APS Observer - Letters

On the whole, I found the July/August isssue of the Observer enlightning as usual. However, I did recoil a bit from [APS President John] Darley's unbridled hubris over the experimental method of psychology. It's not true as he boasts that "experimentation is the one method that can unequivocally establish causality." The findings from all psychological experiments as far as I know are qualified by levels of confidence, and we can't be 100 percent confident of anything other than our own mortality. Good psychological experimentation merely makes some explanations less equivocal. Furthermore, if human nature could be fully explained by science, then there would be no need for jurisprudence, liberal education, the humanities, or theology. I would not want to live in such a sterile world.

There is something ad-hominem in John Furedy's letter on SAT I (Observer, July/August, 2001) in which he lectures Richard Atkinson (who like him is an "experimental psychologist") on the meaning of test validity about which "I recall enough from my undergraduate courses to recognize that the validity of a test. is arrived at. by determining to what extent performance on a test is correlated with some criterion performance." I expect Atkinson is aware of this definition of validity (we served several years together on an NIMH study section on Personality & Cognition), and I know that if Atkinson were to spell out the essence of a validity coefficient in a scientific journal, he would add that even an unusually high validity coefficient of .40 predicts only 16 percent of the total test variance (given that the test has a reliability r of .85), leaving 84 percent of the variance unaccounted for. Furedy correctly intimates this (though he is "far less distinguished" than Atkinson). But Furedy's observations are beside the point. What Atkinson has said publicly and in his Observer letter (April, 2001) is that test results can have a "devastating impact on self-esteem and aspirations and aspirations of young students." This has nothing to do with considerations of validity and reliability coefficients. Therefore, Atkinson is not questioning the usefulness of validity coefficients, and it is impertinent to suggest that he is naive about these constructs. But none of my points (nor those of Furedy) justifies a personal attack on the merits of a colleague's psychometric sophistication, unless of course, that colleague's career accomplishments are irksome.

The Observer welcomes comments and feedback. Letters may be sent by email to apsobserver@aps.washington.dc.us or faxed to 202-783-2083 (Attn: Observer Letters).

The Observer reserves the right to edit for length.