If infections are involved in the aetiology of childhood leukaemia then seasonal variation in the birth or onset dates of the malignancy may be apparent. Previous studies that have examined seasonality of these dates have produced conflicting results. Using population-based data from the National Registry of Childhood Tumours we conducted a larger study than any to date of 15 835 cases of childhood leukaemia born and diagnosed in the UK between 1953-95. We found no evidence of seasonality in either month of birth or month of diagnosis overall or in any subgroups by age, sex, histology or immunophenotype. We did however find a significant (P = 0.01) February peak in month of birth for cases born before 1960 and a significant (P = 0.02) August peak in month of diagnosis for those diagnosed before 1962. Whilst these findings may be due to chance they are also consistent with changes over time in the seasonality of exposure, or immunological response, to a relevant infection. Changes in the seasonal variation in the fatality rate of a pre-leukaemic illness, such as pneumonia, could be another explanation.