We examined the existence of thresholds, cumulative effects and the homogeneity of five air pollutants on the relative risk of three mortality outcomes using data from nine major US cities using data from NMMAPS. Overall, PM(10) (usually 200-day accumulation) and ozone (3-day accumulation) were the two important predictors of outcome but their effect was not uniform across the nine cities. Many models exhibited thresholds (25-45 microm g/m(3) for PM(10), and 10-45 ppb for O(3)). Our preliminary exploratory analyses suggest that the use of a linear, no threshold, model for pollution studies is not consistent with the observed data. The heterogeneity in the risk estimates across the nine cities suggests combining the local risk estimates to obtain a national risk estimate may not be justifiable and the estimate is likely to be confounded.