Introduction

In the previous chapter, I have argued that the classical test theory model is unsatisfying for a number of reasons. Most important is the fact that the attribute to be measured is not adequately represented in the model. The reason for this is that the true score is an operationalist concept, and can only represent a psychological attribute if this attribute is similarly defined in an operationalist fashion. In fact, unless one holds a strongly operationalist view of the measurement process, it is difficult to maintain even that classical test theory is a theory of measurement in the first place.

A view of measurement that does represent the attribute explicitly in the model formulation can be based on latent variable theory. In latent variable models, one sets up a formal structure that relates test scores to the hypothesized attribute, deduces empirical implications of the model, and evaluates the adequacy of the model by examining the goodness of fit with respect to empirical data. Because the latent variable model has to be restricted to make empirical tests possible, a theoretical justification of the model structure is, in general, required. Latent variable theory thus goes beyond classical test theory in that it attempts to construct a hypothesis about the data-generating mechanism in which the attribute is explicitly represented as a latent variable.