Structural Measurement Theory and Cognitive Development

Abstract

The overriding function of scientific theories is to reduce uncertainty about the world we live in by explaining how things work. A basic dilemma in psychological theories, probably the most basic dilemma of all, is that we do not possess what is sometimes called fundamental measurement control over the constructs that we use to explain behavior. On one hand, we regularly resort to notions such as short-term memory capacity, retrieval from long-term memory, attitudes, hypotheses, rule knowledge, preference, attention, motivation, intelligence, and the like when formulating explanations. On the other hand, these notions are not amenable to physical measurement operations such as weighing and counting. They lie somewhere in the uncharted region between true physical reality and metaphysical speculation. Consequently, we are not certain of how to go about quantifying them through experimentation. Here, a familiar catechism is that the most that can defensibly be assumed is that data are related to psychological constructs by unknown but order-preserving transformations. This is the familiar monotonicity constraint on the measurement of psychological constructs. To take a hoary instance, while we can presumably say that a person with a Stanford-Binet IQ of 150 is at least as intelligent as a person with a Stanford-Binet IQ of 100, we cannot say much more than this. And we certainly cannot say that the first person is half again as intelligent as the second.