Classification Experiments For
One-Dimensional Stimulus Sets
Successful participants in discrimination experiments can distinguish two stimulus classes, but in most paradigms they need not be able to name them. In this chapter, we extend detection theory to encompass experiments in which stimuli drawn from large sets are named or classified by the observer. These sets are “one-dimensional, ” that is, they contain stimuli that differ for the participant in just one characteristic. As in earlier chapters, we are interested in sensitivity and bias, but multiple parameters must be estimated, and their interpretation is somewhat different.
One-dimensional classification experiments are among the oldest psychophysical tasks, and they take on many aliases. Accordingly, this chapter has an unusually large number of examples, but one basic strategy for data analysis fits all.
Inclassification experiments, observers use M responses to sort N stimuli into categories. If there are two stimuli and two responses (N = M =2), the task is the familiar one-interval yes-no discrimination. If there are more possible stimuli than responses (N> M), the design is traditionally called category scaling, but is now often called categorization. We consider the important special case in which M= 2 in detail first. When N equals M but both are greater than two, the experiment is absolute judgment, absolute identification, or simply identification', the second part of the chapter concerns this task.
Classification experiments can be modified by the addition of a standard stimulus. The stimuli being judged are called comparisons, and a (standard, comparison) pair is offered on each trial. The presence of standards makes