In developed countries in temperate climates, daily mortality is higher during the winter season than in the rest of the year. In spite of numerous analyses of short-term effects of weather on mortality, questions remain concerning the magnitude and form of the effect of temperature and of weather more generally on total and cause-specific mortality. We have used a two-stage regression approach to address the effects of temperature and dew point on mortality during the period 1973-1994 in 11 large eastern cities: Chicago, Boston, Philadelphia, Baltimore, Washington, D.C., Charlotte, Atlanta, Jacksonville, Tampa, and Miami. In the first stage of the analysis, data for the individual cities were fit with regression models that described the relationship between temperature and mortality within each city. As found by others, a V-shaped relationship was observed between temperature and mortality, but with substantial variation among cities. The V-shaped relationship was characterized by the slope of the relationship at colder temperatures (Cold Slope) and the relationship between temperature and mortality at hotter temperatures (Hot Slope). In the second stage of the analysis, latitude was found to be a significant factor in determining the Cold and Hot Slopes. With increasing latitude, the temperature at which the minimum mortality was observed dropped; that is Cold Slope became less steep, and Hot Slope became more steep. This pattern suggests that the effects of temperature on mortality are differential by latitude, with colder temperatures having less effect with increasing latitude and warmer temperatures having a greater effect.