Examples of controlled averages

from Phil Runkel on 18 April 99
in response to Marken's of same date at 16:35 my same time belt,
concerning the article by Rushton et alii sent by Isaac.

Thanks for your comments. One part especially tickled me:

        Rushton present, in their Figure 2b, an excellent
        description of the expected behavior of alpha (the
        hypothetical controlled variable if subjects are
        controlling retinal target location) as the subject
        walks to the target with prism glasses on. The
        prediction is that alpha will be constant (at 16
        degrees) over the course of the walk. The actual
        behavior of alpha for 5 different subjects is shown
        in their Figure 3; in only one case was the behavior
        of alpha anywhere close to being a constant (and in
        this case alpha was close to zero rather
        than 16 degrees, an indication that this subject was
        _not_ controlling the retinal location of the target).
        Rushton et al. conclude that the subjects were
        controlling alpha (at 16 degrees) because the _average_
        value of alpha (over 5 subjects) over the course of the
        walk was _nearly_ constant at _nearly_ 16 degrees. So
        Rushton et al. are violating the first commandment
        of behavioral research: thou shalt not base conclusions
        about individual behavior on group averages.

Do I read right? Because the average is so-and-so, therefore every
individual wanted to be at that average? That is the kind of thinking I
believe I see over and over again in the literature. I wrote in
Casting Nets about treating correlations like that. But this is the most
explicit statement I have seen yet.

But maybe those are your words, not theirs. I'll go look up their words
in the library.

Thanks again. --Phil R.