BACKGROUND Most previous epidemiological studies have treated testicular cancer as a single entity. However, some investigators suggest that testicular seminomas and non-seminomas may have different risk profiles. We examine the time trends in incidence of the two main histological types separately. METHODS From 1970 through 1995, 7296 cases of testicular cancer were registered in the Canadian provinces of Ontario, Saskatchewan and British Columbia. In addition to analyses of the secular trends by age group and birth cohort, an age-period-cohort (APC) model with standard Poisson assumptions was fitted to the data to assess the time effects. RESULTS The age-adjusted incidence rate for seminomas increased by 53%, from 1.5 per 100 000 males in 1970-1971 to 2.3 per 100 000 males in 1994-1995. Non-seminomas increased by 91%, from 1.1 to 2.1 per 100 000 males over the same period. Non-seminomas were more frequent at young ages whereas seminomas dominated in older ages. In contrast to seminomas, non-seminomas occurred predominantly among adolescent men (15-19 years), with a fourfold increase between 1970-1971 and 1994-1995. Age-period-cohort modelling showed that the increase in the risk of both seminomas and non-seminomas followed a birth cohort pattern, but with differences in birth cohorts in addition to significantly distinct age patterns. CONCLUSIONS Our findings support the hypothesis postulating aetiological heterogeneity in the development of seminomas and non-seminomas. We suggest that epidemiological studies of testicular cancer treat seminomas and non-seminomas separately.