Look at Megan. Not just at her distinc-tively chimpanzee features–her accen-tuated brow ridge, her prognathic face,her coarse black hair–but at the totalityof her being: her darting eyes, her slow,studied movements, the gestures shemakes as her companion, Jadine, passesnearby. Can there be any doubt that be-hind certain obvious differences in herappearance resides a mind nearly identi-cal to our own? Indeed, is it even possi-ble to spend an afternoon with her andnot come to this conclusion? Upon re-flection, you will probably acknowledgethat her mind is not identical to ours.“But surely it’s not qualitatively differ-ent, either,” you will still insist. “I mean,it’s obvious from watching her that weshare the same kind of mind.”

Faced with the overwhelming similari-ty in the spontaneous, everyday behaviorof humans and chimpanzees, how cansomeone like me–someone who has

dedicated his life to studying theseremarkable animals–entertain the pos-sibility that their minds are, in profoundrespects, radically different from ourown? How can I challenge the receivedwisdom of Darwin–con½rmed by myown initial impressions–that the mentallife of a chimpanzee is best compared tothat of a human child?

Actually, it’s easy: I have learned tohave more respect for them than that. I have come to see that we distort theirtrue nature by conceiving of their mindsas smaller, duller, less talkative versionsof our own. Casting aside these insidiousassumptions has been dif½cult, but it hasallowed me to see more clearly that thehuman mind is not the gold standardagainst which other minds must bejudged. For me it has also illuminatedthe possibility of creating a science thatis less contaminated by our deeply an-thropocentric intuitions about the na-ture of other minds.

The best available estimates suggestthat humans and chimpanzees originat-ed from a common ancestor about ½veor six million years ago.1 This is reflected

Dædalus Winter 2004 29

Daniel John Povinelli

Behind the ape’s appearance: escaping anthropocentrismin the study of other minds

Daniel John Povinelli is Louisiana Board of Re-gents Endowed Professor of Science at the Univer-sity of Louisiana at Lafayette, and director of theCognitive Evolution Group and the Center forChild Studies. His latest book is “Folk Physics forApes: The Chimpanzee’s Theory of How theWorld Works” (2000).

© 2004 by the American Academy of Arts & Sciences

1 Galina V. Galzko and Masatoshi Nei, “Esti-mation of Divergence Times for Major Lineagesof Primate Species,” Molecular Biology and Evolu-tion 20 (2003): 424–434.

in estimates of our genetic similarity: weshare, on average, about 98.6 percent ofour total nucleotide sequence in com-mon. This statistic seems impressive.After all, such biological af½nity wouldappear to be the ½nal nail in the cof½n of the notion that there could be anyradical mental differences between them and us: if chimpanzees andhumans share 98.6 percent of theirgenetic material, then doesn’t it followthat there ought to be an extraordinarilyhigh degree of mental similarity as well?This idea has been paraded so frequentlythrough the introductory paragraphs ofboth scholarly journal articles and thepopular press alike that it has come toconstitute a melody of sorts; an anthemthat if not sung raises doubts as to one’sallegiance to the cause of defending thechimpanzee’s dignity.

But what does this 98.6 percent statis-tic really mean? It should be of immedi-ate interest that it is almost invariablymisreported. We do not share 98.6 per-cent of our genes in common with chim-panzees; we share 98.6 percent of ournucleotide sequence. A single nucle-otide difference in a string of four hun-dred may code for a different allele.Furthermore, as the geneticist JonathanMarks has pointed out in lucid detail,the 98.6 percent statistic has so littlegrounding in the average mind that con-fronts it, as to render it essentially mean-ingless.2 We might, after all, share 50percent of our nucleotide sequences incommon with bananas and broccoli. But what on earth does it mean to saythat we are 50 percent the same as a vegetable? I don’t know about you, but I doubt my mind is 50 percent identical to that of the garden pea. And so whatwould it mean, exactly, if we discovered

that our minds were 75 percent chim-panzee?

No, such coarse genetic comparisonswill hardly suf½ce to help us understandthe complex similarities and differencesthat exist between the mental lives ofhumans and chimpanzees. However, in a climate where certain highly visibleexperts have radically anthropomor-phized chimpanzees,3 such statistics areheralded as establishing once and for allthat chimpanzees are, at the very least,mentally equivalent to two- or three-year-old human children, and shouldtherefore be granted human rights.4

A few obvious biological facts may beworth noting here. To begin, it was thehuman lineage, not the chimpanzee one,that underwent radical changes after ourrespective geneologies began to divergefrom their common ancestor. Since thissplit, humans have resculpted their bod-ies from head to toe–quite literally, infact; as our lineage became bipedal, the pelvis, the knee, and the foot were all drastically reshaped, with modi½ca-tions in the hand (including new mus-cles) soon following. To top it all off, we ultimately tripled the size of ourbrain, with disproportionate increasesprobably occurring in the seat of highercognitive function, the prefrontal cortex.Oh yes, and at some point during all ofthis (no one knows exactly when), natu-ral language–perhaps the most notice-

30 Dædalus Winter 2004

Daniel JohnPovinellion learning

2 Jonathan Marks, What It Means to Be 98%Chimpanzee (Berkeley: University of CaliforniaPress, 2002).

3 For examples, see Sue Savage-Rumbaugh,Kanzi: The Ape at the Brink of the Human Mind(New York: John Wiley & Sons, 1994); JaneGoodall, Through a Window (Boston: HoughtonMifflin, 1990); Roger Fouts, Next of Kin (NewYork: William Morrow and Co., 1997).

4 Steven M. Wise, Rattling the Cage: Toward Le-gal Rights for Animals (Cambridge, Mass.: Per-seus Books, 2000); Paola Cavalieri and PeterSinger, eds., The Great Ape Project: Equality Be-yond Humanity (New York: St. Martin’s Press,1993).

able of human adaptations–emerged aswell.

In contrast, chimpanzees have proba-bly changed relatively little from thecommon ancestor they shared with usabout ½ve million years ago. Indeed, of all of the members of the great ape/human group who shared a commonancestor about ½fteen million years ago,none, indeed, has diverged as much ashumans. A simple thought experimentmay help to put this point into perspec-tive: line up all of the species in ques-tion–gorillas, orangutans, chimpan-zees, bonobos, humans–and one ofthem immediately stands out. Guesswhich one?

In fact, the more we compare humansand chimpanzees, the more the differ-ences are becoming apparent. Even ge-neticists are starting to catch up with the reality of these differences. New re-search has shown that rough similarityin our nucleotide sequences obscures the fact that the same genes may havedramatically different activity levels inthe two species. So even where humansand chimpanzees share genes in com-mon, it turns out that there are what canonly be described as major differences ingene expression–that is, whether, when,and for how long genes are actuallyworking to produce the proteins forwhich they code.5 This is the real stuffof genetic comparison, and it casts ourcrude genetic similarity to the gardenpea in a wholly different light.

What makes these differences in geneexpression signi½cant is that they ulti-mately manifest themselves as differ-ences in the bodies–including the

brains–of humans and chimpanzees.So, exactly how similar are the brains of humans and chimpanzees? After all, if we knew that, couldn’t we directly ad-dress the question of their mental sim-ilarity? Well, it would be a start, any-how. Unfortunately, comparisons of thebrains of humans and apes have tradi-tionally been limited to gross considera-tions such as size and surface features(such as lobes and sulcus patterns).Remarkably, the details of the internalorganization of human and great apebrain systems and structures have beenlargely ignored, in part because it’s sodif½cult to study these brains, but alsobecause most neuroscientists have fre-quently assumed that despite great dif-ferences in size, all mammalian brainsare organized pretty much the same.

Fortunately, even this is beginning to change. For example, Todd Preuss,working at the University of Louisiana,recently made a startling discovery while comparing the brains of humansand chimpanzees. Turning his attentionaway from the frontal lobes, his previousarea of research, Preuss decided to take a look at the primary visual cortex (V1),the area of the cerebral cortex that is the½rst way station into the processing ofvisual information. The organization ofthis area of the brain has been assumedto be nearly identical across primates.But there, in the middle of V1, Preussand his colleagues uncovered a distinc-tively human specialization–a kind ofneural architecture not found even inchimpanzees.6 Preuss speculates thatthis specialization involves modi½ca-tions of the pathways related to spatialvision and motion processing. But, re-gardless of what it is for, it suggests that

Dædalus Winter 2004 31

Behind the ape’sappearance

5 Wolfgang Enard et al., “Intra- and Inter-speci½c Variation in Primate Gene ExpressionPatterns,” Science 296 (2002): 341–343; MarioCáceres et al., “Elevated Gene Expression Lev-els Distinguish Human from Non-Human Pri-mate Brains,” Proceedings of the National Acade-my of Sciences 100 (2003): 13030–13035.

6 Todd M. Preuss et al., “Distinctive Compart-mental Organization of Human Primary VisualCortex,” Proceedings of the National Academy ofSciences 96 (1999): 11601–11606.

we need to rethink brain evolution in away that’s consistent with neo-Darwin-ian theory: similarity and differenceamong species as comfortable bedfel-lows; a state of affairs accomplished byweaving in new systems and structuresalongside the old. “If we ½nd such differ-ences in the middle of the primary visualcortex,” Preuss recently remarked to me,“just imagine what we’re going to ½ndwhen we start looking elsewhere.”

Some may be surprised (or evenafraid) to learn of such differences be-tween humans and our nearest living rel-atives. After several decades of being feda diet heavy on exaggerated claims of thedegree of mental continuity between hu-mans and apes, many scientists and lay-persons alike now ½nd it dif½cult to con-front the existence of radical differences.But then, in retrospect, how viable wasthe idea of seamless mental continuity inthe ½rst place? After all, it tended to por-tray chimpanzees as watered-down hu-mans, not-quite-½nished children. De-spite the fact that aspects of this notioncan be traced straight to Darwin, it is anevolutionarily dubious proposition, tosay the least.

If there are substantial differences be-tween the mental abilities of humansand chimpanzees, in what areas are theylikely to exist? Over the past couple ofthousand years, many potential rubiconsseparating human and animal thinkinghave been proposed. Some of these havebeen particularly unhelpful, such as theradical behaviorists’ forgettable proposi-tion that animals don’t ‘think’ at all (ofcourse, these behaviorists were evenskeptical about the existence of humanthought!). And, unfortunately, in thepopular imagination the question stillappears to be, “Can animals think?”7 as

opposed to, “How does thinking differacross species?” (the latter being adecidedly more evolutionarily mindedquestion).

Assuming that chimpanzees and otherspecies have mental states (a point I takefor granted), it seems to me that a moreproductive question to ask is, “What aretheir mental states about?” Or, put an-other way, “What kinds of concepts dothey have at their disposal?” It wouldstand to reason that the mental states ofchimpanzees, ½rst and foremost, mustbe concerned with the things most rele-vant to their natural ecology–remem-bering the location of fruit trees, keepingan eye out for predators, and keepingtrack of the alpha male, for instance.And so surely chimpanzees form con-cepts about concrete things–things liketrees, facial expressions, threat vocaliza-tions, leopards, and the like. But whatabout more abstract concepts? Con-cepts like ghosts, gravity, and God?

Admittedly, to use the term ‘concept’as loosely as I have will require the indul-gence of certain scholars. But perhapssome progress can be made by notingthat every concept is at least somewhatabstract if it extends beyond a particularexample. For instance, if one has a no-tion of an apple that is not limited to asingle instance of that apple, then onehas made a generalization, and thus akind of abstraction. Given that it hasbeen known for decades or more thatchimpanzees and many other speciesform such abstractions,8 this cannot be a de½ning feature of human thinking.

32 Dædalus Winter 2004

Daniel JohnPovinellion learning

7 Eugene Linden, “Can Animals Think?” Time,22 March 1993.

8 Suzette L. Astley and Edward A. Wasserman,“Object Concepts: Behavioral Research withAnimals and Young Children,” in William T.O’Donohue, ed., Learning and Behavior Therapy(Boston: Allyn and Bacon, 1997), 440–463;Tom R. Zentall, “The Case for a CognitiveApproach to Animal Learning and Behavior,”Behavioural Processes 54 (2001): 65–78.

At the risk of oversimpli½cation, letme instead propose a distinction be-tween concepts that refer to objects and events that can be directly observed(that is, things that can be detected bythe unaided senses), versus hypotheticalentities and processes (things that areclassically unobservable). Thus, I wishto separately consider all concepts thatrefer to theoretical things: all the thingsthat are not directly registered by thesenses, but are merely posited to exist on the basis of things we can observe.

Such concepts permeate our common-sense way of thinking: we explain phys-ical events on the basis of things like‘forces’ (supernatural or otherwise) that we have never actually witnessed,and account for the behavior of otherhumans on the basis of mental states we have never seen (e.g., their beliefs,desires, and emotions). These conceptsserve as the bedrock for some of ourmost fundamental explanations for whythe world works the way it does.

Meanwhile, we can directly contrastthese sorts of concepts with ones thatare derived from things that can be di-rectly observed: apples and oranges,trees, flashes of lightning, facial expres-sions–even the raising of a hand or thesound of a train whistle blowing in thedistance. Concepts about these thingsshare at least one property in common:they are all derived from the world ofmacroscopic entities with which the pri-mary senses directly interact. Withoutadditional justi½cation, I am thereforeasserting a distinction between con-cepts that refer to observable objects and events, and ones that refer to strict-ly hypothetical ones.

So, here’s a proposal: the mental livesof humans and chimpanzees are similar,in that both species form innumerable(and in many cases, identical) concepts

about observable things; but, at thesame time, are radically different, in thathumans form additional concepts aboutinherently unobservable things.9

Now, I realize that most people wouldnot be surprised if it were establishedbeyond doubt that chimpanzees lack aconcept of God. But what about other,seemingly more prosaic concepts thatinfest our way of thinking about theworld? Consider the way in which wethink about the social realm. In interact-ing with each other (and with animals,for that matter), we use a dual system of representation: we understand otherbeings both as part of the observableworld (they engage in particular move-ments of their hands and feet, and their lips form particular contortions assounds emerge from their mouths), andas entities with mental properties–un-observable attributes like emotions, in-tentions, desires, and beliefs.

The proposal is that, in contrast tohumans, chimpanzees rely strictly uponobservable features of others to forgetheir social concepts. If correct, it wouldmean that chimpanzees do not realizethat there is more to others than theirmovements, facial expressions, and hab-its of behavior. They would not under-stand that other beings are repositoriesof private, internal experience. Theywould not appreciate that in addition to things that go on in the observableworld, there are forever hidden thingsthat go on in the private life of the mind.It would mean that chimpanzees do notreason about what others think, believe,and feel–precisely because they do notform such concepts in the ½rst place.

Dædalus Winter 2004 33

Behind the ape’sappearance

9 This discussion extends several previousdescriptions of this hypothesis, for example,my article with Jesse Bering and Steve Giam-brone, “Toward a Science of Other Minds:Escaping the Argument by Analogy,” CognitiveScience 24 (2000): 509–541.

Before we get too much further, let me be honest: I recognize that this pro-posal has troubling implications. For one thing, if chimpanzees do not reasonabout unobservable entities, then wewould frequently need distinctly differ-ent explanations for human and chim-panzee behavior–even in situationswhere the behavior looks almost identi-cal. Mind you, we would not need com-pletely different explanations, just onesthat are distinctive enough to capturethe proposed difference. Nonetheless,each time we witnessed a chimpanzeeengage in a complex social behavior thatresembles our own, we would have tobelieve that, unlike us, the chimpanzeehas only one conceptual system for en-coding and reasoning about what is hap-pening: a system that invokes conceptsderived from observable features of theworld. Thus, when chimpanzees deceiveeach other (which they do regularly),they would never be trying to manipu-late what others believe, nor what oth-ers can see or hear, for constructs like‘believing,’ ‘seeing,’ and ‘hearing’ arealready deeply psychological. No, indeciding what to do, the chimpanzeewould be thinking and reasoning solelyabout the abstracted statistical regulari-ties that exist among certain events andthe behaviors, postures, and head move-ments (for example) of others–what wehave called ‘behavioral abstractions.’10

I should note that humans, too, relyheavily upon behavioral abstractions intheir day-to-day interactions. We mustbe doing so: otherwise upon what basiscould we attribute additional, psycho-logical states to others? First, we recog-nize the turn of the head and the direc-tion of the eyes (observable features),then we ascribe the internal experience

of ‘seeing’ (unobservable feature). So,the proposal isn’t that chimpanzees useone system and humans use another;both species are purported to rely uponconcepts about the observable proper-ties of others. Instead, the proposal isthat chimpanzees don’t form additionalconcepts about the unobservable proper-ties of other beings (or the world in gen-eral, for that matter).

So, at face value, the proposal I havemade is worrying. In interpreting whatwould appear to be the exact same be-haviors in humans and chimpanzees indifferent ways, I seem to be applying adouble standard.

But is this implication really problem-atic, or does it just seem problematicbecause it runs counter to some of ourmost deeply engrained–but fundamen-tally flawed–ways of thinking?

Assume, for a moment, that you havetraveled back in time to a point whenthere were no chimpanzees on this plan-et–and no humans, either. Imagine fur-ther that you have come face to face withmembers of the last common ancestor of humans and chimpanzees. Let’s stipu-late that these organisms are intelligent,thinking creatures who deftly attend toand learn about the regularities that un-fold in the world around them. But let us also stipulate that they do not reasonabout unobservable things; they have no ideas about the ‘mind,’ no notion of‘causation.’

As you return to your time machineand speed forward, you will observe newlineages spring to life from this commonancestor. Numerous ape-like species willemerge, then disappear. As you approachthe present day, you will even witnessthe evolutionary birth of modern orang-utans, chimpanzees, and gorillas. Butamid all of this your attention will bedrawn to one particular offshoot of this

34 Dædalus Winter 2004

Daniel JohnPovinellion learning

10 Daniel J. Povinelli and Jennifer Vonk,“Chimpanzee Minds: Suspiciously Human?”Trends in Cognitive Science 7 (2003): 157–160.

process, a peculiar genealogy that budsoff numerous descendent species. Thisparticular lineage has evolved an eye-catching trick: it habitually stands up-right; it walks bipedally. And some ofits descendants build upon this trick,capitalizing upon the new opportunitiesit offers. For reasons that we may neverfully know, tool use and manufacture in-crease exponentially, language emerges,brain size triples, and, as more timepasses, human material and social cul-ture begins to accrete upon the shoul-ders of the lineage’s last surviving mem-ber: Homo sapiens sapiens. Now, imaginethat as part of this process, this lineageevolved new conceptual structures (in-timately connected to the evolution oflanguage) that allow them to reasonabout things that cannot be observed:mental states, physical forces, spiritualdeities.

I have stipulated all of this so we canconfront the following question: If evo-lution proceeded in this quite plausiblemanner, then how would we expect thespontaneous, everyday behavior of hu-mans to compare to that of chimpan-zees? The answer, I think, is that thingswould look pretty much the way they donow. After all, humans would not haveabandoned the important, ancestral psy-chological structures for keeping trackof other individuals within their groups,nor jettisoned their systems for noticingthat something very different happenswhen Joe turns his head toward so-and-so, just depending on whether or not hishair is standing on end. No, in evolving anew psychological system for reasoningabout hypothetical, internal mentalstates, humans would not have (indeed,could not have!) abandoned the ancientsystems for reasoning about observablebehavior. The new system by de½nitionwould depend upon the presence of old-er ones.

Now, is it really troubling to invoke adifferent explanation for what on thesurface seem to be identical units ofbehavior in humans and chimpanzees?If the scenario I have outlined above iscorrect, then the answer must be, no.After all, for any given ability that hu-mans and chimpanzees share in com-mon, the two species would share acommon set of psychological structures,which, at the same time, humans wouldaugment by relying upon a system orsystems unique to our species. The resid-ual effect of this would manifest itself innumerous ways: some subtle (such astightly constrained changes in the de-tails of things to which our visual sys-tems attend), others more profound(such as the creation of cultural artifactslike the issue of Dædalus in which youare now reading these words).

So much for theory. What about theempirical evidence; does it support theproposal I have just offered? Although itwill not surprise you to learn that I thinkit does, I have not always been of thisopinion; I used to believe that any dif-ferences between humans and chim-panzees would have to be trivial. But the results of over two hundred studiesthat we have conducted during the past½fteen years have slowly changed mymind. Combined with ½ndings fromother laboratories, this evidence hasforced me to seriously confront the pos-sibility that chimpanzees do not reasonabout inherently unobservable phenom-ena.

Let me briefly illustrate this evidencewith three simple examples: one fromthe social domain, one from the domainof physics, and one from the domain ofnumerical reasoning.

First, what does the experimental evi-dence suggest about whether chimpan-zees reason about mental states? Al-

Dædalus Winter 2004 35

Behind the ape’sappearance

though the opinions of experts differ(and have swung back and forth over thepast several years), I believe that at pres-ent there is no direct evidence that chim-panzees conceive of mental states, andconsiderable evidence that they do not.As an example, consider the well-studiedquestion of whether chimpanzees rea-son about the internal, visual experi-ences of others, that is, of whether they know anything about ‘seeing.’

To begin, no one doubts that chim-panzees respond to, reason about, andform concepts related to the movementsof the head, face, and eyes of others;these are aspects of behavior that can bereadily witnessed.11 But what about theidea that another being ‘sees’ things,that others are loci of unobservable,visual experiences?

Over the past ten years we have con-ducted dozens of studies of juvenile, ad-olescent, and adult chimpanzees to ex-plore this question. Perhaps the moststraightforward of these studies in-volved examining how chimpanzeesunderstand circumstances under whichothers obviously can or cannot seethem.12 In these studies, chimpanzeeswere exposed to a routine in which they

would approach a familiar playmate orcaretaker to request a food treat usingtheir species-typical begging gesture.Simple enough. But on the crucial testtrials, the chimpanzees were confrontedwith two individuals, only one of whomcould see them. For example, in one condition, one caretaker had a blindfoldcovering her mouth, whereas the otherhad a blindfold covering her eyes. Thequestion was to whom would the chim-panzee gesture.

Not surprisingly, in our trials with hu-man children, even two-year-olds ges-tured to whoever had the blindfold overher mouth (versus the eyes), probablybecause they could represent her inner,psychological state (“She can see me!”).In striking contrast, our chimpanzeesdid nothing of the kind. Indeed, in nu-merous studies, our chimpanzees gavevirtually no indication that they couldunderstand ‘seeing’ as an internal expe-rience of others.

With enough trials of any given condi-tion the chimpanzees were able to learnto select whoever was able to see them;after enough trials of not being handed abanana when gesturing to someone witha bucket over her head, the chimpanzees½gured out to gesture to the other per-son. Did this mean that they had ½nallydiscerned what we were asking them? Innumerous transfer tests in which we pit-ted the idea that the chimpanzees werelearning about the observable cues (i.e.,frontal posture, presence of the face oreyes) against the possibility that on thebasis of such cues they were reasoningabout who could ‘see’ them, the chim-panzees consistently insisted (throughtheir behavior) that they were reasoningabout observable features, not internalmental states, to guide their choices.

In addition to what they learned inthese tests, it also became apparent thatchimpanzees come pre-prepared, as it

36 Dædalus Winter 2004

Daniel JohnPovinellion learning

11 See Daniel J. Povinelli and Timothy J. Eddy,“Chimpanzees: Joint Visual Attention,” Psycho-logical Science 7 (1996): 129–135; Shoji Itakura,“An Exploratory Study of Gaze-Monitoring inNonhuman Primates,” Japanese Psychological Re-search 38 (1996): 174–180; Michael Tomasello,Brian Hare, and Josep Call, “Five Primate Spe-cies Follow the Visual Gaze of Conspeci½cs,”Animal Behaviour 58 (1998): 769–777.

12 Our laboratory’s empirical research ofchimpanzees’ understanding of ‘seeing’ hasbeen summarized in my article, “The Minds of Humans and Apes are Different Outcomes of an Evolutionary Experiment,” in Susan M.Fitzpatrick and John T. Bruer, eds., Carving OurDestiny: Scienti½c Research Faces a New Millenni-um (Washington, D.C.: National Academy ofSciences and John Henry Press, 2001), 1–40.

were, to make sense of certain postures.For instance, in our tests they immedi-ately knew what to do when confrontedwith someone facing them versus some-one facing away, and this ½nding hasbeen replicated in several other labora-tories.13 “But if they make that distinc-tion,” you wonder, “then why do theyperform so differently on the othertests? Is it just because they’re con-fused? How are we to make sense ofsuch a puzzling pattern of ½ndings?”

Actually, these results are not puzzlingat all if the ability to reason about men-tal states evolved in the manner that Isuggested earlier–that is, if humanswove a system for reasoning about men-tal states into an existing system for rea-soning about behavior. After all, if theidea is correct, then chimpanzees maywell be born predisposed to attend tocertain postures and behaviors related to ‘seeing’–even though they knownothing at all about such mental statesper se–precisely because overt featuresof behavior are the tell-tale indicators ofthe future behavior of others. But whensuch features are carefully teased apartto probe for the presence of a mentalisticconstrual of others, the chimpanzeesstare back blankly: this is not part oftheir biological endowment. Thus, if theevolutionary framework I have sketchedis correct, neither the chimpanzees northe results are ‘confused’; that epithetmay fall squarely upon the shoulders ofwe human experimenters and theoristswho are so blinded by our own way ofunderstanding the world that we are notreadily open to the chimpanzee’s way ofviewing things.

Of course, some have challenged thisconclusion, arguing that we need to turnup the microscope and develop moretests that will allow chimpanzees to ex-press their less well-developed under-standing of such concepts.14 So, for ex-ample, researchers at Emory Universityrecently conducted tests in which adominant and a subordinate chimpan-zee were allowed to ½ght over food thatwas positioned in an enclosure betweenthem.15 On the critical trials, two piecesof food were positioned equidistantfrom the animals. The catch was thatone piece of food was placed behind anopaque barrier so that only the subordi-nate could see it. The researchers reportthat when the subordinate was releasedinto the enclosure, he or she tended tohead for the food that was hidden fromthe dominant’s view, suggesting, per-haps, that the subordinate was modelingthe visual experience of his or her domi-nant rival.

But do such tests really help?16 Dothey reveal some weaker understanding

Dædalus Winter 2004 37

Behind the ape’sappearance

13 For example, see Autumn B. Hostetter et al.,“Differential Use of Vocal and Gestural Com-munication by Chimpanzees (Pan troglodytes) inResponse to the Attentional Status of a Human(Homo sapiens),” Journal of Comparative Psycholo-gy 115 (2001): 337–343.

14 Michael Tomasello et al., “Chimpanzees Un-derstand Psychological States–The Question Is Which Ones and to What Extent,” Trends inCognitive Science 7 (2003): 153–156, esp. 156.

15 Brian Hare et al., “Chimpanzees Know What Conspeci½cs Do and Do Not See,” Ani-mal Behaviour 59 (2000): 771–785; see also M.Rosalyn Karin-D’Arcy and Daniel J. Povinelli,“Do Chimpanzees Know What Each OtherSee? A Closer Look,” International Journal ofComparative Psychology 15 (2002): 21–54.

16 In a recent analysis of the diagnostic poten-tial of these and other tests, Jennifer Vonk and I (see footnote 10) argued that the logic of cur-rent tests with chimpanzees (and other ani-mals) cannot, in principle, provide evidencethat uniquely supports the notion that they arereasoning about mental states (as opposed tobehavior alone), and we advocated a new par-adigm of tests that may have such diagnosticpower. An alternative point of view is providedin the companion piece by Tomasello and col-

of mental states in chimpanzees? Theseare precisely the situations in whichchimpanzees will be evolutionarilyprimed to use their abilities to form concepts about the actions of others toguide their social behavior. So, for exam-ple, they can simply know to avoid foodthat is out in the open when a dominantanimal is about to be released. “Butstill,” the skeptic within you asks,“that’s pretty smart, isn’t it? The chim-panzees would have to be paying atten-tion to who’s behind the door, and whatthat other individual is going to do whenthe door opens, right?”

Fair enough. But that, in the end, is the point: chimpanzees can be intelli-gent, thinking creatures even if they do not possess a system for reasoningabout psychological states like ‘seeing.’If it turns out that this is a uniquely hu-man system, this should not detractfrom our sense of the evolved intelli-gence of apes. By way of analogy, the fact that bats echolocate but humansdon’t, hardly constitutes an intellectualor evolutionary crisis.

In the ½nal analysis, the best theorywill be the one that explains both datasets: the fact that chimpanzees reasonabout all the observable features of oth-ers that are associated with ‘seeing’–and yet at the same time exhibit a strik-ing lack of knowledge when those fea-tures are juxtaposed in a manner thatthey have never witnessed before (i.e.,blindfolds over eyes versus over themouth). I submit that, at least for thetime being, the evolutionary hypothesisI have described best meets this criteri-on.

A second example of the operation ofwhat may be a uniquely human capacityto reason about unobservables comesfrom comparisons of humans’ andchimpanzees’ commonsense under-standing of physics. Humans–even very young children–seem disposed toassume that there’s more to the physi-cal world than what meets the eye. Forexample, when one ball collides withanother, stationary one, and the secondspeeds away, even quite young childrenare insistent that the ½rst one caused the second to move away. Indeed, asMichotte’s classic experiments revealed,this seems to be an automatic mentalprocess in adult humans.17 But what isit, exactly, that humans believe causes themovement of the second ball? As Humenoted long ago, they do not merely rec-ognize that the objects touched; that’sjust a re-description of the observedevents.18 No, the ½rst one is seen as hav-

38 Dædalus Winter 2004

Daniel JohnPovinellion learning

17 Albert Michotte, The Perception of Causality(New York: Basic Books, 1963).

18 David Hume, Treatise of Human Nature, vols.1–2, ed. A. D. Lindsay (London: Dent, 1739;1911).

Table 1Theoretical causal constructs and their observ-able ‘ambassadors’

Theoretical Paired observable concept ‘ambassador’

gravity downward objecttrajectories

transfer of force motion-contact-motion sequences

strength propensity for deformation

shape perceptual form

physical connection degree of contact

weight muscle/tendon stretch sensations

leagues. However, I believe that this view dra-matically underestimates the representationalpower of a psychological system that formsconcepts solely about the observable aspectsof behavior.

ing transmitted something to the secondobject, some kind of ‘force.’ But where isthis force? Can it be seen? No, it is a the-oretical thing.

In an initial ½ve-year study of ‘chim-panzee physics,’ we focused our apes’attention on simple tool-using prob-lems.19 Given their natural expertisewith tools, our goal was to teach themhow to solve simple problems–tasks in-volving pulling, pushing, poking, etc.–and then to use carefully designed trans-fer tests to assess their understanding ofwhy the tool objects produced the effectsthey did. In this way, we attempted todetermine if they reason about thingslike gravity, transfer of force, weight,and physical connection, or merely formconcepts about spatio-temporal regular-ities. To do so, we contrasted such con-cepts with their perceptual ‘ambassa-dors’ (see table 1), much in the same waythat we had contrasted the unobservablepsychological state of ‘seeing’ againstthe observable behavioral regularitiesthat co-vary with ‘seeing.’

To pick just one example: we exploredin detail the chimpanzee’s understand-ing of physical connection–of the ideathat two objects are bound togetherthrough some unobservable interactionsuch as the force transmitted by themass of one object resting on another, or the frictional forces of one objectagainst another; or conversely, the ideathat simply because two objects arephysically touching does not mean thereis any real form of ‘connection.’ We pre-sented our chimpanzees with numerousproblems, but consider one test in whichwe ½rst taught them to use a simple toolto hook a ring in order to drag a platformwith a food treat on it toward them. Al-though they learned to do so, our real

question was whether, when confrontedwith two new options, they would selectthe one involving genuine physical con-nection as opposed to mere ‘contact.’Consistent with our ½ndings in othertests, they did not. Instead, ‘perceptualcontact’ seemed to be their operatingconcept. The observable property ofcontact (of any type) was generally suf-½cient for them to think that a tool couldmove another object.

Finally, consider the chimpanzee’s nu-merical understanding. Over the pastdecade or so, it has become apparentthat many species share what StanislasDehaene has called a ‘number sense’–the ability to distinguish between largerand smaller quantities, even when thequantities being compared occupy iden-tical volumes.20

In an attempt to explore the questionof numerical reasoning in animals, sev-eral research laboratories have trainedapes to match a speci½c quantity ofitems (say, three jelly beans) with theappropriate Arabic numeral.21 That theycan accomplish this should not be theleast bit surprising: humans and chim-panzees (and many other species) sharethe ability to visually individuate ob-jects. After extensive training, further-more, the most apt of these pupils havegone on to exhibit some understandingof ordinality (the idea that 5 represents a

Dædalus Winter 2004 39

Behind the ape’sappearance

19 Daniel J. Povinelli, Folk Physics for Apes (Ox-ford: Oxford University Press, 2000).

20 Stanislas Dehaene, The Number Sense (Ox-ford: Oxford University Press, 1997).

21 For this discussion, I rely heavily on thedetailed results from Ai, a twenty-½ve-year-oldchimpanzee whose numerical abilities havebeen studied since she was ½ve by a team led byTetsuro Matsuzawa in Kyoto, Japan. See DoraBiro and Tetsuro Matsuzawa, “ChimpanzeeNumerical Competence: Cardinal and OrdinalSkills,” in Tetsuro Matsuzawa, ed., Primate Ori-gins of Human Cognition and Behavior (Tokyo:Springer, 2001), 199–225.

larger quantity than 4, for example). So,isn’t this evidence that chimpanzeeshave a solid grasp of the notion of thenumber?

Let us scratch the surface a bit, to lookat these ½ndings from the perspective Ihave been advocating. First, do thesechimpanzees possess a dual understand-ing of numbers–both as associates ofreal object sets and as inherently theo-retical things–such that every succes-sive number in the system is exactly ‘1’more than the previous number? Thetraining data even from Ai, the mostmathematically educated of all chim-panzees, suggests that they do not. Forexample, each time the next numeralwas added into her training set, it tookher just as long to learn its associationwith the appropriate number of objectsas it took with the previous numeral. Inother words, there appeared to be littleevidence that Ai understood the symbolsas anything other than associates of theobject sets. Furthermore, even her dedi-cated mentors suggest that she was not‘counting’ at all: with quantities of up to three or four objects, she performedlike humans, using an automatic process(‘subitizing’) to make her judgments;but with larger quantities, instead ofcounting, it appears as if she was simplyestimating ‘larger’ or ‘smaller.’

What about ordinality? When ½rsttested for her understanding of the rela-tive ordering of numbers, Ai exhibitedno evidence that this was part of herconceptual structure. That is, when pre-sented with pairs of numbers, 1 versus 8,for example, she did not seem to haveany notion that the value of 1 is smallerthan the value of 8–even though she hadbeen correctly matching these numeralsto object sets for years! Of course, afterextended training, Ai did eventually ex-hibit evidence of this ability, and now,after more than ½fteen years of training,

when confronted with a scrambled ar-ray of the numerals 1 to 9, she has theremarkable ability to select them inascending order.

But what does it mean that under theright training regime we can guide achimpanzee like Ai into a performancethat looks, in many but not all respects,like human counting? One possibility is that a basic number sense–a systemgrounded to individual macroscopicobjects–is widespread among animals,and that apes (and other animals) canuse this ability (in concert with theirother cognitive skills) to ½gure out ways to cope with the ‘rules’ thathumans establish in their tests. In con-trast, the human system for counting (aswell as other mathematical ideas) couldbe seen as building upon these older sys-tems by reifying numbers as things intheir own right–theoretical things. Thismay seem like a subtle and unimportantdistinction for some tasks, but it may beone that leaves the ape mysti½ed whenfacing questions that treat numbers as things in their own right.

As a striking example of the distinc-tion I have been trying to draw, considerzero, surely one of the purest examplesthat exists of an inherently unobservableentity. If I am right, then zero ought tobe virtually undetectable by the chim-panzee’s cognitive system. And indeed,the data seem to bear this out.22 For allof her training, even Ai does not appearto have learned to understand zero inthis sense. True, she (and other animals)have quickly learned to pick the numeral0 in response to the absence of objects(something easily explained by associa-tive learning processes). But tests of or-

40 Dædalus Winter 2004

Daniel JohnPovinellion learning

22 Dora Biro and Tetsuro Matsuzawa, “Use ofNumerical Symbols by the Chimpanzee (Pantroglodytes): Cardinal, Ordinals, and the Intro-duction of Zero,” Animal Cognition 4 (2001):193–199.

dinality involving zero (choosing wheth-er 0 is greater or lesser than 6, for ex-ample) have consistently revealed what I believe might be best described as thevirtual absence of the concept. Althoughthis training has gradually forced her‘understanding’ of zero into a positionfurther and further down the ‘numberline,’ even to this day, after thousands oftrials, Ai still reliably confuses 0 with 1(and in some tasks, with 2 or 3 as well).However one wishes to interpret such½ndings, they are certainly not consis-tent with an understanding of the veryessence of zero-ness.23

Our work together is done. To the best of my ability I have laid out the casefor believing that chimpanzees can bebright, alert, intelligent, fully cognitivecreatures, and yet still have minds oftheir own. From this perspective, it may be our species that is the peculiarone–unsatis½ed in merely knowingwhat things happen, but continuallydriven to explain why they happen, aswell. Armed with a natural language thatmakes referring to abstract things easy,we continually pry behind appearances,probing ever deeper into the causalstructure of things. Indeed, some testswe have conducted suggest that chim-panzees may not seek ‘explanations’ atall.24

And yet I cannot help but suspect thatmany of you will react to what I havesaid with a feeling of dismay–perhapsloss; a sense that if the possibility I havesketched here turns out to be correct,then our world will be an even lonelierplace than it was before. But for the time being, at least, I ask you to stay thisthought. After all, would it really be sodisappointing if our ½rst, uncontaminat-ed glimpse into the mind of anotherspecies revealed a world strikingly dif-ferent from our own; or all that surpris-ing if the price of admission into thatworld were that we check some of ourmost familiar ways of thinking at thedoor? No, to me, the idea that there maybe profound psychological differencesbetween humans and chimpanzees nolonger seems unsettling. On the con-trary, it’s the sort of possibility that has,on at least some occasions, emboldenedour species to reach out and discovernew worlds with open minds and hearts.

Dædalus Winter 2004 41

Behind the ape’sappearance

23 One might retort that the numeral 0 ap-peared quite late in human history. But here’s a thought experiment. Return to our imaginarytime machine (see above) and travel back tothose civilizations that predate the invention of the numeral 0. How dif½cult would it be toteach those adult humans the position occupiedby the symbol for zero?

24 Daniel J. Povinelli and Sarah Dunphy-Lelii,“Do Chimpanzees Seek Explanations? Prelimi-nary Comparative Investigations,” CanadianJournal of Comparative Psychology 55 (2001):187–195.