Measures of compensatory and noncompensatory models of decision behavior: Process tracing versus policy capturing

While a variety of techniques have been used to infer compensatory versus noncompensatory decision making, few studies have used multiple measures in order to evaluate their validity. Using a sample of 48 college students, this study examines two measures from a search task (the variability and pattern of search on an information board) and three measures from a rating or judgment task (nonlinear regression modeling, ANOVA measures of interaction and curvilinearity). The validity of these measures is assessed by their sensitivity to a manipulation of information load and their extent of convergence with one another. The variability of search on the information board and the ANOVA measures of interaction and curvilinearity all indicated an increase in noncompensatory decision making under high information load, while the regression modeling measure did not. There was some convergence between the regression and ANOVA indices, but no relation between the search task and rating task measures. It is concluded that the ANOVA measures of interaction and curvilinearity are more sensitive measures than the nonlinear regression procedure. The information board and ANOVA measures are apparently both valid indices of noncompensatory decision making; they may lack convergence because they represent different parts of the decision process (information acquisition vs combination) or require different responses (choice vs judgment).