Stretching the Brain

I would like to take a few moments to discuss an autism research paper that was recently published online by the journal Human Brain Mapping. The paper is entitled Enhanced Visual Processing Contributes to Matrix Reasoning in Autism (Soulières, Dawson, Samson, Barbeau, Sahyoun, Strangman, Zeffiro and Mottron, 2009, hereafter referred to as EVP). As best as I can tell, the full text of EVP, like too much of autism research, is not being made reasonably available to the public; however, for those who are interested, an email request to the lead author will likely net you a pdf copy (that is how I obtained mine). As per usual, I am not entirely comfortable commenting at length about a research article that has not been made accessible to all, but as with other works produced by members associated with Laurent Mottron's research team, I feel these efforts are far too valuable and far too relevant to be left unconsidered. (Maybe one day the Mottron team will be in a position to remove itself from the autism publishing grid and present its insightful work more directly to the public, an action I would certainly recommend.)

Before I summarize the experiment at the heart of EVP and its corresponding results, let me note that in doing so I cannot do justice to the paper itself. One of EVP's main strengths is that it is extremely well produced—very thorough in detail, quite readable for a highly technical subject, and generally even-handed in its interpretations and judgments. For those who have not been exposed to the latest techniques in neuroimaging science, I can highly recommend EVP as an excellent example of what the discipline currently has to offer.

In this particular study, Dr. Soulières' team divided the experimental participants into two groups, 15 autistic participants in one group and 18 non-autistic participants in the other group, the two groups otherwise matched on factors of age, sex, Wechsler IQ scores, and manual preference. Soulières' team also rigged up a test-taking mechanism inside an fMRI scanner, with each participant being made to take first a simple pattern-matching test and then a Raven's Progressive Matrices (RPM) test within that mechanism. Soulières' team then compared the test results and the brain scanning images across the two study groups and interpreted the findings.

The experiment and its findings would have remained straightforward if it were not for an unexpected anomaly in the results. Although the autistic and non-autistic groups scored with similar accuracy on the RPM test, the autistic group was about forty percent quicker on average in completing the test. This led to some discussion about how to interpret the surprising result, and it also spurred the Soulières team to look at the fMRI images in two different ways, one designed to factor out the faster performance effect. In the final analysis, the anomaly was taken as a potentially significant, but not yet fully understood phenomenon, and the fMRI images were judged to be mostly unaffected by the between-group differences in performance times. Finally, getting back to what the experiment was originally designed to measure, the Soulières team judged that the fMRI images revealed some small, but significant neuronal differences between the autistic and non-autistic groups when answering RPM questions, and in particular judged that the autistic group seemed to be displaying a potential bias towards increased (enhanced) use of visual processing mechanisms to aid with reasoning tasks.

When I first read the press reports that came out upon publication of EVP, my initial surprise was not that the autistic group was forty percent quicker in answering RPM questions, my initial surprise was that the autistic group was not more accurate in answering RPM questions. The press reports (indeed, the abstract too) suggested that the autistic group had been matched to the non-autistic group on Wechsler scores and yet had performed with similar accuracy on the RPM questions. This would run counter to previous Mottron team findings, in particular those of The Level and Nature of Autistic Intelligence (Dawson, Soulières, Gernsbacher, and Mottron, 2007), which showed that autistic individuals tend to perform relatively better on RPM tests than on Wechsler exams. The details of EVP, however, appear to reveal a slightly different story than is suggested by the press reports and the abstract. On average, the autistic group scored about 101 on the Wechsler full scale, while the non-autistic group scored about 106, and the autistic group scored with about 76% accuracy on RPM, while the non-autistic group scored around 74%. It is not clear to me if these differences add up to statistical significance (and of course the recruitment procedures of EVP might also have skewed the results—although there is nothing in the paper to suggest this possibility), but an eyeball estimate suggests that the findings of Dawson, et al. still hold, and I suppose if we add in the speedier performance of the autistic group in answering the RPM questions, it seems reasonable to say that autistic individuals continue to show relatively better performance on RPM than would be suggested by their full scale Wechsler scores.

Now, about that faster performance finding.

In some respects, the finding that the autistic group was forty percent faster than the non-autistic group in answering RPM questions is a bit of a distraction within EVP. I can certainly understand why the authors included the finding and discussed it at some length—the between-group difference is too large to simply ignore. On the other hand, there are many factors that make it difficult, if not impossible, to draw any meaningful conclusions from the result.

The first problem of course is that RPM is not a timed test. RPM test takers are instructed to take as much time as is deemed necessary to feel reasonably certain about their answers, and these instructions were the ones given to the EVP participants. Furthermore, all the statistical information ever gathered around RPM has been compiled under conditions of a non-timed test, so in a certain sense, it is a violation of the spirit of RPM to even measure the amount of time it takes a participant to answer the questions, let alone report on those measurements. (Or to put it in perspective, a test taker who gets 35 questions correct in an hour is deemed to have performed better than a test taker who gets 34 questions correct in two minutes flat—speed has simply never been a factor in judging RPM performance.) Indeed, perhaps the first thing to do in determining whether EVP's faster performance finding is meaningful would be to repeat the experiment with a different set of instructions—instructions designed to make it clear that time is being measured and that speedier performance is somehow going to be judged as “better”—and see if those instructions affect the overall results. Such an experiment would of course be an even greater violation of the spirit of RPM, although at this point, I would have to say, in for a penny, in for a pound. None of this is meant as criticism, by the way—again, I understand the reasons the EVP authors reported on the finding—but I also want to make it clear that the entire circumstances of the finding are already resting on shaky ground.

If, however, we put aside those concerns momentarily, we can still discuss tentatively what the faster RPM performance from the autistic group is suggesting, and indeed that is the approach the EVP authors have taken. In short, the EVP authors posit that the faster performance might be indicative of an underlying processing “advantage” in the autistic group, one loosely tied to the visual processing differences seen in the fMRI results. At the same time, the authors are upfront about admitting that it is impossible to rule out, without further investigation, other plausible explanations—explanations that would have little to do with any advantages in reasoning skill. I would tentatively concur with such an analysis, but I would also like to consider a few more details than were discussed in the paper itself.

The test-taking conditions in the EVP experiment are highly unusual. Most people do not take an RPM test within the confines of an fMRI scanner. Such conditions are generally uncomfortable—loud and often claustrophobic—and certain individuals, at least after awhile, might feel highly motivated to answer questions quickly and bring the test to an end (God knows, I would feel that way). If it turns out, for instance, that autistic individuals feel this urge more strongly than do non-autistic individuals, that would explain at least in part their relative haste, but it would also have very little to say about their reasoning skills.

At the other end of the significance scale, one plausible explanation for the between-group time difference relates to a finding from another paper several of the EVP authors participated in: Cognitive Differences in Pictorial Reasoning Between High-Functioning Autism and Asperger's Syndrome (Sahyoun, Soulières, Belliveau, Mottron, and Mody, 2009). In short, that paper demonstrates that autistic individuals appear to favor visuospatial processing strategies over semantic or linguistic strategies, more so than would be the case for non-autistic individuals (and more so than would be the case for Asperger's individuals, but it is important to keep in mind that Asperger's individuals were excluded from the EVP study). Although between-group response times were similar on the simple pattern matching test of EVP, one can surmise that such simple and straightforward visual patterns are going to be processed quickly and similarly by all. The RPM questions, however, are an entirely different story. RPM questions are complex and multi-dimensional, enough so that they can be tackled effectively through a variety of strategies. One approach, for instance, would be to try literally to see the patterns emerge; that is, tackle the problems visually. Another approach would be to talk one's way through the problem; e.g., “there are three dots in the upper left in the first square, there are just two dots in the middle in the second square, and so there must be one dot in the lower right in the last square, …”; that is, tackle the problem with a linguistic or semantic strategy (this is the manner in which I would approach RPM problems, for instance). Both strategies, and perhaps others as well, can be effective in determining the correct answer, but given that the RPM test is visually set, it would seem reasonable to surmise that those using a visually based strategy might have a built-in advantage for being quicker, because there is less cognitive translation required for such an effort. (This, by the way, would justify the decision to set RPM as a non-timed test.)

Such an explanation would actually fit in quite nicely with EVP's thesis that autistic individuals rely more heavily on visual processing abilities in the performance of reasoning tasks, but such an explanation would also present the EVP authors with a potentially confounding issue—namely, is the EVP experiment picking up a true autistic versus non-autistic reasoning difference, or is it instead picking up a visuospatial style versus semantic style difference? That question would extend to the fMRI results as well, so it is important to give it some consideration. The fact that autistic individuals may be more visual in their approach to problem solving than non-autistic individuals is certainly interesting and significant, but I am not sure that by itself it captures what is frequently meant by autistic versus non-autistic reasoning skills. If, for instance, visually-oriented non-autistic individuals show similar performance and fMRI patterns as autistic individuals, then the whole notion that EVP is capturing a true autistic versus non-autistic reasoning difference becomes far more doubtful.

In the end, the 40% greater efficiency finding, although much ballyhooed in the press reports, still represents little more than a tempting sideshow at the moment, with a whole host of possibilities lined up to account for the surprising result. Perhaps Dr. Soulières' team will in the future have the opportunity to explore the discovery in greater detail, and maybe then we can draw some more definitive conclusions.

And now back to the regularly scheduled programming.

The main purpose of EVP is to compare autistic versus non-autistic brain activity during the performance of a complex reasoning task. This is in keeping with the vernacular that the autistic brain is somehow “wired” differently, and in particular, EVP explores the hypothesis that autistic individuals make greater use of perceptual (especially visual) neural processing mechanisms during cognitive tasks.

There is an interesting historical background to this hypothesis that I want to take some time to recall. In 2006, an unusual paper was published in the Journal of Autism and Developmental Disorders called Enhanced Perceptual Functioning in Autism: an Update, and Eight Principles of Autistic Perception (Mottron, Dawson, Soulières, Hubert, Burack, 2006, hereafter referred to as EPF). I say unusual, because EPF is perhaps the most schismatic research paper I have yet to encounter. If one were to remove principles 6 and 7 (and to some extent, principle 5) from that paper, all that would remain is a straightforward description of autistic cognition as biased towards local processing and detail discrimination, possibly caused by a different brain structure biased towards enhanced perceptual processing. This is an interesting and somewhat novel approach, yet it is not all that different in kind from other research-oriented descriptions of autistic behavior and cognition, and is not all that applicable in a broader and anthropological sense. But injected right into the middle of this rather dry description of autistic psychiatry and neurology comes suddenly a fascinating use of the characteristics of savant syndrome to highlight the distinctive perceptual characteristics of autistic thinking, replete with an enlightening new emphasis on the role that environmental patterns, symmetries and structure can play in the development and presentation of savant and autistic cognition. Furthermore, this discussion of principles 6 and 7 takes on an entirely different tone from the rest of the paper, for suddenly gone is the over reliance on psychiatric and experimental literature, and in its place is substituted a more conceptual description of autistic perception and cognition. In the modern sense of the word, this discussion is less “scientific” than the rest of EPF, but it is also demonstrably more fruitful and broadly applicable, to the point of being anthropologically stunning. In EPF, principles 6 and 7 appear like a bolt from the blue.

I have written elsewhere about these two aspects of EPF (one aspect which I have clearly enjoyed, one aspect which I am less thrilled about), but it is important to note that if in the year 2006, EPF was like a cell getting ready to divide, by 2009 that division has been thoroughly accomplished. Members associated with Laurent Mottron's research team have published two significant papers this year. First, in Enhanced Perception in Savant Syndrome: Patterns, Structure, and Creativity (Mottron, Dawson, Soulières, 2009, hereafter referred to as EPSS), the team has put together a more complete reflection upon EPF's principles 6 and 7, backed now by a richer catalog of case studies, but still essentially a conceptual description of autistic perception, as opposed to being a scientific hypothesis supported by a plethora of experimental evidence. In contrast, they have also published EVP, the paper under discussion here, one that more clearly aligns to those aspects of EPF that were concentrated on the psychiatry and neurology behind enhanced perceptual processing, an effort now being supported in the most direct manner possible by the techniques of experimental science.

Of course, it is possible the members of the Mottron team do not appreciate my attempts to divide their work so thoroughly down the middle—I suspect they might see their two recent papers as simply different views onto a model they conceive of as constituting a whole. But I want to insist upon making the distinction, because from my point of view I see one of these approaches as holding the potential for providing abundant knowledge about the nature of autism—as well as providing abundant knowledge about the nature of mankind—whereas I see the other approach leading mostly to a scientific dead end. Since I have already written abundantly and effusively about the unlimited value to be found inside the pages of EPSS, let me now correspondingly add my reasons for believing that the findings of EVP hold a high likelihood of leading nowhere.

The problem is not with EVP itself. As I have said, EVP is extremely well done, and in many respects we owe a debt of gratitude to Dr. Soulières and her team for demonstrating the leading edge of what neuroimaging has to offer for providing an affirmative description of the characteristics of the autistic brain. That I believe these efforts ultimately reveal very little is to be attributed more directly to the shortcomings of neuroscience itself—and neuroimaging in particular—disciplines that, despite their widely advertised promise, are constructed upon logical foundations that are dubious at best. Therefore let me use the findings of EVP to explore the current state of neuroimaging science, and let me see if the discipline really has that much to offer in the way of helping us narrow in on the qualities of autistic intelligence and reasoning. There are many logical and conceptual problems I might list, but let me start with just these:

The neural activity of autistic and non-autistic individuals is far more similar than it is different, a fact that is not given its true weight in assessments of the brain's role in human cognitive tasks. EVP admits to the large overall similarity between autistic and non-autistic neural activity during RPM tasks, but since it does so with only a single sentence or two, and then devotes its remaining sentences to a thorough analysis of the subtle distinctions, I think it is important to take a step back and put things in proper perspective. If we are to believe the assumption that the brain is the primary source of reasoning/intelligence activities, then all neural activity observed during a reasoning task must be given its due weight, and on that basis alone, we would have to conclude that autistics and non-autistics cognate in essentially the same manner, the differences being slight at best. That is what the overall fMRI evidence strongly suggests.

Of course, there are known domains in which marginal differences can be important, but that tends to happen only when the entity under study leverages a much larger context. That larger context, however, is exactly what modern neuroscience does not consider. Cognitive scientists study the human brain so intensely because for them the human brain is the alpha and omega of human cognitive activity—all the secrets are to be found within the confines of the human skull. In consequence, cognitive scientists are left with an overriding conundrum of how to explain wholesale cognitive distinctions on the basis of only minor-scale neural differences. That conundrum is on full display in EVP.

Neuroimaging technology is still extremely crude, its expense notwithstanding. A nice touch within the EVP paper is its inclusion of a set of fMRI composites, pictures that might easily dazzle the reader if they were not by comparison—comparison to the pictures of a disposable camera, for instance—so incredibly crude. Indeed, such a comparison is extremely apt, for a tourist taking snapshots of the Parthenon, let's say, can record a wealth of accurate detail about human intelligence and reasoning—structure, pattern, style, history, and the like; and although neuroscientists no doubt desire to reveal a similar set of cognitive characteristics, even at a more nuanced level of detail, it is hard to make those kinds of distinctions when dealing with little more than a collection of colorful blobs.

I suspect the reason most people are dazzled by fMRI pictures, and the reason most scientists place such incredible faith in their fMRI results—despite the obvious coarseness of the data—is because the technology is so darned sophisticated and new, not to mention so darned expensive; technology costing that much and qualifying as a genuine modern medical marvel must surely have something significant to impart. But of course such reasoning is not an example of brilliant science: it is instead a textbook example of a logical fallacy. The price and sophistication of the equipment has nothing to do with the value of its output; only the quality of the resulting data can have a bearing on its importance, and on that basis, neuroimaging still has an incredibly long ways to go.

It is easy to confuse neural difference with neural causation, a mistake neuroscientists will make at the drop of a hat. Let me demonstrate with an analogy. Imagine a tribe of people in which a small minority of children are trained from birth to become high jumpers. They practice, they work with weights, they compete—a good portion of their early life is devoted to gaining an extra defiance of the planet's gravity. Then a naïve group of researchers decides to study this tribe of people, and in particular wishes to discover what turns some of these people into such unusually prodigious leapers. They do this study by performing muscle scans on comparison groups of high jumpers versus non-high jumpers as the participants progress through a series of physical activities, and the published findings reveal that the high jumping group has some small but significant differences in their quadricep, hamstring and calf muscle systems.

Now if the researchers stopped there, there would be no problem: they are simply reporting the data. But of course we know the researchers will not stop there, we know precisely what is coming next. The researchers will then announce, with great fanfare, that the muscle distinctions they have discovered in the high jumping group are the actual cause of the high jumping phenomenon. (And if we were to extend this analogy to autism science, the researchers would next propose some surgical procedures whereby to remove the muscle differences and return the high jumpers back to normal.)

The crux of the Mottron team's enhanced perceptual functioning hypothesis turns on whether that hypothesis is meant as a description or an explanation. A description I would not mind at all, but an explanation I object to vehemently. It is not always clear to me where the Mottron team stands on this distinction, but if the title of EVP is to be given any weight, then I suspect the team is tilting in the wrong direction.

Few people would disagree that autistics and non-autistics cognate differently, and so neural differences are certainly to be expected (indeed, perhaps the most surprising finding in EVP is the neural similarity observed on the simple pattern matching task). But of course autistics cognate differently right from birth, and so when we take fMRIs of their brain activity many years later, we have no way of easily assessing whether we are observing neural causes or neural effects. Under the circumstances, I would think the latter would get initial preference, but then again, I am not a neuroscientist.

The lack of any plausible mechanism connecting neural activity to observable cognitive behavior allows neuroscientists unfettered creativity in explaining their results. Scientists, quite rightly, dismiss the vaccines-causes-autism hypothesis by noting there is no plausible mechanism connecting the ingredients and actions of vaccines with the observable behaviors of autistic individuals. Then those very same scientists will turn right around and promote the output of their neuroimaging studies as explanations for certain types of observable cognitive activity. Simply amazing.

Near the end of EVP, in the section entitled “Origin of Neural Differences in Matrix Reasoning Between Autistics and Non-Autistics,” the EVP authors attempt to explain how their neural findings might give rise to certain types of autistic cognitive behavior. They do this mostly by appealing to the findings of other neuroscience studies, studies focused on such things as white matter microstructure and functional connectivity differences, studies which have found that autistic individuals seem to have less neural connectivity between, for instance, the prefrontal and occipital regions of the brain. The authors then lend their support to some theorizing that suggests this reduced connectivity produces compensatory activity in perceptual mechanisms, leading perhaps to a unique autistic cognitive signature. All this sounds scientific enough, but note what would happen if, for the sake of argument, the other neuroscience studies had found something entirely different, for instance that autistics had an overabundance of connectivity between the prefrontal and occipital regions. Would this turn everyone's theorizing around 180 degrees? Well, of course not, not at all. What would then happen is that we would get some kind of explanation about how the abundant connectivity was too much, too overwhelming, producing the equivalent of a neural traffic jam, and the compensatory activity of the perceptual mechanisms was therefore like taking an alternative route to work. In point of fact, it would not matter if autistics displayed under-connectivity, over-connectivity, or had their synapses tied together with pink, curly bows, neuroscientists would use whatever they found to cook up some type of explanation about the differential cognitive activity of autistic individuals. And what, you ask, could possibly give rise to such an unlimited degree of explanatory freedom? Well, what gives rise to this phenomenon is that neuroscientists have no plausible mechanism linking the neural activity they measure in their studies to the observable cognitive behavior of actual human beings. The conceptual gap here is at least as wide as that between thimerosal and autism—perhaps much wider—and into the space of that gap neuroscientists feel free to insert whatever convenient explanation they like. And boy do they ever!

Neuroscience remains blissfully ignorant of human history, and in particular remains blissfully ignorant of the Flynn effect. This seems to be a blind spot for all of neuroscience, but is particularly relevant for the experiment conducted in EVP. The type of intelligence being measured with RPM has undergone a rapid increase through the latter half of the twentieth century, an increase almost certainly experienced by both the autistic and non-autistic populations (indeed, experienced by all human populations). Furthermore, as I have pointed out elsewhere, there is no reason to believe the Flynn effect is restricted to the twentieth century alone: how well can we expect an average human to have scored on an RPM-type test say ten thousand years ago? No, the Flynn effect has been with humanity since at least the time of the great leap forward, and so any direct observations being made of human intelligence and reasoning—even neuroimaging observations—are observations of a phenomenon that, historically speaking, has been exceptionally non-static.

Thus do we want to insist that we can literally see human reasoning and intelligence within the biology of the human brain? How are we to accommodate evolutionarily static brain biology to an historically non-static Flynn effect? And note that these problems are actually doubled in EVP, which by positing two different neural reasoning mechanisms for the human brain, one autistic and one not, leaves the biological problem of the Flynn effect now to be answered times two.

Yes, there is observable neural activity during the performance of an RPM task, but does that neural activity equate to human reasoning and intelligence? The evidence of the Flynn effect votes a resounding no.

Nearly all the above-listed problems can be traced to one overriding logical fallacy, namely the unquestioned acceptance that human brain activity is sufficient to explain human cognitive functioning. The widespread belief in this assumption competes only with the notion that evolution explains every biological and cultural process for being modern science's greatest logical misstep. Such widespread belief is, to put it simply, a clear and massive instance of assuming what needs to be shown.

I have tried to demonstrate in several places (for instance here and here) that much of human cognition—intelligence, learning, language, etc.—can be more accurately and more fruitfully described by appealing not to the workings of the human brain, but instead by appealing to the fast-changing, self-constructed form of the human environment (and the members of the Mottron team too have been making that demonstration within the pages of EPSS—whether they realize it or not). I am open to a valid criticism of that argument, just as I am open to any effective demonstration of neural causation for human cognitive activity. But when the debate is always conducted as though the answer is already known, I find it nearly impossible to achieve any meaningful progress.

I am not certain where this unquestioned reliance on the human brain has come from. Perhaps it has originated out of those case studies where someone has been damaged in a section of their brain (Broca's Area, for example) and has correspondingly lost a portion of their cognitive functioning (their speech, for instance). But I assure you, if I yank someone's heart from out of their chest, they will also lose a good portion of their cognitive activity, but that does not justify my making the cardiovascular system the ultimate explanation for human reasoning and intelligence. It is hard to believe so many scientists cannot make the proper distinction between the concepts of necessity and sufficiency, but sad to say, when it comes to human cognition and the human brain, that mistake is nearly universal.

In conclusion, I confess to some ambivalence when I see Dr. Soulières, Dr. Mottron, and the rest of their colleagues listed as authors on an autism-specific neuroimaging study. On the one hand, they are doing me an immense favor, for a brain-specific approach to describing autistic cognition presents a sharp challenge to many of my own ideas about autism, a challenge that, in the interest of achieving greater acuity, I gladly welcome. And since I do not have the resources or means (or desire, for that matter) to conduct such studies myself, I am pleased to see them being conducted by scientists whose intelligence and integrity I can trust. The members of the Mottron team have built a rich history of providing positive, creative and informative insights into the nature of autism, and to the extent any neuroimaging approach might advance our understanding of the autistic brain, I suspect that team will achieve the goal as well as any other. EVP's authors are to be congratulated for their careful and considered effort.

But on the other hand, from a more practical point of view, I know time and resources are limited, and there are still so many autistic individuals in need of a greater understanding. The poignancy here is that the Mottron team has already developed an alternative approach to describing autism that possesses nearly unlimited potential for providing greater insight. Outside the confines of mainstream autism science, and outside the confines of the human skull, the Mottron team has been advancing an understanding of autistic individuals through an appeal to the autistic cognitive environment, an environment best described through the surrounding presence of pattern, structure and form. In my opinion, that effort possesses no discernible bounds: it holds promise for describing autistic individuals as they truly are, and it holds promise for revealing the remarkable influence of autistic cognition upon the entire human species.

If the Mottron team wishes to supplement these excellent ideas by exploring neuroscience as well, it is not for me to make an objection. But I hope they will at least consider my challenges to the foundations of neuroscience; I hope they will at least consider that a brain-based approach is almost certainly going to be more limited. The glory of autistic individuals, including the glory of their surprising impact upon the human population, is to be discovered in the abundance of their surrounding cognitive circumstances, and not in what exists inside their heads.