In environmental epidemiology, short-term effects of ambient air pollution on mortality are explored using time-series analysis including various definitions of the exposure variable. In order to find out if and to what extent the magnitude of air pollution-related relative risk of death depends on the definition of the exposure variable, we analyzed data on daily mortality and air pollution concentrations obtained in the city of Katowice in 2001-02 (range of daily number of deaths: 17-76; range of 24-hour concentrations in μg/m³: PM₁₀ = 11.2-421.3, SO₂ = 10.5-239.8, NO_x = 15.7-287.7). The modeling results confirmed the dominant role of SO₂ among the monitored ambient air pollutants, after adjustment for meteorological variables. The value of SO_2- related relative risk of death (total mortality) depended on the definition of exposure variable – for same-day concentrations of SO₂ it was 1.007, and for a three-day moving average it was 1.012. The largest values of risk estimates were provided by exposure variables expressed as a 40-day moving average (SO₂-related relative risk = 1.022). Our findings highlight the importance of the choice of the model (including definition of exposure variables) in exploring time-series mortality data. On biological grounds our findings suggest that people at risk of death (i.e. elderly with cardiorespiratory disorders) could be more affected by an accumulating burden of exposure (expressed by average air pollution levels over a longer period) than by acute exposures to increasing air pollution levels.