Causality is the cognitive basis for the acquisition and use of categories and concepts in children. The previous papers in this conference have discussed whether causality has a single underlying mechanism for this ability (Watson), two systems (Reboul), or three domains (Inagaki & Hatano). Watson suggests that a single mechanism could be used to reason about objects and people and that there is no need for these to be divided. These views are in line with work by Onishi and Baillargeon (2005) which will be discussed below. Reboul highlights comparisons between human and non-human causal reasoning and suggests that the onset of language might be a critical difference between arbitrary causal knowledge and natural causal knowledge. Some of my work coincides with this view and I will describe data that relate to these points. Finally, the core domains of physics, psychology, and biology presented by Inagaki and Hatano present compelling data describing the different signatures that each of these domains present. Their views overlap with Spelke’s (2000) core domains, with the caveat that the domain names are slightly different and there may be more than three.

In this paper I take a developmental approach and investigate the origins of physical causality in infancy. By looking at change and continuities in infants’ abilities it may be possible to gain a better understanding of the fundamental aspects of causal reasoning. Another advantage of looking at young infants is that we can gain insight to the nature of representation abilities prior to the onset of language. If we can characterize the prelinguistic state of causal categories and compare it to adult abilities, then we can better understand the impact that language makes on our cognitive processes. I propose that physical causality is evident early in infancy and it develops in a category-specific pattern. This ability is likely to be universal among human and non-human primates and language alters our conceptual categories.

Knowledge about physical causality - defined as the way that objects behave and interact - is a central issue in cognitive development (Haith & Benson, 1998; Piaget, 1952, 1954; Spelke & Newport, 1998). The classic view of object representation was Piaget’s (1952, 1954) depiction of object permanence developing over the first 18 months of life. Qualitative shifts in cognitive concepts about objects were central to Piaget’s theory (e.g., out of sight is out of mind to the A not B error). More recently, theories of continuity and elaboration are used to explain developmental changes in object representation. Mandler describes a dual representation model where perceptual and conceptual capacities develop in parallel (Mandler, 2004). Baillargeon (2004) suggests that infants have some innate principles and over the first year they identify category-specific variables that allow them to predict the outcome of events more accurately. Spelke (2000) suggests that signature characteristics in object representation are ontogenetic - the characteristics of early object representation are still observable in adult object representation.

Experiments that test knowledge in preverbal infants rely on infants’ tendency to habituate to repeated events and look longer at events that they perceive as novel or unexpected. Baillargeon and her colleagues have used looking time methods for over twenty years to map the developmental trajectory of infants’ knowledge about physical causality. In this time we have amassed data that describe what infants know at different ages. More recently the agenda has switched to finding out how infants learn about objects (Baillargeon, 2004).

If the task before the infant is to learn about how objects behave and interact. The solution that infants seem to use is to divide the world into smaller categories of events and learn within each of these categories. This strategy implies that knowledge is bounded and doesn’t generalize to other categories. Baillargeon (2004) has revealed this pattern of learning for a wide variety of physical events including, support, collision, occlusion, covering, and containment.

To illustrate I will describe the developmental trajectory of containment events in infancy. Our first study investigated whether infants had different expectations about occlusion and containment events. To test this we recorded infants looking times to the two displays depicted below. The perceptual features of the two displays were very similar the only difference between them was whether the checkered object was lowered behind the container (an occlusion event) or inside the container (a containment event). There was an important conceptual difference across the displays. The container event has and unexpected outcome because the checkered object appears to pass through the side of the container. If infants detect this difference they should look significantly longer at the containment event compared to the occlusion event. Two-month-old infants looked significantly longer at the containment compared to the occlusion event suggesting that very young infants have specific expectations about these events (Hespos & Baillargeon, 2001b).

Further experiments investigated how these initial expectations change over time. Baillargeon’s (2004) model predicts that infants identify variables that allow them to predict the outcomes of events more accurately over time. For example, by 4 months of age infants have expectations about how much of an object should become hidden when it is lowered behind an occluder (Hespos & Baillargeon, 2001a). When infants were shown the events depicted below, they looked significantly longer at the short compared to the tall occluder event.

To investigate whether infants generalize their expectations to perceptually similar containment events we tested a new group of infants in conditions where the occluders were replaced with containers (see below). We tested 4-month-old infants in the container condition and they looked equal amounts of time at both events suggesting that they did not discriminate the short event as more unexpected than the tall event.  Next we tested new groups of 5- and 6-month-old infants and got the same result.  It wasn’t until the infants were 7.5-months old that they looked significantly longer at the short compared to tall containment event.  These findings suggest that infants’ knowledge about physical causality is category specific and does not generalize broadly across perceptually similar events. Further evidence of context-specific limitations were found for categories of containment vs. covers and covers vs. tubes (Wang, Baillargeon, & Paterson, 2005). In addition, the same developmental patterns have been revealed in reaching tasks testing knowledge of occlusion, containment, and support events (Hespos & Baillargeon, 2005, in prep).

The work of Baillargeon and her colleagues has revealed a reliable pattern in infants’ knowledge about physical causality for categories of occlusion, containment, support, and covering events. These findings beg the question of what constitutes and event category boundary? One possible answer begins with the observation that each of the events studied has a word in English that describes the spatial relationship (occlusion - behind, containment - in, support - on, covering – under).  This is not a novel idea to linguists, who know that languages vary in how they describe spatial relationships(Bowerman, 1996; Levinson, 1996; Sinha & Jensen de Lopez, 2000). One well-studied contrast is between English and Korean. For example, when Koreans say that one object joins another, they specify whether the objects touch tightly or loosely. English speakers, in contrast, say whether one object is in or on another. The book is on the table in English and is held loosely by the table in Korean.

Given the cross-linguistic differences, there are two possibilities for how these distinctions emerge. It is possible that infants do not have any categories until after they acquire language and their language creates the initial categories. Alternatively, languages may selectively enhance or diminish distinctions that are already there. Given the vast array of evidence that Baillargeon presents demonstrating that infants have expectations about physical causality as early as 2 months of age, it seemed unlikely that infants wait for language to create the initial categories. We decided to test this empirically by looking at preverbal infants and their knowledge about containment events.

The experiment used 5-month-old infants that were from a monolingual, English-speaking environment. We investigated infants’ ability to discriminate a categorical boundary that is captured in Korean but not English. More specifically, we tested infants’ categorization of tight-fitting versus loose-fitting containment relations using a habituation-dishabituation paradigm. First, infants saw a narrow cylindrical object lowered into a series of loose-fitting, medium sized containers on a series of trials until their looking time declined (see below). Next, the infants were presented with 6 test trials in which the same cylindrical object was lowered, in alternation, into a wide container (1.5 times wider, hence also a loose fit) and into a narrower container (1.5 times more narrow, a tight fit). If infants made a language-independent categorical distinction between tight- and loose-fitting containment events, then they were expected to look significantly longer at the tight-fit trials. Our results confirmed this prediction (Hespos & Spelke, 2004). Additional conditions replaced the checkered object in habituation trials with a wider object that was a tight-fit with the container and the reverse pattern of looking was revealed in test trials. We concluded that sensitivity to this distinction develops in the absence of any relevant linguistic experience.

A remaining question is whether speakers lose sensitivity to the conceptual distinctions that are not captured by the lexical semantics of their native language. To begin to address this question, we presented the same containment events to English-speaking adults and asked them to rate the similarity between the habituation and test events. In contrast to the infants’ patterns of preferential looking, the adults rated the two test events as equally similar to the habituation event. The adults, therefore, did not appear to make the same categorical distinction as the infants.

If the categories are not coming from linguisitic input, the question is where do these event categories come from? One possibility is that they come from knowledge about physical causality – the mechanical principles that govern how objects behave and interact. Because tight- and loose-fitting containment place different constraints on the motions of objects, it is possible that the principles governing infants’ representations of objects and their motions could also lead infants to categorize these spatial relationships differently. In the first experiment, we used a paradigm similar to Baillargeon’s experiments described above. First, infants saw a narrow cylindrical object lowered into a wide container until their looking time declined. Next, the infants were presented with 6 test trials that alternated between a move-separately event and a move-together event. In the move-separately event, the cylindrical object was lowered inside the wide container and then the container remained stationary and the object moved back and forth inside the container (expected event). In the move-together event, the cylindrical object was lowered inside the wide container and then both the object and the container moved horizontally as a unitary object (unexpected event). If infants expected the loose-fitting container to allow the object to move with some independence then they were expected to look longer at the move-together event. Our results confirmed this prediction.

In a second condition, we similarly tested infants’ expectations for the effects of motion on tight-fitting containment relations. Infants saw the same cylindrical object lowered into a narrow container during the habituation and test trials. In the test trials, infants saw the object inside the container move back and forth horizontally. On alternative trials, the object and container moved together (expected events) or separately (unexpected events). If infants appreciated that the tight-fitting container more strongly constrains the motion of its contained object, then infants were expected to show the opposite looking preference from those in the loose-fitting condition and look longer at the move-separately  event compared to the move together event. The results confirmed this prediction, 5-month-old infants have different expectations for horizontal movement in tight- and loose-fit containment.

In summary, this series of experiments suggest an interaction between language and physical causality. Five-month-old infants parse a continuum of the spatial variation into categories of spatial relationships between objects. Infants are sensitive to spatial distinctions that are lexicalized in non-native languages. These findings stress the theme of human universals that underneath all the things that vary across humans, (e.g, the language we speak, the meanings we convey) are a set of perceptual and conceptual capacities that are common to everyone and rich enough to capture the core meanings expressed by any language. The developmental trajectory is one where infants have more flexibility than adults because language has not influenced their causal categories. Taken together, these findings suggest that there is ontogenetic continuity in the development of physical causality but language may alter the category boundaries.

The evidence for ontogenetic continuity complements evidence for phylogenetic continuity in the capacity to represent objects and physical causality. In particular, monkeys represent objects similarly to human infants both in preferential looking and object search tasks (Antinucci, 1990; Hauser, MacNeilage, & Ware, 1996). In fact monkeys progress through Piaget’s stage sequences more rapidly than human infants. These findings fit well with the view that basic mechanisms of object representation are consistent over much of evolution and ontogeny, and that their expression depends in part on the developing precision of representations and that this development occurs at different rates for different species (Spelke, 2000).

This paper has focused on evidence for physical causality in human infants. Could the same mechanisms be applied to other domains of knowledge? Most of the infant data pertains to physical causality and investigates expectations about inert objects. However, there is a growing body of research about psychological causality in infants. Psychological causality is distinct from physical causality in that it involves agents (e.g., people or other living things) and the fact that these creatures can have goals and intentions that guide their behavior.

Woodward and her colleagues have a variety of studies that show infants’ expectations about people are different than their expectations about objects (Guajardo & Woodward, 2004; Woodward, 1998, 2003). For example, infants were habituated to a human hand grasping a bear on a stage. The bear was on one side of the stage and a doll was on the other side. In test trials, the position of the bear and the doll were reversed. They measured infants’ looking time to events where the goal (grabbing the bear) was the same or the action (grabbing the object in a specific location) was the same. Infants looked significantly longer when the human’s goal changed. Interestingly, when the experiment was repeated with a mechanical arm instead of the human arm the looking pattern was the reversed, suggesting that infants expect humans to have goal-directed actions but mechanical arms do not have goal-directed actions (Woodward, 1998). Further experiments demonstrate that infants will link goals with human actions at 7 months of age but that infants do not link eye gaze with goals until 12 months of age (Woodward, 2003). Recent evidence that slightly old infants can use eye gaze information comes from a paper that demonstrates that 15-month-old infants succeed at a modified theory of mind task (Onishi & Baillargeon, 2005).

In conclusion I have given a brief tour of evidence for physical causality in infancy demonstrating that as early as 2 months of age infants have expectations about the way that objects behave and interact. In addition there are context-specific patterns in the acquisition of physical causality. These patterns can help us better understand the fundamental aspects of causal reasoning. The current evidence suggests that there is ontogenetic continuity in our causal categories, this ability may be shared with non-human primates, and language may alter our category boundaries. There is a growing literature on the development of psychological causality. In future research it will be interesting to compare the developmental characteristics of physical and psychological reasoning abilities.

References

Antinucci, F. (1990). The comparative study of cognitive ontogeny in four primate species. In K. R. Gibson, & Parker, Sue Taylor (Ed.), "Language" and intelligence in monkeys and apes: Comparative developmental perspectives. (pp. 157-171). New York, NY: Cambridge University Press.

Baillargeon, R. (2004). Infants' Physical World. Current Directions in Psychological Science, 13(3), 89-94.

Bowerman, M. (1996). Learning how to structure space for language: A crosslinguistic perspective. In (1996). Peterson, Mary A (Ed), et al. Bloom, Paul (Ed), Language and space. (pp.385 436). Cambridge, MA, US: The MIT Press.

Guajardo, J. J., & Woodward, A. L. (2004). Is Agency Skin Deep? Surface Attributes Influence Infants' Sensitivity to Goal-Directed Action. Infancy, 6(3), 361-384.

Haith, M. M., & Benson, J. B. (1998). Infant Cognition. In W. Damon, Kuhn, D., & Siegler, R. S. (Ed.), Handbook of Child Psychology: Cognition, perception, and language (5 ed., Vol. 2). New York: Wiley.

Hauser, M., MacNeilage, P., & Ware, M. (1996). Numerical representations in primates. Paper presented at the Proceedings of the National Academy of Sciences of the United States of America.

Hespos, & Baillargeon, R. (2005). Décalage in infants' knowledge about occlusion and contaiment events: Converging evidence from action tasks. Cognition, Manuscript accepted for publication.

Hespos, & Baillargeon, R. (in prep). "Which toy can I get?": Converging evidence from action tasks for violations-of-expectation findings.

Hespos, S., & Spelke, E. (2004). Conceptual precursors to language. Nature, 430, 453 - 456.

Hespos, S. J., & Baillargeon, R. (2001a). Infants' knowledge about occlusion and containment events: A surprising discrepancy. Psychological Science, 121(2), 141-147.

Hespos, S. J., & Baillargeon, R. (2001b). Reasoning about containment events in very young infants. Cognition, 78(3), 207-245.

Levinson, S. C. (1996). Relativity in spatial conception and description. In S. C. Levinson & J. J. Gumperz (Eds.), Rethinking linguistic relativity. New York: Cambridge University Press.

Mandler, J. M. (2004). The foundation of mind: Origins of conceptual thought. New York: Oxford University Press.

Onishi, K., & Baillargeon, R. (2005). Do 15-month-old infants understand false beliefs? Science, 308, 255-258.

Piaget, J. (1952). The origins of intelligence in children: Oxford, England: International Universities Press. (1952).

Piaget, J. (1954). The construction of reality in the child. Oxford, England: Basic Books.

Sinha, C., & Jensen de Lopez, K. (2000). Language, culture and the embodiment of spatial cognition. Cognitive Linguistics, 11(1-2), 17-41.

Spelke, E., & Newport, E. L. (1998). Nativism, empiricism, and the development of knowledge. In R. E. Learner (Ed.), Handbook of child psychology: Theoretical models of human development (Vol. 5th Ed Vol. 1). New York: Wiley.

Spelke, E. S. (2000). Core knowledge. American Psychologist, 55(11), 1233-1243.

Wang, S. h., Baillargeon, R., & Paterson, S. (2005). Detecting continuity violations in infancy: A new account and new evidence from covering and tube events. Cognition, 95(2), 129-173.

Woodward, A. L. (1998). Infants selectively encode the goal object of an actor's reach. Cognition, 69(1), 1-34.

Woodward, A. L. (2003). Infants' developing understanding of the link between looker and object. Developmental Science, 6(3), 297-311.