Using ZIP code-level mortality data, the association of cardiovascular mortality with PM(2.5) and PM(10-2.5), measured at a central monitoring site, was determined for three populations at different distances from the monitoring site but with similar numbers of deaths and therefore similar statistical power. The % risk and statistical significance for the association of mortality with PM(2.5) fell off with distance from the monitor, as would be expected if exposure error increased with distance. However, the % risk for PM(10-2.5) increased in going from the population in Central Phoenix, where the monitoring site was located, to a population in a Middle Ring around Phoenix and fell off in an Outer Ring population. The % risks for the Outer Ring were low for each of the six lag days (0-5) and for the 6-day moving average. The lag structures for PM(2.5) and PM(10-2.5) also differed for the Central Phoenix and Middle Ring populations. These differences led us to examine the socioeconomic status (SES) of the populations. On the basis of education and income, the population in Central Phoenix had a lower SES than the Middle Ring. Thus, the differences between Central Phoenix and the Middle Ring may be due to effect modification by SES and differences in exposure error. However, the effect modification by SES may be different for thoracic coarse particulate matter (PM) than for fine PM. This study provides new information on the association of PM(10-2.5) with cardiovascular mortality. In the Middle Ring, the % risk per 10 microg/m3 increase in PM(10-2.5) concentration (lower and upper 95% confidence levels) for lag day 1 was 3.4 (1.0, 5.8) and for the 6-day distributed-lag was 3.8 (0.3, 7.5). The differences in lag structure for PM(2.5) and PM(10-2.5) provide evidence that the two particle size classes have health effects that are different and independent. This study also helps explain the high % risks for PM(2.5) found for Central Phoenix, 6.6 (1.1, 12.5) for lag day 1, and 11.5 (2.8, 20.9) for the 6-day moving average. The smaller area may have a lower exposure error, and the lower SES population may be more susceptible to fine PM as compared to the larger areas and more heterogeneous populations used in many studies.