Michael M. Cohen Dominic W. Massaro Program in Experimental Psychology University of California Santa Cruz, CA
Most chapters in this book describe probabilistic multidimensional models of perceptual and cognitive tasks. This fact acknowledges several important properties of human information processing. First, performance is not deterministic, but is variable or probabilistic. A subject responds in one way to a stimulus on one trial and responds in another way on another trial. Thus, performance is often characterized by some probability value representing overall response probability to a repeated presentation of a stimulus. Probabilistic performance might result from probabilistic differences in processing, probabilistic differences in the physical stimulus information from trial to trial, or probabilistic representations of prototype items in memory. Second, the term multidimensional refers to the multiple sources of information that influence perception and cognition ( Massaro & Cohen, 1991; Massaro & Friedman, 1990).
In a previous paper, Massaro & Friedman ( 1990) presented and compared various existing models of how multiple sources of information influence perception and decision. The question they addressed was how individuals process two or more sources of information that may reinforce or conflict to various degrees. The models were analyzed in terms of a prototypical pattern recognition task and the application of extant models to this task. The central concerns were the processes assumed by the models, the similarities and differences in predictions of the models, their optimality properties, and empirical validity.
Our goal in this chapter is to extend the analyses carried out by Massaro and Friedman ( 1990) by comparing several additional classes of models. The models analyzed and tested by Massaro and Friedman were the fuzzy logical model of