The discussion presented so far in this conference has nicely articulated what are, in my view, the critical issues surrounding the study of causal understanding in non-verbal animals. Anne Reboul’s contribution, in particular, is consistent with my own approach to the problem. The “Unobservability Hypothesis” (Bering & Povinelli, 2003; Povinelli, 2000, 2004; Vonk & Povinelli, in press) addresses a key difference between human and non-human cognition in general that is also relevant for differences between human and non-human causal understanding in particular. If humans alone are capable of reasoning about abstract theoretical entities that cannot be directly perceived through the senses, then non-humans will fail to demonstrate causal understanding because causal forces are unobservable theoretical entities.

The preceding discussion, however, has raised the question of whether all causal forces are necessarily unobservable[1]. If not, it is possible, even if the Unobservability Hypothesis is correct, that some animals might be capable of causal understanding in cases in which the cause of an event can be directly perceived. If we accept this tenuous assertion that causal forces can be both observable and unobservable, a clear prediction can be made from the Unobservability Hypothesis; animals (and probably humans) should perform better on tasks requiring them to reason about the former, compared to the latter, and may not be able to reason about the latter at all. But, is it ever the case that the causal force itself is directly observable? It is clear that we need to carefully define what we mean by ‘causal force’ and ‘causal understanding’. If ‘causal force’ is defined as a mediating variable between a cause and an effect (c.f. Tomasello, 1998, it is most probably of an unobservable nature. For example, one can directly observe the effect of pouring liquid on to a dry object. It seems reasonable to attribute causal understanding to an individual who appreciates that the liquid causes the dampening of the object, but it is debatable whether the process by which this happens, (diffusion), is considered observable or unobservable. In the case of Call’s (2004) experiment, the presence of food is directly observable as the cause of the noise produced by shaking the cup.

Much of human knowledge about causal forces is derived from explicit teachings. If one knows that an ice cube will melt when it is hot does that mean that one understands that the heat causes the ice cube to melt? In order to demonstrate causal understanding, would one need to understand precisely how the heat causes the ice to melt? If that is the case, then humans often fail to demonstrate true causal understanding. How many of us truly understand the principles of gravity or force transfer? How many fewer of us would respond affirmatively to that question without the benefit of explicit tutoring on these subjects? So perhaps simply understanding the direction of the relation between the cause and the effect (it is the heat that causes the ice to melt) should be sufficient to constitute causal understanding. But clearly the representation has to be more than an association between the two (when it is hot, ice melts; when it is cold, ice does not melt). How do we distinguish between these two representations of the event – one that relies on reasoning about underlying causal forces, and one that relies on associating a cause and effect- in a nonverbal being?

It may be impossible to make this distinction both in cases in which the causal force is observable and unobservable. This is true because it will be exceedingly difficult, if not impossible, to distinguish a representation of the causal forces at work, from an ability to predict effects from causes where the two have been observed to occur together. Even unobservable forces invariably result in observable regularities. These forces and their observable manifestations will rarely lead to contrasting predictions as they are necessarily correlated. So we return to the question posed at the end of the preceding paragraph; how do we disentangle the two possible representations of events? We, as scientists, can create situations in which the two events (the cause and the effect) have not been previously observed to occur together, but should be logically expected to co-occur if one appreciates the causal force at work. The difficulty of course is in providing subjects with enough information that they could conceivably make the logical inference about the causal force.

Elsewhere we have attempted to clearly articulate this empirical challenge with the goal of encouraging the development of clever, new, more diagnostic methods for testing our hypothesis. We stressed the danger of confounding predictions based on observable correlates of underlying causes with predictions based on the cause itself. In addition, we called for an overhaul of the current research paradigms being used to address these issues specifically in the context of theory of mind research (Povinelli & Vonk, 2003, 2004; Vonk & Povinelli, in press). Mental states are only one particular type of unobservable causal forces. In our previous writings we did not outline a clear strategy for tackling the same problems within the context of testing non-humans for their understanding of physical causality. But, perhaps too subtlety, we acknowledged the strength of focusing on explanatory versus predictive paradigms because we believe that one key distinction between humans and non-humans, as relates to causality, is that humans alone strive to explain as well as to predict events (see also Andrews, in press; Bering & Povinelli, 2003; Reboul, this conference, Vonk & Povinelli, in press). If this is true, then non-humans should fail at tasks requiring them to posit explanations for observed events[2]. However, if they show no interest in positing explanations for events, would this disinterest constitute evidence that they are not capable of doing so? A vast array of situations, in which a solution to a problem depends upon an ability to explain effects, must be called upon to address this question.

Clearly animals can reason from observed events (causes) to resulting events (effects), in other words, to make predictions based on observable regularities in their environments. But do they also reason backward from effects to causes? Very little empirical research has directly addressed this question (see Povinelli & Dunphy-Lelii, 2001, for one exception).  Indeed, what has been lacking to this point in this conference, although it has certainly been touched upon, is a clear articulation of a program of empirical research designed specifically to address such questions. In the remainder of this brief paper, I would like to suggest some approaches for testing the hypotheses that have influenced the participants of this conference.

Here are the critical questions as I see them:

1. Do humans alone strive to explain as well as to predict the behavior of objects and other beings[3]?

2. Are humans the only species capable of reasoning about strictly theoretical entities (i.e. The Unobservability Hypothesis)?

3. Do all causal forces belong to the category of strictly theoretical, unobservable constructs?

4. If the answer to the first two questions turns out to be affirmative, why did this come to be the case? To what extent do these abilities rely upon the human capacity for language?

 5. What advantages are conferred to a species that a) reasons about unobservables, and b) has an explanatory drive?

The latter three questions are theoretical and can not be adequately addressed in this brief exposition (and in the absence of answers to the first two questions). The first two questions are empirical and should be tested. We are pleased that others have embraced the Unobservability Hypothesis as a valid model for framing the uniqueness of human cognition, but we would like to encourage tests of our hypothesis. We would also like to emphasize the utility of explanatory versus predictive paradigms for elucidating non-human understanding of the causes underlying events. I will focus on the latter hypothesis as it more specifically addresses the question of representations of causality in non-humans.

Researchers have attempted to address non-humans’ understanding of causality a number of ways. Most notably, they have tested other species’ (mostly primates’) use of tools, and examined whether tool use reveals an understanding of the causal structure of a task. As with many areas of research, there is some dispute as to the extent to which apes, but not monkeys, grasp the causal significance of various attributes of the given tasks (Boesch & Boesch, 1990; Cacchione & Krist, 2004; Fujita, Kuroshima & Asai, 2003; Hauser, 1997; Hauser, Kralik & Botto-Mahan, 1999; Hauser, Pearson & Seelig, 2002; Hauser, Santos, Spaepan & Pearson, 2002; Kralik & Hauser, 2002; Kohler, 1925; Limongelli, Boysen & Visalberghi, 1995; Matsuzawa, 1996, 2001; Povinelli, 2000; Santos & Hauser, 2002; Santos, Miller & Hauser, 2003; Visalberghi, 1997, 2002; Visalberghi, Fragaszy & Savage-Rumbaugh, 1995; Visalberghi & Limongelli, 1994, 1996; Visalberghi & Tomasello, 1998; Visalberghi  & Trinca, 1989). However, an extensive research program in our own laboratory has given us reason to doubt that chimpanzees appreciate various causal forces underlying folk physics, such as physical connection, gravity, rigidity, etc. (Povinelli, 2000). In these tasks, chimpanzees were presented with problems that they should have been able to solve from the very first trial forward if they had an accurate representation of the causal forces at work in the task. For instance, if they understood the effect of gravity they would not attempt to retrieve a food reward by dragging it across an empty space in a “trap-table”, and yet all but one subject did so just as often as they chose to drag the food across an unbroken surface (Povinelli, 2000, Chapter 5, Experiment 3).

Recently, Josep Call (2004, in press, and see his comments in response to Anne Reboul’s paper) found that apes chose cups containing hidden food more accurately when cues as to the food’s location were of a causal versus an arbitrary nature. (Although please see Anne Reboul’s comments on why these experiments do not address whether apes are capable of diagnostic versus predictive learning). Let us conduct a thought experiment in order to illuminate some of the issues raised by this sort of paradigm. Suppose an ape is presented with two identical liquid containers, the contents of which are hidden. An experimenter moves the two containers to two new locations. The ape is required to point to the container that he wants the experimenter to give to him. Only one of the containers has left a water mark in its former resting place. If the ape points to the container that formerly held the position where now there is a water stain, is this because he has evoked an explanation for the stain – that it must be caused by condensation caused by the presence of liquid in that particular container? Or is he predicting that liquid will be found (or is more likely to be found) in the presence of containers that leave marks behind, versus those that do not? Of course the ape would have to have had the relevant experience with such containers – those that are full and leave stains, and those that are empty and do not, in order to make both the non-causal prediction and the causal inference/explanation.

In a related vein, others have sought to determine whether chimpanzees imitate arbitrary, irrelevant actions as well as causally relevant actions after watching a demonstrator retrieve food from a closed container (Horner & Whiten, 2005). The basic idea is that, if animals are privy to the mechanism in a mechanical task, they should learn faster, or at least, fail to imitate irrelevant actions, compared to conditions in which they are not privy to the mechanism involved. Although the experiment is clever, and the results intriguing, there are some flaws in the design that prevent us from accepting the authors’ conclusions that access to causal knowledge allowed the apes to ignore irrelevant actions on the box. For instance, it is possible that the chimpanzees performed the relevant actions, because they were directed toward the food reward, and, in the ‘causal’ condition, they were aware of the food’s location in the box. They may have failed to imitate irrelevant actions only because they were not directed towards the food’s location. We do, however, believe that the experiment could be modified to answer the kinds of questions we have posed here.

Researchers have also made extensive use of the violation of expectancy paradigms first applied to the study of human infants (Baillargeon, Spelke & Wasserman, 1985; Leslie, 1982; Spelke, 1985). This paradigm can be exploited to determine what nonverbal subjects understand about the likelihood of various events, by measuring whether they look longer at events that violate natural laws (Gergely, Nadasdy, Csibra & Biro, 1995; Hauser, 1998; Hauser & Carey, 1998; O’Connell & Dunbar, 2005). If they look longer at outcomes that would only be surprising given an understanding of the causal force at work, then we might conclude that they make predictions based on an understanding of that causal force. Of course it is possible that an event violates an individual’s expectations, not because they have a deep understanding of the causal forces at work, but, instead because they have gained, through previous experience, an appreciation of the likelihood, or commonality of various events. This problem can be easily overcome by designing scenarios in which the unexpected events in a particular context are generally as common as the expected events. This paradigm could possibly be recruited to examine animals’ diagnostic abilities. Imagine a procedure whereby subjects are shown first the final result of an event, and subsequently were shown the preceding action under conditions in which the action could and could not have resulted in the observed end result. For example, in a physical causality study, if a collision has just occurred; can the subjects guess as to which object must have been used to strike another, based on the resulting impact and end state of the objects involved? So, can an object’s current resting position be used as a cue as to which event previously occurred in both collision and deformation paradigms?

Only one study, to our knowledge, has directly tested whether non-humans seek causal explanations (Povinelli & Dunphi-Lelii, 2001). This study was briefly described in Anne Reboul’s paper so will not be reiterated here. We recommend more experiments along these lines. These experiments will be most informative when they measure subjects’ attempts to seek explanations for unexpected events, especially when an event violates expectations based on presumed causal understanding. For instance, our lab has begun to investigate chimpanzees’ reactions to apparent “magical” (arbitrary) causality. For example, chimpanzees might be trained to associate arbitrary cues with the functionality of a tool. Once they have learned to associate such cues with success in a food retrieval task we could reverse the contingencies such that the cues previously associated with the ‘correct’ or ‘functional’ end of a tool are now associated with the ‘incorrect’ end and vice versa. If chimpanzees are surprised by this unexpected change one might expect them to examine the tools to look for some anomaly that would explain their failure in this task.

Some other contexts in which this experimental approach might be appropriate are as follows: subjects have been misled by cues provided by experimenters as to the location of hidden food – do they check back to make sure they read the cue correctly? Do they look for signs of an event that would explain another’s actions? For example, do they search for a predator or another alarming object if another animal has just displayed a fearful reaction? Can animals guess what another animal has just done from cues as to a preceding act as easily as they can predict what another animal will do? In other words, one could contrast subjects’ ability to explain versus predict another’s actions. Do animals question why another individual reacted in the manner that they did, particularly when this reaction was directed towards the subject in question? Humans are very sensitive to behaviors directed towards themselves. If someone appears to be angry or fearful towards us we wonder why. We often analyze our own behavior in an attempt to understand whether we are to blame for this unexplained reaction. If we recognize some component of our behavior that may have precipitated the aversive behavior of another, we typically attempt to change our own behavior in an effort to avoid the aversive reaction of the other individual. Do animals adjust their own behaviors accordingly? If they do so, can we conclude that they have “explained” the source of the aversive behavior? If the individual’s changed behavior is a direct result of the realization that his own behavior caused the behavior of another, we can attribute causal understanding to that individual. But how can we determine whether the change came about as a result of reasoning about the cause of the behavior, as opposed to reasoning in a predictive manner about how others will react to one’s own behavior? Can we construct an empirical test of this question?

Imagine that two chimpanzees are placed in adjacent enclosures. One chimpanzee (the actor) has the opportunity to choose between two food trays – however, he can not see the contents of either of the trays prior to making his choice. Prior to the choice, the chimpanzees have been trained to understand that only one tray contains a food reward on every trial. The first tray chosen by the chimpanzee will be offered to the neighboring chimpanzee (the partner). The actor will not be able to see what was in the tray given to the partner, but will be able to see the partner’s reaction. Based on the reaction of the partner the chimpanzee should be able to determine whether the chosen tray contained a food reward, and to deduce that, if the partner received a reward, the remaining tray must be empty. Therefore there would be no point to making an additional choice. However, if the partner does not appear to have received a reward, the baited tray must remain available. Therefore it would benefit the actor to make a second selection, because the actor is always offered the reward, if present, from the second tray chosen. The frequency of second selections should vary according to whether the partner expresses pleasure or displeasure. If the actors consistently choose a second tray only when the partners do not receive a food reward, does this mean that they have deduced the contents of the chosen tray from the partner’s expression? I suspect, based on Anne’s summary of Cheney and Seyfarth’s (1990) findings that vervets do not infer the presence of predators from observable evidence as to their whereabouts (for example, the tracks of a python, or the carcass from a cheetah’s kill etc.) that monkeys will not demonstrate evidence for seeking causal explanations. It is possible that apes will differ on these measures, although Povinelli & Dunphy-Lelii’s (2001) findings suggest otherwise.

Data from the numerous animal language projects might get us closer to this goal of evaluating animals’ diagnostic abilities. Because there is some evidence that sign-language-trained chimpanzees, at least, answer “wh” questions appropriately (Van Cantford, Gardner & Gardner, 1989; Gardner, Van Cantford & Gardner, 1992), these chimpanzees might to some degree understand the meaning of the signs for such questions. If there is evidence that they utilize these same signs to ask questions of others, particularly “how” or “why” questions, we would be very interested. Of course such a finding should not be taken to imply that only language-trained animals are capable of, or interested in, searching for explanations, or understanding causality[4]. It may be that only language allows a creature to express such a drive in a manner that we, as humans, can interpret as such. This circularity is precisely why we need to invent diagnostic tests for studying causal understanding in creatures who can not express themselves through language.

Within all of these experiments, it is possible to contrast cases in which the cause can be described as emanating from an unobservable process (such as transfer of force, gravity, beliefs, intentions) versus an observable event (a baited cup makes a sound, an unbaited cup does not, as in Call, 2004). It should be clear from these examples that there is much work to be done before we can draw any final conclusions as to which other species might share our desire to explain events, and which are capable of reasoning about causal forces, either observable or not, in order to do so. Based on the data so far, it appears that only humans reason about causal forces that are not directly observable, while other species appear to reason solely about the observable properties of events and object interactions. It also seems to be the case that non-humans reason about causes only insofar as to predict the actions of others and objects in their environments, and are not driven to explain events that have already occurred.

References

Andrews, K. (in press). Chimpanzee theory of mind: Looking in all the wrong places?  Mind and Language.

Baillargeon, R., Spelke, E.S., & Wasserman, S. (1985). Object permanence in five- month-old infants. Cognition, 20, 191-208.

Bering, J.M. & Povinelli, D.J. (2003). Comparing cognitive development. In D. Maestripieri, Ed. pp. 205-233. Primate psychology: Bridging the gap between the mind and behavior of human and nonhuman primates. Cambridge, MA: Harvard University Press.

Call, J. (2004). Inferences about the location of food in the great apes. Journal of Comparative Psychology, 118, 232-241.

Call, J. (In press). Descartes’ two errors: Reason and reflection in the great apes. In S.

Hurley and M. Nudds (Eds.) Rational Animals? OxfordUniversityPress.

Cheney, D.L., & Seyfarth, R.M. (1990). How Monkeys See the World: Inside the Mind of Another Species.Universityof ChicagoPress, Chicago.

Boesch, C. & Boesch, H. (1990). Tool use and tool making in wild chimpanzees. Folia Primatologica, 54, 86-99.

Cacchione, T. & Krist, H. (2004). Recognizing impossible object relations: Intuitions about support in Chimpanzees (Pan troglodytes). Journal of Comparative  Psychology, 118, 140-148.

Fujita, K., Kuroshima, H. & Asai, S. (2003). How do tufted capuchin monkeys (Cebus apella) understand causality involved in tool use? Journal of Experimental Psychology: Animal Behavior Processes, 19, 233-242.

Gardner, R. A., Van Cantford, T.E., & Gardner, B.T. (1992). Categorical replies to categorical questions by cross-fostered chimpanzees. American Journal of Psychology, 105, 27-57.

Gergely, G. Nadasdy, Z., Csibra, G. & Biro, S. (1995). Taking the intentional stance at 12 months of age. Cognition, 56, 165-193.

Hauser, M.D. (1997). Artifactual kinds and functional design features: What a primate understands without language, Cognition, 64, 285-308.

Hauser, M.D. (1998). A nonhuman primate’s expectations about object motion and destination: The importance of self-propelled movement and animacy. Developmental Science, 1, 31-37.

Hauser, M.D. & Carey, S (1998). Building a cognitive creature from a set of primitives: Evolutionary and developmental insights. In D. Cummins & C. Allen (Eds.) The Evolution of Mind (pp. 52-106). Oxford, England, OxfordUniversityPress.

Hauser, M.D., Kralik, J. & Botto-Mahan, C. (1999). Problem solving and functional design features: Experiments on cotton-top tamarins (Saguinus oedipus). Animal Behaviour, 57, 565-582.

Hauser, M.D., Pearson, H.M. & Seelig, D. (2002). Ontogeny of tool-use in cotton-top tamarins (Saguinus oedipus): Innate recognition of functionally relevant features. Animal Behaviour, 64, 299-311.

Hauser, M.D., Santos, L.R., Spaepen, G.M. & Pearson, H.E. (2002). Problem solving, inhibition and domain-specific experience: Experiments on cottontop tamarins, (Saguinus oedipus). Animal Behaviour, 64, 387-396.

Horner, V. & Whiten, A. (in press). Causal knowledge and imitation/emulation switching in chimpanzees (Pan troglodytes) and children (Homo sapiens). Animal Cognition. Published online Nov. 11, 2004. DOI: 10.1007/s10071-004-0239-6

Kohler, W. (1925). The Mentality of Apes. Liveright, New York.

Kralik, J.D. & Hauser, M.D. (2002). A nonhuman primates’ perception of object relations: Experiments on cottontop tamarins, Saguinus oedipus. Animal Behaviour, 63, 419-435.

Leslie, A.M. (1982). The perception of causality in infants. Perception, 11, 173-186.

Limongelli, L., Boysen, S.T. & Visalberghi, E. (1995). Comprehension of cause-effect relations in a tool-using task by chimpanzees (Pan troglodytes). Journal of ComparativePsychology, 109, 18-26.

Matsuzawa, T. (2001). Primate foundations of human intelligence: A view of tool use in non-human primates and fossil hominids. In: Primate Origins of Human Cognition and       Behavior (Ed. by T. Matsuzawa). Pp. 3-25. Tokyo, Springer-Verlag.

O’Connnell, & Dunbar, R.I.M. (2005). The perception of causality in chimpanzees (Pan spp.). Animal Cognition, 8, 60-66.

Povinelli, D.J. (2000). Folk physics for apes: The chimpanzee's theory of how the world works.   Oxford: OxfordUniversity Press [Reprinted with revisions, 2003].

Povinelli, D.J. (2004, Winter). Behind the ape’s appearance: Escaping anthropocentrism in the study of other minds, Daedalus, 29- 41.

Povinelli, D.J. & Dunphy-Lelii, S. (2001). Do chimpanzees seek explanations?Preliminary comparative investigations. Canadian Journal of Experimental  Psychology, 55,93-101.

Povinelli, D.J. and Vonk, J. (2003). Chimpanzee minds: Suspiciously human? Trends in Cognitive Science, 7, 157-160.

Povinelli, D.J. and Vonk, J. (2004). We don’t need a microscope to explore the chimpanzee mind. Mind and Language, 19, 1-28.

Santos, L.R. & Hauser, M.D. (2002). A non-human primate’s understanding of solidity:

Dissociations between seeing and acting. Developmental Science, 5, F1-F7.

Santos, L.R., Miller, C.T. & Hauser, M.D. (2003). Representing tools: How two non-human primate species distinguish between the functionally relevant and irrelevant features of a tool. Animal Cognition, 6, 269-281.

Spelke, E.S. (1985). Preferential looking methods as tools for the study of cognition in infancy. In G.Gottlieb & N. Krasnegor(Eds.) Measurement of audition and vision in the first year of postnatal life (pp. 85-168). Norwood: Ablex.

Tomasello, M. (1998).  Uniquely primate, uniquely human. Developmental Science, 1, 1-16.

Van Cantford, T., Gardner, B.T. & Gardner, R.A. (1989). Developmental trends in replies

To Wh-questions by Children and Chimpanzees. In R.A. Gardner, B.T. Gardner, & T. E. Van Cantford (Eds.) Teaching sign language to chimpanzees, (pp. 198-239). Albany, NY: SUNY Press.

Visalberghi, E. (1997). Success and understanding in cognitive tasks: A comparison between Cebus apella and Pan troglodytes. International Journal of Primatology, 18, 811-830.

Visalberghi, E. (2002). Insight from capuchin monkey studies: Ingredients of recipes for, and flaws in capuchins’ success. In Bekoff, M. & Allen, C. (Eds.) The Cognitive Animal: Empirical and Theoretical Perspectives on Animal Cognition. MIT Press, Cambridge, MA, pp. 405-411.

Visalberghi, E., Fragaszy, D.M. & Savage-Rumbaugh, S. (1995). Performance in a tool-using task by common chimpanzees (Pan troglodytes), Bonobos (Pan paniscus), an Orangutan  (Pongo pygmaeus), and capuchin monkeys (Cebus apella). Journal of Comparative Psychology, 109,52-60.

Visalberghi, E., & Limongelli, L. (1994). Lack of comprehension of cause-effect relations in tool-using capuchin monkeys (Cebus apella). Journal of Comparative Psychology, 108, 15-22.

Visalberghi, E., & Limongelli, L. (1996). Acting and understanding: Tool use revisited through the minds of capuchin monkeys. In Russon, A.E, Bard, K.A. (Eds), Reaching Into Thought: The Minds of the Great Apes. New York, Cambridge University Press.

Visalberghi, E. & Tomasello, M. (1998). Primate causal understanding in the physical and  psychological domains. Behavioral Processes, 42, 189-203.

Visalberghi, E., & Trinca, L. (1989). Tool use in capuchin monkeys: Distinguishing between performing and understanding. Primates, 30, 511-521.

Vonk, J. & Povinelli, D.J. (in press). Similarity and difference in the conceptual systems of  primates: The Unobservability hypothesis. In E.Wasserman and T. Zentall (Eds.)  Comparative Cognition: Experimental Explorations of Animal Intelligence.