BACKGROUND: Recent studies indicate that cancers of the salivary gland are increasing, and the factors responsible for the increase are unknown. Artefactual changes, such as shift in classifying cancers of the floor of the mouth to cancers of the salivary gland, could affect the time trend for salivary gland cancer. METHODS: The current study examined the time trends for cancers of the salivary gland and for cancers of the floor of the mouth and lower gum by using Connecticut Tumor Registry data for the time period 1935-1992. A regression model was used to identify the components of birth cohort, period and age as determinants of the observed time trend. RESULTS: Cancers of the salivary gland have recently increased in Connecticut, with a relative risk of 1.48 (95% CI: 1.06-2.08) for females in 1990-1992 compared to 1980-1984, and a comparable relative risk of 1.60 (95% CI: 1.16-2.22) for males. The increase was found in all age groups 40 and over, particularly among those aged 70 and over. The results from age-period-cohort modelling show a recent upturn in the trend for period slopes, with no clear increase from recent birth cohorts, which is consistent with the results from univariate analyses suggesting no clear increase among those under 40 years of age. CONCLUSION: Our results suggest that artifactual changes, such as a shift in designation of cancer sites, increasing use of the needle aspirate biopsies, and greater access to medical care for the elderly, may have largely contributed to the rising trend. The known risk factors, radiation exposure and a history of a prior cancer, can hardly explain the observed increase. The Epstein-Barr virus infection has only been associated with certain types of rare squamous cell carcinomas of the salivary gland in the Eskimo population. The AIDS epidemic also cannot explain why older age groups have accounted for most of the increase in incidence of the disease. An examination of the incidence rates for cancers of the salivary gland from other populations may help to clarify the issue.