Around the world, daily variations in ambient air pollution have been consistently associated with variations in daily mortality. The aim of the study presented here was to assess the effects of ambient air pollution on daily mortality during a period of tremendous changes in air quality in the city of Erfurt, in eastern Germany, from October 1991 to March 2002. Data on particle size distributions were obtained from September 1995 to March 2002 at a research monitoring station. For particles from 0.01 microm to 2.5 microm in diameter, number concentrations (NCs)* and mass concentrations (MCs) were calculated. Particles with diameters less than or equal to 0.10 microm are defined as ultrafine particles (UFP). Data on the gaseous pollutants NO2, CO, SO2, and O3 and on PM10 (particulate matter [PM] with aerodynamic diameter less than or equal to 10 microm) were obtained from a government air-monitoring station. Data on changes in energy consumption, car fleet composition, and population were collected from local authorities. Death certificates of persons living in and dying in Erfurt were abstracted, and daily mortality counts were calculated. Poisson regression models were used to analyze the data, applying penalized splines (also known as P-splines) to model nonlinear relationships in the confounders. Model selection was done without air pollutants in the models, based on a combination of goodness-of-fit criteria and avoidance of autocorrelation in error terms. Final models included P-splines of time trend, meteorologic data, and influenza epidemics as well as day of the week with an indicator variable. Results are presented as change per interquartile range (IQR), i.e., change in the relative risk of mortality associated with a change in the concentration from the 25th to the 75th percentile of a given pollutant. Air pollutants were considered both as linear terms and as P-splines to assess the exposure-response functions. Changes in effect estimates over time were calculated using fully Bayesian time-varying coefficient models. This method was selected over four other approaches tested in simulation studies. Air-pollution concentrations decreased substantially in Erfurt during the decade under observation. The strongest changes were observed for SO2, for which annual concentrations decreased from 64 microg/m3 in 1992 to 4 microg/m3 in 2001. Concentrations of PM10, PM2.5 (particulate matter with aerodynamic diameter less than or equal to 2.5 microm), and CO decreased by more than 50%. NO2, O3, and ultrafine particles also decreased, though to a lesser extent. Based on visual inspection of the data on the changes in ambient air-pollution concentrations during the study period, we defined three study subperiods: A first subperiod from 1991 to 1995; a second, transitional subperiod from 1995 to 1998; and a third subperiod from 1998 to 2002. Generally, air-pollution concentrations decreased substantially from the first subperiod to the second, and some additional decreases occurred from the second subperiod to the third. During the second, transitional subperiod, natural gas replaced coal as the main energy source in Erfurt. In addition, the number of cars with catalytic converters increased over time, as did the number of cars in general. To facilitate the interpretation of the results, we organized the air pollutants into four groups: (1) NO2, CO, and ultrafine particles, (2) PM10 and PM2.5, (3) SO2, and (4) O3. We observed a 1.6% increased risk for daily mortality (CI, -0.4% to 3.5%) for an increase of 19.7 microg/m3 in NO2 (lag day 3), a 1.9% increased risk (CI, 0.2%-3.6%) for an increase of 0.48 mg/m3 in CO (lag day 4), and a 2.9% increased risk (CI, 0.3%-5.5%) for an increase of 9743/cm3 in ultrafine particles (lag day 4). No consistent associations were observed for PM10, PM2.5, or SO2. For O3, a 4.6% increased risk for daily mortality (CI, 1.1%-8.3%) was associated with a 43.8 microg/m3 maximum 8-hr concentration of O3 per day (lag day 2). For all four pollutants, exposure-response functions suggested no deviation from linearity. However, in time-varying models the strongest associations were observed for NO2, CO, and ultrafine particles during the transition subperiod, from 1995 to 1998, when O3 concentrations were lowest. Changes in source characteristics or ambient air-pollution concentrations were not able to explain these observations in a straightforward manner. However, the observations suggested that changes such as the introduction of three-way catalytic converters in cars and the substitution natural gas for coal might have been beneficial. Overall we concluded that: 1. Economic and political changes and the adoption of new technologies in eastern Germany resulted in distinct improvements in ambient air quality; 2. Urban air pollution in Erfurt changed within one decade from the eastern mixture toward that of western Europe ("western mixture"), which is dominated by concentrations of NOx, O3, fine particles, and ultrafine particles with low concentrations of SO2; 3. There was an association between daily mortality and ultrafine particles and combustion-related gases (lag days 3 or 4); 4. Ultrafine particles seemed to be the best pollution indicator and to point to the role of local combustion in the pollution mixture; 5. Regression coefficients showed variation over time for NO2, CO, ultrafine particles, and O3 that could not be explained by nonlinearity in the exposure-response functions; 6. Mortality associated with pollution was lower at the end of the 1990s than during the 1990s, except for mortality associated with O3; and 7. Mortality associated with pollution was strongest in the second, transitional subperiod, from 1995 to 1998, when changes in source characteristics had taken place but the benefits of improved ambient air quality had not yet been completely achieved.