On relationships between numbers and behavior

Abstract

My title is a somewhat free paraphrase of the expressed aim of the Psychometric Society, namely, to further "the development of psychology as a quantitative rational science." I think we are all agreed that psychology has to do with behavior, that a quantitative discipline involves numbers, and that a rational science requires broad generalizations. The premises underlying psychometrics seem to be the following: 1. That behavior can be represented numerically; 2. That the resultant numbers can be manipulated according to mathematical principles; and 3. That the results of these manipulations can yield new and valid representations of behavior. These assumptions are made implicitly every time we quantify a set of observations, compute a descriptive statistic, and generalize to a population of cases not yet encountered. They are so much part and parcel of the psychometrician's kit of tools that only on rare occasions do we bother to scrutinize them. It is somewhat surprising, however, that the first premise, namely, that behavior can be represented numerically, is generally accepted with relatively few questions. The triumphs of mensuration in almost countless fields of human endeavor have led us to believe that in psychology, too, measurement has had or will have its triumphs. In this we have been encouraged by early successes: the measurement of reaction time, of mental age, of aptitudes, and of attitudes. Let us consider briefly, but not too rigorously, the three specific processes: 1. The representation of some aspect of behavior in a numeric code; 2. The summarization of the coded information in terms of one or more descriptive statistics; and 3. The utilization of the summary in the prediction and control of behavior.