Studies of space-time clustering of cases of childhood leukaemia have yielded equivocal results. This might be because the disease has a long and variable latent period, in which case the usual statistical tests for such clustering are inappropriate. A new statistical method is described which allows for such latent periods. For each patient, periods of "susceptibility" and "infectivity" are defined in which it is assumed he respectively "caught" and could "transmit" the disease. The measure of clustering is taken as the number of patients who were in the "right" place at the "right" time to "catch" the disease from another patient. This test is applied to childhood acute lymphoblastic leukaemia (death before age 6) in Greater London in the period 1952-65. Cases are postulated to be "susceptible" at various times before clinical onset of leukaemia, including in utero, and "infective" at various times around onset. Their effective "contacts" at these times are defined as circles of radius up to 4 km around their places of residence at these times. Slight evidence of clustering was found associated with certain of the defined times and distances, but the degree of clustering was small and could reasonably be attributed to chance. It is suggested, however, that this method of analysis might usefully be applied to other sets of such data. No evidence was found to add to our previously reported finding of space-time clustering of the dates and places of birth of children with leukaemia.