Do six-month-old infants perceive causality?

The idea of cause and effect is often assumed to originate in prolonged learning. However, the present findings suggest that 27-week-old infants may already perceive a cause-effect relationship. Reversal of an apparently causal event (direct launching) produced more recovery of attention following habituation than the reversal of a similar but apparently non-causal event (delayed reaction). In both cases the changes in the spatiotemporal properties of the stimuli were identical. Hence the infant's percept of direct launching may involve more than an encoding of its spatiotemporal properties. Since the same kind of stimulus gives rise to a causal illusion in adults, it may be that the additional factor at work is the perception of a causal relationship. This finding may be significant in terms of the modularity of the infant visual system and the later development of causal understanding. The idea of cause and effect lies at the heart of both commonsense and scientific thought. The question of its origins in psychological development has long been a topic of speculation (e.g., Gibson, 1984; Gibson & Spelke, 1983; Hume, 1740; Kant, 1781; Michotte, 1963; Piaget, 1955). Recent experimental studies of children show that even 3-year-olds employ fairly sophisticated causal ideas in understanding mechanical interactions (Bullock, 1985; Bullock, Gelman, & Baillargeon, 1982; Kun, 1978; Shultz, 1982). The origins of causality must, therefore, lie further back in development, perhaps in infancy. *We thank Uta Frith, John Morton, Graeme Halford and Jean Mandler for helpful comments on earlier drafts. Earlier versions of this paper have also formed portions of talks given to the Departments of Psychology at the Universities of St. Andrews, York and Manchester, to the 5th Biennial International Conference on Infant Studies, Los Angeles, April 1986, and to the Salk Institute, University of California, San Diego. Reprint requests should bc addressed to Ahm Leslie, MRC Cognitive Development Unit, 17 Gordon Street, London WC1H (IAH. U.K. 0010-0277/87/$7.70 © 1987, Elsevier Science Publishers B.V. 266 A.M. Leslie and S. Keeble Hume (1740) argued that only the spatial and temporal arrangement of events, and not their causal connections, could be sensed. Since+causal relations are not known by force of logic, our belief in a causal worl'd could only be the result of "imagination" on our part: a natural response of our minds to prolonged experience of events which occur constantly together and which are closely connected in time and space. If Hume had ever considered infancy, he would no doubt have thought that infants, lacking any substantial experience of the world, would only be able to sense the spatial and temporal arrangement of events, and have little or no knowledge of causality. Piaget (1955) speculated that infants might be sensitive to the feelings of effort that accompany action. He thought that this, together with detecting "statistical" associations between events or stimuli, might jointly be the basis for later causal understanding. Evidence has since accumulated on one of these points: infants can indeed detect contingencies between their own actions and events in the world (see Watson, 1984 for a review). There is also recent evidence that infants can perceive and remember the internal spatiotemporal structure of at least two objective events that appear causal to adults (Leslie, 1982, 1984a, 1984b, 1984c). For example, the spatial relation of contact between a hand and a doll while the hand picks up the doll, appears to be important to 6-month-olds in a way that contact between another similarly moving inanimate object and doll is not (Leslie, 1984a). Infants of this age also appear to be able to remember the degree of spatiotemporal continuity between the movements in collision events (Leslie, 1984b). The young infant's sensitivity to spatiotemporal correlates of causality, like contingency and continuity, is certainly suggestive. But there has been no direct evidence so far from these or any other studies that infants are able to perceive a specifically causal relation. Among the traditional approaches to causal perception, only Michotte (1963) has suggested that infants might have a direct impression of cause-effect as a sort of perceptual gestalt (cf. Rock, 1983, p. 134-138). This arises for adults from certain kinds of collision events, as, for instance, when one billiard ball launches another by striking it. But as Michotte showed, a causal percept can also be obtained with quite abstract stimuli such as marks on paper or coloured lights, so long as the movement pattern is right. Michotte argued that such stimuli gave rise to a perceptual illusion since the effect appeared to be obtained immediately, repeatedly and despite the observer's knowledge of how the display was actually produced. Since then, further work, while questioning a number of the details, has tended to support this central finding of Michotte's (Beasley, 1968; Gemelli & Cappellini, 1958; Powesland, 1959). Causality perception in young infants 267 Infantile perceptions: Michotte versus Hume Experimental test of these ideas with infants must proceed in a number of steps, each contributing to the overall picture. We have used the habituationdishabituation of looking technique for this purpose with sets of cinematic stimuli depicting a red object colliding with a green object in a variety of ways. Parsing subcomponents In the first of these studies, Leslie (1982) showed that infants can distinguish the continuous motion of a direct launching from similar but discontinuous events. Given this, the next question to ask is, Can they distinguish the submovements in the continuous direct launching or is this perceived simply as a single unanalysable "whoosh" from one side of the screen to the other? This was tested in the following way. We reasoned that if direct launching is seen as an event with a particular internal structure (i.e., composed of submovements), then reversing the event, by playing the film backwards, should rearrange that structure. If, however, an event has no submovements, then reversing it would only affect those properties, such as spatial direction, which do not depend upon structured subcomponents. The general idea here can perhaps be grasped by considering a linguistic example, like the word "houseboat". This word has lexical subcomponents which can be reversed to produce "boathouse". But where there are no lexical subcomponents (as in "vehicle"), reversal can affect only lower level (e.g., phonemic) structure. The idea then was to use reversal to probe for the infant's perception of internal structure in direct launching. To do this, Leslie compared the effect of reversing direct launching with the effect of reversing a single movement made by a single object (see Figure la & b). Since a single movement has no subcomponents, reversal will change only its spatial direction. Using an habituation-dishabituation of looking technique, one can predict the following from the subcomponent hypothesis; a group of infants habituated to direct launching and tested on its reversal will recover their looking more than a group habituated to a single movement and tested on its reversal. Leslie (1984b, Experiment 1A) used the above design. Both groups of infants were equated by presenting the same spatial direction change, but the direct launching group was hypothesised to see, in addition to this, a reversal of an internal relationship like temporal order. Of course, it was possible that both groups would recover their looking to ceiling level, since both changes might be discriminable. However, given that discrimination may be a neces268 A.M. Leslie and S. Keeble Figure 1. Illustration of sequences used in Leslie (1984b; Experiment 1 A & B). The open square represents the red brick, the shaded square the green brick. Each brick moved consecutively for I s (24frames) in direct launching and for 2.17s in the single movement fihns. Differences in sequence duration were compensated for by adjusting the stationary periods at the beginning and end of the sequence. Films were formed into loops for continuous projection. For more details about how these stimuli are constructed see Leslie (1984b) and below. Reversal of the top two sequences was produced by turning the projector #1tO reverse. (a) D I R E C T L A U N C H I N G E3 • [] • C ! O • O • (b) S I N G L E M O V E M E N T