OK, to clarify what I did with the eyes, which is derived from Simon-Baron Cohens test.
First, I just run the original test in Aspie Quiz to see if it had any relevance for adults. The result on that trial is more or less that it lack relevance, and that the Aspie and neurotypical group gave almost similar answers.
Next, I thought I might get better results if I selected better alternatives, as I suspect that Simon-Baron Cohen just selected non-correct alternatives more or less randomly. Thus, I created a list of 85 different emotions, and let people select any emotion for any eye-picture. Then I eliminated alternatives by using the alternatives that best discriminated between the groups in an iteractive process. I got down to 16, 10, 8, 6, 4 and eventually 3 alternatives. While I got better relevance, and up to 16% differences between the groups on some eye-pictures, the final version really didn't do much better than the original. About half of the eye-pictures got "staring" as an alternative (all these alternatives had higher scores in the Aspie population).
The score does not depend on which alternatives people selected on the eyes part. That's also why they are not part of the pdf. And since none of the eyes had any correlation to ordinary Aspie Quiz questions, there will be no eyes in Aspie Quiz. The initial conclusion that such a test lack relevance for adults stands even with optimal alternatives.