Wednesday, March 18, 2009

An Odd Conclusion

The Journal of Autism and Developmental Disorders has recently published the paper The Level and Nature of Autistic Intelligence Revisited (Bölte, Dziobek, and Poustka, 2009, hereafter referred to as Bölte). Bölte is in response to the paper The Level and Nature of Autistic Intelligence (Dawson, Soulières, Gernsbacher, and Mottron, 2007, hereafter referred to as Dawson) and serves as an effort to replicate Dawson’s findings. There is much that can be praised about Bölte—it lays out its methodology in good detail, and it is up front and honest about the various ways in which it does and does not match the techniques of Dawson—but there are also some very troubling aspects about Bölte. In the first place, the motivation for doing this follow-up study has been made somewhat suspicious by the Bölte authors setting up a fictitious target in Dawson for which to attack. But even more concerning, the main conclusion of Bölte can only be described as something extremely odd, puzzling in the highest degree, given that it is directly contradicted by Bölte’s own findings.

The alarm bells regarding Bölte’s motivations are set off in its introduction, during a discussion about the various merits of the Wechsler Intelligence Scales (WIS) and the Raven’s Progressive Matrices (RPM). Bölte claims that Dawson “suggests that while RPM do allow fair IQ testing in autism, the WIS do not, which would dispute the utility of the WIS as the standard IQ measure for clinical or research purposes.” The problem with this claim is that I have been through Dawson at least twenty-five times (and went through it again just to make sure), and this so-called “suggestion,” let alone a direct statement, is never made. Absolutely nowhere in Dawson does the question of suitability of various intelligence scales for autism research and practice ever get raised—Bölte’s claim is no more than a canard. And although it might seem like a quibble of a canard, I have far too many times seen modern scientists approach their research work with purpose and outcome already set in stone, and here it appears as though Bölte’s authors have approached this particular study with the express intent of defending the use of WIS in autism research and practice, and are going to have that result no matter what, even if no one else has bothered to raise the issue, even if it means the facts be damned.

Which leads to the troubling aspect of Bölte’s main conclusion: Bölte’s facts be damned.

Bölte’s authors summarize their findings in the following way: “in conclusion, the claim that intelligence has been underestimated in autism seems somewhat premature.” They base this conclusion primarily on two pieces of evidence from their own study: 1. the average difference in WIS and RPM scores for autistic individuals, while significant, is less in Bölte than it is Dawson; and 2. in Bölte (unlike in Dawson), the difference in WIS and RPM scores for autistic individuals is noticeable only for those individuals with WIS scores less than 85.

Neither of these pieces of evidence supports the Bölte conclusion—in fact, exactly the opposite.

That the average difference in WIS and RPM scores for autistic individuals is less in Bölte than it is in Dawson is certainly interesting and warrants further investigation—it would be nice to obtain an accurate reading upon this number—but the far more important piece of information is that the difference continues to prevail in Bölte, and is statistically significant. So I would like to pose a question: just exactly how many times will Dawson need to be replicated before we can say it is not premature to claim that intelligence has been underestimated in autism? If the study gets re-performed and re-affirmed at least a hundred times, with the numbers coming out not exactly the same in each instance, will it still be necessary to hesitate about the Dawson conclusion? Do the Bölte authors volunteer to perform this study again and again and again? At what point do they propose to remove the word “premature”?

And if that were not bad enough, the other piece of evidence offered in support of Bölte’s summary, namely that the difference in WIS and RPM scores for autistic individuals is significantly large only for those individuals with WIS scores less than 85—well, that piece of evidence is even more damaging to the Bölte conclusion. Assuming that Bölte’s findings prove to be accurate, and that it is only for this particular autistic population that the WIS/RPM differences are significantly different, it is precisely this population that finds itself most at risk for having its intelligence underestimated. Dawson notes the precarious circumstances of this population in its Figure 3 chart, highlighting the cluster of autistic individuals in the upper left hand quadrant, the ones who by WIS measures would be labeled mentally impaired, but who by RPM scores are demonstrating strong fluid intelligence. Bölte does not provide the data to construct a similar chart, but based upon Bölte’s general findings, it is almost certain its upper left hand quadrant would display a similar cluster. So again, the question must be posed: do Bölte’s authors mean to claim that these autistic individuals are having their intelligence accurately described by such terms as “low-functioning” and “mentally retarded”? Exactly when would a reconsideration of such assessments not be “premature”?

Autism research never ceases to amaze me. I am told again and again that the purpose of such research is to help us better understand the autistic individuals within our midst and to allow us better to serve their needs. But here, on an occasion when the data (from two studies, no less) points to an instance where we have been clearly misunderstanding the characteristics of autistic individuals, the response from the autism research community seems to be a languorous effort to tamp those findings down, a ho-hum call to continue on as we were. Dawson is not an attempt to promote the use of RPM over WIS in autism research and practice; that at best would be a gross misreading. Dawson simply demonstrates that the level and nature of autistic intelligence has been poorly understood up to the present time—not just underestimated, but indeed mostly overlooked. In Bölte, there was an opportunity to contribute meaningfully to this discussion by not only refining and elaborating on the Dawson data, but also by noting how Bölte in general validates the Dawson theme—an opportunity Bölte’s authors apparently decided to forgo, because they must have had a different sort of purpose in mind.

For Dawson, et al., the Bölte report, despite its defiant tone, serves essentially as a confirmation of Dawson’s original findings. The difference in data is certainly worthy of note, and is probably deserving of a follow-up check; but Dawson’s assertion that the level and nature of autistic intelligence is poorly understood has now been verified (in more ways than one) by another team’s work.

And for Bölte, et al., although they have my thanks for having made the effort and for having been thorough in the reporting of methods and findings, my suggestion would be that for their next paper, before they move on to writing the conclusion, they should probably take the time to read over their own results.


Fleecy said...

That does sound like a pretty odd conclusion. Good post. I'm always glad to see new posts on this blog. :)

jonathan said...

One of the problems with your analysis is that Boelte used a different population than Dawson did. Dawson's work had a 12:1 ratio of males:females while Boelte had a 3:1 ratio. The population of Boelte were lower functioning lower IQ than those of Dawson. Had the population of Boelte been similar to that of Dawson there would have been no significant difference between the scores of this group between the Wechsler and RPM.

Dawson goes much further than stating that her work shows intelligence in autistics is underestimated. She refuses to acknowledge any limitations of her findings as makes the claim that she hopes her nonreplicated work will have the effect of canadian society not regarding autistics as write-offs.

Alan Griswold said...

I think I am going to let the Dawson, et al. study, the Bölte, et al. study, and Jonathan's comment (which in several instances does not seem to match what was said in the studies) speak for themselves. I can't think of anything useful to add.