This article comments on the article "Behavioral Implications of Taxation," D. Larry Crumbley, published in the October 1973 issue of the journal "The Accounting Review." Crumbley considers some behavioral aspects of income taxation. He makes two major points. First, he suggests that research should be undertaken into the effects of tax laws on behavior. Second, he suggests that accountants should become involved. With respect to the first point, the purpose of this article is to indicate that Crumbley is recommending two distinctly different types of research. He does not always maintain a clear distinction between these two types in his discussion of behavioral tax research. With respect to the second point, the purpose of this article is to examine Crumbley's reasoning for his suggestion that accountants become involved in behavioral tax research. It has been suggested that the primary aim of behavioral science is to identify underlying regularities in human behavior and to determine what antecedent conditions give rise to them and what consequences follow from them. However, research which focuses primarily upon antecedent conditions and human behavior in an effort to predict consequences is very different from research which focuses primarily upon consequences and tries to infer the behavior that led to those consequences.

This article focuses on various models of auditing. One property of accounting numbers which has received a great deal of attention in the past few years is that of predictive ability. Research conducted to date has tended to view predictive ability as an impersonal criterion for the evaluation of accounting numbers. Behavioral reactions to particular sets of accounting numbers by particular decision-makers have typically not been investigated by predictive-ability researchers. However, a committee of the American Accounting Association recently introduced a model which, according to the committee, provides the potential for evaluating internal accounting numbers by examination of the utilization of those numbers by decision-makers. The purpose of this article is to provide a more complete discussion of the proposed model and, more importantly, to indicate potential hazards and other considerations in the implementation of the committee's proposal. The discussion draws heavily upon certain portions of the psychology literature which deal with the manner in which individuals utilize numerous pieces of data in making complex judgments or predictions.