The aim of this study was to investigate seasonal variation in the incidence of cancer in children aged 0-14 years. Details of 2959 primary malignant cases (1659 males, 1300 females), diagnosed during the period 1968-2005, were extracted from a specialist registry (the Northern Region Young Persons' Malignant Disease Registry). Seasonal variation was analysed with respect to month of birth and diagnosis. The chi-squared heterogeneity test was used to test for non-uniform variation. Poisson regression analysis was used to fit sinusoidal (harmonic) models to the data, using month of birth and month of diagnosis, respectively, as covariates in separate models. There was significant sinusoidal variation based on month of birth for acute lymphoblastic leukaemia (ALL) aged 1-6 years (P = 0.04; peak in March). For 0- to 14-year-old boys, there was statistically significant sinusoidal variation in month of birth for acute non-lymphocytic leukaemia (P = 0.04; peak in September) and astrocytoma (P = 0.03; peak in October). Based on month of diagnosis, there was statistically significant sinusoidal variation in girls for all lymphomas (P = 0.048; peak in March) and Hodgkin lymphoma (HL) (P = 0.005; peak in January), and in boys for osteosarcoma (P = 0.049; peak in October). This study confirms previous findings of seasonal variation around the month of birth for childhood ALL (at the peak ages) and provides further evidence of seasonal variation around month of birth for astrocytoma and around month of diagnosis for HL. The results are consistent with a role for environmental factors in the aetiology of these diagnostic groups. Further studies are needed to examine putative candidate agents.