Comparison Between Stochastic Models for Visual Selection

Abstract

A number of stochastic models for visual selection were formulated and tested against data from a partial report experiment. The data comprised 96 probability distributions of the numbers of correctly reported targets, one for each of 12 combinations of the number of targets and the number of distractors at each of 8 exposure durations. First, in modeling the limitations of the human visual information processing system, a fixed-capacity parallel model and a serial model were tested against the data from target-only displays. The former was found to give better fits. Second, in modeling visual selectivity, six fixed-capacity parallel models were tested against the partial report data. The best fitting model was the exponential fixed-capacity independent race model (eFIRM) proposed by Shibuya and Bundesen (1988) that assumes independent parallel processing, exponentially distributed processing times, limitations in both processing capacity and storage capacity, and time-invariant selectivity. The formulation of eFIRM was shown to be simpler and more useful than formulations of the serial models that are equivalent to eFIRM in the predictions.

I wish to thank Claus Bundesen and Axel Larsen for their suggestions and for encouragement throughout all phases of the preparation of this article. I also wish to thank the two anonymous reviewers for their valuable comments on an earlier version.