How can we explain and predict human behaviour and, in particular, behavioural regularities in a social setting? Adequate answers to questions of this type have been of interest to influential traditions in the social sciences and in particular to sociology, economics, anthropology, and psychology.² The main concern of the present paper is a particular type of answer — let us call it a "rationalistic answer' — to the above question. Rationalistic answers have been especially influential in economics, where an elaborate theoretical structure was built on a particular conception of human behaviour. The coherence, elegance and apparent problem-solving ability of the theoretical framework of neoclassical economics is probably responsible for the longevity and relative immunity of the rationalistic approach from incisive criticisms which have been levelled against it in the last few decades.³