Theory of Mind

Much has been written about the so-called lack of "Theory of Mind" in autism. While we are organizing the site, I've decided to feature one article (there will be more, as many Autistics have written quite eloquently on the subject). Phil Schwarz has given permission to reprint his writing on the topic of Autism, Memory, and Personality, first posted here:


June 24, 2006

I think there is a serious problem with most attempts to study performance on tasks as complex as recognition of subtle personality traits, or imputation of emotional state and intent in others.

Experimental models for studying these tasks are necessarily simplified, in part not the least to control for potentially confounding factors.

But then the results often do not support the generality of the interpretations that are applied to them.

In studies of “theory of mind”, that is, the ability to represent some aspects of the mental state of others, the skill tested is almost always the ability to represent what (sub)set of *facts* (and specific beliefs based on directly known facts) is possibly known to the other individual.

Correctly representing the subset of facts known to an individual in a given scenario, and beliefs likely held by that individual based directly on those facts, is a *much* less complex task than generally imputing emotional state, or motive.

To prognosticate about the latter task based on results obtained regarding the former is wrong.

The theory-of-mind hypothesis has been called into question for a number of reasons, including, for example, Morton Gernsbacher’s question about whether the experimental designs purporting to test it properly controlled for variance in the subjects’ understanding of the necessary subtleties of language.

But there is one additional question that must be asked of the ToM hypothesis, that I have not seen asked, and I think that question may be relevant to Klein’s hypothesis as well.
The question is whether non-autistics are actually successfully performing the more general tasks that the hypothesis supposedly finds autistic people failing at.

In the case of theory-of-mind, I think there is plenty of reason to believe that the non-autistic majority is not particularly good at imputing emotion or intent to individuals significantly different from themselves. They get by, by imagining *themselves* in the other person’s situation, and imputing their own likely reactions and motives to the other person. If the other person is similar enough to them in enough relevant ways, their guess will be correct. *My* hypothesis — and I would love to see design and execution of experiments to test it — is that this is in fact what the majority does. And a critical (mis)interpretation of the theory-of-mind experiments is that they are doing anything more than that.

Now think about the autistic individual, who is *dissimilar* enough from the vast majority of surrounding people in enough relevant ways, that putting themselves in the other person’s shoes will often enough *not* yield correct results. What an autistic person who sets out to master “theory-of-mind” skills must do, is to *construct* an abstraction of the majority’s reactions and motives — an abstraction in some ways quite alien to themselves. The relevant question is not “what would *I* do in their shoes?” but rather “what would ‘Joe Normal’ do in their shoes?”, where ‘Joe Normal’ is a name for that abstraction.

I would argue that autistic people who *do* manage to navigate non-autistic society well enough to hold jobs, manage a household, and in some cases marry and raise kids (there will be a whole panel of us at the Autism Society of America’s 2006 annual conference next month), have to do just that, and in doing so, actually succeed at the business of accumulating expertise about the personality traits of others — at least those details that are relevant to their own successful navigation — to a much more exacting degree than the majority ever has to.
And that the inability of other autistic people to do so is not a factor unique to autism, but is actually shared by a significant portion of the non-autistic majority, who can get by by substituting the simpler task of comparison-to-self for the one purportedly being measured or studied, by sheer numerical advantage of being part of a majority sufficiently homogenous to do so.

So I’d love to see an experimental design that controls for that, by seeing how well both autistics and non-autistics perform on the tasks involved in learning the traits of a personality with an underlying emotional and motivational calculus quite different from their own. Autistic people are surrounded by people like that. Non-autistic people generally aren’t, and their performance in such environments is, I think, largely untested scientifically.

In real life, one situation in which non-autistic people are in fact surrounded by people significantly dissimilar to them in emotional and motivational calculus, is as employees or overseers of facilities constructed for the housing of such people: group homes, special needs schools, inpatient psychiatric facilities, and so on. And in those institutions, there are far too many cases in which non-autistic people with *inadequate* understanding of the dissimilar emotional and motivational calculus of the residents/students are placed in charge of decision-making or supervision that requires such understanding.

So I think the design and execution of experiments to measure the *majority’s* range of ability to construct a rich enough representation of personalities *significantly* dissimilar from their own, is not just an idle question. I think there is a very real need to be doing such research, and perhaps to use it to develop tools to (a.) measure the *exceptional* proficiency at such tasks that really should be a criterion for working with significantly dissimilar people, particularly in positions that confer tremendous power over other people’s lives, and (b.) to help *teach* and *develop* the skills involved in constructing sufficiently rich models of significantly different personalities, both to help more and more autistic people navigate the sea of non-autistic people they find themselves in, and to help train professionals working with our population to better understand us and to surmount the disconnects that lead to so many iatrogenic consequences in the lives of people on the spectrum.

– Phil Schwarz
Vice-president, Asperger’s Association of New England
AS father of an autistic son