科技文献：黑猩猩具备心理理论吗？

e r Pla able not only to react to what others are doing but also to Review others in many situations based on past experience (and perhaps specialized cognitive adaptations), but they do not which chimpanzees begged indiscriminately from humans facing them and others with buckets over their heads (see also Ref. [7]). There was also one negative study on chim- panzees’ understanding of false beliefs [8]. All of these data led Tomasello and Call [1] to the general conclusion that chimpanzees and other non- human primates do not understand the psychological states of others. That is, they can predict the actions of anticipate what theywill do. Oneway of accomplishing this is by observing what others do in particular situations and deriving a set of ‘behavioral rules’ (or, in some cases, having those built in). This will enable behavioral prediction when the same or a highly similar situation arises again. But anotherway to do it – indeed amore flexible way to do it – is to discern directly what the other is trying to do, what state of the environment he is trying to bring about and what his goal is. This enables behavioral prediction not only in previously observed or highly similar situations but also panzees’ understanding of visual perception, especially from the well-known studies of Povinelli and Eddy [6] in tive evidence also accrued during the 1990s about chim- group, highly social animals, such as chimpanzees,must be Does the chimpanz mind? 30 years late Josep Call and Michael Tomasello Max Planck Institute for Evolutionary Anthropology, Deutscher On the 30th anniversary of Premack and Woodruff’s seminal paper asking whether chimpanzees have a theory of mind, we review recent evidence that suggests in many respects they do, whereas in other respects they might not. Specifically, there is solid evidence from several different experimental paradigms that chimpan- zees understand the goals and intentions of others, as well as the perception and knowledge of others. Never- theless, despite several seemingly valid attempts, there is currently no evidence that chimpanzees understand false beliefs. Our conclusion for the moment is, thus, that chimpanzees understand others in terms of a percep- tion–goal psychology, as opposed to a full-fledged, human-like belief–desire psychology. Introduction In 1978 Premack and Woodruff asked, ‘Does the chimpan- zee have a theory of mind?’ In this brief review we attempt to answer this question based on much research that has been conducted in the 30 years since that time, particularly in the last decade or so (see Tomasello and Call [1] for the state of the art as of the mid-1990s). The answer will not be a simple yes or no, however, because part of the progress that has been made in recent years is a recognition that there are many different ways in which organisms might understand the psychological functioning of others. A brief history Premack and Woodruff’s [2] original study was actually about chimpanzees’ understanding of human goals. But soon therewas new research suggesting that perhaps these results were experimental artifacts [3,4] and other research on social learning that suggested chimpanzees did not have an understanding of human goals [5]. Nega- Corresponding author: Call, J. (call@eva.mpg.de). 1364-6613/$ – see front matter � 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.tics.2008.0 e have a theory of tz 6, D-04103 Leipzig, Germany go beneath the surface to an understanding of the goals, perceptions, knowledge and beliefs that guide action (nor to an understanding of underlying physical forces either). But not all of the data existing at that time were consistent in suggesting this conclusion, and moreover, chimpanzees had been observed doing things that would seem to require more than just an understanding of surface-level behavior, for example, tactical deception [9]. The story since the late 1990s has been one of exper- imenters finding better ways to tap into what chimpanzees know about the psychological states of others – and so getting many more positive results. In most cases this has been guided by attempts to model the experiments more closely on situations that chimpanzees routinely encounter in their natural environments, for example, presenting them with problems not in situations in which they must cooperate with others but, rather, in situations in which they must compete with others (see Box 1). Skeptics still abound, as represented most prominently by Povinelli and colleagues [10,11]. They cling to the hypothesis that chim- panzees understand only surface-level behavior (forming ‘behavioral rules’), and indeed this explanation is almost always possible for any single experiment. But there are now in many cases multiple experimental paradigms all aimed at a single psychological state – each presenting chimpanzees with a highly novel problem – that makes the positing of learned behavioral rules a difficult explanatory strategy [12]. Here, we review current experimental evi- dence about chimpanzees’ understanding of: (i) the goals and intentions of others and (ii) the perception, knowledge and beliefs of others. Understanding goals and intentions To compete and cooperate effectively with others in their in novel situations. The methodological problem is that when an actor acts toward a goal successfully, then what he is trying to do and 2.010 187 Box 1. Assessing visual and auditory perspective taking in chimpanzees Hare et al. [26] had chimpanzees compete for food with a human competitor who was inside a glass booth (see Figure I). In the main study chimpanzees preferentially chose to approach the food that was on the side of the booth with an opaque barrier so that the human could not see them approaching. They, thus, attempted to influence what the human could see, and they did so from the very Review what he actually does are the same, and so it is very difficult, if not impossible, to distinguish when an observer is reading the actor’s behavior versus reading his goals. But there are situations in which what an actor does does not match his goal, specifically, in unsuccessful attempts and accidents. The best evidence that an observer under- stands goals, therefore, is when she reacts specifically to the actor’s goal and not to his overt behavior when he is trying unsuccessfully or having an accident. first trials. Melis et al. [27] went one step further by having a booth with opaque barriers on both sides, but chimpanzees had to choose between reaching through a clear tunnel, in which case the human competitor could see their reaching arm, and an opaque tunnel, in which case she could not. Even though they could not see the human under any conditions, the chimpanzees reached more through the opaque tunnel than the clear tunnel – presumably imagining the perspective of the human. In an extension of this, Melis et al. [27] used this same booth arrangement but with two clear tunnels leading to food. Here the experimenter lowered her head between her knees so that she was unable to see the chimpanzee. In this case each tunnel had a small door midway through, one of which was very noisy and one of which was silent (they had learned about the doors’ properties earlier in a different context). If the human heard a door opening, she would raise her head and grab the food before the chimpanzee had a chance, but not if she could not hear a door opening. Chimpanzees in this case preferentially reached through the silent door, and again they did so from the very first trials. Figure I. Experimental setup for the studies of Hare et al. [26] and Melis et al. [27]. 188 Box 2. Chimpanzees infer a human’s intentions Buttelmann et al. [28] used the Gergely et al. [29] method to test six human-raised chimpanzees in the so-called rational-imitation para- digm. The chimpanzees were shown how to operate an apparatus to produce an interesting result (e.g. lights or sounds), and then they were given a turn. The most natural behavior for them in all cases was to operate it with their hands. But this obvious behavior was never demonstrated for them; they always saw a human manipulate the apparatus in a novel way with some other body part. The idea was that in some cases the physical constraints of the situation dictated that the human (referred to as ‘E’ in the figure) had to use that unusual body part; for example, he had to turn on a light with his head because his hands were occupied holding a blanket or he had to operate a light with his foot because his hands were occupied with a heavy bucket (see Figure I). When the chimpanzees saw this forced use of the unusual body part, they mostly discounted it and used their hands as they normally would (because the constraints were not present for them). However, when they saw the human use the unusual body part when there was no physical constraint dictating this, they quite often copied the unusual behavioral means themselves. If we interpret this experiment the way it is interpreted Trends in Cognitive Sciences Vol.12 No.5 Table 1 lists ten separate studies suggesting that chimpanzees understand others’ goals. At least six of these aremodeled closely on studies with human infants (see Box 2), which are routinely used as evidence of goal under- standing. These ten studies involve several different situ- ations that chimpanzees must understand and several different response measures as well. Thus, in the first seven studies in Table 1 (briefly described there, points 1–7), chimpanzees react not toward the actual behavior of the human but toward his goal, as they attempt to get a reward or anticipate his behavior. For example, they react differently to very similar behaviors when the human is refraining from giving food because he is unwilling versus unable to do so (Table 1, points 1 and 3), when he is doing something on purpose versus by accident (Table 1, point 2) or else when he is an innocent bystander (Table 1, point 7). They also discern that another’s goal is to reach an out-of- reach object and then they either help him to reach it (Table 1, point 4) or grab it first if it indicates the location of food (Table 1, point 5). And they anticipate a human’s for human infants, the conclusion is that the chimpanzees under- stood not only what the experimenter was trying to do (his goal) but also why he was doing it in the way he was doing it – the rationality behind the choice of the plan of action toward the goal. According to Tomasello et al. [30], an understanding of the action plan chosen toward a goal constitutes an understanding of the intention. Figure I. Turning on the light with the foot because (a) E wanted to or (b) E had to (because his hands were occupied) in the Buttelmann et al. [28] experiments. nd op imi bo ive stan impending actions when he stands and turns in a certain direction based on what goal he is probably pursuing (Table 1, point 6). Importantly, in two of these studies (Table 1, points 3 and 6) the behavior at the time the chimpanzee must react is identical in experimental and control conditions, with the only difference being in the immediately preceding context (e.g. given that the human has been opening boxes, the current ambiguous behavior is probably an attempt to open a new box, whereas without this preceding context the human’s goal is unknown). One could of course attempt to explain any one of these results in terms of behavioral rules that chimpanzees are either born with or learn, and this might be plausible in a particular case. But this explanatory strategy is not plaus- ible across all of the seven studies because of the diversity of situations and reactions required. Moreover, to explain the studies in which the human’s behavior is identical in experimental and control conditions at the time of reaction, one would need to also posit something like ‘contextual rules’ because there is no differential behavior to read in the two situations. Table 1. Studies on chimpanzees’ and human infants’ understa Studies Getting/finding food 1. Leave earlier and beg more intensely from an E who is unwilling as (behavior similar in the two cases) 2. Select the box acted on intentionally versus accidentally (behavior s 3. Leave earlier when E is playing with as opposed to trying to open a two cases) Reacting to a partner’s actions 4. Give the object that the E is trying to reach 5. Take the food that a competitor is trying to reach 6. Anticipate where E is going based on potential goals available 7. When food is stolen retaliate against thief, not against innocent rece Imitation 8. Produce target action based on observing a failed attempt 9. Copy intentional actions more often than accidental actions 10. Selectively copy freely chosen acts but not those forced by circum a(D. Buttelmann et al., personal communication). Gaps in the table indicate no information available. Review In three further studies in Table 1 (points 8–10) expla- nations in terms of behavioral or contextual rules are not possible. This is because these all use an imitation para- digm in which the chimpanzee subject actually acts out in her own behavior what she understands the other to be attempting to do, and this does not always correspond to the overt actions (note that all three of these studies use human-raised chimpanzees, who are very probably the only ones capable of copying actions in specific ways [13]). In these three studies (all modeled on similar studies with human infants), chimpanzees imitated what the human was trying to do (not what he did) (Table 1, point 8), his purposeful rather than his accidental actions (Table 1, point 9), and they even selectively imitated actions based on an understanding of why the actor chose this particular action (Table 1, point 10), which might be construed as understanding his intention, in the sense of the action plan he chose for pursuing his goal. Behavioral and contextual rules only help a subject to predict what an actor will do; they provide no guidance whatsoever for social learning. We believe that there is only one reasonable conclusion to be drawn from the ten studies reviewed here: chimpan- zees, like humans, understand the actions of others not just in terms of surface behaviors but also in terms of the underlying goals, and possibly intentions, involved. Beha- vioral or contextual rules might be concocted to explain the results of any one of the seven studies in which the chim- panzees react to or predict the behavior of others, but this requires many different ad hoc behavioral and contextual rules for which there is absolutely no positive evidence. Indeed consistent use of this explanatory strategy would also deny human children an understanding of goals and intentions because most of the chimpanzee studies are modeled on child studies. Moreover, the three imitation studies would not seem to be amenable to behavioral-rules explanations at all. Understanding perception and knowledge To understand how another works as a goal-directed agent, an observermust understand not only his goals but also his perceptions because what he sees and knows helps to ing of goals and intentions References Chimpanzees Infants posed to unable to deliver food [31] [32] lar in the two cases) [33] [33] x with food (behavior identical in the a [34,35] [34] [36] a r of stolen food [37] [38,39] [40] [38] [41] ces [28] [29] Trends in Cognitive Sciences Vol.12 No.5 determine what he does. Here, we examine what chimpan- zees understand about what an actor sees – not just what he is oriented to, but what he registers from the environ- ment in ways that affect his actions – and about what that actor knows in the sense of information he has previously registered that still affects his current actions (e.g. his knowledge of where food is even though he cannot see it now). Understanding false beliefs is the special case in which an observer predicts or explains the behavior of an actor based on a judgment of what that actor believes to be the case, not what really is the case as the observer knows it (e.g. the actor believes the food is in one place when the observer knows that it is really in another). Again in the current case, the main alternative hypoth- esis that we must consider is that chimpanzees either are born with or learn certain behavioral or contextual rules that determine how they respond to others’ surface beha- viors (orienting behaviors in this case) without any un- derstanding of their perception or knowledge. And analogous to the case with goals, the best evidence that an individual understands another’s perception is when 189 nd on indo hem ctio f b ood the two individuals’ perceptions differ (e.g. they see different sides of a barrier), and the observer acts on the basis not just of what she herself sees or knows but also on the basis of what the other sees or knows. Table 2 lists 16 different studies relevant to the question of whether chimpanzees understand what others see and know. The first set of studies (Table 2, points 1–7) involve gaze-following behaviors, which are, admittedly, not such powerful evidence of an understanding of perception. The key fact here is that chimpanzees follow the gaze direction of others behind themselves, around barriers and past distractors; they check back with the gazer if nothing interesting is to be found in the indicated direction, and Table 2. Studies on chimpanzees’ and human infants’ understa Studies Gaze following 1. Follow gaze to distant locations behind self 2. Follow gaze on the basis of both face and eye direction 3. Check back with gazer if nothing relevant at the target location 4. Stop looking after a few trials if nothing relevant at the target locati 5. Ignore distracting objects on the way to the target location 6. Move to the side of opaque barriers to view the target location 7. Understand that gaze stops at an opaque barrier - unless it has a w Gestural communication 8. Use visual gestures mostly when conspecifics or E are oriented to t 9. Position oneself to gesture in front of others 10. Both face and eye orientation of recipient determine gesture produ Food competition 11. Pick the food that the E is not looking at 12. Pick the food that a dominant individual or E cannot see because o 13. Visually conceal approach to food (using barrier) 14. Auditorially conceal approach to food (choosing silent door) 15. Take food that a dominant individual did not see being hidden 16. Understand that if competitor picks first, he will have chosen the f being hidden a(J. Kaminski et al., personal communication). Gaps in the table indicate no information available. Review they eventually stop following if an individual always gazes at nothing interesting. These are all characteristics of gaze following shared with human infants, and even though human infants use the eyes, as opposed to the head, more often than chimpanzees do in following gaze direc- tion, chimpanzees do use the eyes to some extent as well (Table 2, point 5). The second set of studies involves gestural communi- cation. The general finding is that chimpanzees take into account the visual orientation of the recipient when ges- turing (Table 2, point 8), even moving themselves in front of the recipient to gesture when necessary (Table 2, point 9). Of particular importance, Kaminski et al. [14] found chimpanzees muchmore sensitive to the recipient than did Povinelli and Eddy [6] in a similar paradigm when the human was actually in a position to deliver food (i.e. facing them bodily). It is true that chimpanzees are not as sensi- tive to the eyes as are human infants, but still they under- stand when someone is or is not in a position to receive their communicative act (Table 2, point 10) – again in a manner very similar to human infants. By far the most power