The Relationship of Problem-Solving Styles to the Factor Composition of Tests1

WHEN an ability or achievement test is constructed, its author usually assumes, or a t least hopes, that his creation will measure the same thing for all examinees. An often talked about exception to this desirable quality of a test is the case of a fourth grader and a college student who take the same test of addition. We think such a test is one of reasoning for the younger subject, while it is one of speed for the older one. Some specific data on this appears in Balinski (1941). Large age differences, however, are not the only differences that affect the nature of the factor being measured by a test. Some tests of higher mental processes are solved in one way by some subjects and in another may by other subjects (Bloom and Broder, 1950, Ch. 3; Lucas, 1953). This means that the tests may be measuring different abilities for some subjects than for others, and so it follows that tests mill have different correlations with one another for different kinds of groups of subjects. Therefore, since factor loadings of the tests depend on these correlations, “the factor loadings cannot be expected to be invariant from one population to a different population” (Thurstone, 1947, p. 360). It seems possible that the factorial composition of test problems involving higher mental processes often appears complex, not only because the problems require several different kinds of abilities in their solution, but also because they