Aetiological factors known to influence the devel opment of leukaemia in man include ionizing radiation and chromosomal abnormalities (Miller, 1964). Viral agents have been demonstrated in laboratory animals and although virus-like particles and mycoplasma have been found in affected marrow cells (Negroni, 1964; Grist and Fallon, 1964) attempts to show a direct causal relationship have so far been unsuccessful in man. Epidemiological studies have become of increasing importance, and interest in the distribution has been stimulated by reports of clustering within various communities. The outbreak in one parish in Niles, Illinois, in which eight children developed the disease within 4 years, was thought to be due to an infectious cause, especially as there was an outbreak of a rheumatic-like illness during the same period (Heath and Hasterlik, 1963). Other reports of smaller clusterings (Wood, 1960; Pinkel and Nefzger, 1959) were more open to interpretation as being due to chance distribution, although infection could not be excluded as a contributory factor. Studies of varia tion in San Francisco (Mustacchi, 1965) supported the view that cases tend to occur in a non-random way referable to the presence of previous cases. As the incidence of leukaemia in the population is very low, it is necessary to develop statistical techniques of suitable sensitivity and validity to detect clustering, if it exists, especially within communities where the overall incidence is not obviously different from that expected. Ideally the same method should be applied to as many sets of comparable data as possible. The procedure of Knox (1963), which was first used to study the epidemiology of childhood leukaemia in Northumberland and Durham (Knox, 1964) and later in Oregon (Meighan and Knox, 1965), has recently been assessed indepen dently (David and Barton, 1966) and shown to be an effective method of revealing space-time interaction. The object of the present study is to examine data from the large conurbation centred on the port of Liverpool where clustering was not apparent.