Migrant populations comprise substantial numbers of individuals who have undergone a change in their environment, sociocultural and physical. The corresponding changes in risk for different cancers have, therefore, been widely used to infer the relative importance of environmental factors versus inherited predisposition in cancer aetiology. The uncontrolled experiment of migration also provides an indication of the possible effects of certain preventive interventions at the population level--especially with respect to diet. In the past, there has been a surprising lack of attention to analytical methods for migrant data, and we review the epidemiological methods available to best bring out the relevant differences in risk. The major sources of bias which confuse interpretation are also described. Migrant studies are classified into four groups, in a hierarchy corresponding to the amount of information which they can provide, and examples of each type are provided.