Animal Cognition: Monkey Looks Contradict Hume

2
Saccharomyces cerevisiae. Mol. Cell Biol. 21, 827–839. 18. Yang, H.C., and Pon, L.A. (2002). Actin cable dynamics in budding yeast. Proc. Natl. Acad. Sci. USA 99, 751–756. 19. Schirenbeck, A., Bretschneider, T., Arasada, R., Schleicher, M., and Faix, J. (2005). The Diaphanous-related formin dDia2 is required for the formation and maintenance of filopodia. Nat. Cell Biol. 7, 619–625. 20. Higgs, H.N., and Peterson, K.J. (2005). Phylogenetic analysis of the formin homology 2 domain. Mol. Biol. Cell 16, 1–13. Departments of Molecular Genetics and Cell Biology and of Biochemistry and Molecular Biology, University of Chicago, Chicago, Illinois 60637, USA. E-mail: [email protected] DOI: 10.1016/j.cub.2006.06.039 Animal Cognition: Monkey Looks Contradict Hume Wild rhesus monkeys have been found to attend to causal anomalies, suggesting that they make inferences about possible and impossible physical transformations in the absence of direct relevant experience. Juan-Carlos Go ´ mez Causality and how we understand it have been the subject of intense scholarly debate. The eighteenth century Scottish philosopher David Hume (see [1,2]) influentially argued that all we know is that one event repeatedly occurs after another — for example, a ball moves after being hit by another — and we call this causation, but we cannot see or logically demonstrate cause–effect connections, the impression of causation derives from repeated experience. In contrast, others [1,3] have suggested that the notion of cause is a primitive of the human mind through which we make sense of experience, including novel events. Indeed, humans know that certain things will or will not happen without having experienced them. For example, I know that by throwing my mobile phone, but not my wallet, I may break a window glass, without having actually tried this ever. We can predict this on the basis of the physical properties of the objects involved — glass is fragile and breakable, my mobile is hard, whereas my wallet is soft. In some cases our ability to predict novel outcomes likely derives from the vast number of experiences, with an extraordinary variety of objects, we accumulate over a lifetime. But some impressions of causality, such as the launching effect of a ball hitting another, might require little or no experience — indeed, 6-month-old human infants show awareness of such effects [4]. Do other animals have such a sense of causality? Ever since scientists started to study animal minds, they debated whether animals blindly learn to associate actions with outcomes through repeated experience, or react insightfully to new situations. In effect, the animal intelligence debate is about whether animals can or cannot understand the causal connections between actions and objects [5]. Some suggest that even chimpanzees, with their well-known tool-using skills, commit too many glaring errors in experimental tasks to be credited with genuine causal understanding [6]. Others argue that chimpanzees show a natural sense of causality, even outside the tool-using domain: for example, they choose containers that make a noise when shaken as the ones that must contain food, but find it difficult to learn an arbitrary cue, such as a tapping sound, associated with food containers, even after repeated experience [7]. This suggests that they spontaneously pay attention to causal relations. This debate has just been enriched by Hauser and Spaulding’s [8] report that rhesus monkeys — a non-tool-using species with little interest in objects which is relatively distant from humans — understand causal relations involving tools. The authors showed wild rhesus monkeys sequences such as those illustrated in Figure 1. In one example, an experimenter placed an apple on a platform, erected a screen, then lowered and lifted a knife behind the screen. When the screen was removed, the apple appeared cut in two halves. Alternatively, the same sequence — from intact apple to halved apple — was shown, but instead of a knife the experimenter lowered a drinking glass behind the screen. In both cases the experimenter surreptitiously replaced the whole apple with a halved one. Humans would assume that, in the first sequence, the experimenter cut the apple with the knife, but they would find surprising the same outcome in the sequence with the glass. We know that knives, but not glasses, have the causal power of cutting. How can we tell what the monkeys thought of these sequences? Hauser and Spaulding [8] applied a technique that developmental psychologists use with human babies: they measured how long the monkeys looked at the cut apple in each sequence, on the assumption that monkeys, like babies, would look longer at anything violating their expectations. And they found that the monkeys indeed looked longer at the ‘impossible’ outcome of an apple cut with a glass than at the ‘possible’ outcome involving the knife. Moreover, they also looked longer at the impossible event of a cut apple reappearing as a whole apple after using a knife behind the screen, which suggests that monkeys were not acting on a loose association between knives and cut pieces of food, but paid attention to the order of events. Similar results were obtained with a different transformation: a white piece of cloth dyed with blue paint versus the same effect involving a knife. The news is not just that rhesus monkeys may show causal understanding of a type until now Current Biology Vol 16 No 14 R538

Transcript of Animal Cognition: Monkey Looks Contradict Hume

Page 1: Animal Cognition: Monkey Looks Contradict Hume

Saccharomyces cerevisiae. Mol. Cell Biol.21, 827–839.

18. Yang, H.C., and Pon, L.A. (2002).Actin cable dynamics in budding yeast.Proc. Natl. Acad. Sci. USA 99,751–756.

19. Schirenbeck, A., Bretschneider, T.,Arasada, R., Schleicher, M., and Faix, J.

(2005). The Diaphanous-related formindDia2 is required for the formation andmaintenance of filopodia. Nat. Cell Biol. 7,619–625.

20. Higgs, H.N., and Peterson, K.J. (2005).Phylogenetic analysis of the forminhomology 2 domain. Mol. Biol. Cell 16,1–13.

Departments of Molecular Genetics andCell Biology and of Biochemistry andMolecular Biology, University ofChicago, Chicago, Illinois 60637, USA.E-mail: [email protected]

DOI: 10.1016/j.cub.2006.06.039

Current Biology Vol 16 No 14R538

Animal Cognition: Monkey LooksContradict Hume

Wild rhesus monkeys have been found to attend to causal anomalies,suggesting that they make inferences about possible and impossiblephysical transformations in the absence of direct relevant experience.

Juan-Carlos Gomez

Causality and how we understandit have been the subject of intensescholarly debate. The eighteenthcentury Scottish philosopher DavidHume (see [1,2]) influentiallyargued that all we know is thatone event repeatedly occurs afteranother — for example, a ballmoves after being hit byanother — and we call thiscausation, but we cannot seeor logically demonstratecause–effect connections, theimpression of causation derivesfrom repeated experience. Incontrast, others [1,3] havesuggested that the notion ofcause is a primitive of the humanmind through which we makesense of experience, includingnovel events.

Indeed, humans know thatcertain things will or will not happenwithout having experienced them.For example, I know that bythrowing my mobile phone, but notmy wallet, I may break a windowglass, without having actually triedthis ever. We can predict this on thebasis of the physical properties ofthe objects involved — glass isfragile and breakable, my mobile ishard, whereas my wallet is soft. Insome cases our ability to predictnovel outcomes likely derives fromthe vast number of experiences,with an extraordinary variety ofobjects, we accumulate overa lifetime. But some impressions ofcausality, such as the launchingeffect of a ball hitting another,might require little or noexperience — indeed, 6-month-old

human infants show awareness ofsuch effects [4].

Do other animals have sucha sense of causality? Ever sincescientists started to study animalminds, they debated whetheranimals blindly learn to associateactions with outcomes throughrepeated experience, or reactinsightfully to new situations. Ineffect, the animal intelligencedebate is about whether animalscan or cannot understand thecausal connections betweenactions and objects [5]. Somesuggest that even chimpanzees,with their well-known tool-usingskills, commit too many glaringerrors in experimental tasks to becredited with genuine causalunderstanding [6]. Others arguethat chimpanzees show a naturalsense of causality, even outsidethe tool-using domain: forexample, they choose containersthat make a noise when shakenas the ones that must containfood, but find it difficult to learnan arbitrary cue, such as atapping sound, associated withfood containers, even afterrepeated experience [7]. Thissuggests that they spontaneouslypay attention to causal relations.

This debate has just beenenriched by Hauser andSpaulding’s [8] report that rhesusmonkeys — a non-tool-usingspecies with little interest in objectswhich is relatively distant fromhumans — understand causalrelations involving tools. Theauthors showed wild rhesusmonkeys sequences such as thoseillustrated in Figure 1. In one

example, an experimenter placedan apple on a platform, erecteda screen, then lowered and lifteda knife behind the screen. When thescreen was removed, the appleappeared cut in two halves.Alternatively, the samesequence — from intact apple tohalved apple — was shown, butinstead of a knife the experimenterlowered a drinking glass behind thescreen. In both cases theexperimenter surreptitiouslyreplaced the whole apple witha halved one. Humans wouldassume that, in the first sequence,the experimenter cut the apple withthe knife, but they would findsurprising the same outcome in thesequence with the glass. We knowthat knives, but not glasses, havethe causal power of cutting.

How can we tell what themonkeys thought of thesesequences? Hauser and Spaulding[8] applied a technique thatdevelopmental psychologists usewith human babies: they measuredhow long the monkeys looked at thecut apple in each sequence, on theassumption that monkeys, likebabies, would look longer atanything violating theirexpectations. And they found thatthe monkeys indeed looked longerat the ‘impossible’ outcome of anapple cut with a glass than at the‘possible’ outcome involving theknife. Moreover, they also lookedlonger at the impossible event ofa cut apple reappearing as a wholeapple after using a knife behind thescreen, which suggests thatmonkeys were not acting on a looseassociation between knives and cutpieces of food, but paid attention tothe order of events. Similar resultswere obtained with a differenttransformation: a white piece ofcloth dyed with blue paint versusthe same effect involving a knife.

The news is not just that rhesusmonkeys may show causalunderstanding of a type until now

Page 2: Animal Cognition: Monkey Looks Contradict Hume

DispatchR539

reported only in highly trainedlaboratory chimpanzees [3]. Moresurprising is that the monkeys hadno direct experience of usingknives, dyes or other tools. And,according to the new report [8],their vicarious experience of seeinghuman visitors cut food withknives, drink with glasses or dyewith paint was very limited, if any atall. It seems that the monkeys haveeither a fabulous capacity forvicarious learning of causalrelations in one-off events or, asHauser and Spaulding [8] suggest,a natural tendency to draw novelcausal inferences from theirgeneral knowledge of objects,including objects they have neverpersonally handled. Although thereport may downplay the amountof causal experience of themonkeys — surely they spendmuch time daily picking andchewing food items, andexperiencing the resultingtransformations — theirperformance in the looking testsremains impressive.

But can we really infer so muchfrom looking times? Does anaverage extra second and a half ofattention justify the attribution ofsophisticated causal knowledge tothe monkeys? In developmentalpsychology, looking time has beenused to argue for complex physicaland social knowledge in very younginfants incapable of complexactions [9], but there is somecontroversy about the significanceof such measures. Most agreethat looking time does reflectsome form of early knowledge,but there is disagreement abouthow similar this is to theknowledge used in explicitadaptive action [10].

Looking time methods posea special problem when appliedto non-human primates.Perception–action mismatches inhuman infants can be understoodas transitional phases in thedevelopment of mature skills. Butwhat is the point of permanentperception–action mismatches inadult monkeys? Why doesa species that rarely uses objectshave such an astonishing ability tounderstand the causal mechanicsof objects used by others? What isthe evolutionary point of acquiringknowledge that is not put to

Possible

Possible

Impossible

Impossible

Current Biology

Figure 1. Some of the pos-sible and impossible se-quences shown by Hauserand Spaulding [8] to rhesusmonkeys.

The monkeys looked signif-icantly longer (3.65 secondson average) to the impossi-ble outcomes than to thepossible ones (1.96 sec-onds on average) during the10 seconds following thewithdrawal of the screen.(Adapted with permissionfrom [8].)

practical use? [11]. Is this aby-product of more generalabilities that do have an adaptiveimpact? Perhaps attending to ‘odd’events is advantageous in otherareas in which the monkeys dodevelop adaptive actions, but herethey just look longer withoutknowing why their attention iscaught by sequences such as‘apple-glass-cut apple’? But thenwhat exactly catches their eye if notthe causal anomalies of glassescutting apples and knivescolouring clothes?

Hauser and Spaulding’s [8] studyis an important and excitingcontribution to the body ofevidence contradicting Hume’sassumptions about causalcognition. Causal understanding ofnovel object relations may be anevolutionary primitive of cognitionrequiring a relatively modestamount of experience with objects,and not the result of generalisingfrom human-like massive objectexperience. But it is alsoa challenging reminder of the needto address the riddle of the contrastbetween perceptual knowledgeand knowledge for adaptive action,which now matters not only incognitive development but also incognitive evolution.

References1. Cheng, P. (1999). Causal reasoning. In The

MIT Encyclopedia of the CognitiveSciences, R.A. Wilson and F.C. Keil, eds.(Cambridge, Mass.: MIT Press), pp. 6–108.

2. Hume, D. (1739/1987). A Treatise ofHuman Nature, Second Edition (Oxford:Clarendon Press).

3. Premack, D., and Premack, A.J. (2003).Original Intelligence (New York: McGraw-Hill).

4. Leslie, A.M., and Keeble, S. (1987). Dosix-month-old infants perceive causality?Cognition 25, 265–288.

5. Gomez, J.C. (2004). Apes, Monkeys,Children and the Growth of Mind.(Cambridge, Mass.: Harvard UniversityPress).

6. Povinelli, D. (2000). Folk Physics for Apes.(Oxford: Oxford University Press).

7. Call, J. (2004). Inferences about the locationof food in the great apes (Pan paniscus,Pan troglodytes, Gorilla gorilla, and Pongopygmaeus).J.Comp.Psychol.118, 232–241.

8. Hauser, M., and Spaulding, B. (2006). Wildrhesus monkeys generate causalinferences about possible and impossiblephysical transformations in the absenceof experience. Proc. Natl. Acad. Sci. USA103, 7181–7185.

9. Bremner, J.G. (2001). CognitiveDevelopment: Knowledge of the PhysicalWorld. In Infant Development, G. Bremnerand A. Fogel, eds. (Oxford: Blackwell),pp. 99–138.

10. Karmiloff-Smith, A. (1992). BeyondModularity: A Developmental Perspectiveon Cognitive Science (Cambridge, Mass.:MIT Press).

11. Gomez, J.C. (2005). Species comparativestudies and cognitive development.Trends Cogn. Sci. 9, 91–158.

School of Psychology, University ofSt. Andrews, St. Andrews KY16 9JP, UK.

DOI: 10.1016/j.cub.2006.06.040