Exemplar similarity, study list homogeneity, and short-term perceptual recognition

Kahana and Sekuler (2002) conducted short-term perceptual recognition experiments and modeled the data with a noisy exemplar similarity model. They found model-based evidence that list homogeneity (i.e., the degree to which exemplars on a study list are similar to one another) exerted a significant impact on recognition performance--a finding that is not predicted by standard global familiarity models. A potential limitation of their experiments is that they tested complex stimuli in which psychological similarities among exemplars may have been misspecified. Also, the relative importance of list homogeneity was not compared with that of alternative forms of parametric variation in the model. We conducted conceptual replications of their experiments, using a simpler set of stimuli in which interexemplar similarities could be more precisely measured. Extensive model-based comparisons reveal, in accord with the results of Kahana and Sekuler, strong evidence for a role of listhomogeneity on old-new recognition performance. We suggest that subjects systematically adjust their response criteria on the basis of the homogeneity of the study list items.