This paper introduces procedures for plotting and analysing partial residuals for binary logit choice models. Seen within the general context of a movement away from purely inferential statistical model-building towards more ‘data-analytical’ or exploratory approaches, these graphical methods are presented as offering significant advantages over established methods of identifying the functional forms of discrete choice models. An empirical example is developed in the context of revealed tenure choices in London, England, in which a number of revisions are made to the functional form in the light of the analysis of partial residual plots. These results are discussed in relation to the dynamic and policy-dominated British housing market.