This study outlines the derivation of a behavioural model intended to underpin thinking on questions concerning the design of information retrieval systems for academic social scientists. The historical background of information retrieval research is reviewed and the behavioural assumptions made in that research, characterised as the information retrieval model, are examined. Developments in the research tradition following from the early empirical tests on information retrieval systems are discussed, and problems experienced in the attempt to develop a coherent and practically useful research programme around the notion of relevance as a quantitative concept are analysed. A major alternative to work employing the information retrieval model has been the cognitive approach to information retrieval system design. In research so far undertaken in this approach the system has had to construct a cognitive model of the searcher's requirements in order for retrieval to take place. It is argued that the characteristics of social science information present particular difficulties for building such a model, and that the step of the system building a cognitive model of the searcher's requirements as a pre—requisite for retrieval is unnecessary. Instead, a behavioural approach to system design is recommended. The information seeking patterns of a variety of academic social scientists, derived from transcripts of interviews, were analysed and broken down into six characteristics, starting, chaining, browsing, differentiating, monitoring, and extracting. These characteristics seemed sufficient to exhaust the different generic features of the various patterns, and to provide a flexible behavioural model to underpin thinking about information retrieval system design. The features of an exploratory retrieval system for social scientists based on the characteristics of their information seeking patterns are outlined, the extent to which such features are available in existing systems, and the requirements for implementing the model on an operational system are considered. The argument of the study is that the behavioural approach, and the analysis of information seeking patterns into their characteristics, represents a coherent and practically useful basis for the design of exploratory information retrieval systems for academic social scientists.