ABSTRACT
In recent years there has been growing uneasiness with the use of statistical methods in psychological research when applied to groups of subjects in a search for causal relations between initial conditions and the subsequent outcomes of subjecting people to those conditions. This has come about for various reasons. First of all there has been the growing awareness of the prevalence of elementary statistical fallacies in the interpretation of results. For instance it is quite common to find a conclusion about individual propensities drawn from a statistical distribution of attributes over a population to which those individuals belong. Then there has been the turn from looking for group averages towards the search for individual cognitive procedures, whether in the form of abstract AI models or of more concrete cases of symbol using. Complaints about the loss of idiographic data have been voiced for at least the last twenty years. Finally there has been the changing character of psychological “experiments”. The trend has been more and more to the use of questionnaires and commentaries on vignettes as ways of exploring many aspects of psychological functioning. The traditional practice of interpreting correlations revealed by factor analysis of the answers to questionnaires as indicative of causes has been strongly criticised. However, we believe that a judicious combination of statistical analyses using data expressed in numerical form, and semantic and narratological interpretations can be a very powerful method of revealing the sources of regularities in psychological phenomena, combining the virtues of numerical analysis while avoiding the errors of a blanket and unexamined assumption of a causal metaphysics.
