Special Problems in Modern Autism Rhetoric: Relativity

I was reading Lawrence M. Krauss’s Fear of Physics during my last migraine (as you do), and I realized modern autism nonsense has a relativity problem.  It goes like this:

1.)  Claim: “Autistic people lack theory of mind.  This is why they are bad at humanities-related things, like relationships and literature.”

2.)  Claim:  “Autistic people are unusually good at science-related things, like physics.”

Claim 1, when it is based on anything at all, is usually based on things like Baron-Cohen et al’s results in administering the Sally-Anne test to autistic children (handy infographic, minimal science at the link).  I don’t know what folk nonsense Claim 2 is based on.

Baron-Cohen et al’s Sally-Anne results, and their interpretation in cognitive theories of literature and in popular conversation to things like Claim 1, serves as a particularly good illustration of what I’m calling the “autism rhetoric relativity problem.”  The Sally-Anne test of “theory of mind” is a test that is less about reading other people’s “thoughts” or emotional states correctly as it is about correctly interpreting the viewpoints of various observers.  Getting the final question correct depends on being able to predict what Sally did or did not physically view relative to Anne, the marble, the box, and the basket.  In fact, to get the correct answer at all, one must abstract out information considered by the test to be irrelevant, such as any previous understandings between Anne and Sally regarding the marble.

The practice of physics tends to demand that physicists get good at (among others) two things: (1) grasping concepts like “what changes when we observe this physical phenomenon from some other point?” and (2) abstracting out irrelevant information.  In fact, one might argue – and as a humanities scholar I am going to argue – that these skills are even more important for physics than they are for the humanities.  Ergo, if the Sally-Anne test “proves” Claim 1 is true, it also “proves” that Claim 2 is false.  A test that “proves” autistic people suck at relative positioning and ignoring irrelevances also “proves” that autistic people suck at physics, because relative positioning and ignoring irrelevancies are exactly the things that make physics work.

The Spherical Cow in the Room

The reason Claim 1 and Claim 2 can so happily coexist in the popular imagination, of course, is that the popular imagination doesn’t see that the “theory of mind” the Sally-Anne test-takers are thought not to possess is neither the ability to consider relative physical position (important in physics) nor the ability to predict other people’s thought-feelings (important in the humanities).  It’s actually a hybrid.

The hybrid is summed up pretty well in my mention above of “abstracting out irrelevancies” in order to arrive at the “right” answer: that Sally is going to assume the marble hasn’t moved from the box while she’s been out.  The skill, “abstract out irrelevancies,” is crucial to solving any basic problem in physics.  The answer it demands in this context, however, is essentially a humanities matter – because the “irrelevancies” in question are human differences.  They’re questions of agreements, social contracts, relationships, communication, and assumptions; they all have to do with what Sally knows about Anne (“she steals a marble every chance she gets”), what Sally thinks she knows about Anne (“she’ll leave my marble where she found it”), what Sally thinks Anne knows about Sally (“she thinks I won’t mind if she takes my marble”), and so on.

My hypothesis is that autistic test-takers get tripped up at this point not because we are unusually bad at abstracting out irrelevancies.  Rather, we struggle with it when we do not default to the assumption that human differences are irrelevant – an especially difficult default when all our lives we have been constantly reminded that our human differences are deeply and problematically relevant, that other people think our differences are relevant, that we are expected to account for the fact that other people think our differences are relevant (“eye contact is important because it makes other people feel more comfortable”), etc.

I assert that these “irrelevancies” – ironically, the very same irrelevancies that social contact demands we consider relevant – are what make the Sally-Anne question fundamentally unanswerable.  Not all the facts required are in evidence.

The ability to array the human differences in a question, sort them, and target the “relevant but missing” pieces is what has made me a good lawyer, and it has been invaluable to my study of literature.  I can’t say what it might have done for me in a career in physics.  I can say that, insofar as we think the Sally-Anne test teaches us anything at all about any autistic person’s potential career path, we’re on the wrong path.


9 thoughts on “Special Problems in Modern Autism Rhetoric: Relativity

  1. 1. I like all of what you said. There’s lots of cool research getting done on salience processing/marking as processes of multi-sensory spatial and temporal reality-building in neurochemistry and systems neuroscience. Like, for instance: dopamine receptor agonists, like amphetamines, cause (NT) people to consistently over-estimate how much time has elapsed in an interval, whereas dopamine receptor antagonists, like haloperidol, cause people to consistently under-estimate how much time has elapsed. Which is pretty baller, considering that dopamine plays a huge role both in working memory and motor coordination/timing. To be fun and reductive for a moment: The more dopamine there is available to the prefrontal cortex, the more incoming sensory information gets marked as salient…And it seems a lot like time-intervals are internally estimated based on the amount of salient sensory information that has accumulated. So like, if you give a normal person an amphetamine, their brain normally experiences a salience-marking rate of, say, approximately 5 bits of sensory data per “second”-like interval. And suddenly they have way more dopamine than normal, and in the same time that they would normally have only marked, say, 25 bits of data salient, they’ve now marked 100 bits salient! Their brain tries to estimate time, and, using it’s old metric, determines that four-times more second-intervals have elapsed than they normally would. And then if you’re me, and have no sense of time whatsoever ever at all ever, and without amphetamine medications for ADHD, constantly both over- and under-estimate how much time has passed because LITERALLY WHAT IS TIME…You react to amphetamine dosing by still being terrible at estimating how much time has passed, but doing, on the whole, LESS over-estimation than you do without meds, because you’re finally able to like…engage in tasks like normal people do. And task-engagement/action compresses people’s sense of time, because they’re supposedly marking only the task-relevant information as salient. I, of course, am still terrible at knowing what is an isn’t salient, and at knowing time, or sensing time, or estimating any time at all even down to seconds (unless I literally count them off out loud). Because my brain is a glitter-sparkle-circus-parade from Neptune. I have the articles if you want to peep–they’re pretty nifty.

    2. Aaaaaand…I mean…The Sally-Anne Test discrepancy is almost definitely caused by the ridiculous level of linguistic complexity and high number of nested clauses in the Sally-Anne Test questions? Gernsbacher sasses well on this in one of her reviews–I can try and find it, but I know it’s available on her lab website. Bonus of physics: don’t need to be able to remember/understand all of a sentence read aloud to you containing multiple phrases like: “Where does X think that Y thinks that Z thinks…” Instead…numbers and pictures and calculations and spatial models. I could do the easy Sally-Anne questions, and probably could do the whole test if it was given in text-on-paper form (though slowly), but I would have a lot of trouble with the harder questions if read out loud to me, because my auditory-verbal memory capacity is barely enough to remember a clause at a time. Three-dimensional images, though? And physics models? Totes fine. Actually, better than fine. I’m also really good at descriptions of complicated three-dimensional sensory environments and at navigation/spatial imagery–it’s the temporal/sequential nature of complex sentences read aloud, and of many-small-step narratives, that screws me over.

    Hi, my name is Emma, and I’m a verbose-a-holic.

    Liked by 1 person

    1. Here’s an interesting thing about the multiple-phrase nested assumptions in “theory of mind” tests (that I don’t know if Gernsbacher covers): nobody can handle more than about five minds at a time. Nobody. Regardless of neurotype. It would appear humans can do “1 thinks X,” “2 thinks that 1 thinks X,” “3 thinks that 2 thinks that 1 thinks X,” and even “4 thinks that 3 thinks that 2 thinks that 1 thinks X” (although this is where breakdown starts to show up for most people), but throw in Person 5 and study after study implies it all goes to hell for everyone. So asking autistic subjects what Joy thinks Maria thinks Kate thinks Sally thinks Anne knows about the marble is kind of crap, because actually most people will get that wrong whether they’re autistic or not.

      This, according to Lisa Zunshine, is why if you’re at a party and having a conversation in a group of four people, things go fine, but if you add a fifth person, usually the group will dissolve into a “conversation of 3” and a “conversation of 2”.

      (TRUE STORY: I actually have a significant question about this, which is: Are we actually under-counting the number of minds by 1, because the count always fails to take into consideration one’s own experience of one’s own mind? Or am I just unusually self-conscious about my own mind’s doings, so bringing an “extra mind” into the conversation (“I think that Sally thinks Anne knows where the marble is” is a thing I do but most other people tend not to)? This strikes me as relevant in the Sally-Anne test context, because to me, the answer to “Where does Sally think the marble is?” (“I don’t know enough about Sally to guess”) is different from the answer to “where do you think Sally thinks the marble is?” (“in the box, probably”). From *my* point of view, I’m aware that I’m taking a test in which the information given is probably the only information the test administrator thinks I need to answer the question. From *Sally’s* point of view, I’m aware that there could be a bunch of stuff between her and Anne, or about the context of the place they’re in, that would affect the answer. What I don’t know is whether that’s a me thing (and if so, why?) or if other people do it too (and if so, why?).


      1. I think I was in the middle of replying to this soon after I saw your response, and then deleted what I originally wrote because I was like “Emma, in the interest of your academic/intellectual career, you should probably not say rude things about prominent scholars in your field on the internet…” But here I am once again. I’m probably going to say some rude things.

        With regard to your first point, I think that there’s more nuance to the “nesting” thesis than you’re necessarily giving it credit for. For instance, there’s a serious interaction between “nesting” levels and simple verbal memory/comprehension difficulty. A sentence with multiple nesting clauses, for me, often requires me to refrain from constructing a mental representation of the sentence meaning until I’ve read all the nested clauses, while at the same time keeping the already-read parts of the sentence in my memory long enough for me to understand all the parts as a whole. I just can’t do that a lot of the time–not because of some wishy-washy conception of rhetorical complexity (I’m just loving how, in this context, rhetorical complexity is explicitly coded as social complexity–what a coincidence!), but because by the time I manage to hear or read the second or third nested clause, I’ve forgotten what the first one said. I can maintain non-verbal representations of clauses well, but when the clauses only have representable meaning when they’re combined into a rhetorical whole, I’m going to have a harder and harder time comprehending the sentence the more and more nested clauses it has.

        And I mean, the Sally Anne tests are a piece of shit because all the TOM tests are pieces of shit. End of story. The Sally Anne test norms for typical adults are laughably variable and pathetically low.

        As much as I’d love this not to be true, it’s hard for me to take a sentence that starts with “…according to Lisa Zunshine” seriously. Really though. When was the last time you observed a conversation with five or more people in it that actually involved all five or more people talking to the other five or more people? I’m going to bet you money that it was actually a meeting, and required turn-taking and non-conversational social structures/rules. Because an actual five person conversation is almost always just five people alternating between engaging in, and listening to, a varying array of two- or three-person dialogues. Also I mean, just to reiterate, I can’t take sentences that start with “…according to Lisa Zunshine” seriously.

        I think your question concerning the level of “meta” expected in those tests is a valid one–I can say that I’m probably functionally as meta as you describe yourself being, but phenomenologically I’m distinctly more concrete/direct (probably an impulsivity thing, lesbihonest). So like, I also have the problem of people asking me questions like “Where does Sally Anne think the marble is?” and my natural response being “I can’t possibly know that…” Except in my case, instead of it being a process of thinking “I don’t know enough about Sally to guess,” it’s one of just thinking “I have no fucking idea. Why would I know.” I think that the degree of verbal meta-self-awareness one experiences, and the degree of observer fallibility one assumes, are semi-independently variable factors. Normal people are often highly meta-aware, while also being very lacking in a sense of observational fallibility/limitation, whereas I think a lot of autistic people who are even more concrete than I am tend to be less verbally meta-aware, while maintaining an unusually high degree of presumed observational fallibility/limitation. There’s a very interesting/not necessarily new argument about mental states, predictive coding and uncertainty to be made here, but I’m a bit too sleepy to make it now.

        Liked by 1 person

  2. One thing I learned in studying physics was that (when done properly) it is important to explore the whole spectrum of possible causes of the effect under investigation.

    One of the problems I have with the Sally-Anne test is that the subject’s underlying reasoning is never investigated. For example, in some test settings there are either non-visual clues that the marble has been disturbed (e.g. Sally could overhear Anne moving things), or small visual indications that something has changed between Sally leaving the room and re-entering. By failing to determine how the subject reached their conclusion, the testers are ignoring the very real possibility that there exists a valid chain of deduction behind the “wrong” answer.

    Liked by 1 person

    1. Yes!

      And I think you sum up my problem with the Sally-Anne test a lot more succinctly than I did, lol: The test demands that we pretend any number of things either aren’t relevant or aren’t happening, when in fact we don’t know, and not knowing affects the answer.

      It’s a bit like insisting that the correct answer to “if I drop this feather and this bowling ball, which one will hit the ground first?” is always only ever “they hit at the same time,” when in fact they might not if the medium in which they are dropped is not a vacuum, or if one is dropped several seconds after the other, or they are “dropped” into an orbit (and thus never hit the ground), or something. We need more information, because that information affects the answer. And being able to recognize that strikes me as distinctly related to questions of perspective-taking, which things like the Sally-Anne test claim to deal with but do a really crappy job of actually dealing with.

      Liked by 1 person

  3. Good points.

    We also have a healthy/unhealthy tendency to make assumptions about what will be the outcomes, given ambiguous situations. It is sometimes said “we all have to ignore reality or we would never get into a car, for riding in a car presumes the other drivers will obey the rules, which is demonstrably false. If the penalty for potentially being wrong is death, and the evidence is that some drivers do ignore the rules (hence traffic tickets) it makes no sense to risk death by climbing into a car.

    However, our society could not function the same way unless we all agree to make this irrational assumption: The next driver I see coming toward me at a combined 100 mph will in fact obey what I have learned to be the traffic rules.

    I would suggest that instead we jointly create a social view of reality that is different from the logical/rational/physics-based view. In this view we intentionally disregard some of what we know to be fact so that we all can get along. We learn this first by deferring to the statements or actions of our authority figures–parents–and later by learning the exceptions that these rules create.

    In a sense it becomes a one-tailed problem: Both those with limited social processing capability–who have learned the general rules but who cannot then unlearn them to make room for the exceptions that are taught–and those with exceptional social processing capability–who see not only the original rules and the taught exceptions, but also can critique those exceptions, recognizing their arbitrariness–face difficulty in understanding the situation “correctly.” On the other hand, those in the middle, with enough social capability to recognize the exceptions but without the ability to do more than rote memorize those social cues, get along just fine.

    The implications include that children with autism traits have a lower level of social processing capability, and are not disadvantaged in all situations. In fact, in those settings in which the expectations typically taught as “normal” are violated (such as with illusions) both those with lower and higher levels of social processing ability should do better at recognizing “the right answer” than those in the middle.

    A second implication is that we could all be arranged, as is the case with most classificatory characters, in a vast continuum from the person with the least capability to the one with the greatest capability for social processing and that societally we have drawn arbitrary lines as to where “too little” and “too much” begin. This locates the “normalcy” within our social setting, rather than within the measurable physics of neuron potential. In short, we are rendered “abnormal” by where society draws the lines, not by our inherent lack of an ability. For having too much can be as much a problem as having too little.

    A third implication is that any ‘treatment” of this deficiency probably should not rest with correcting whatever neurochemical or neurophysiological processes may be delimiting an individual from “normal” but rather in providing extra attention–tutoring–in either doing better at teaching the exceptions that those with too little social processing capacity possess were unable to pick up the first time through the normal channels; or at teaching those who are so good at processing social cues they can see how the exceptions typically taught are no more to be logically preferred than any other group of arbitrary characteristics, but to learn to recognize and to “act as if” they lacked the knowledge they do possess about the source and character of these exceptions.


  4. I think the phenomenological approach is interesting. Here, they argue that neurotypicals perceive mental states more directly and that using a theory of mind is not as primary as it is made out to be. In fact, it is autistic people, along with people with schizophrenia, who must rely primarily on theory of mind and it is often found wanting:


    This was published in the journal _Research in Developmental Disabilities_.


Comments are closed.