Week 12: 11/19-11/25
This week, I finally sat down and poured over Simon Baron-Cohen book, Mindblindness. In this book, Baron-Cohen (BC) attempts to theorize autism by approaching it as a constellation of deficits related to certain evolutionary components of innate sociability mechanisms. He is clearer than I am.
BC starts his project off asserting that neurotypical people are “mindreaders,” that they have the “capacity to imagine or represent states of mind that we or others might hold” (2). BC believes that this mindreading capability is evolutionary and that it is necessary to function as a fully social being: “…a person with mindblindness [the inability to “mindread”] is thrown back on temporal-regularity accounts or on routine-script explanations…or is forced to use unwieldy things resembling the ‘reinforcement schedule’ explanations that behaviorist psychologists construct. None of these seem very useful here. The first two are too limited in their application to the constantly changing social world; the third takes too long to compute. In the heat of a social situation, it pays to be able to come up with a sensible interpretation of the causes of actions quickly if one is to survive to socialize another day. Non-mentalistic explanations are just not up to the job of making sense of and predicting behavior rapidly. Instead, a person with mindblindness is left confused: Just what are Joe and Tim up to? In the meantime, the mindreader seizes up the situation instantly” (4).
It is important to note that BC prefaces the book with the caveat that “children and adults with the biological condition of autism suffer, to varying degrees, from mindblindness. For reasons to be explored, they fail to develop the capacity to mindread in the normal way” (5). BC also hedges about intuition quite a bit in the early chapters. For example, BC wants to create an analogy between expert chess players and neurotypical socializers: “Expert chess players may feel that they intuitively know what the next best move is, and their skill at chess may be an excellent metaphor for how we routinely make judgments during social interaction. Like the chess expert, we are social experts. Our social reasoning process has become automatic and effortless—possibly as a result of years of daily practice, possibly also because, right from the beginning of life, the human brain is programmed to automatically and effortless interpret social behavior in this way, as a result of millions of years of evolution” (19). BC spends much of the start of his book with “maybes.” This makes it easy to tell other stories, other counterexplanations. More on that later…
BC dos not think that behaviorist approaches to behavior (and language) are as powerful or economical as mindreading because mindreading can take place when there are no behavioral cues. BC cites a friend not calling for a number of months as a potential example. In this example, BC argues, one might take the silence as evidence of offense caused and attribute the intentional state that one’s friend is offended at a previous action (25-26). It’s not clear how mindreading is the best explanation here. One could tell another counter-story dependent on contextual actions.
BC is not only concerned with the relationship between behavior and mindreading—he seems to primarily be concerned with the relationship between mindreading and communication. BC argues that mindreading is what helps us connect intentional states and semantics, that it fills in the gaps that words and their arrangements leave. For example, borrowing from Stephen Pinker’s work, BC provides the following dialogue that, he believes, implies mindreading:
Woman: I’m leaving you.
Man: Who is he?
In this example, BC argues, “the man must have thought the woman was leaving him for another man. When we make this attribution to the man, the dialogue hangs together perfectly. If we did not make it, the dialogue would seem disconnected, almost a random string of words. As mindreaders, we perceive the man’s sentence as far from random. Presumably, a person with mindblindness would struggle in vain to find any relevance in this exchange” (28-29). But this doesn’t sound exactly right. There seems something conventionalized about this exchange. Here is another experiment (not BC’s):
Woman: I’m leaving you.
Man: I’ll put the seat down.
In this example, a mindreader has to work a little harder, but they can read intentional states (the man believes the woman is leaving him because he leaves the toilet seat up). This ascription is predicated on a causal linkage: the second utterance is assumed to be causally connected to the first both in why it is offered and in the content (the content is assumed to be related to the first). Because of this, I would argue we can ascribe the intentional states to the man. This second example might show how the causal relationship can then imply context as the key to intentionality rather than any sort of mindreading device.
Once BC gets to the fourth chapter of his book, he begins to systematically put together his theory of mindreading. He does this by scaffolding upward, uniting four mechanisms that are responsible for mindreading.
The first mechanism is the “Intentionality Detector” (ID): “a perceptual device that interprets motion stimuli in terms of the primitive volitional mental states of goal and desire. I see these as primitive mental states in that they are the ones that are needed in order to be able to make sense of the universal movements of all animals: approach and avoidance…The basic idea is that this device is activated whenever there is any perceptual input that might identify something as an agent…This could be anything with self-propelled motion…It works through the senses (vision, touch, and audition), and its value lies in its generality of application: it will interpret almost anything with self-propelled motion, or anything that makes a non-random sound, as a query agent with goals and desires” (32-34). Of course, there is a learning curve here, and experience will teach one that certain things—like clouds—are not agents (35).
The second mechanism is the “Eye-Direction Detector” (EDD): “I suggest that in the human case EDD has three basic functions: it detects the presence of eyes or eye-like stimuli, it computes whether eyes are directed toward it or toward something else, and it infers from its own case that if another organism’s eyes are directed at something then that organism sees that thing. This last function is important because it allows the infant to attribute a perceptual state to another organism (such as ‘Mummy sees me’)” (38-39).
BC argues that ID and EDD can form dyadic intentional representations such as “Agent-wants-the food” (ID) or “Agent-is looking at-the door” (EDD): “these representations can be described as dyadic, since they only specify the intentional (i.e., mentalistic) relation between two objects (Agent and Object, or Agent and Self)” (44). Such representations have their use, but they do not allow for communication about a shared reality, the necessary step to avoid an “autistic universe” where “you would have sensations, and you would have images of people doing things and even wanting and seeing things, but you would have no way of knowing that what you and another person were seeing or thinking about was the very same thing” (44). This is where the third mechanism comes in.
The Shared-Attention Mechanism (SAM) builds triadic representations: “Essentially, triadic representations specify the relations among an Agent, the Self, and a (third) Object. (The Object can be another Agent, too.) Included in a triadic representation is an embedded element which specifies that Agent and Self are both attending to the same object” (44-45). The triadic representation is built, then, when an agent recognizes another agent’s perceptual state. “It then computes shared attention by comparing another agent’s perceptual state with the self’s current perceptual state. It is like a comparator, fusing dyadic representations about another’s perceptual state and dyadic representations about the self’s current perceptual state into a triadic representation” (46). BC is quick to point out that using SAM in this way is very limiting because the only way to verify mutual attention is by being mutually present to that which each agent was attending (47).
The final mechanism is the Theory-of-Mind Mechanism (ToMM), which “has the dual function of representing the set of epistemic mental states [pretending, thinking, knowing, believing, imagining, dreaming, guessing, and deceiving] and turning all this mentalistic knowledge into a useful theory” (51). A unique and invaluable part of this theory involves “referential opacity”: “ToMM allows the referential opacity that is a key property of epistemic mental states. Referential opacity (or non-substitutability) is the property of suspending normal truth relations of propositions” (52). For example:
“Snow White thought the woman selling apples was a kind person” may be true even though “Snow White thought her wicked stepmother was a kind person” is false (52-53).
BC goes on to say, “My idea is that ToMM is triggered in development by taking triadic representations from SAM and converting them into M-Representations. In its strongest and clearest form, my claim is that without SAM ToMM cannot get started” (55). “M-Representations” involve propositions + intentional stances towards them such as “Adam believes ‘monkeys are magic.’” The proposition within the M-representation may be false, but the intentionality is what determines the truth value of the whole (51-52). This ability to misrepresent (in a sense) makes ToMM more versatile and powerful than the other three mechanisms (56).
BC feels that ID and EDD can both be functionally intact in children with autism, but he feels that there is a “massive impairment in the functioning of SAM in most children with autism” (63-66). BC argues that the joint-attention mechanisms (that form triadic relationships) are absent or underdeveloped in children with autism (66). Even hand-over-hand actions—such as when an autistic child may put an adult’s hand on an object to get the adult to operate it—are considered instrumental and not indicative of shared interest (68-69). This is highly questionable, and BC does not offer any reason why this must be the case.
BC goes on to argue that deficits in SAM lead to impairments in ToMM. BC’s litmus test of this is “false belief tests.” While there are a number of variants on the test, the “Smarties test” is perhaps the easiest to understand. In this test, a researcher shows a child a cylindrical tube with a Smarties candy label on the outside. The researcher then asks the child what the child thinks is in the tube. The child responds “Smarties.” The researcher then opens the tube, showing the child that it is actually filled with pencils. After this sequence, the researcher then asks the child what he or she initially believed was in the tube. A neurotypical child will respond “Smarties,” while autistic children tend to respond “pencils.” The researcher will then ask what the next person who looks at the tube will think is in it. The neurotypical child will respond “Smarties,” while the majority of autistic children will respond “pencils.” From this, BC infers that the autistic children “answered by considering their own knowledge of what was in the box rather than by referring to their own previous false belief or to someone else’s current false belief. The robustness of this finding suggests that in autism there is a genuine inability to understand other people’s different beliefs” (71).
This impairment in ToMM, BC hypothesizes, should also manifest an impairment in pretend play (as this is predicated, in a sense, on false beliefs), and research seems to bear this out (76-77). BC also speculates that autistic children should have trouble understanding belief-based emotions and argues that his previous research supports such a view (78-79). Finally, BC argues that children with autism have difficulty distinguishing between appearances and reality, with much of their world being constituted by immediate perceptions (82).
Near the end of Mindblindness, BC explicitly points out that his theory points shows that “individuals with autism may be delayed or deviant relative to the normal course of development of the mindreading system at any point” (136). He goes on to say that individuals may, indeed, develop compensatory measures and even develop some mindreading ability, though such theory of mind will differ from typical theory of mind (just as an ESL-speaker’s English will differ from a native speaker’s, even if such difference is “remarkably subtle” (136-37). And herein lies the crux of the issue (at least for my purposes).
Baron-Cohen's theory seems quite compelling: it has significant predictive power, it is intuitively appealing, and it is backed by a good deal of research. The only major limitation is one that Baron-Cohen himself points out: it is developmental. In other words, he leaves open the possibility that individuals can develop either a ToMM or compensatory measures that would mirror the work that a ToMM might do (such as an intentional calculus). Anecdotal evidence suggest that both of these may be the case sometimes.
One additional limitation in BC's work is that it lumps "autism" into a uniform disorder (more or less), and variants of autism might have quite divergent SAM and ToMM. To be fair, by classifying his theory as both developmental and "deviant," BC does leave room for his theory to accommodate this aspect of autism.
All told, I tend to buy into much of BC's theory as a developmental one. I do think that ToMM needs to be studied longitudinally, and I think there is vital research to be done in terms of adults on the spectrum and how they negotiate ToM. I'll have more to say on this in my seminar paper, I think...