(APA Eastern Div. Symposium on Children's Theory of Mind,
12/27/95, as delivered)
Robert M. Gordon
Philosophers have been debating the nature of folk or commonsense psychology for three decades. We ask: What are the resources that enable us to navigate the social world, anticipating what others do, explaining what they’ve done, and perceiving them--and ourselves--as selves, subjects, persons, with beliefs, desire, perceptions, and feelings? Unlike traditional philosophy of mind, instead of directly confronting the mind-body problem and subproblems such as intentionality and qualia, we step back and look at the resources that give us the concepts that get us into these knots.
Recently this philosopher’s debate has linked up with a debate that’s been going on in developmental psychology about the way folk psychology develops in young children. Much of the discussion truly cuts across disciplinary boundaries. Most of us involved in the interdisciplinary debate think you’re not in a position to understand the nature of folk psychology[RMG1] unless you look at the developmental evidence, and you can’t adequately interpret this evidence without reference to philosophical theories of mind.
I’m honored to join a symposium with one of the leading developmental psychologists in the debate. But I shouldn’t be the only philosopher up here. To mention 3 others: Martin Davies, Alvin Goldman, and Steve Stich. I must say something about the role Stich has played. One might think he was just Defender of the Status Quo. He’s also been the Socratic Midwife: Long before his article with Shaun Nichols, he was the first to see simulation as a serious challenge to widely accepted assumptions, including some of his own. Yet he assisted generously in developing and promoting the challenge.[RMG2]
The title of this talk mentions philosophical data. This needs an explanation. Questions about the conceptual development of children are matters to be settled, it would seem, by empirical research. Yet in the recent debate philosophy has played a remarkably active role. For one thing, the major theories in contention all descend from philosophical accounts of the nature of mature folk psychology. This shouldn’t be surprising, because what children are developing is after all our mature folk psychology. But there is a further philosophical involvement that may be more surprising. Much of the debate centers on features of mind that fall within the province of traditional philosophy of mind: intentionality, referential opacity, and the peculiarities of self-awareness and self-ascription—features that set mental state concepts off from nonmental concepts, and thus constitute the traditional barriers to resolution of the mind-body problem. I can think of two good reasons for this. Precisely because these features set mental state concepts off from nonmental concepts, they provide good litmus tests of concept acquisition. A child who doesn’t understand that a belief may be contrary to fact isn’t yet capable of genuine belief ascription. Likewise, a child whose inferences from belief ascriptions violate referential opacity.[RMG3] Thus when it comes to discriminating genuine mental state ascriptions from cheap imitations, philosophy serves in the role of expert witness. The distinctive properties of mind provide good tests, not only of genuine mental state ascription, but also of an adequate developmental theory. Such a they should explain how children develop a working mastery of peculiarities of mind such as intentionality and the asymmetries of first person and third person. Philosophy serves as expert witness on what these peculiarities are as well as on what counts as mastering them.
I’ll present a brief capsule of a debate that has many ramifications. I will restrict the topic to children’s understanding of belief. This isn’t a severe restriction, because belief has been the central topic for over a decade. In a 1978 paper Premack and Woodruff reported experimental results suggesting that chimpanzees attribute goals or purposes on the basis of their observations of another's behavior. In independent commentaries, Jonathan Bennett, Daniel Dennett, and Gilbert Harman each proposed what they thought would be a more persuasive test of mental state attributions. If chimps failed to appreciate that agents sometimes act on the basis of a false belief, then under certain conditions they would make tell-tale errors in anticipating the actions of others. Why not set up such conditions and see whether they do make such errors?[RMG4]
This general idea, as well as some specific suggestions, was soon picked up by two developmental psychologists at the University of Salzburg, Heinz Wimmer and Josef Perner. Their subjects were, not chimpanzees, but human children of ages 3 through 5 years.[RMG5] The subject hears a story, illustrated with puppets, about Maxi, a popular Austrian cartoon character. Maxi puts his chocolate away in a box on the puppet stage (I’ll call it Location A) and goes out to play. While he is out, his mother transfers the chocolate to the cupboard (Location B.). The subject is then asked where Maxi look for his chocolate when he comes back. Notice that the experiment--this part of it, at least--doesn’t rely on a child’s ability to use the word believe or its equivalent. It asks only for a prediction of behavior.
The child knows something that Maxi presumably wouldn’t know, because he was outside at the time: that the chocolate moved from A to B. As a consequence of Maxi’s ignorance or failure to know, we adults would attribute to him a false belief as to the present location of the chocolate. So we’d predict that, acting on that belief, he will look for it in the place where he had left it, Location A, even though the chocolate is now at Location B.
What about children? By the time they’re about 4 years old they make the same prediction. Three-year-olds, on the other hand, nearly always get it wrong: they indicate Location B. The Wimmer and Perner experiment has been replicated many times, with a number of variations, and the results have held up very well. Evidently children youmger than about 4 fail to adjust for the fact that Maxi wasn’t in a position to know that the object was relocated. They treat all the facts presented to them in the story as accessible to Maxi as a basis for action. It’s as if they were attributing knowledge, but haven’t yet learned to attribute mere belief. I say “as if” because they seem to lack an understanding of the ways in which one can fail to have knowledge: so they are not making a genuine attribution of knowledge.
What’s going on in 3-year-olds that accounts for their getting it wrong? I’ll limit my discussion to 3 representative answers. I’m sure Alison will help fill in the picture. First I’ll focus on Perner’s account, in his 1991 book, Understanding the Representational Mind. Then I‘ll present an alternative, the Simulation Theory. Later, when I ask how children with autism fare in such an experiment, I’ll briefly discuss a third approach.[RMG6]
Perner theorizes that 3-year olds have a psychological understanding that embraces desire but not belief. For them some aspect of the actual situation or state of affairs performs the role of belief. Thus actions are caused by desires in conjunction with some aspects of the actual situation. Perner calls this the “Situation Theory.” Perner adds that 3-year-olds can entertain alternative models of the world: They can pretend something they know to be false, and they can attribute a desire for a non-existent state of affairs. What they lack, Perner says, is a conception of a model or representation. Four-year-olds pass the false belief test, he argues, because they have the idea that people act on the basis of mental models, and that these may be true or false. This he calls the “Representation Theory.”[RMG7]
Perner goes on to say that the 4 year old’s theory doesn’t really replace the 3-year-old’s situation theory; it only amends it to cover problem cases.[1] Even as adults, he says, “we stay situation theorists at heart. We resort to a representational theory only when we need to.”[2] The 3-year-old’s conception, which does not cite the agent’s beliefs, is all we usually call upon, because it is typically all we need. This comes to saying that explaining and predicting actions in terms of actual situations or facts is our default mode of explanation and prediction.
I agree. This is reflected in our folk psychological idiom.
Consider the forms of speech we normally use to explain the actions of others
or to specify what their emotions are about.
Why is Mary running? “Because
she’s late.” What’s Tom upset
about? “Mary’s being late.” For most
emotions, sentences that say what someone is or was emoting about— excited
about, upset about, embarrassed about, and so forth—ordinarily cite what is explicitly
or implicitly a fact: Tom was upset about Mary’s being late/ upset about the
fact that Mary was late/ upset that Mary was late. Suppose he was wrong, she
wasn’t late. Then we’re forced to use a circumlocution: “He was upset because
he thought she was late.” With regret we don’t even have the circumlocution,
“She regretted because she believed she was late.” So we resort to something
even more circum, such as, “She felt regret because she thought she was late.”
In such cases there is really no direct answer to the questions, “What did Mary
regret?” and “What was Tom upset about?” because there is nothing in the actual
world that fills the bill. Likewise, when specifying someone’s reason for
acting, we standardly specify what it is about the world that is her reason:
“She did it because ....” where the “because” is followed by an embedded
sentence, such as, “She’s late,” and the presupposition is that the sentence is
true. This form of explanation, which picks out a fact rather than something
merely believed by the agent to be a fact, is the default form. We deviate from
it only when we have a reason for doing so. If we do mention the agent’s
belief, there is a Gricean implicature to the effect that there is a reason for
avoiding the default form.
If Perner is right, when we adults shift our focus from the actual facts to what the agent believes to be the facts, we recapitulate in our everyday thinking the development from a 3-year-old to a 4-year-old understanding of people. But we are happier being 3-year-olds. Explanations in terms of beliefs are second-class, faute de mieux explanations within folk psychology. Therefore it’s inaccurate to characterize folk psychology as “belief-desire psychology.” “Knowledge-desire psychology” would be a step closer to the truth.
The simulation theory (ST) would agree with much of what Perner says.To decide where Maxi will look for his chocolate, you play the role of Maxi. Thus you decide where Maxi will look by deciding where to look. In the default mode of simulation, you make no adjustments, other than in the reference of indexical terms such as “I” and “here.” The world that furnishes the basis for your decision in the role of Maxi is the simply the world--the one that furnishes the basis for your own decision-making. It’s as if you were implicitly attributing to Maxi full knowledge of that world. In effect you are projecting your own beliefs (aside from essentially indexical beliefs) onto Maxi. Yet you are not using the concept of belief.
Where I believe Perner is wrong is in his account of what 4-year-olds and adults have that enables them to pass the false belief test: namely, a theory that the information on which people’s actions and emotions are based comes from mental models that represent the world. [3] He thinks the experimental results show that 4-year-olds move from thoughts about the actual world—say, about where the chocolate actually is—to thoughts about these models. Because they think Maxi’s model falsely represents his chocolate as being at Location A, we expect him, acting on the basis of that representation, to look for his chocolate at Location A.[4] This of course leaves open many questions. Why do they think Maxi’s model is false? Perhaps because they accept a generalization that people outside a house can't see what is going on inside, and a generalization that what one can't see one isn't aware of, unless by other means. And why would Maxi look for his chocolate at Location A? Perhaps because a desire to find x and a belief that x is at location l typically causes the behavior of going to location l.
What does the Simulation Theory, or for short ST, have to say? According to ST, to decide where Maxi will look for his chocolate, I don’t move from thinking about about where the chocolate is to thinking about representational intermediaries inside Maxi. I continue to think about the situation, the location of the chocolate. But I change the way I think about it. I locate myself mentally with Maxi. Or as we say, I “put myself in his place.” One of the consequences is that, when Maxi is out of doors, I mentally place myself out of doors. I might imagine a stereotypical wall that occludes everything indoors--or I might not even have to.[RMG8] In any case, I block out any new information about what goes on indoors—including a crucial part of the story that I have been privy to, the moving of the chocolate from A to B. Now, when I am asked, “Where will Maxi look?” I will answer in Maxi's place. I will decide where to look—but without benefit of the crucial information about the move from A to B.[RMG9]
According to ST, then, even when dealing with problem cases, we go on explaining or predicting in terms of facts, but at the 4-year-old stage we can mentally relocate ourselves and adjust the facts accordingly. We also flag the adjusted facts: they are recognized, not as actual facts, but as facts embedded within an alternative point of view: they are facts to Maxi. This isn't to say they are represented by a model inside Maxi: the mechanics, whatever they may be, aren’t what folk psychology is about. Point of view talk would be unilluminating to people who are not capable of transporting themselves to alternative viewpoints and altering the facts within that context. As I’ll note shortly, children with autism appear to lack at least one of these capacities.[RMG10]
With ST in mind, consider a recent variant of the Maxi experiment. In the original experiment the child subject is explicitly asked where Maxi will look for his chocolate., This time Wendy Clements and Josef Perner used a video camera to monitor the child’s nonverbal response to the story. Here the story is that Sam the mouse places his cheese in the blue box (Box A). While he is asleep another mouse moves it to the red box (Box B). Sam awakens and announces, “I feel very hungry now. I’ll go and get the cheese.”[5] The camera reveals something interesting at this point: Without explicitly being asked where Sam will look, children are very likely to look at one of the two boxes. But which one? Four-year-olds, as might be expected, look at Box A, even though they know the cheese isn't in that box. What about 3-year olds? The amazing finding is that 80% of them looked at Box A, too. But when asked where Sam will look, most of them indicate Box B, the classic 3-year-old response.
About 50% of the 3-year olds fix their eyes on the right box and then give the wrong answer. Why? Perner, in a new paper,[6] presents the following explanation as “an interesting possibility”: 3-year olds look at the right box because they already tend to identify empathically with story protagonists such as Sam. “In parallel with the real story events,” Perner writes, “they keep track of how the world looks from Sam's perspective: the cheese is still in the old location.” But they give the wrong verbal answer because mere empathic identification falls short of simulation. Simulation, Perner argues, requires pulling back from Sam’s world and labeling it: “That’s the way it is to Sam.” Because the children aren’t yet pulling back and labelling, they don't correctly predict what Sam will do.
It’s a fascinating interpretation. But has Perner changed his mind and accepted ST? Not quite. For one thing, he presents this account not as the best explanation but only as an interesting possibility that awaits further empirical testing. For another, it’s recently emerged that there are two significantly different versions of ST, as I’ll explain in a moment, and Perner continues to be critical of one of them. He does now make the important concession that what’s crucial for passing the false belief test is the capacity to appreciate that there are diverse points of view with regard to the facts—points of view that one can simulate—rather than the ability to conceive of a representational mechanism that makes this diversity physically possible. But he thinks there are some data that can be explained only if we impute to children a theory as well as a capacity to simulate. Thus his present view is a hybrid simulation-plus-theory theory. Significantly, though, he holds simulation to be the basic bedrock on which any theoretical elements must stand. One reason is that any theorizing must build on the 3-year-old’s practice of explaining and predicting in terms of actual facts or situations. Because these explanations and predictions presuppose that the other is aware of the relevant facts or situations, we are in effect projecting onto the other our own beliefs and ways of “entifying” and categorizing the world.[7] We are simulating in default mode, that is, without making any adjustments.
The version of ST that I favor holds such projection to be the starting point for simulation: the child’s conception of the non-mental world gets projected onto others by default. To be sure, this non-mental world is an emotionally and motivationally charged world: the so-called tertiary qualities, the frightening quality of the thunderstorm and the soothing quality of a parental embrace, haven’t been purged from the world and relegated to mere appearance. So in a sense the mental is already “out there” in the environment, though not yet conceptualized as mental.
The other version of ST takes the starting point for simulation to be our knowledge of our own mental states and processes. It assumes that we have the concepts of various mental states and the ability to recognize instantiations of these states in ourselves, say, by their qualitative character. On the assumption that others are similar in relevant respects, we then have a basis for inferring the mental states of others.[8] Simulation is understood as an imaginative procedure founded on the old argument from analogy. Alvin Goldman and Paul Harris have promoted such a view in several of their writings, and I’ve argued against it.
Following Harris, developmental psychologists supposed ST to be an essentially Cartesian theory. Criticisms were mounted by Janet Astington, Alison Gopnik, and Henry Wellman, as well as Perner. Along with other problems, the epistemological priority assigned to self-knowledge seemed to imply a chronological priority. As Astington and Gopnik say, “a central prediction of such a view would be that children's understanding of their own minds, of their own beliefs, desires, and so on, would consistently precede their understanding of the minds of others.” But the evidence, they argue, doesn’t bear out this prediction. An important series of experiments by Astington and Gopnik revealed that young children have problems understanding and recognizing false belief not only in others, but also in themselves. Suppose you show a child a box of Smarties (a well-known Canadian and British candy). Ask what it contains and the child will probably tell you, “Smarties.” Now open the box and show that it contains nothing but pencils. Ask, “What did you first think was in the box?” Four-year-olds get it right: “Smarties.” But 3-year-olds will tell you, “Pencils.” Even in their own case, they don't attribute false beliefs. So how can first person ascription of false beliefs be the basis for third person ascription?
Paul Harris responds that these experiments are irrelevant, because they concern only past beliefs, not current ones.[9] For on his view it is the awareness of one’s own current mental states that is the basis for attribution of mental states to others. But reporting our own current beliefs doesn’t require countenancing the possibility of false belief. We never run into a current belief that we currently believe false.[10] Indeed, we don't need the concept of belief in order to report our beliefs fairly reliably. If I’m asked, “Where do you believe the chocolate is now?” I simply say where the chocolate is. That is, I answer questions about my own current beliefs just by answering questions about the facts—within certain constraints, such as limits on processing time and on the procedures employed to get the answer. In the first person case, then, there is a split. Getting first person reports right is the easy part; understanding them is the hard part.[11]
Harris is wrong in thinking that the basis on which children attribute beliefs to others is their awareness of their own beliefs as such, quâ beliefs. They won't have this awareness until they are ready to attribute false beliefs, and this will arise in regard to others, or to themselves at another time. Genuine present-tense self-ascriptions of belief are likely to come last of all. For understanding these requires conceptually prising one’s own present beliefs apart from the facts, so that, like the beliefs of another, they may be false or at variance with the facts. More or less in sympathy with George Herbert Mead, I think it’s through role-taking that I come to see myself as an individual mind. And what I now conceptualize as the world gets reconceptualized as the world from a point of view. I discuss this in my article, “Simulation Without Introspection or Inference from me to you,” in the Blackwell Mental Simulation volume.
Children afflicted with autism often give the impression of treating people as objects. It would be a reasonable guess that such children would have trouble understanding that people sometimes act on the basis of a false belief. Simon Baron-Cohen, working with Alan Leslie & Uta Frith, tested this hypothesis. They used a false belief test based on Wimmer and Perner with three groups of children: children diagnosed with autism, children with Down's syndrome, and clinically normal children. The normal children were 4 or 5 years old; each of the other two groups averaged between 10 and 11. Nonverbal tests showed the average mental age of the Down’s syndrome group to be 5. People with autism are usually retarded, but this was a high-functioning group whose performance on nonverbal tests was on average that of a 9-year-old, fairly close to their chronological age.[12] The results? Nearly all the children with Down’s syndrome passed the test, as did the normal 4 to 5 year olds. 80% of those with autism failed; they gave the 3-year-old response. There have been some replications, and children with autism have been found to fail other tests that require an understanding of false belief.[13] They also fail Astington and Gopnik’s Smartie’s test: asked what they first thought was in the box, they use their current knowledge of its contents. If performance on false belief tests were mainly a matter of general intelligence, one would expect the autism group, with an average nonverbal mental age almost double that of the other groups, to do as well as the others or better. The experimenters concluded that autism involves a specific deficit in this domain of understanding.
One of the other abnormalities characteristic of autism is the failure to engage in spontaneous pretend play, whether in conjunction with others or alone. Role play and joint pretend play are completely absent. Perhaps it isn't just a coincidence that the same children do poorly on false belief tests. It clearly isn't a coincidence if ST is right. Mentally relocating ourselves and identifying with Maxi is a kind of role play. It would call on many of the same imaginative abilities used in overt pretend play. If these abilities are necessary for passing the Maxi test, then someone who is incapable of such exercises of imagination would not pass it.
There is another theory that suggests that understanding and attributing false belief calls on abilities that are also needed for pretending. Alan Leslie, one of Baron-Cohen’s mentors and a coauthor of his ground-breaking 1985 paper, noted that in pretend play we are permitted to portray the world in contrary-to-fact ways, referring to things that don’t exist and making false predications of things that do exist. We are also allowed such license, he argues, when we refer to the content of anothers’ beliefs: for example, Maxi’s belief may have the content, “The chocolate is at Location A,” when in fact the chocolate is at Location B. The same is true of the content of other propositional attitudes, too, according to Leslie: the content of wishes and desires, for example, is typically counterfactual.
Leslie doesn’t regard the abilities common to pretending and belief attribution as imaginative abilities. He portrays them as representational abilities. Like many cognitive scientists he conceives the brain as a system that processes information in the form of sentence-like representations in a Language of Thought. His concern is that such a system would have to be modified in certain ways if it is to allow the representational abilities needed for either pretending or the ascription of propositional attitudes. It requires a special mechanism, Leslie argues, which creates and manipulates a special set of representations, which he sometimes characterizes as “metarepresentations.” The details of Leslie’s account are too complicated for a brief summary and also, to my mind, puzzling and obscure[RMG11]. So I’ll focus on what I consider an important virtue of Leslie’s account: It recognizes that the sort of license we’re allowed in pretending and in specifying the content of another's belief is potentially dangerous: unless these operations are somehow cordoned off or quarantined, they are liable to corrupt our own everyday thinking. Simulation Theorists recognize this, too. They speculate that the quarantining is done by a special mechanism that allows various mental functions to operate in partial isolation from their normal input and output systems: “offline” was the word Steve Stich suggested to me 10 years ago. Such a mechanism would be needed not only in pretending and the attribution of mental states but also in counterfactual reasoning, action planning, ethical evaluation, decision-making, and a number of other mental operations. Even within a Language of Thought framework, it is hard to see why one couldn’t get everything Leslie wants with, not a separate representational system, but a single system used twice, once on-line and once partially off-line. In fact, there is no reason why a Language of Thought system couldn’t represent and reason about the mental lives of others chiefly or even exclusively by simulation. A model of a simple Language of Thought system that does just that is ProtoThinker, an AI program written by a philosopher, John Barker.
Whatever the major resource that underlies folk psychology--whether something resembling a scientific theory, a special Language of Thought Mechanism, the capacity for simulation, or something else--it’s become increasingly clear that this resource doesn’t work alone. It is assisted by a number of special-purpose mechanisms. In his recent book Mindblindness, Simon Baron-Cohen posits 3 such mechanisms. One of these automatically interprets anything that looks like self-propelled motion as purposive approach or avoidance. Another detects eyes and the direction in which they are gazing. A third mechanism gets us to turn our eyes toward the target of the other's gaze, enabling us to identify what the other is looking at. Even 1-year-olds will then glance back and forth between the other’s eyes and the target, as if to confirm that they are looking at the same thing.[14] This is obviously important in ostension and language acquisition. There is behavioral and neurophysiological evidence suggesting that human and other primate brains are endowed with such special-purpose mechanisms.
Normally,
children develop a certain bag of tricks, which, in turn, together with
experience, predisposes them to develop this conception of a common shared
world. Despite the very likely influence of experience, however, it is
implausible to suggest that they form this conception because they find it
justified. I do not think they have justification, and I definitely do not
think it is justification that leads them to hold this conception. One might
want to argue that through the agency of natural selection they form this
conception because it is right. I take no stand on this.
But to speak of justification here presupposes that
they already have a conception of a multiplicity of selves that may or may not
have access to a common world: that may be windowless monads, for instance. And this is evidently not the
case. It is from the common world that we begin to pull apart separate selves
with an inner life.
There are
specific
reliability issues: for example, isn't this a case in which adjustment
must be made for O’s poor vision, for O’s not looking in the right direction,
etc. But the only general reliability issue for simulation is whether the very idea of a common world is a sound one, this
hard-won achievement of normal children and probably not those with autism:
that you and I both see the same things, and see that the same things
are so, etc--the presupposition underlying, e.g., my
attempt to find in the environment the cause of my fellow hiker’s behavior.
If one accepts this, then one accepts the general reliability of
simulation.
In searching the environment for an explanation of your friend’s action, you are projecting your own beliefs about the environment onto your friend. The field of search consists in a subset of those features you yourself believe (after looking around) to comprise the actual world. And the objects within that field are fixed by your own ways of “entifying” and categorizing the world (Goldman 1989: 179-80.)
My only criticism is that Baron-Cohen fails to notice how much these mechanisms must be complemented by projection. For example, just what objects our eyes will fix on as a result of following another's gaze will usually depend on what is salient along the other’s line of sight. Ordinarily this would depend on our own interests, attitudes, and emotional dispositions--unless we make adjustments, as in viewing the environment through the eyes of a child or a foreigner. Likewise, how we conceptualize what we see will depend on the concepts, categories, and background information we bring to the job--again, unless we make adjustments. Such projection is the starting point, the default mode, for simulation; and, as Perner says, for any theorizing as well.
Another set of mechanisms, posited by Andrew Meltzoff and Alison Gopnik, may be involved in picking up the emotions of others from their facial expressions--although I think they would grant that their hypothesis is more speculative. It’s well established that our facial muscles automatically mimic the expressions of others--even from birth, as Meltzoff established. And there is some evidence that adopting certain facial expressions tends to make us feel the corresponding emotions. Working together, these two mechanisms might cause us to internalize another's emotion, and thereby put us in position to recognize it. These mechanisms alone would not tell us what the emotion is about. But if you throw in Baron-Cohen’s mechanisms, our attention will automatically be shifted from the other’s face to what the other is looking at, and this is likely to be what the emotion is about.
These are almost certainly not the only special purpose mechanisms that are relevant to folk psychology. From our early years on, we view members of our own species with an arsenal of specially attuned responses: to the sounds of human speech and crying, to the sight of faces, facial expressions, and bodily postures and movements. As a result, we never get the opportunity to view human faces and bodies dispassionately, as we might other complex objects. People get under our skin, they get to us, in ways that nothing else does. We find ourselves both mimicking them in superficial ways--in direction of gaze, facial expression and muscular set--and mirroring them in deeper ways, in what we see and feel.
I said at the beginning that in this folk psychology debate we look at the resources that give us the concepts that get us into the traditional conceptual knots about mind-body. My personal belief is that the more naturalistically we approach folk psychology, with due attention to the mechanisms that cause us to resonate in these special ways to human-like faces and bodies, the less plausible the prospect of a naturalistic reduction of the mental will appear. A closer look at such mechanisms will show us why practical and intellectual difficulties are inevitable when we try to view anything simultaneously as a person and as a mechanism, or, more broadly, as an object.
Perner compares this to extending an older
theory, giving as an example the extension of classical genetics by molecular
genetics, as described by Philip Kitcher (1984, 370)
[3] Perner thought internal
representations were the way to capture the intentionality of propositional
attitudes. He acknowledges influence by philosophers such as Hartry Field and
Jerry Fodor. But Field and Fodor were concerned, as I understand it, not with
the accurate depiction of commonsense conceptions, but with the possibility of
a naturalistic semantics for the propositional attitudes.
[4] Notice that in order to pick
out the location where Maxi will look—point to it, for example—older children
will have to interpret Maxi’s model of the world. And they will have to give it
a de re interpretation. They will have to locate in the actual world the
chocolate as it is represented in Maxi’s model of the world. She will have to
think, “To Maxi, this (with an ostension to the chocolate at Location B) is
over there (with an ostension to location A).” They might do this by imagining
at Location A the chocolate that she knows to be at Location B. This imagining
need not involve visual imagery; it might just involve the thought, “The
chocolate is there,” where “there” is identified on a mental map of the
environment. It is the interpretation of Maxi’s variant model, or more
precisely, Maxi’s variant cognitive map, that would allow children to indicate
the location where Maxi look for his chocolate. What makes this an account of
where the chocolate is to Maxi is the fact that I understand myself to be
interpreting a model inside Maxi, a model that Maxi’s behavior uses for its
guidance. This comes close to what I say, except that it preserves the
requirement that children think in terms of a model, a requirement Perner
insisted on because he thought it was the only plausible grounding for the idea
that beliefs may be true or false, and may have certain other intentional properties. But just relativizing to Maxi
can do the same.
[5] In some trials, The
experimenter prompts the child: “I wonder where he’s going to look?” But
(personal communication from Perner) the effect is the same with or without the
prompt.
[6] Josef Perner, “Simulation as
Explicitation of Predication-Implicit Knowledge: Reassessing its Value for
Explaining the Development of Mental State Attributions,” in Chapter for P.
Carruthers & P. K. Smith (Eds.),
Theories of Theories of Mind. Cambridge: Cambridge University Press,
1996.
[7] Perner is also persuaded by
Jane Heal’s argument that to predict what a chemist will come to believe about
some chemical matter when given certain evidence, we don't use a sub-theory of
chemical beliefs; we don't import all of our beliefs about the subject matter
into a theory of mind. Rather we just
use our own beliefs about the subject matter.
[8] See also J.S. Mill: "I am conscious in myself of a series of
facts connected by an uniform sequence, of which the beginning is modifications
of my body, the middle is feelings, the end is outward demeanour. In the case of other human beings I have the
evidence of my senses for the first and last links of the series, but not for
the intermediate link....by supposing the link to be of the same nature as in
the case of which I have experience,...I bring other human beings, as
phenomena, under the same generalizations which I know by experience to be the
true theory of my own existence." An Examination of Sir William Hamilton's
Philosophy. 6th edition. London,
1869.
[9] Harris writes: “The
simulation account, far from being embarrassed by the difficulties that
three-year-olds have in acknowledging an earlier false belief, explains them in
exactly the same way as it explains three-year-olds’ difficulties in working
out another's false belief.” He thinks the two cases are on a par because both
require “an imaginative construction of a counterfactual situation.”
[10] Nor for that matter do we
ever encounter any proposition that we recognize as true and yet fail to
believe.
[11]In other work I
claim that the same is true of other mental states, including desire and even
pain. By this I mean both understanding them to be reports
about oneself, and understanding what one is reporting about
oneself.
[12] The experiment, I should
mention, was part of Simon Baron-Cohen’s doctoral work at the University of
London.
[13]It should be noted, however,
that other findings are not as consistently supportive as in the case of Wimmer
& Perner 1983.
[14] (Vervet monkeys do this,
too, for example, when seeking a friend’s help in warding off an aggressor.)
Around the same age children also begin to engage in protodeclarative pointing,
directing another's attention by pointing and checking the other's eyes to make
sure they are looking in the right direction.
[RMG1]commonsense thinking about the mind
[RMG2]***Here’s a little
background. Standard accounts of folk psychology deemed it a theory, comparable
to a scientific theory, or an innate information base similar to a Chomskian
universal grammar. These views originated with philosophers and were taken up by
developmental psychologists. What brought philosophers and developmental
psychologists into a common debate was the challenge posed to these standard
views by the Simulation Theory. This holds that our chief resource for
navigating the social world is our capacity to simulate others: roughly, to
project onto them our own mental states, using imagination to make adjustments
when these are warranted by the evidence.
***An article of mine in the journal Mind and Language had put forward the simulation approach in 1986, and also suggested that it seemed to explain the results of certain experiments in developmental psychology. Alvin Goldman wrote a sympathetic essay that developed the view further. Paul Harris, a developmental psychologist at Oxford, made the theory a central thesis of his book, Children and Emotion. The debate began with a special issue of Mind and Language in 1992, under the editorship of Martin Davies. In an important article in that issue, Steve Stich and Shaun Nichols both touted the view as a major challenge and criticized it on a number of grounds. Other philosophers wrote articles pro and con, and a roughly equal number of developmental psychologists set out the results of relevant experiments, some of them designed explicitly to test the theory. The contents of that issue, along with a lot of other material, appear in two collections edited by Davies and Tony Stone that were recently published by Blackwell.
[RMG4]Harman’s suggestion was the most explicit: “Suppose that a subject chimpanzee sees a second chimpanzee watch a banana being placed into one of two opaque pots. The second chimpanzee is then distracted while the banana is removed from the first pot and placed in the second. If the subject chimpanzee expects the second to reach into the pot which originally contained the banana, that would seem to show it has a conception of mere belief.”[1] That is, mere belief as opposed to knowledge.
[RMG5]In place of Harman’s second chimpanzee they substituted a second child, represented by a puppet. (The child was called Maxi, after a popular Austrian cartoon character.) For the banana they substituted a piece of chocolate.
[RMG6]I should say that Alison’s view is closer to Perner’s than to the others, though there are important differences. Along with Henry Wellman, they defend one or another version of a view familiar to philosophers, the so-called “theory” theory, which holds that folk psychology resembles a scientific theory in important respects. Leslie’s view is in the Language of Thought tradition; it portrays folk psychology as chiefly the product of an innate module akin to a universal grammar. The simulation theory holds that we represent other human beings, and explain and predict their behavior, by simulating them: using our own cognitive, motivational, emotional resources, we project onto them our own mental states, and using imagination we make adjustments when these are warranted by situational or behavioral evidence.
[RMG7]DELETED
FROM PRESENTATION:
Many
philosophers would find fault with Perner’s account of 3-year-old folk
psychology. Some would object that one could not have a concept of desire
without a concept of belief. Responding to a similar suggestion regarding
chimpanzees, Harman wrote: “This cannot be right. . . . A chimpanzee can have a
conception of desire or purpose or goal only if it also has a conception of
knowledge or belief.” Bennett too thought it “impossible”: “An animal’s
actions ... show what it wants only on assumptions about what it thinks.”
A
second objection, raised by Colin McGinn, is that genuine communication, with
speaker meaning and its pickup by an audience, require that participants be
capable of attributing complex beliefs and intentions to others. Assuming that
young children do engage in genuine communication, they must be making such
attributions. (Yesterday a paper by Mark Risjord did a modus tollens on this:
Because research shows that young children do not attribute beliefs,
such attributions aren’t necessary conditions for meaning.)
To
this Perner replies that
Infants can be competent
communicators because normal communication is characterized by transparency [or
“openness”] ... The infant operates with this openness by default, ignoring the
underlying intentions and beliefs. As the young child starts to understand
these intentions and beliefs the reasons for breakdown in social interaction
become understood . . .
One
reason young children get by without such sophistication is that their peers
lack it as well.
Perner
goes on to claim that the 4 year old’s theory doesn’t really replace the old
theory; it only amends it to cover problem cases.[1]
Even as adults, he says, “we stay situation theorists at heart. We
resort to a representational theory only when we need to.” This is why
communication, and the solution of coordination problems in general, are
possible. Each would seem to require not just knowing the other's intentions, but
mutual knowledge, an infinite series of knowings. “Yet in normal life,” Perner
suspects, “we do not bother about these problems at all, because we are
situation theorists.”
As
I would put it, our adult sophistication in attributing beliefs and intentions
becomes manifest, not in our understanding of normal open communication—which
to us appears to us as the simple condition—but only when there is a breach of
openness and mutual knowledge. Unlike young children, we are sensitive to the
complex ways in which communication can fall short of the norm, and we’re
capable of exploiting these ways.
I
would say the same of knowledge in general. Our adult
epistemological sophistication becomes manifest, not in our ability to
recognize that Jack knows it’s snowing, but in our sensitivity to the complex
ways in which he can fail to know. Consider a simple tale of perception,
emotion, and action:
Jack saw that it was snowing.
Jack was glad that it was snowing. Jack went out because it was snowing.
This
simple story presupposes that in fact it was snowing, and also that Jack knew
it was. Yet we don't have
Knowledge
is supposed to entail belief. 3-year-olds do not seem to have the concept of
belief. Yet they seem to understand such a story.
3-year-old
can understand this. And for adults to understand it, they need only call on a
3-year-old’s conceptual equipment.
I
used the standard or default form of narration. To avoid the presupposition of
factivity or of knowledge, I would tell a complex story:
To Jack it looked as if it
were snowing. Jack was glad because he believed that it was snowing. Jack went
out because he believed that it was snowing.
I
wouldn’t tell this complex story unless there were some reason to avoid some
commitment implicit in the simple story. Notice, by the way, that the
convoluted story doesn’t just remove the presuppositions of factivity and
knowledge, in some ways it even does violence to the sense of the simple story.
Consider the shift from, “Jack is glad that it's snowing,” to, “Jack is glad because he believes that it's
snowing.” Strictly speaking, we are no longer saying what Jack is glad about.
Instead of assigning semantic content to Jack’s gladness, we assign content
only to a belief that is a cause of Jack's gladness.
Many
philosophers would find fault with Perner’s view that 3-year-olds have a
concept of desire without a concept of belief. Responding to a similar
suggestion regarding chimpanzees, Harman wrote: “This cannot be right. . . . A
chimpanzee can have a conception of desire or purpose or goal only if it also
has a conception of knowledge or belief.” Bennett too thought it “impossible.”
“An animal’s actions ... show what it wants only on assumptions about what it
thinks.”[1]But I believe
Perner is right about this, and the philosophers wrong. The conceptual and
epistemological roles of beliefs can just as well be played by the actual
world, the facts. Indeed, better.
[RMG8], except what is revealed through windows
[RMG9]Until and unless children develop the capacity and the motivation to make these imaginative adjustments, we should expect them to explain and predict as if everything they themselves count as fact were accessible to the other as a basis for action and emotion. We should also expect them to allow no false propositions into the other’s data base: no false beliefs, in other words.
[RMG10]There are other ways of locating oneself mentally with another person. Many of our tendencies to emotion and action are specially keyed to egocentric locations and vectors: they depend, for example, on whether something is approaching or moving away: from me, that is. If I use imagination to shift the spatial center of my egocentric map to where you are standing, then the ball I throw to you would be imagined as an “approaching” ball. Or change the example from a ball to a bomb. Imagining an approaching ball or bomb tends to elicit some of those egocentrically keyed emotions and actions I mentioned. Rather than allowing these emotions and actions to reach their normal overt expression, I use them in interpreting your behavior.
[RMG11]As several commentators have remarked, in his effort to put pretending into the same mold as propositional attitude attribution, Leslie claims that pretending, say, that the banana is a telephone would require a metarepresentation, “I pretend ‘the banana is a telephone.’” That’s too easy, as well as false: it just makes pretending into a kind of propositional attitude attribution.