According to the message, this conclusion was based on the results of a test that sounded an awful lot like the MMPI.
Beyond the clinical malpractice involved (the MMPI shouldn't be used as the sole basis for a diagnosis), there's the issue of the test's validity in assessing autism. Specifically, it isn't any more valid for that than it is for assessing brain damage.
I can say that with confidence now, thanks to Michelle Dawson. When I commented on the matter, explaining the MMPI and how it works, I added a comment that I wasn't aware of any attempts to validate the test for an autistic population. She then mentioned that there was a paper on the MMPI-2 and autistics and provided me with the citation. When I mentioned that my university didn't have access to the 2005 issues of the journal it was in (NSU does have access to the 2003 and 2004 issues), she e-mailed me a copy.
I've done a citation search and a reverse citation search on the paper as well -- and a keyword search in PsycInfo for good measure -- so I can say with a good degree of confidence that it's the only published paper on the topic. There may be more in press, of course, but that's not reference material that a clinical practitioner would have access to, barring highly unusual circumstances.
I estimate the odds of these circumstances (such as the clinician being on the research team for a follow-up study) to be somewhere between the odds of me getting struck by lightning within five minutes of finishing this blog entry and the odds of me winning the next lottery.
For the record, I haven't purchased any tickets.
The study itself is pretty interesting. I have my fair share of issues with the intro, of course, mainly due to the purely psychiatric/psychopathological approach that it takes to explaining what autism is. Others have written extensively on this topic and the sort of harm that this sort of thing can do, however, so I won't go there.
The experiment itself was actually pretty simple. They gave the MMPI-2 to 21 "high-functioning" autistics and Aspies, then did the same to a control group. Following this, they compared the resulting profiles.
I have my fair share of criticisms of what they did. For one thing, the sample sizes were far, far too small. The initial autistic group was composed of 21 adults (a mix of "high-functioning" autistics -- defined as those who could score in the normal range on the WASI -- and Aspies), which isn't even enough to expect a decent standard distribution, much less an adequate comparison between groups. Of these, one was disqualified from consideration (he left too many questions blank) meaning that the final autistic sample was only 20 people.
Similarly, the initial non-autistic sample consisted of 25 college students, recruited from an introductory psychology class. Of these, one was disqualified... apparently for trying to just answer everything with the same response regardless of what the question said. The TRIN scale (one of the MMPI's validity scales) measures attempts to do that.
The experimentors did try to match the samples, at least, and did so based on WASI results, age, and gender. Unfortunately, college students aren't representitive of the general population, and students from an intro psychology class are less so than most. I am not certain how this may have effected the study, mainly because I'm not sure about differences between college students and the general population on the MMPI. I will note, however, that this fact introduces confounds.
Fortunately, however, the authors acknowledge most of the above in their discussion.
Judging by the results, it looks like my initial concern (that the correlational relationships that the MMPI is based on might not hold true for an autistic population, thereby rendering the test results invalid) isn't the case, at least for adult, "high-functioning" autistics and for adult Aspies. That said, the reporting conventions are one area where I have... issue... with the study. Simply put, no statistical significance testing was conducted.
Fortunately, there's enough data reported that you could, theoretically, go back and conduct significance testing yourself. Unfortunately, this is a pain in the butt when you have to do it for 61 sets of data. I suppose that I could use a spreadsheet to do it for me, but it's still an annoyance... especially since it's standard practice to report this stuff.
The explanation given in the paper is as follows:
Because the sample sizes in this study were modest and this investigation is exploratory and descriptive in nature, effect sizes (e.g., the magnitude of group differences), ds, are reported, rather than statistical significance.
This makes sense, after a fashion, but it makes replication with a larger sample essential... and doesn't really excuse the annoyance. Given the way most statistical packages are set up, reporting the relevant data is roughly five minutes' work.
On the plus side, this has rather forcibly reminded me that I'm getting somewhat out of practice with calculating statistics and solving equations by hand.
So, that taken care of, what were the findings? On the core scales of the MMPI, the autistic sample scored higher on the tests for depression (even after five autistic participants who had a diagnosis of major depression were excluded), social introversion, and schizophrenia. This last would be more meaningful were it not for the fact that the MMPI's schizophrenia scale really measures social alienation and ways of thinking that are different from the norm. Autistics also typically scored lower on the hypomania scale.
One issue that this brings up is the issue of exogenous versus endogenous depression. The autistic sample was recruited from the University of Utah Child and Adolescent Specialties Clinic, meaning that all of them were treated there. This means that they all had at least one life event in common, and, in fact, the article mentions a second: all of them were diagnosed by either Dr. Sally Ozonoff or Dr. Janet E. Lainhart. The possiblity of a relationship is not considered in the paper... and the MMPI does not make a distinction between the two types of depression. Although, as a personality test, it is presumably designed to focus on depression as a personality trait, I know of at least one study that uses the MMPI to evaluate non-trait depression in the form of postpartum depression. While vulnerability to postpartum depression may be a trait, it is difficult to consider the depression itself to be such.
This is not to accuse Dr. Ozonoff or Dr. Lainhart of malpractice. I know very little about either of them... and less about their clinical practice. This is, in fact, precisely why I point out the above possiblity. The paper does not even begin to address this challenge.
On the content scales, the largest group difference was that autistics scored higher on the social discomfort scale. This... should not be a surprise. Their higher scores on the low self-esteem scale is probably related to a combination of environmental effects and the depression difference noted above. This is also backed up by an elevation of the negative treatment indicators scale, which measures physical and emotional distress.
I'd also like to mention an elevation in the work interference scale, which measures personal difficulties that interfere with work. This, too, is hardly news to the autistic community.
For the supplementary scales, the autistic group scored higher on the repression and anxiety scales. Again, this can easily be explained by life events. Certain nasty things, like bullying, tend to happen to us.
On the other hand, the autistic sample scored lower on the MacAndrew Alcoholism-Revised scale and the Addiction Potential scales, both of which measure vulnerability to substance abuse. Way to go, us!
On the other hand, this may be because we're painfully aware of the dangers thereof.
Then, we come to the ego strength scale, which is a bit tricky to explain. While it was originally intended to asess the degree to which a client can benefit from psychotherapy, this is by the standards of psychoanalytic theory. As such, it seems to measure adaptability, personal resources, tendency to seek help, ability to deal with others, and good physical health... among a host of other things. The autistic group scored lower, overall, on this measure... but, given the sheer variety of things that it assesses, I'm reluctant to ascribe meaning to this.
The rest of the results pertain to the revised clinical (RC) and PSY-5 scales. Unfortunately, I lack critical knowledge needed to properly interpret this and will be seeking information over the next week or so as I try to avoid spending the eighty bucks or so needed to buy a proper assessment book. I have an old copy of Graham's textbook on the MMPI (I picked it up at a book-sale), but it doesn't provide information on them. As such, I'll need to do some more searching.
Another interesting set of data provided is the percentage of the sample who provided clinically significant scores on the various scales. 40% of the sample provided clinically-significant elevations in the schizophrenia scale (I discuss the meaning of this briefly above, but suffice it to say that I doubt it means that the description of autism as "childhood schizophrenia" was even partially right)... and the post-traumatic stress disorder scale.
I don't believe I need to explain why that last bit is worrying.
The discussion for the paper was, overall, well-researched as far as I can tell (I'm not familiar with all of the papers cited), but every bit as... off... in tone as the intro.
And, of course, the article does absolutely nothing to change my opinion that the MMPI is completely inappropriate as a diagnostic measure for autism.