Following recent claims in social psychology, it is suggested that many hypotheses in the discipline do not lend themselves to empirical investigation. Certain studies in attribution theory are particularly vulnerable to this criticism. From this perspective, it is claimed that Heider's naive analysis of action should be taken as an a priori explication of common sense accounting practices, rather than as a basis for experimental study. A computer model, NANA CT, of the naive analysis of action is presented and this is used to demonstrate that several attribution theory results follow logically from Heider's analysis. It is argued that analytic investigation into such a system is preferable to the empirical investigation currently fashionable. Furthermore, the increasing availability of computers and modeling software means there is now an appropriate representational technology within which to carry out such investigations.