Like Richard Atkinson, I am an experimental psychologist (though a far less distinguished one) rather than a differential psychologist who has specialized in psychological test theory. However, I recall enough from my undergraduate courses to recognize that the validity of a test is assessed not by speculating that it "can have a devastating impact on self-esteem and aspirations of young students" ("Use of SAT I 'Compromises Education System' Says UC President", Observer, April 2001), but by determining to what extent performance on the test is correlated with some defined criterion performance (here, academic success in a prestigeous university).
Nor is this correlation with criterion performance expected to be perfect, so that there may well be factors other than sheer cognitive ability in analogical reasoning (factors such as socio-economic class, home environment, and, of course, motivation - recall that living organisms and not computers are being tested) that contribute to test performance. In terms of this normal, scientific criterion of validity, the SAT I, to my knowledge, is a useful instrument, and specialists in psychological test construction have, over the years, improved its validity, though not to any level of perfect prediction. So from the perspective of psychological test theory, I see no rational grounds for Atkinson's recommendation to abandon the SAT I.
Atkinson also advances a more general, educational argument for dropping the SAT I. He avers that it "compromises the education system", and, besides the SAT II (which, he feels, is a better measure than the SAT I - to my knowledge he advances no systematic evidence for this comparative empirical claim about two psychological tests), he suggests that selectors should rely on "grade point average, activity records, and other more 'holistic' measures of students' achievement".
I cannot help noting that the latter two aspects appear to be more related to how well a student can get along with others, rather than to what extent s/he has been able to master various academic disciplines.
Moreover, the North American high school system lacks state-wide standard examinations as exist, for example, in Australia. Grade points, then, are at least partly determined by how much individual teachers like individual students, and hence, in more crude terms, may simply indicate sucking-up, rather than academic, ability. In my view, it is the use of these more subjective and "holistic" measures of student achievement (together with race- and sex-based quotas intended to produce 'diversity') that really "compromise the education system".