Developmental Science 7:4 (2004), pp 488 - PDF document

Developmental Science 7:4 (2004), pp 488Ð498© Blackwell Publishing Ltd. 2004, 9600 Garsington Road, Oxford OXMain Street, Malden, MA 02148, USA.Blackwell Publishing, Ltd. Unwilling versus unable489© Blackwell Publishing Ltd. 2004Sarah did more poorly on. In addition, Savage-Rumbaughpresented two different language-trained chimpanzeeswith a match-to-sample procedure in which they wereshown, for example, a picture of a foot and asked tochoose between pictures of a shoe and a key. Theanimals had no training in this task, but both chim-panzees chose the closely associated objects on 25 of 28trials.Subsequent research on chimpanzeesÕ understandingof intentions has produced mixed results. Thus, Premack(1986) reported on an attempt to train Sarah to discrim-inate between videotaped sequences that depicted inten-tional actions versus those that depicted non-intentionalactions. Sarah never learned the discrimination. Sim-ilarly, Povinelli, Perilloux, Reaux and Bierschwale (1998)presented six captive juvenile chimpanzees with twoalternatives. In one, a human experimenter, on the wayto deliver juice to the chimpanzee, accidentally spilledthe juice. In the other, the experimenter intentionallypoured the juice on the ßoor. When chimpanzees werelater asked to choose between experimenters from whomthey would receive juice, none of the six chimpanzeesshowed a preference for the ÔclumsyÕ over the ÔmeanÕhuman experimenter. If chimpanzees understood theintentions of the two actors, they should have chosen toeceive the juice from the well-intentioned one. However,since they received no juice from any experimenter, it isunclear whether there was any motivation to choosebetween any of them. Chimpanzees, however, avoided athreatening experimenter, but this could simply be aesult of the negative emotional charge produced byggressive behaviors, rather than its intentional or accid-ental nature.There are also two attempts to study understandingof othersÕ unfulÞlled intentions in chimpanzees usingMeltzoffÕs (1995) behavioral re-enactment procedure,ut both yielded ambiguous results. Myowa-Yamakoshiand Matsuzawa (2000) and Call, Carpenter and Toma-sello (submitted) found that chimpanzees performed thetarget action equally as often when they saw a failedttempt as when they saw the completed action. How-in both studies, chimpanzees also performed thetarget action at high levels in a baseline condition, lim-iting what we can conclude about their understanding ofothersÕ intentions from these studies.Call and Tomasello (1998) trained chimpanzees andorangutans to identify a marker placed on top of one ofthree opaque buckets as an indicator for the location ofhidden food. During training the apes never saw thehuman actually placing the marker on the bucket, butthe marker was already on top of one of the bucketshen they were presented to the ape. On test trials ahuman experimenter then placed (or dropped) themarker on one of the buckets intentionally, but eitherbefore or after this he let the marker fall accidentallyonto one of the other buckets. The marker was removedthe time of choice for the ape, so for test trials the apeas faced with a choice in which one bucket had beenmarked with the marker intentionally and the otheraccidentally. Apes as a group chose the bucket that wasmarked intentionally (with the intentional drop showingthe strongest effect), although no individual except alanguage-trained orangutan was above chance on hisecently, Suddendorf and Whiten (2001) have indic-ted that there is evidence that apes recognize certainmental states such as attention and intention. Besidesthe experimental studies previously cited, these authorsused observational data to support this idea. Althoughsuggestive, these positive results (both observational andxperimental) must be interpreted with caution. Observa-tional data can be problematic because isolated observa-ons can often be interpreted in multiple ways. Onthe other hand, experimental data can be problematicbecause usually there is a fairly extensive amount oftraining involved before or during the test. Moreover,both observational and experimental approaches havethe potential problem that because intentional and accid-ental actions differ from one another in a number ofys, it is possible that the apes were focused on one ofthese superÞcial characteristics rather than the under-ying mental states. In the current study, therefore, wesought a method for investigating apesÕ understandingof intentional action that involved (1) no training, (2) thepresentation of multiple conditions to minimize successusing a single superÞcial cue (thus creating a triangu-lation procedure) and (3) which used apesÕ natural reac-tions as responses.The basic set-up was that a human experimenter (E)began giving food normally to an ape through a hole ina Plexiglas wall. Occasionally during this feeding rout-ine, E conducted a test trial in which the food transferas delayed: E brought out another piece of food andeither refused to give it to the ape or else attempted togive it to the ape unsuccessfully. E was thus unwilling orunable, respectively, to give the food during test trials.There were three trios of different conditions, with oneunwilling and two unable conditions in each trio. Weused two unable conditions for each unwilling conditionto obtain a broader comparison between conditions.ithin each trio all conditions were matched for thegazing behavior of the experimenter and, as closely as 490Josep Call © Blackwell Publishing Ltd. 2004possible, for general body motions and position of theood. However, it was of course necessary that eachcondition presented speciÞc behaviors so that conditionscould be differentiated from each other. We used mul-tiple conditions (each depicting different behaviors) toguard against the use of single superÞcial cues as anxplanation for our results. For instance, a difference ina particular unwillingÐunable pair could be explaineda single superÞcial cue (rather than its underlyingintentions), but such a single cue would lose some of itsxplanatory power if individuals still differentiated be-tween other unwilling and unable pairs in the absence ofthat speciÞc superÞcial cue. Using multiple and dispar-te tests that share a common psychological feature hasbeen previously used in other studies (e.g. Meltzoff, 1995;Premack & Woodruff, 1978). The experimental questionas whether the apes would behave differently when thehuman was unwilling to give them food as opposed tohen he was unable to do so, for example, by waitingpatiently during his well-intentioned attempts but tryingto spur him into action or leaving when he simply wasefusing to give the food.elve captive born chimpanzees (4Ð26 years old; mean15.4 years) participated in this study. All chimpanzeeshoused together in several large enclosures at theolfgang Kšhler Research Center, Leipzig Zoo, Germany.le 1 presents the age, sex and rearing history of eachchimpanzee. All the chimpanzees had been moved to theLeipzig Zoo only 6 months prior to testing and werestill becoming accustomed to their new home. They hadnever completed any cognitive experiments before thisA chimpanzee was allowed into a set of three testingooms, and moved into one of the extreme rooms 130 cm). The doors between the rooms wereleft open. Once the chimpanzee was in position, E(always the same in all experiments) sat facing the chim-panzee at eye level separated by a transparent Plexiglaspanel with three feeding holes cut into the bottom(6 cm in diameter and 23 cm apart), through which foodcould be given. A small transparent Plexiglas panel 69 cm) with smaller holes (3.5 cm diameter)as placed over the holes in the Plexiglas panel so thatchimpanzees could only stick their Þngers through theholes toward E. In front of E there was a small table 30 cm) ßush against the Plexiglas panel under-neath the feeding holes. On this table there was asmall wooden ramp (66 cm 30 cm) that was also ßushgainst the Plexiglas panel and angled down toward E(see Figure 1).ithin a session there were motivational and testtrials. Before and after each test trial there were a set ofmotivational trials, in which the chimpanzee receivedbetween two and six grapes in succession (counterbalancedacross subjects). In test trials, which lasted 30 seconds,the chimpanzee did not receive any food. There were twotypes of test trials, unwilling and unable, represented bythree and six conditions, respectively. In the six unableconditions E was unable to transfer the food becausesomething in the situation prevented it (e.g. the feedinghole was too small) or he was distracted, whereas in thethree unwilling conditions E simply refused to transferthe food for a variety of reasons.The nine different conditions were organized into threetrios of conditions. Each trio contained one unwilling(years)Sexobert26MNurseryeit25FNurseryNatascha24FNurseryDorien22FNurseryaukje26FNurseryUlla22FNurseryahaga8FMotherFiÞ8FMotherSandra8FMotherGertruida8FMotherodo8MMothertrick4MMother Figure 1 The testing situation. Unwilling versus unable491© Blackwell Publishing Ltd. 2004condition and two unable conditions. The trios weremed based on the similarity of EÕs surface movementsacross conditions. Within each trio, EÕs looking behaviorand body movements were highly similar, and the foodas placed in identical locations as well. In each condi-tion, E repeated or continued the action for 30 seconds.The three trios were the following:(i). Tease trio (E alternated gaze between the food andthe SÕs face; food moving back and forth):Unwilling Tease: E held up a grape for the chimpanzeeto see and then moved it toward the hole in the Plexiglasto transfer it to the chimpanzee. Once the chimpanzeettempted to take the grape with her lips or Þngers, Epulled the grape back from the hole toward himself andlooked up at the chimpanzee (the grape was held abovea designated spot on the table).Unable Clumsy: E held up the grape for the chimpanzeeto see and moved it toward the hole in the Plexiglas totransfer it to the chimpanzee. However, when E tried totransfer the grape he accidentally dropped it after touch-ing the chimpanzeeÕs protruding mouth or Þnger or afterhitting the Plexiglas panel above the feeding hole. Thegrape rolled down the ramp on the table toward E andE looked up at the chimpanzee.Unable Blocked Hole: At the end of the motivationaltrial E quickly reversed the small Plexiglas panel so thatit almost completely blocked the three feeding holes.E held up the grape for the chimpanzee to see andttempted to push the grape through the small hole leftin the Plexiglas. After unsuccessfully trying to push thegrape through, E pulled it back toward himself andlooked up at the chimpanzee.ii). Refuse trio (E looked at SÕs face; food remained still):At the beginning of this session a small (22 cm opaque occluder was attached to the ramp on the tableso that any grape that might roll down the ramp couldbe seen by the chimpanzees but not by E.Unwilling Refuse: E placed a grape on top of theopaque occluder so that both chimpanzee and E couldsee the grape, and then E stared at the chimpanzee.Unable Distracted: E placed a grape on top of theopaque occluder so that both chimpanzee and E couldsee the grape, and then E crossed his arms and vigor-ously scratched his arms, shoulders and back whilestaring at the chimpanzee.Unable CanÕt See: While giving the last grape in amotivational trial, E surreptitiously dropped anothergrape so that it rolled under his arm into the opaqueoccluder (8 cm below the top of the container so that thechimpanzee could see the grape but E could not). E thenstared at the chimpanzee.(iii). Eat Trio (E looked down at what he was doing;ood remained in front of E):In all of these trials E was holding a bucket of grapesbetween his legs.: E took a grape from the bucket, bit offa small piece, and then held the grape just above thetable surface for 4Ð5 seconds while continuing to chewhile staring at the table.Unable Search: E took a grape from the bucket andplaced it on the table while looking at and searching insidethe bucket with both hands in order to get more food.Unable Stuck: E took a transparent plastic tube con-taining a trapped grape, held it over the table, andttempted to dislodge the grape by putting his Þngerinto the tube, hitting the tube or shaking it slightly whilestaring at the tube.Each subject received three test sessions, one for eachtrio (see Table 2). Test sessions took place on differentdays. Each session contained two trials per condition fora total of six test trials per session. Thus each subjecteceived 18 trials in all. The order of the conditions wascounterbalanced across subjects.All trials were videotaped and scored from the tape. Weecorded two dependent measures during the 30 secondsof each test trial: behavioral rate and participation.Behavioral rate consisted of the frequency of relevantbehaviors divided by the time the subject was in view ofthe camera. For instance, if during a trial the subject wasin view 20 seconds and produced behavior A Þve timesand behavior B once, the corresponding behavioralte would be 0.3. The behaviors considered were thosecommonly used by chimpanzees to request food fromhumans. Thus they include a mixture of begging andassertive/coercive acts. In particular, we coded the fol-lowing behaviors:Þnger poking: the rate at which the chimpanzee put itsÞnger(s) through any of the three feeding holes;move apparatus: the rate at which the chimpanzeepushed the Plexiglas panel or wooden ramp on the tableout of its normal position; Session 1Session 2Session 3rial 1ClumsySearchDistractedrial 2TeaseEatRefuserial 3Blocked HoleStuckCanÕt Seerial 4ClumsySearchDistractedrial 5TeaseEatRefuserial 6Blocked HoleStuckCanÕt See 492Josep Call © Blackwell Publishing Ltd. 2004knocking: the rate at which the chimpanzee used anypart of its body to make an audible noise by striking anobject (e.g. by striking its hand against the Plexiglaspanel). Vocalizations and hand clapping were not in-cluded because they appeared at very low rates.articipation consisted of the amount of time thechimpanzee remained at the testing station (since theycould leave any time during the trial because the doorsleft open). We used the latency (in seconds) to leavethe station once the trial had started.A second coder, unaware of the study rationale andlind to the experimental conditions, scored 20% oftrials to assess inter-observer reliability. Reliability wascellent (behavioral rate: Spearman 0.97; participa-tion latency to leave: Spearman 1.0). In all analysesused two-tailed Wilcoxon and Friedman tests.Overall, chimpanzees produced a higher behavioral ratein the unwilling than in the unable conditions (Wilcoxon .008; unwilling: mean .11, SEM unable: mean .08, SEM .02). However, this was notthe case equally across trios. Figure 2 presents thebehavioral rate for each of the three trios separately.There were also signiÞcant differences across conditionsin the Tease trio by itself (Friedman test 17.64, df .001) and the Eat trio by itself (Friedman test 10.17, df .006), but not in the Refuse trio (Fried- 1.27, df .53). Multiple comparisonswithin the Tease trio revealed that chimpanzees signi-Þcantly increased their behavioral rate in the Unwillingease condition compared to the Unable Blocked Holecondition (Wilcoxon T .003) but notthe Unable Clumsy condition (Wilcoxon T .18). Multiple comparisons within the Eat trioealed that chimpanzeesÕ behavioral rate was signi-Þcantly higher in the Unwilling Eat condition than inthe Unable Stuck condition (Wilcoxon T .008) but not the Unable Search condition (Wilcoxon .20). Thus, in two of the three trios,chimpanzees had a higher behavioural rate in the unwill-ing condition than in one of the two unable conditions.This pattern shows that differences were especiallylarge when we distinguished between conditions in whichE actually acted on the grape (i.e. Tease, Clumsy, BlockedHole, Eat, Stuck), rather than just ignored it while hedid something else (i.e. Refuse, Distracted, CanÕt See,Search; see Figure 3). There were signiÞcant differencesbetween these two groups of conditions (Friedman test 13.0, df .005). Pairwise comparisons revealedthat chimpanzees produced more behaviors in the 00.10.20.30.4 Mean behavioral rateUnwilling ClumsyRefuseCan’tDistractedEatStuck 00.050.10.150.2E acted on grape Mean behavioral rate UnwillingUnableYesNoMean behavioral rate (SEM) for the unwilling and unable conditions as a function of whether E acted on the grape during the trial. Unwilling versus unable493© Blackwell Publishing Ltd. 2004unwilling than in the unable conditions when E acted onthe grape (Wilcoxon T .002) but nothen he did not act on the grape (Wilcoxon T .75). Likewise chimpanzees produced morebehaviors in those unwilling conditions in which E actedon the grape compared to those in which he did not(Wilcoxon T Overall, chimpanzees left the testing station earlier in theunwilling than in the unable conditions (Wilcoxon T .023; unwilling: mean 17.5, SEM unable: mean 21.7, SEM 1.2). Once more, this wasnot the case equally across trios. Figure 4 presents thelatency to leave the station for each of the three triosseparately. There were signiÞcant differences acrossconditions in the Tease trio by itself (Friedman test 12.76, df .002), but not in the Refuse (Friedman 2.4, df .30) or the Eat trios (Friedman 2.0, df .37). Multiple comparisons withinthe Tease trio revealed that chimpanzees left earlier inthe Unwilling Tease condition compared to the UnableClumsy condition (Wilcoxon T ut not the Unable Blocked Hole condition (Wilcoxon .25). An analysis of other participa-tion variables such as total time at the station producedidentical results.Again, differences were especially large when we dis-tinguished between conditions in which E actually actedon the grape (i.e. Tease, Clumsy, Blocked Hole, Eat,Stuck), rather than just ignored it while he did some-thing else (i.e. Refuse, Distracted, CanÕt See, Search).Figure 5 presents the mean latency to leave the stationor the unwilling and unable conditions as a function ofhether E acted on the grape during the trial. TheresigniÞcant differences between conditions (Fried- 9.66, df .022). Pairwise comparisonsealed that chimpanzees left earlier in the unwillingcompared to the unable condition when E acted on thegrape (Wilcoxon T .003) but not whenhe did not act on the grape (Wilcoxon T Overall, chimpanzees produced more behaviors and leftearlier in those conditions in which E was unwilling asopposed to unable to transfer the food. However, thisdifference in interpreting unwilling and unable actionsxisted only in those conditions in which E made contactwith the food to be transferred.This difference between unwilling and unable conditionsas observed even within trios, in which EÕs looking 051520 Unwilling UnableBlockedHoleClumsyRefuseCan’tDistractedEatStuck 0510152530E acted on grapeMean latency (sec.) illin UnableYesNo 494Josep Call © Blackwell Publishing Ltd. 2004pattern and the movement pattern of the food werematched as closely as possible across conditions. Althoughsurface behavioral cues such as the pattern of foodmotion or eye contact cannot easily account for theseesults, it is still possible that other factors may explainthe differences between unwilling and unable conditions.One possibility is that instead of chimpanzeesÕ differentialbehavior reßecting communicative attempts to changeEÕs intention, the differences observed were simply theesult of a general feeling of frustration at not receivingthe food right away (e.g. knocking the Plexiglas) and/orsimply reßected the chimpanzeeÕs attempts to get theood on her own (e.g. Þnger-poking through the hole).This hypothesis could be tested by presenting a non-social condition in which E left the room after placingthe grape on the platform. Alternatively, chimpanzeesmay have reacted to certain movements independentlyof what E was trying to achieve or what was happeningto the grape. This could be tested by presenting thesame movements with and without grapes. The nextstudy was designed to address these two potentialxplanations.The purpose of this experiment was to establish with twocontrol tests whether chimpanzees behaved the samey when some of the key elements in the situation Ð thepresence of E or the presence of food Ð were removed.First, to assess whether the behaviors were directed at Eor purposes of communication rather than being justthe result of a frustrating situation, we used a non-socialtest in which we compared a condition in which E hadleft the room with the Unwilling Refuse and the UnableCanÕt See conditions from the previous study. Theeason for choosing these two conditions was that theydid not involve movements and therefore were the bestmatches for E not being present. Second, to assesshether the behaviors were interpreted the same way inthe absence of food Ð that is, to test whether chimpan-ees were responding based simply on something aboutEÕs bodily movements and not his goal-directed actionsÐ we presented again three of the conditions that hadproduced clear results in the previous study (Unwillingease, Unable Clumsy and Unable Stuck) with andwithout food being involved.The same chimpanzees served as participants.The general procedure was identical to that of Experi-ment 1, with the only difference being the conditions(i). Social Control testUnwilling RefuseUnable CanÕt See: E placed the grape on the platformas in the other conditions and then left the room for 30seconds.(ii). Food Control testUnwilling TeaseNo Food Tease: E performed the same actions (andgaze alternation) as in the Unwilling Tease condition butwithout a grape in his hand.Unable ClumsyNo Food Clumsy: E performed the same actions (andgaze alternation) as in the Unable Clumsy condition butwithout a grape in his hand.Unable StuckNo Food Stuck: E performed the same actions (whilestaring at the tube) as in the Unable Stuck condition butwithout a grape inside the tube.Half of the subjects received the Food Control condi-tions Þrst and the other half received the Social Controlconditions Þrst. Each group of conditions was tested ona different day. Within a session chimpanzees receivedtwo trials of each condition, for a total of 12 test trials inthe Food Control session and six test trials in the SocialControl session. The order of conditions was counter-balanced across subjects. All trials were videotaped.Same as in Experiment 1. Inter-observer reliability (basedon 20% of the trials) was excellent (behavioral rate:Spearman .96; participation latency to leave: Spear-Figure 6 presents the behavioral rate for each of the threeconditions. There were signiÞcant differences across con-ditions (Friedman test 8.58, df .014). Multiplecomparisons across conditions revealed that chimpan-eesÕ behavioral rate was signiÞcantly lower in the NoExperimenter condition compared to the Unwilling Refuse(Wilcoxon T .051) and the Unable CanÕtSee (Wilcoxon T .038) conditions. As inExperiment 1, there were no signiÞcant differences between Unwilling versus unable495© Blackwell Publishing Ltd. 2004the Unwilling Refuse and Unable CanÕt See conditions(Wilcoxon T .72). Thus, chimpanzeesproduced more behaviors when E was present.Figure 7 presents the latency to leave the testing sta-tion across conditions. There were signiÞcant differencesacross conditions (Friedman test 11.49, df .003).airwise comparisons revealed that chimpanzees leftearlier in the No Experimenter condition compared tothe Unwilling Refuse (Wilcoxon T and Unable CanÕt See (Wilcoxon T conditions. As in Experiment 1, there were no signiÞcantdifferences between the Unwilling Refuse and UnableCanÕt See conditions (Wilcoxon T Figure 8 presents the behavioral rate for each of thesix conditions in this test. There were no signiÞcant dif-ferences across conditions (Friedman test 9.0, df Figure 9 presents the latency to leave the testingstation across conditions. There were signiÞcant differ-ences across conditions (Friedman test 26.03, df .001). Pairwise comparisons within each pair revealedthat chimpanzees left earlier in the absence of food thanhen food was present for the Unwilling Tease (Wilcoxon .04) and the Unable Clumsy (Wilcoxon .003) conditions but not for the UnableStuck condition (Wilcoxon T .95). Thus,in two of the three conditions tested, chimpanzees leftearlier when E acted without food. Focusing on the foodconditions, as a replication of Experiment 1, there weresigniÞcant differences across conditions (Friedman test 15.40, df .001). Pairwise comparisons revealedthat, as in Experiment 1, chimpanzees left earlier in theUnwilling Tease condition compared to the UnableClumsy condition (Wilcoxon T There were no differences between the Unwilling Teaseand the Unable Stuck conditions (Wilcoxon T 0.10.2Mean behavioral rateNoExperimenterRefuseCan’t SeeMean behavioral rate (SEM) for each of the conditions in the Social Control test. 0510RefuseCan’t SeeMean latency (SEM) to leave the testing station for each of the conditions in the Social Control test. 00.10.20.3Mean behavioral rate NoFoodTeaseStuck Figure 8Mean behavioral rate (SEM) for each of the conditions in the No Food Control test. 051015202530Mean latency (sec.) TeaseStuck Figure 9Mean latency (SEM) to leave the testing station for each of the conditions in the No Food Control test. 496Josep Call © Blackwell Publishing Ltd. 2004ith the conditions that were identical across Experi-ments 1 and 2, we replicated our previous results. Moreinterestingly, the Social control test showed that chim-panzees produced less behaviors and left earlier when Ehad left the room than in those conditions in which Eemained in the room. This suggests that in Experiment1 their behavior was directed at E, and it is unlikely thatthe chimpanzeesÕ responses were solely a result of fru-stration or an attempt to reach the food on their own.Instead, we argue that chimpanzeesÕ responses constitutedcommunicative attempts to encourage E to transfer thegrape similar to those reported in previous studies (Call& Tomasello, 1994; Leavens & Hopkins, 1998).Furthermore, the No Food control test showed thatchimpanzees left earlier in the no-food than the foodcondition in two of the three conditions investigated.There were no differences in the mean behavioral rateacross conditions with or without food, however. Thusthis test offered partial support to the idea that theesults of Experiment 1 were not solely based on the useof certain movements independently to what E was tryingto achieve or what was happening to the grape.The current study provides suggestive evidence thatchimpanzees spontaneously (i.e. without training) aresensitive to othersÕ intentions. Observing the behavior ofa human not giving them food, chimpanzees demon-strated in their spontaneous behavior that they recognizeda difference between cases in which he was not givingood because he was unwilling to or because, for variouseasons, he was unable to. Chimpanzees thus did not justperceive othersÕ behavior, they also interpreted it. Theydid this with no training of any kind, and despite the factthat EÕs gazing pattern was identical within trios and witha variety of behaviors indicating that E was unwilling orunable, thus making discrimination on the basis of speciÞcsuperÞcial cues highly unlikely. Control tests suggestedthat chimpanzees were trying to inßuence EÕs behaviorwith communicative gestures and were not respondingsolely as a result of frustration or expectation.It is notable that the Þndings were especially strongor those individual experimental conditions in which Edirected an overt behavior of one kind or another to theood (i.e. as opposed to conditions in which he did nottouch the food). It is thus possible that chimpanzeesunderstand intentions most readily when they are moreor less directly perceivable in behavior Ð for example,through signs of effort, frustration, accident and so forthÐ what Searle (1983) calls Ôintention in actionÕ. Althoughcertain actions such as those in which the experimenteracted on the grape may be especially important toinfer intentions, the presence of these actions alone can-not be the whole explanation for the differences betweenunwilling and unable conditions. Note that in some casesthe experimenter acted on the grape both in unwillingand unable conditions and subjects still reacted differ-ently across conditions (see Figure 3). Chimpanzees mayor may not understand the intentions of others whenthese can only be discerned more indirectly Ð what Searlecalls Ôprior intentionÕ. Alternatively, it could be thatchimpanzees perceived EÕs behavior in the unable condi-tions in which he did not act on the grape as unwillinginstead of unable. That is, in the Distracted and Searchconditions, E conceivably could have stopped scratchingor searching just for a moment to give the chimpanzeethe grape, and in the CanÕt See condition if E did not seethe concealed grape, he should have moved on to the nexttrial. It would be interesting to test other unable conditionsin which E does not touch the grape but is truly unableto give it, for example because his arms have been tied tohis sides by an assistant. Future studies of apesÕ under-standing of othersÕ prior intentions are also needed.Because chimpanzees only discriminated betweenunwilling and unable conditions when E acted on theood, there is a possible alternative explanation that doesnot involve understanding of othersÕ intentions. Perhapschimpanzees had learned from their previous experienceto expect that certain actions usually result in themeceiving food and certain actions usually result in themnot receiving food (Behne, personal communication).or example, normally after humans drop a piece ofood on the way to giving it to the chimpanzee, they pickit up and give it to the chimpanzee, whereas normallyhen humans are eating they do not share their foodwith the chimpanzee. Based on their perception of thehumanÕs action, chimpanzees thus might be more patientin the Þrst instance and more likely to leave or beg forthe food in the second instance, simply based on theirxpectation of the probability of receiving food. How- if chimpanzees were using their previous experienceof EÕs actions to decide how to react, they would havehad to have a separate learning history for each of theÞve conditions in which they discriminated successfully.This is unlikely because some conditions, at least, arguablynovel to the chimpanzees and because these chim-panzees had little experience with experimenters or test-ing situations in general because they were new to thefacility. Note that chimpanzees could not have devel-oped such an expectation during the test because theynot differentially rewarded in the experimental con-ditions and there was no training involved. Unwilling versus unable497© Blackwell Publishing Ltd. 2004Another possibility is that chimpanzees were reactingto the physical constraints of the situation rather thanthe mental states of the experimenter. For instance,chimpanzees may have reacted to whether the action ofthe experimenter to get them the grape was impeded bysome physical barrier. This could explain the differencebetween the Tease and Blocked Hole conditions and theEat and Stuck conditions. In both unable conditions(Blocked Hole and Stuck), the grape could not be trans-ferred to the chimpanzee because there was a physicalbarrier (a hole which was too small or a plastic tube)that prevented the grapeÕs free movement. However, thisxplanation cannot account for all our results because(1) counterintuitively, subjects stayed longerpath of the reward was blocked in the unable conditionsand (2) subjects still left 11 seconds earlier in the Teasecompared to the Clumsy condition even though thereas no barrier blocking the path of the reward in eitherbelieve that chimpanzees were using the actions ofthe experimenter not just as superÞcial discriminativecues but as a way to determine his goal. This is analogousto WhitenÕs (1994) invocation of the notion of interven-ing variable to explain disparate behavioral acts with anunderlying mental cause. Goal would be the interveningariable governing the expression of those superÞcialcues. Whether ÔgoalÕ refers to a physical target (food inthe mouth of the chimpanzee), a mental desired endstate of affairs (E wants the chimpanzee to have the food),or an intention understood as a plan of action (E istrying to give the food to the chimpanzee) is a questionthat deserves further study. That is, intentions but notgoals involve the actorÕs actions, and speciÞcally theknowledge that the actor can choose one means amongseveral possible means to achieve an end (Piaget, 1952;omasello, 1999). Note that Meltzoff (1995) also calledttention to similar distinctions when discussing hisesults with 18-month-old children. In the current study,believe that chimpanzees are using more than asuperÞcial understanding of EÕs physical goal but wecannot determine from this study whether they have anunderstanding of EÕs mental goal or intention. In theunable conditions we believe that chimpanzees waitedlonger and gestured less frequently because they believedthat E would continue to act until the transfer of theood was achieved. In the conditions in which E actedon the grape, he used slightly different means (e.g. in theBlocked Hole condition he Þrst tried to insert the grapewith one hand position and then rotated his hand orused more force with his further attempts), so this infor-mation was available to chimpanzees. To adult humans,this type of action would appear more unable/intentional/goal-directed than repeatedly pushing the grape into thehole with the exact same movements. Further study ofdifferent levels of understanding of intentions (includingprior intentions and communicative intentions; Toma-sello, 1999) is needed in chimpanzees, children and otheranimals.In sum, the current study suggests that chimpanzeesdo not simply perceive the behavior of others, they alsointerpret it. In particular, they can distinguish betweenan experimenter that is either unwilling or unable to givethem food. These data together with recent studies onchimpanzeesÕ knowledge about what others can or can-not see (Hare, Call, Agnetta & Tomasello, 2000; Hare,Call & Tomasello, 2001; see also Suddendorf & Whiten,2001) show that chimpanzees have more knowledge aboutsome psychological states of others than previouslybelieved by some authors (Heyes, 1998; Povinelli &Eddy, 1996; Tomasello & Call, 1997) and answers thecall for more experimental evidence on this topic madeothers (Whiten & Byrne, 1991). Currently, the debateon this issue is still open (see Povinelli & Vonk, 2003,and Tomasello, Call & Hare, 2003a, 2003b, for a recentdiscussion). Future studies should be devoted to pin-pointing the level of sophistication of each of these typesof social-cognitive understanding and to charting apesÕknowledge about other mental states such as belief,desire and attention that fall under the umbrella of the-ory of mind research.thank Chikako Suda for help with coding and TanyaBehne for helpful comments.Call, J., Carpenter, M., & Tomasello, M. (submitted). Focusingon outcomes and focusing on actions in the process of sociallearning: chimpanzees (an troglodytes) and human childrenCall, J., & Tomasello, M. (1994). Production and comprehen-sion of referential pointing by orangutans (ongo pygmaeusournal of Comparative Psychology, 307Ð317.Call, J., & Tomasello, M. (1998). Distinguishing intentionalfrom accidental actions in orangutans, chimpanzees, andhuman children. ournal of Comparative PsychologyHare, B., Call, J., Agnetta, B., & Tomasello, M. (2000). Chim-panzees know what conspeciÞcs do and do not see. Hare, B., Call, J., & Tomasello, M. (2001). Do chimpanzeesknow what conspeciÞcs know and do not know? , 139Ð151. 498Josep Call © Blackwell Publishing Ltd. 2004Heyes, C.M. (1998). Theory of mind in nonhuman primates.Behavioral & Brain SciencesLeavens, D.A., & Hopkins, W.D. (1998). Intentional commun-ication by chimpanzees: a cross-sectional study of the use ofeferential gestures. Developmental Psychology, 813Ð822.Malle, B.F., Moses, L.J., & Baldwin, D.A. (Eds.) (2001). tions and intentionality: Foundations of social cognition.Cambridge, MA: MIT Press.Meltzoff, A.N. (1995). Understanding the intentions of others:e-enactment of intended acts by 18-month-old children.Developmental Psychology, 838Ð850.Myowa-Yamakoshi, M., & Matsuzawa, T. (2000). Imitation ofintentional manipulatory actions in chimpanzees (an trog-ournal of Comparative PsychologyPiaget, J. (1952). The origins of intelligence in children. Newvinelli, D.J., & Eddy, T.J. (1996). What young chimpanzeesknow about seeing. Monographs of the Society for Researchin Child Developmentvinelli, D., Perilloux, H., Reaux, J., & Bierschwale, D.(1998). Young chimpanzeesÕ reactions to intentional versusaccidental and inadvertent actions. Behavioral Processes205Ð218.vinelli, D.J., & Vonk, J. (2003). Chimpanzee minds: suspici-ously human? ends in Cognitive SciencesPremack, D. (1986). Gavagai!Cambridge, MA: MIT Press.Premack, D., & Woodruff, G. (1978). Does the chimpanzeehave a theory of mind? Behavioral and Brain Sciences515Ð526.Savage-Rumbaugh, E.S., Rumbaugh, D., & Boysen, S.T.(1978). SarahÕs problems in comprehension. Behavioral andBrain Sciences, 555Ð557.Searle, J. (1983). ersity Press.Suddendorf, T., & Whiten, A. (2001). Mental evolution anddevelopment: evidence for secondary representation in chil-dren, great apes, and other animals. Psychological Bulletin, 629Ð650.The cultural origins of human cognition.Cambridge, MA: Harvard University Press.omasello, M., & Call, J. (1997). New York:Oxford University Press.omasello, M., Call, J., & Hare, B. (2003a). Chimpanzeesunderstand psychological states Ð the question is which onesand to what extent. ends in Cognitive Sciences, 153Ð156.omasello, M., Call, J., & Hare, B. (2003b). Chimpanzeesersus humans: itÕs not that simple. ends in Cognitive Sci-, 239Ð240.hiten, A. (1994). Grades of mindreading. In C. Lewis &ChildrenÕs early understanding of mind47Ð70). Hillsdale, NJ: Lawrence Erlbaum Associates.hiten, A., & Byrne, R.W. (1991). The emergence of metarep-esentation in human ontogeny and primate phylogeny. InA. Whiten (Ed.), Natural theories of mind(pp. 267Ð281).Oxford: Blackwell.Zelazo, P., Olson, D., & Astington, J. (Eds.) (1999). Developingtheories of intention.Cambridge: Cambridge University Press.eceived: 2 January 2003 Accepted: 8 November 2003