Modeling nonstationarity in perceptual matching

Patricia Alda Van Zandt, Purdue University

Abstract

Findings have been reported in the area of perceptual matching to suggest that the process by which information accumulates is nonstationary, i.e., changes with time. In particular, information about the similarities between two stimuli may build up more rapidly than information about their differences. This early-"same" hypothesis is quantified by representing the "same"/"different" response selection process as a race between two nonstationary Poisson processes. Three experiments are presented to determine the appropriateness of this model. Despite the findings in the literature, no empirical evidence was uncovered to support the early-"same" hypothesis. It was discovered that a model in which the rates of information accumulation were constant, i.e., stationary, fit the data as well as or better than the nonstationary models that instantiated the early-"same" hypothesis. The superiority of the stationary model was evident even in conditions where nonstationary processing was forced. The stationary model was able to account for families of RT data across conditions of bias by manipulating only bias parameters, and similarly across perceptual changes by manipulating only rate parameters. The stationary model is therefore a viable model of response selection in perceptual matching, and generalizations of this model appear to be unnecessary.

Degree

Ph.D.

Advisors

Proctor, Purdue University.

Subject Area

Psychology|Experiments

COinS