Population migration occurs for many reasons, or combinations of reasons. Subsequently, it is often useful to distinguish streams of migrants, within a given migration system, which are likely to exhibit similar characteristics in their pattern of search and choice of residential destination. This has often been achieved by using distance cutoffs to delimit the various flows into categories. In this paper, an approach is suggested which does not disaggregate migrant flows by distance criteria alone, but incorporates measures relating to the nature of the origins and of the destinations. Relatively homogeneous groups of migrants are therefore distinguished within the county of Hereford and Worcester with information from three dimensions rather than one. In comparison with a standard single-stream regression model this method improves the fit substantially, and the variability in the resulting parameter estimates for each of the eight streams supports the need for identification of distinct migrant streams. In particular, flows over short distances between wards with high population densities are estimated more effectively.