem of galactose in the diet. Whatever the proximate biochemical cause, there is on this thesis an underlying genetic defect, The genetic analysis of schizophrenia has been a propensity to be schizophrenic. pioneered by such men as Slater and, although the original single-gene models are now relatively disfavored by comparison with polygenic effects, all the classical apparatus of the clinical biochemists, in their search for bizarre metabolites in schizophrenic brain, blood, urine or sweat, and of the human population geneticists with their hunt for monozygotic and dizygotic twins and heritability estimates, have been employed (for reviews, see, e.g., Iversen and Rose, 1974). By contrast sociological and social psychological explanations of schizophrenia stress the social and familial environment and the personal history of the individual; they point, for instance, to the different distribution of schizophrenia with family type and across social class. This approach is most commonly associated at the present time with the names of Laing, Esteson and Cooper. (e.g., Esteson and Laing, 1970). All three deny any biological role in schizophrenia, essentially adopting an anti-reductionist position of the type described above as emphasizing the autonomy of a higher-level science. The dichotomy between the two approaches is complete ; the paradigms meet only at the level of mutual abuse, as witness the extraordinary

‘Do not adjust your mind, there is a fault in reality’

489

correspondence that ensued in Science following the publication there of Rosenhan’s now-classic paper ‘On being sane in insane places’ (Rosenhan, 1973; Rosenhan et al., 1973). The ideological components within the reductionist paradigm are apparent; the inborn view of schizophrenia at once refuses to admit criticism of social structures, such as the family and alienated work forms, whilst at the same time encouraging a manipulative view of treatment. This is even more apparent when we look at the respective analyses, biochemical or social, as applied to the affective disorders such as depression. Those who argue a biochemical cause of depression, such as Sargent (1967), look for treatment by way of anti-depressant drugs; treatment is effective if it adjusts the depressed individual (typically a woman post-natally or around menopause) back into an acceptable social role, such as that of the good housewife or mother. The stability and appropriateness of the social order is taken as a natural given in this situation, and the job of the psychopharmacologist and clinician is to chemically fit people to it; it is not surprising to learn that 50 million patients were given chlorpromazine within the first decade of its use, or that 12 million barbiturate and 16 million tranquillizer prescriptions are issued a year in Britain. Note that we do not argue that the drugs themselves do not ‘work’ in the sense that they affect an individual’s responses and performance, often, though not always, in the predicted and hoped for (by the clinician) direction. However, even here there are some difficulties because one consequence of the reductionist mode of thought is that drugs are supposed to have a single site of action; effects other than those sought for are seen as ‘side-effects’ to be eliminated. The complexity of drug-behavior interactions, which has been so revealingly brought out even in the context of an apparently much less controversial agent, L-Dopa, in the treatment of Parkinsonism, tend to be dismissed by the reductionists. But this is only a relatively marginal point; the crucial issue is that the ideology which reflects itself in a reductionist model of the cause of schizophrenia and depression, and upon which is based a vast research and development program at all levels, from the universities to the drug house, also finds expression in an output at the social level, which essentially regards individuals as objects, to be manipulated into requited social patterns. Compare the average clinical research paper’s description of the behavior of patients exposed to given agents with the multi-leveled account of the patients to whom Sacks gave L-Dopa, in his book Awakenings (1973). For each of his 20 patients Sacks provides an account of the workings of L-Dopa, in terms of both minutely observed behavior asssociated with different levels and occasions of giving the drug, and also of the personal histories and present relationships of the patients. Throughout there is a constant emphasis on the integration of all the effects of the drug into the complex individual situation. Sacks’ model is clearly an inter-

490

Steven P. R. Rose and Hilary Rose

actionist,

dialectical

one, truly scientific rather than ideological.

The search for a biological rationale for problems of the social order has reached new heights in recent years with the development of a new clinical concept, that of ‘minimal brain dysfunction’ (MBD), which, an almost unrecognized category a few years back, was recently warranted a full symposium of the New York Academy of Sciences (Cruz, Fox and Roberts, 1973). Minimal brain dysfunction is essentially a behavior-defined syndrome; that is, the concept of brain dysfunction is invoked to explain a particular pattern of socially disapproved behavior, although no brain abnormality can in fact be detected by physiological techniques. In general, minimal brain dysfunction is supposed to be a disease of childhood and has derived from an extension of the concept of ‘hyperkinesis’, a disease state believed to characterise an overactive child. In Britain, there are estimated to be a few hundred children categorised as hyperkinetic. A large proportion are institutionalised; they are described as being unable to sit still without forcible restraint. In the USA, the diagnosis of hyperkinesis has become much broader, to cover a very large group of children who show behavior problems at school, being poor learners, inattentive in class and disrespectful of authority. Amongst the diagnostic signs for minimal brain dysfunction are, according to Wender (1971, p. 20) being ‘agressive socially . . . playing with younger children, and, if a boy, with girls’. For all these patterns of behavior, treatment with amphetamine, or its congener Ritalin, is proposed. Indeed, Wender waxes euphoric about the effects of Ritalin treatment, going so far as to claim that minimal brain dysfunction children may be regarded as suffering from ‘hypoamphetaminosis’. Under the drug ‘children often begin to talk about and behave in a manner consistent with their parents formerly unheeded “oughts” and “shoulds”. . . One bright 8-year old referred to d-amphetamines as his “magic pills which make me into a good boy and make everybody like me”‘ (Wender, pp. 90-91). The child turns from a ‘whirling dervish’ into being ‘quiet, compliant’ and with an ‘improved class behavior, group participation and attitudes to authority’ under medication (Wender, p. 94). What is more, Ritalin is cheaper than ‘expensive, non-organic therapies’ (Wender, p. 131).1 Small wonder that Ritalin is now prescribed, at a daily dose of 5-40 mg and on the basis of school reports, to 250,000 U.S. school children daily. Here indeed is a conspicuous success for a reductionist research and development program that discards any alternative explanation for a child’s inattentiveness in class or poor attitude to authority; not even other biological factors which may have apparently similar consequences, such as nutrition, are considered, still less that inattentiveness in class 1. In Britain, amphetamine prescription is generally discouraged by the BMA. Connoisseurs of British children’s comics will also

note the strong resemblance of Wender’s disapproved of hyperkinetic child to the weekly Beano hero Billy Whizz.

‘Do not adjust your mind, there is a fault in reality’

491

may reflect poor teaching or an irrelevant educational program, or that disrespect for authority may represent a more socially appropriate response to oppression than does subservience. Biochemistry is not the only brain discipline whose reductionism has both ideoPhysiology and anatomy have shown similar logical and direct social significance. tendencies. Over recent years it has become increasingly apparent that cellular activity in particular brain regions is associated with specific behavior patterns, so that, for instance, when certain regions of the hypothalamus are electrically stimulated by implanted electrodes in the cat or rat, then, depending on the particular cells stimulated, the animals show hunger, thirst, satiety, anger, fear, sexual arousal or pleasure. Surgical removal of these regions is associated with a behavioral effect reciprocal to that of cellular stimulation. The reductionist interpretation of these experiments is that the firing of particular cells in the hypothalamus causers anger, sexual arousal, etc., and, as with the biochemists, the social technologies which have emerged, notably in the hands of Delgado (1971) have been human experimentation in which schizophrenics and low-IQ patients have permanently implanted electrodes to the hypothalamus, remotely radio controlled by the doctor/experimenter. Passing current through the electrodes is associated with sharp mood changes in the patients. Once again, the patient’s anger, arousal and so on is seen as a consequence of the functioning of particular brain cells; the cells can be manipulated and so the patient can be manipulated, irrespective of the external circumstances which might be expected to affect the individual’s mood. According to Delgado, implanted electrode studies and their utilisation in practice can be expected to develop substantially in the next few years. Still more revealing is the recent growth in popularity of psychosurgical techniques in the U.S.A., and also in Britain, Japan and other countries (Breggin, 1972). The protagonists of these techniques argue that particular behavioral patterns are associated with malfunction or hyperfunction of particular brain regions, so that the appropriate medical strategy is the removal of these regions, a surgical approach which is a modification of the old prefrontal lobotomy popular for use of schizophrenics in the early 1950s but more recently a relatively declining treatment. Increased knowledge of the hypothalamic centres and related regions of the limbic system, associated with fear, anger and similar emotional responses, has led to a considerable ramification of these techniques. Surgical removal of such brain regions as the amygdala has been both proposed and practiced to deal with individuals suffering from behavior problems without any obvious ‘organic’ brain dysfunction. Such psychosurgery is intended as a pacifier, producing better adjusted individuals, easier to maintain in institutions or at home. In the U.S.A. the commonest groups of patients are claimed to be working-class blacks and women. A book by two

492

Steven P. R. Rose and Hilary Rose

psychosurgeons (Mark and Ervin, 1970), has drawn on the circumstances of the revolt in the U.S. cities to query whether there may not be brain abnormalities present in ghetto militants, to be cured by amygdalectomy. Their estimates argue that between 5 and 10% of Americans might be candidates for such treatment. Nor is this discussion purely theoretical; psychosurgical research has been supported by law enforcement agencies in the U.S. Whilst in a recent Detroit court case, proposed brain surgery on a prisoner was ruled illegal despite his consent, the number of actual operations conducted continues, according to Breggin, to increase. An indication of the candidates for such operations is provided by an exchange of correspondence between the Director of Corrections, Human Relations Agency2 (sic) Sacramento, and the Director of Hospitals and Clinics, University of California Medical Center, in 1971 (Opton, 1973). The Director asks for a clinical investigation of selected prison inmates, ‘who have shown aggressive, destructive behavior, possibly as a result of severe neurological disease’ to conduct ‘surgical and diagnostic procedures . . . to locate centers in the brain which may have been previously damaged and which could serve as the focus for episodes of violent behavior’, for subsequent surgical removal. An accompanying letter describes a possible candidate for such treatment, whose infractions whilst in prison include problems of ‘respect towards officials’, ‘refusal to work’ and ‘militancy’, he had to be transferred from prisons because of ‘his sophistication . . . he had to be warned several times . . . to cease his practicing and teaching Karate and judo. He was transferred . . . for increasing militancy, leadership ability and outspoken hatred for the white society . . . he was identified as one of several leaders in the work strike of April 1971 . . . Also evident at approximately the same time was an avalanche of revolutionary reading material’. To which request, the Director of Hospitals and Clinics replies, agreeing to provide the treatment, including electrode implantation ‘on a regular cost basis. At the present time this would amount to approximately $1000 per patient for a seven-day stay . ..’ Clearly such cases are those to whom the label ‘these animals are dangerous; when attacked they bite’ might well be attached. Once again, the reductionist slogan is the reverse of that painted on the Oxford College wall: ‘Do not adjust your mind; there is a fault in reality.’ In such examples of molecular reductionism we see an amalgam of all those features of ideologically based science discussed above. The research paradigms do not merely dictate the experimental operations conducted, such as the search for abnormal metabolites or particular ‘centers’ in the brain, but have an ideological significance which lies both in determining scientific directions and in providing a 2. What need for a Ministry of Love?

‘Do not adjust your mind, there is a fault in reality’

493

powerful scientific rationale for particular social interests. But, not only do these paradigms provide ideological support for the existing social order (it is your brain that is at fault if you are disaffected); they provide a set of social technologies which help maintain precisely the same social order.

Genetic determinism Genetic determinism represents a particular paradigm within the broad framework of molecular reductionism. Its research program is based upon the premise that all human behavioral characters can be analysed as representing the algebraic sum of two components - a contribution from genetics and a contribution from the environment, with a further separable balancing item for interaction included. From this premise follows the belief that experiments can be devised to answer the question: ‘How much does environment, and how much does heredity, contribute to the difference in behavior between individuals or between populations?’ Note that such a question with its implicit claim that behavioral characters can be teased apart and reduced into elemental components which summate is intrinsically reductionist. While genetic analysis of particular behavioral traits has been attempted with non-human animals, it is with humans that most work is concerned. Heritability studies have been performed on traits such as schizophrenia and other mental disorders; there have been attempts to explore the genetic basis of criminality (for instance, the attention paid to proposed relationships between XYY chromosomal abnormality and violent crime). But most interest has undoubtedly been associated with the work on human behavior and its supposed class or racial distribution. There is a long tradition of such studies, going back even beyond their generally regarded progenitor Galton, but the contemporary controversies are indissolubly linked with the names of Eysenck, Jensen and Hernnstein. Perhaps what is most interesting in this resurrected debate is the fact that relatively little at the core of the present arguments is fresh, despite Eysenck and Jensen’s claim to scientificity. A historic-sociological survey of the contribution of science to racist or class supremacist ideology would be valuable, particularly of the thirties, but certainly also during the expansion of the British Empire, together with its recrudescence during the present period when black and working-class people question the legitimacy of the stratification system (Rose, Hambley and Haywood, 1973). The fatal flaws in the arguments advanced by Eysenck and Jensen have been analysed at length, in this journal and elsewhere (e.g., Chomsky, 1972; Layzer, 1972). Indeed, the job was effectively done in the 1930s by Hogben and Haldane (e.g., Haldane 1938). There is no need to repeat the analysis here; our interest is to examine the propositions on which Eysenck-Jensen-Hernnstein position is based. Amongst many

494

Steven P. R. Rose and Hilary Rose

reductionist strands, we may include (1) the reduction of the sociological concept of race to a biological concept. Whereas the social definition of race depends on social ascription based on real or presumed cultural or physical differences which differ from time to time and society to society, the biological use of race refers to the relative discontinuities that occur in the distribution of a character within a species, corresponding to differences in gene frequencies, which need not be reflected in observable physical differences. ‘Biologism’ to use Hogben’s term of the 193Os, claims that socially defined races can be reduced to biologically defined ones, a trap into which Jensen and his followers persistently fall (vide Shockley’s emphasis on the percentage of Caucasian genes in the U.S. Black population). (2) The reduction of the broad concept intelligence to that of a single character (Spearman’s g) measurable linearly by means of an IQ test and scale. This is really a category reduction of the type referred to below. The IQ scale is then given the significance and stability of a measuring instrument more appropriate to the physical sciences. (3) The reduction of behavioral characters, such as intelligence (or rather IQ), to separable genetic components. It is because this reductionist framework is fundamentally ideological that experiments or observations conducted within it are bound to produce results which are themselves not scientific but ideological, however much they are dressed out in the trappings of scientificity (thus the issues have nothing to do with the questions of scientific responsibility or freedom to pursue truth - the forms in which they are frequently posed). The utility of Eysenck, Jensen and Hernnstein is, however, quite another question. As with the other examples ot molecular reductionism that were discussed above, their role is manifestly that of providing support and justification for the existing social order. Just as ideas derived from the social order have clearly provided the framework within which Eysenck, Jensen and Hernnstein set their research program (of what relevance or interest is the ‘inheritance’ of ‘IQ’ except in a class and race-stratified society?) so the ‘results’ of such research help armor the social order against attack from those who would change it, providing an apparent biological inevitability for the facts of social stratification. As for the eugenicists of the 193Os, for whom unemployment ran in the genes (and who programmatically advocated sterilization for workers on the dole), so today genetic determinism is used to justify (explicitly or implicitly) a meritocratic structure, or the consignment of Black children to schools for the educationally subnormal and a subsequent fate of unskilled labor in which upward progress is blocked. At the same time, the race/IQ argument performs another useful ideological role; it is by its nature divisive, not merely sharpening the class division of society but also, within the working class, helping exploit the division between black and white and weakening opposition to oppression.

‘Do not adjust your mind, there is a fault in reality’

495

Evolutionary reductionism

Amongst the several reductionist strands present within genetic determinism, some perhaps belong more properly to that paradigm-group we have classed as evolutionary reductionism. The most clearcut and prominent examples of this paradigm are provided by certain tendencies within ethology. The development of ethology, evolving as it did in reaction to the sterility of behaviorism, has certainly provided a new approach to an understanding both of patterns of behavior and of relationships between individuals of a species, which has enriched neurobiological understanding of the complexities of social behavior. However, it has also manifest within it some very obvious and vulgar reductionist models, typified by, for instance, Desmond Morris’ Naked ape (1968) in which he argues that human conduct is most fruitfully interpreted, predicted and controlled in the light of studies of other primates. Whilst Morris’ more extreme books, or for that matter Ardrey’s Territorial imperative (1971) or the Russell? Violence, monkeys and man (1968) are by and large deplored by professional ethologists as being oriented towards the lay rather than the professional audience, they are nonetheless influential in determining research both within ethology and in neighboring areas (e.g., Tiger and Fox 1972). What is particularly apparent in these publicist accounts of ethology is the clarity with which they articulate some of the central dogmas of ethological authority. Thus the innate aggressiveness of humans is claimed directly by such distinguished experimentalists as Lorenz and Eibl-Eibesfeld (1970) whilst Ardrey’s exposition of territoriality in man derives sustenance from Wynne-Edward’s (1962) studies on territoriality in red grouse on Scottish moors, extrapolated to the human world. Reductionist paradigms, in which a mode of operation becomes elevated by some invisible hand into a principle, is like goal displacement in organizations; a kind of explanation displacement occurs so that research which may provide an elegant account of certain aspects of animal behavior is displaced into a total account of the whole human condition. Scarcely surprisingly, if humans are interpreted as illsupressed bundles of aggressive instincts, the formulation for social policy relates to control rather than liberation. Thus an ethologically based legitimation for conserving the social order is provided by the dominance hierarchy (‘pecking order’) studies; stratification is not associated with specific societies and cultures but reflects a genetically laid-down necessity. The limitations of this particular type of ethological approach have been criticized by Bateson (1974) who has pointed out that not only do studies of pecking orders and dominance hierarchies relate merely to particular species examined under particular conditions, but, in addition, even within a group, the pecking order itself is not rigidly ordained but relates rather precisely to a

496

Steven P. R. Rose and Hilary Rose

particular type of experimental situation; in other situations quite different orders may obtain, so that different hierarchies may be apparent between, for example, eating activities and sexual activities; reductionist analyses of hierarchies, even in non-human species, must be replaced by interactionist ones which take into account the environmental circumstances of individuals and their past experience. But a reductionist ethology is one which, by definition, appropriates a set of linear and pared-down analyses of particular situations and therefore is far more prone to extract out from the richness of the experimental data the simplistic and linear concept of a pecking order or a dominance hierarchy. Insofar as the social and political beliefs of such ethologists ale apparent from their writings, there are few areas of contemporary neurobiology in which ideology stands out so sharply as in the work of ethologists such as Lorenz, Eibl-Eibesfeld or Morris. In this respect, reductionist ethology in the 1960s and 1970s has played and is playing the same role that Darwinism in the form of social Darwinism played in the nineteenth century. Then, Victorian capitalism was interpreted as obeying the iron laws of biology; the struggle for existence and the survival of the fittest demanded a laissez-faire economy at home and legitimated imperialism and colonialism abroad. Today, managerial capitalism, bureaucracies and social stratification, and social conflict of all types from football hooliganism through class war to struggles of national liberation and wars between national states, are seen as the inevitable result of human evolution from the primate. Ignoring the entirely new dimensions to human behavior generated by the human capacity for communication, social existence and above all production, economic and sociological explanations of conflict or social structures become diminished to the working out of the evolutionary imperative.

Category reductionism

This type of reductionism may be distinguished in its intentions and the models it uses, from the molecular and evolutionary reductionism paradigms we have outlined above. Broadly, it functions by proposing a research program whereby the complexity and richness of human activity is compressed into a model system. The properties of this model system are studied and then extrapolated back into the human situation once more. As the ambitions of the research program expand, so there is an increasing tendency for its protagonists to attempt to define more and more of the human situation in terms of the model; the model comes to replace or redefine the reality. As with other classes of reductionism, the research program of category reduction is ideological in that the models proposed are derived from within the expectations and experience of the dominant class; insofar as they lead to proposals for action, the

‘Do not adjust your mind, there is a fault in reality’

497

outputs from the research program are inevitably in their turn ideological. The most familiar example of category reductionism is Behaviorism; the framework into which all human behavior is to be reduced is that of contingencies of reinforcement. The scientific vacuity and ideological role of this reduction has been amply argued by Chomsky (1972) and does not need further demonstration here. Instead we will refer to another example of category reduction which has so far received rather less attention. Mechanistic reductionism Mechanistic reductionism replaces the evolutionary reductionist’s use of animal models for human behavior by models derived from the behavior of machines. There is a long history to this approach (golems were mediaeval concepts, Frankenstein’s monster a nineteenth-century one), hydraulic, clockwork or telephone exchange models of brain function had a long run for their money. The field of artificial intelligence proper, however, derives from the post-war development of cybernetics and the technology of computers. The aims are to simulate particular brain and behavioral activities using computers or related models, sometimes coupled to robot outputs, the objectives being twofold: To gain greater understanding and predictive power over brain function by providing logical models thereof and, in so doing, to generate machines with properties which will allow them to substitute for humans in a variety of ways. Assessment of this area has not been helped by the rather dramatic pronouncements of many of its protagonists, who have described it as artificial intelligence and claimed not only that adequate brain models were just around the corner but that before very long computers would take over the world as a higher evolutionary form than man, reducing humans to nothing other than a temporary symbiotic role (a computer’s way of making another computer). The onward march of computers in this conception was supposed to be aided by humans in the name of either a technological imperative, or, as sometimes conceived, a sort of evolutionary imperative, to make better and higher forms of intelligence. The nature of these claims should not be allowed to obscure the relatively modest advances that have actually been made in computer modelling: A few robots which can laboriously build towers of bricks, teaching machines, chess-playing computers and a series of largely unsuccessful attempts to produce pattern-recognizing devices and translating machines. This relative lack of success has had the useful consequence of prompting modifications of existing theories of pattern recognition and of concentrating effort into linguistic analysis. But even the enthusiasts now see it as a long time off before computers will be given the vote, a prospect once suggested by Sutherland (1968).

498

Steven P. R. Rose and Hilary Rose

This modest performance may itself help focus attention on the adequacy of the paradigms of the computer modellers; essentially, they rest on the claim that human behavior can be modelled by a set of ‘logical neurons’ interacting in particular networks. It would be non-materialist to deny that, if one could build a network of logical neurons of the size and scale of the human brain (that is, with 1O’O units interconnecting by way of some 1014 synaptic junctions) one could generate a system with properties resembling the brain. But the logical networks of the modellers, with their few hundreds or thousands of units, each with a very small number of possible interactions, is so different in scale as to be on a level which is largely irrelevant to that of the brain; models can be made only of the brain’s most straightforwardly ‘logical’ functions ; rapid arithmetic, chess playing or even pattern recognition - but not emotional responses, creativity, or, for that matter, poker playing. One point to be borne in mind is that the model is only as good as its program. If the program cannot handle the richness of environmental interactions and their essentially processlike rather than static quality, it can never replicate the human experience. But within an overall reductionist theoretical framework, a valid model cannot be produced. The value of a model lies only in its capacity to test between possible alternative theories of particular functions. When it is not easy to predict mathematically which of two possible variants of a model will produce a given output from a given input, to actually produce the model in hardware may resolve this difficulty. But first, a theoretical insight into the nature of the processes under examination and a model of adequate complexity to deal with these processes is required. Where the models used have been essentially reductionist, as in the field of artificial intelligence over the last two decades, it becomes necessary to look at the validity of the approach in terms of its sponsorship and its ideological potential. When this is done, it cannot fail to be noted that in this field above all, there has been a preponderance of support for research from military and space agencies. This is far more strikingly true here than in any other area of neurobiology. The possible value to the military of translating machines, sophisticated computers or automated surveillance devices goes without saying; it is arguable that one aspect of the technology which has been derived from investment in this and related areas since the Second World War has been the development of the so-called automated battle-field in Vietnam, and new outputs along these lines are not hard to envisage. The ideological outputs of the field of artificial intelligence derive primarily from its over-simplistic approach to human behavior and performance. It is not that to make models of behavior is itself an ideological act, but to make models of behavior which are based on reductionist assumptions either about the behavior to be modelled or the nature of the model which will predict the behavior is by definition ideological. To regard man as a machine, a soft computer, is by definition to stress his most

‘Do not adjust your mind, there is a fault in reality’

499

machine-like aspects and to make him, as with machines, more biddable, programmable, subservient to the demands of the machine-maker or programmer. This must be so, until such time as the machines are really as complex and as interactive as are humans. If human intelligence is to be delimited by an analogy with the limited capacity of machine intelligence, then the prescription for the treatment of humans becomes related to the treatment of machines. They may be well serviced and maintained, but they should not get above themselves, or else, by extension, it becomes perfectly proper to give machine tools, IBMs - or even automatic lifts - the vote.

5. Conclusions: Towards a non-ideological

neurobiology

Throughout this account of reductionism in neurobiology, we have attempted to show both that reductionism is more than merely ‘bad science’, in the Anglo-Saxon sense mentioned above, but that it is bad science because it is ideological, that is, its research program and organizing paradigms are permeated with those ‘ruling ideas’ which express class interest and that the technologies that they generate are essentially defensive of that class interest, serving to protect it both physically, by manipulating and pacifying would-be protesters, and culturally, by providing an apparent biological justification for the social order. But as the analysis has proceeded, we have counterposed to reductionist models those provided by interactionism or dialectical materialism. We have not dwelt at any length with the autonomist paradigms. One strand of autonomism has a long history which ranges back through three centuries of past struggles between science and non-scientific forms of thought, such as religion. But more significant are the newer versions of autonomism which, as with Laing, represent a rejection of biology that is but one aspect of the broader rejection of ‘science’ advanced by Roszak. It is easy to see how this rejection has emerged in its revulsion from the oppressive role of reductionism. Nonetheless, whilst one may sympathize with the attempt to counter the manipulation and objectification of the individual which reductionism imposes, and to replace the human subject as the centerpiece of events and actions, we must reject autonomism as inadequate; biology (and the natural sciences in general) cannot be disregarded, and to attempt to do so is to run the risk of developing counterideologies which in their consequences are just as oppressive for the mass of the people as those which autonomism rejects or opposes (Rose and Rose, 1969, 1972). An interactionist or dialectical model avoids these dilemmas, as it avoids many older philosophical questions of the relations of mind and brain, like dualism, epiphenomenalism and psycho-physical parallelism (see Payne, 1968; Graham 1971). It recognises both the biological and the cultural and social uniqueness of the individual

500

Steven P. R. Rose and Hilary Rose

by way of its version of the identity hypothesis. It permits the development of hypotheses within which, for example, the consequences of drug administration are considered in the context of a system which includes not only the individual to whom the drug is administered but also his/her past history and hence unique brain states, and the social context in which that drug is administered. ‘Side-effects’ are no surprise in a dialectical model (though the term is rejected) nor are the extremely variable effects of taking alcohol, marijuana or amphetamine, in the context of an open, interacting, cross-hierachical system model of the brain and its relationships. Nor, within such a context, would the analysis of biological correlates of human performance or even intelligence (were it felt that such a term was useful) be necessarily ideologically oppressive. We cannot, in this paper, attempt to provide a full justification of the case for regarding an interactionist or dialectical neurobiology as inherently less oppressive, more ideology-free, than reductionism, and hence ‘good science’ as distinguished from ‘bad science’. There have been several attempts to spell out this position more articulately in recent years (Ratner, 1971; Riegel, 1972, 1973) which represent steps in this direction. But we recognize the difficulty of discussing the possibility of a non-ideological science when we live in a society in which, because of its class structure, our own thinking is likely to remain ideologically saturated. Thus we do not believe that it is possible to do good neurobiology in an oppressive society - but by analysing the ideological components with our thinking, we can at least hope to do better neurobiology; human history, and the history of science, is not static but capable of creative and conscious transformation; a new neurobiology is present in the womb of the old.

REFERENCES Ardrey, R. (1971) The territorial imperative. New York, Dell. Bateson, P. P. G. (1974) Are hierarchies necessary? London, Brain Research Association. Breggin, P. R. (1972) U.S. Congressional Record H.R. 118 (26), Washington. Chomsky, N. (1972) Psychology and ideology. Cog., 1, 11-46. Crick, F. H. C. (1973) Project K: The complete solution of E. Coli., Perspect. Biol. Med.,

17, 67-70.

Cruz, F. F. de la., Fox, B. H., and Roberts, R. H. (1973) Minimal brain dysfunction. Ann. N. Y. Acad. Sci. 205 (entire volume).

Delgado,

J. M. R. (1971) Physical control of the mind: Towards a psychocivilized society. New York, Harper and Row. Eibl-Eibesfeldt, I. (1970) Ethology, the biology ofbehavior. New York, Holt, Rinehart and Winston. Esteson, A., and Laing, R. D. (1970) Sanity, madness and the family. Harmondsworth, Penguin. Graham, L. (1971) Science and philosophy in the Soviet Union. New York, Knopf. Habermas, J. (1971) Towards a rational society. London, Heinemann. Haldane, J. B. S. (1938) Heredity andpolitics. London, Allen and Unwin.

‘Do not adjust your mind, there is a fault in reality’

Hessen, B. (1931) The social and economic roots of Newton’s Principia. In N. Bukharin (Ed.) Science al fhe crossroads. Pp. 146-212. [Reprinted London, Cass, 19711 Iversen, L. L., and Rose, S. P. R. (1973) Biochemistry andmentaldisorder. London, The Biochemical Society. Kamin, L. (1973) Heredity, intelligence, politics and psychology. Paper to Eastern Psychological Association Convention. Layzer, D. (1972) Science or superstition? Cog., 1(2-3), 265299. Lewontin, R. C. (1969) The bases of conflict in biological exploration, J. hist. Biol. 2, 35-46. Mark, V., and Ervin, F. (1970) Violence and the brain. New York, Harper and Row. Marx, K., and Engels, F. (1846) The German ideology. Eng. Trans., 1947, New York, International Publishers. Morris, D. (1968) The naked ape. London, Cape. Monod, J. (1973) Chance and necessity. London, Cape. Musgrave, I., and Lakatos, A. (1970) Criticism and the growth of knowledge. Cambridge, Cambridge University Press. Opton, E. M. (1973) Documents circulated at Winter Conference on Brain Research, Vail, Colo. Osmond, H., and Smythies, J. R. (1952) Schizophrenia, a new approach. J. ment. Sci., 98, 309-315. Payne, T. R. (1968) S. L. Rubinstein and the philosophical foundations of Soviet psychology. Dordrecht, Reidel. Putnam, H. (1973) Reductionism and the nature of psychology. Cog.., 2(l), 131-146. Ratner, C. (1971) Principles of dialectical Psychology. Telos, 9, 83-109. Riegel, K. F. (1972) Influence of economic and political ideologies on the development of developmental psychology, Psychol. Bull.,

501

78 129-141. (1973) An epitaph for a paradigm, Hum. Devel., 16, 1-7. Rose, H., and Rose, S. (1969) Science and society. London, Allen Lane. and Rose, S. (1972) The radicalisation of science. In R. Miliband and J. Saville (Eds.) Socialist register. London, Merlin. Pp. 105-132. Rose, S. (1973) The conscious brain. London, Weidenfeld & Nicolson. Hambley, I., and Haywood, J. (1973) Science, racism and ideology. In R. Miliband and J. Saville (Eds.) Socialist register. London, Merlin. Pp. 235-260. and Rose, H. (1973) Can science be neutral? Perspect. biol. Med. 16, 605-623. Roszack, T. (1969) The making of a counterculture. London, Faber & Faber. Rosenhan, D. L. (1973) On being sane in insane places. Sci., 179, 250-258. et al. (1973) Correspondence, Sci., 180, 356-369. Russell, C., and Russell, W. M. S. (1968) Violence, monkeys and man. London, Macmillan. Sacks, 0. (1973) Awakenings. London, Duckworth. Sargent, W. (1967) The unquiet mind. London, Heinemann. Sutherland, N. S. (1968) Machines like men. Sci. J. (September). Tiger, L., and Fox, R. (1972) The imperial animal. London, Seeker & Warburg. Wender, P. H. (1971) Minimal brain dysfunction in children. New York, Wiley Interscience. Whitley, R. (1974) Socialprocesses of scientific development. London, Routledge, Kegan Paul. Wynne-Edwards, V. C. (1962) Animal dispersion in relation to social behaviour. London, Oliver & Boyd.

-

502

Steven P. R. Rose and Hilary Rose

R&urn&

Ce texte s’attache ii la relation entre la connaissance scientifique et le contexte social dans lequel la science est pratiqube. Etant donne la relation existant entre la science et les hommes de science dune part, et l’Etat d’autre part, il se peut que les paradigmescientifiques soient ideologiques et que beaue coup de ce qui est fait au nom de la scienc, soit en fait ideologiquement sature. L’ideologie dominante en neurobiologie est celle du reductionnisme, et de fait ses partisans soutiennent que le reductionnisme constitue la methont scientifique, avec des resonances moraled alors que les opposants a la science citent le reductionnisme comme illustration de la nature inevitablement oppressive de la science. Au sein de la neurobiologie, plusieurs types de reductionnisme sont pris en consideration. Le r6ductionnisme mokhlaire recherche des causes biologiques a des kenement observes dans la societe, comme par exemple la schizophrenic ou l’ttat depressif. Ses consequences, au plan du ‘dysfonctionnement minimal du

cerveau’ et de la psychochirurgie sont exminesLe diterminisme gdnbtique lie, par exemple, a l’intelligence (Q.I.) constitue une autre forme de ce type de reductionnisme. Le rkduction. nisme &oZuti,cherche a expliquer le comportement humain par le comportement des primates ou d’autres animaux, visant ainsi a justifier l’ordre social existant comme dependant dune base biologique. Le rkductionnisme c&goriel comprend le ‘behaviorisme’ et le rtductionnisme mtcaniste. Dans ce dernier cas, on tente de reduire le cerveau aux interactions de reseaux de neurons logiquement modellisables de facon simple. A l’encontre du reductionnisme, I’autonomisme en tant que type de paradigme distinct, est bribvement etudie et rejete. Un paradigme non ideologique, et done scientifique et nonen une version de oppressif, consisterait l’interactionnisme: le materialisme dialectique, mais une telle science ne peut pleinement voir le jour que dans une sock36 transform&e.