Increases in morbidity and mortality have been observed consistently and coherently in association with ambient air pollution. A number of studies on short-term effects have identified ambient particles as a major pollutant in urban air. This study, conducted in Erfurt, Germany, investigated the association of mortality not only with ambient particles but also with gaseous pollutants and indicators of sources. Part I of this study concentrates on particles. Data were collected prospectively over a 3.5-year period from September 1995 to December 1998. Death certificates were obtained from the local authorities and aggregated to daily time series of total counts and counts for subgroups. In addition to standard data for particle mass with diameters < or = 2.5 microm (PM2.5)* or < or = 10 microm (PM10) from impactors, a mobile aerosol spectrometer (MAS) was used to obtain size-specific number and mass concentration data in six size classes between 0.01 microm and 2.5 microm. Particles smaller than 0.1 microm were labeled ultrafine particles (three size classes), and particles between 0.1 and 2.5 microm were termed fine particles (three size classes). Concentrations of the gases sulfur dioxide (SO2), nitrogen dioxide (NO2), and carbon monoxide (CO) were also measured. The daily average total number concentration was 18,000 particles/cm3 with 88% of particles below 0.1 pm and 58% below 0.03 microm in diameter. The average mass concentration (PM2.5) was 26 microg/m3; of this, 75% of particles were between 0.1 and 0.5 microm in diameter. Other average concentrations were 38 microg/m3 for PM10, 17 microg/m3 for SO2, 36 microg/m3 for NO2, and 600 microg/m3 for CO. Ambient air pollution demonstrated a strong seasonality with maximum concentrations in winter. Across the study period, fine particle mass decreased, whereas ultrafine particle number was unchanged. The proportion of ultrafine particles below 0.03 microm diameter increased compared with the proportion of other particles. During the study, concentrations of SO2 and CO also decreased, whereas the concentration of NO2 remained unchanged. The data were analyzed using Poisson regression techniques with generalized additive modeling (GAM) to allow nonparametric adjustment for the confounders. Both the best single-day lag and the overall association of multiple days fitted by a polynomial distributed lag model were used to assess the lag structure between air pollution and death. Mortality increased in association with level of ambient air pollution after adjustment for season, influenza epidemics, day of week, and weather. In the sensitivity analyses, the results proved stable against changes of the confounder model. We saw comparable associations for ultrafine and fine particles in a distributed lag model where the contribution of the previous 4 to 5 days was considered. Furthermore, the data suggest a somewhat more delayed association of ultrafine particles than of fine particles if single-day lags are considered. The associations tended to be stronger in winter than in summer and at ages below 70 years compared to ages above 70 years. Analysis of the prevalent diseases mentioned on death certificates revealed that the overall association for respiratory diseases was slightly stronger than for cardiovascular diseases. In two-pollutant models, associations of ultrafine and fine particles seemed to be largely independent of each other, and the risk was enhanced if both were considered at the same time. Furthermore, when the associations were summed for the six size classes between 0.01 and 2.5 microm, the overall association was clearly stronger than the associations of the individual size classes alone. Associations were observed for SO2, NO2, and CO with mortality despite low concentrations of these gases. These associations disappeared in two-pollutant models for NO2 and CO, but they remained stable for SO2. The persistence of the SO2 effect was interpreted as artifact, however, because the SO2 concentration was much below levels at which effects are usually expected. Furthermore, the results for SO2 were inconsistent with those from earlier studies conducted in Erfurt. We conclude that both fine particles (represented by particle mass) and ultrafine particles (represented by particle number) showed independent effects on mortality at ambient concentrations. Comparable associations for gaseous pollutants were interpreted as artifacts of collinearity with particles from the same sources.