REVISED AND EXPANDED EDITION 43
niques (still vigorously in vogue and eminently useful): factor analy-
sis. I had learned this procedure as an abstract mathematical theory
and had applied factor analysis to the study of growth and evolution
in various fossil organisms (for example, my Ph.D. thesis, published
in 1969, on Bermudian land snails; and one of my first papers,
published in 1967, on growth and form in pelycosaurian reptiles—
those peculiar creatures with sails on their backs, always included in
sets of plastic dinosaurs, but really ancestors of mammals and not
dinosaurs at all).
Factor analysis allows one to find common axes influencing sets
of independently measured variables. For example, as an animal
grows, most bones get longer—so general increase in size acts as a
common factor behind the positive correlations measured for the
length of bones in a series of organisms varying from small to large
within a species. This example is trivial. In a more complex case,
subject to numerous interpretations, we generally measure positive
correlations among mental tests given to the same person—that is,
in general and with many exceptions, people who do well on one
kind of test tend to do well on others. Factor analysis might detect a
general axis that can, in a mathematical sense, capture a common
element in this joint variation among tests.
I had spent a year learning the intricacies of factor analysis. I
was then historically naive, and never dreamed that such a valuable
abstraction, which I had applied only to fossils with minimal political
import, might have arisen in a social context to tout a particular
theory of mental functioning with definite political meaning. Then
one day I was reading, quite aimlessly and only for leisure, an article
about the history of mental testing, and I realized that Spearman's
g—the central claim of the unitary theory of intelligence, and the
only justification that such a notion has ever had (The Bell Curve is
fundamentally a long defense of g, explicitly so stated)—was noth-
ing more than the first principal component of a factor analysis of
mental tests. Moreover, I learned that Spearman had invented the
technique of factor analysis specifically to study the underlying basis
of positive correlation among tests. I also knew that principal com-
ponents of factor analyses are mathematical abstractions, not empir-
ical realities—and that every matrix subject to factor analysis can be
represented just as well by other components with different mean-
ings, depending on the style of factor analysis applied in a particular