Andrew Tomarken, PhD Vanderbilt University School of Medicine
The study of individual differences has historically been a major focus of psychological research, with recent extensions to the domains of cognitive and perception psychology. An implicit if not explicit assumption of such work is that there are stable differences between people in overall levels of behavior that represent the individual’s “central tendency”. I will discuss our recent research on individual differences in binocular rivalry that is designed to test several overriding hypotheses: (1) There are stable, subject-specific distributions of behavior; (2) Non-parametric measures of distributional overlap can be successfully used to assess such individual differences; (3) Increasing aggregation serves to sharpen predictions from sub-groups to the individual; and, (4) Distributional features represent a composite, subject-specific signature that contributes to accurate identification and classification at the individual level. A major focus of discussion will be on several methodological issues raised by the study of individual differences in distributions. These include the measurement of distributional overlap, coping with several consequences of an intensive repeated measures design, and the selection of an appropriate strategy for classification when the ratio of number of possible predictors to sample size is relatively high. .