This paper considers the possibility of prediction in land use planning, and the use of statistical research methods in analyses of relationships between urban form and travel behaviour. Influential writers within the tradition of critical realism reject the possibility of predicting social phenomena. This position is fundamentally problematic to public planning. Without at least some ability to predict the likely consequences of different proposals, the justification for public sector intervention into market mechanisms will be frail. Statistical methods like regression analyses (...) are commonly seen as necessary in order to identify aggregate-level effects of policy measures, but are questioned by many advocates of critical realist ontology. Using research into the relationship between urban structure and travel as an example, the paper discusses relevant research methods and the kinds of prediction necessary and possible in spatial planning of urban development. Finally, the political implications of positions within theory of science rejecting the possibility of predictions about social phenomena are addressed. (shrink)