Models of integration given multiple sources of information.
Dominic W. Massaro, Daniel Friedman · Psychological Review · 1990
Several models of information integration are developed and analyzed within the context ofa prototypical pattern-recognition task.The central concerns are whether the models prescribe maximally efficient (optimal) integration and to what extent the models are psychologically valid.Evaluation, integration, and decision processes are specified for each model.Important features are whether evaluation is noisy, whether integration follows Bayes's theorem, and whether decision consists of a criterion rule or a relative goodness rule.Simulations of the models and predictions of the results by the same models are carried out to provide a measure of identifiability or the extent to which the models can be distinguished from one another.The models are also contrasted against empirical results from tasks with 2 and 4 response alternatives and with graded responses. Conceptual FrameworkThere is a growing consensus that behavior reflects the influence of multiple sources of information.Auditory and visual perception, reading and speech perception, and decision making and judgment are modulated by a wide variety of influences (Anderson, 1981;Bruno & Cutting, 1988;Falmagne, 1985; Massaro, 1987a Massaro, , 1988a;;Oden, 1981; Perkell & Klatt, 1986;Welch & Warren, 1980).Until only recently, psychological inquiry was aimed at studying the relationship between behavior and a given single source independently of other sources of information.The common strategy was to eliminate or to hold constant all potential sources of information except the source of interest.This research strategy was most apparent in psychophysics but was also pervasive in perception, memory, and learning.The single-factor experiment was the dominant mode of investigation when one-dimensional functional relationships were the primary goal.Trying to understand behavior when multiple sources of information are available poses additional problems.Factorial experiments seem to be the most promising approach, and we have witnessed immense methodological and theoretical progress in this domain.Specifically, the additivefactor method developed by Sternberg (1969) and Anderson's (1970, 1981) functional measurement are milestones that will not be easily surpassed.Without these methodologies, there would have been a plethora of idle psychologists in the last couple of decades.True, Fisher (1935) bequeathed the statistical tools for factorial designs long before Anderson, Sternberg, and