OBJECTIVE To study the seasonal variation in mortality for all causes, grouped according to age and sex, within the city of Valencia during the period 1976-1990 and its relationship with air temperature. METHODS This is an ecological study using monthly global mortality and mean atmospheric temperature statistics for the city of Valencia, Spain between 1976 and 1990. The principal variable has turned out to be mortality for all causes. Mortality rates have been determined for specific age groups (0-4, 35-49, 50-64, 65-74 and > or = 75), as well as for both genders, within the city of Valencia for each of the 180-month period of investigation. Average monthly mortality data has been obtained and the percentage variation has been calculated. Graphical analysis has been used to examine the seasonal variation in mortality and in order to discern the nature of any relationship between atmospheric temperature and mortality. The regression line has been adjusted using the Loess method (smoothed regression motive, locally pondered). The association between mean temperature and mortality has calculated by means of multiple regression analysis controlling for trend, seasonality and period effect. RESULTS A clear seasonal pattern emerges with mortality peaking during the Winter months and dipping during the Summer and early Autumn. The increase in mortality seen during the month of January came to 27.7% of the total mortality for females and 19.5% of the total mortality for males. By age groups, this increase was greater in the 75-plus age group. Graphically, the relationship between temperature and mortality has a V-shape appearance, with a lower mortality level when the mean daily temperature for the month in question is approximately 23 degrees Celsius. Based on this relationship, two complementary.