Many studies have examined associations between the geographical patterns of disease and possible causal factors. Underlying such studies is the assumption that residence in an area can be equated with exposure to conditions that pertain there. Migration upsets this assumption. Using data on self-reported morbidity from the 1981 Census for Great Britain it is shown that the health status of migrants differs considerably from that of non-migrants. Young migrants, particularly those moving longer distances are relatively healthy. Area of net out-migration of this age group are therefore likely to become characterised by a less healthy population whereas the reverse will be true of areas of net in-migration. Amongst older people migration tends to be selective of those in poor health moving shorter distances to avoid environmental health hazards or to be closer to medical care. This has the curious effect of increasing morbidity and mortality rates in areas with favourable environmental conditions and good medical services. Migration is therefore likely to be a significant source of error in geographically based studies of the association between disease and the environment. Sometimes the effect will be to obscure real environmental causes of disease. In other circumstances spurious association will be produced. These problems will be greatest for diseases with long latent periods which allow plenty of time for mobility of the population. Problems are also likely to increase with the growing mobility of the population.