A cohort of 72,740 persons for whom information on household air-conditioning was available was monitored for mortality via the National Death Index from April 1980 through December 1985. A total of 2,275 deaths occurred among the members of this cohort. The basic question addressed was whether persons in households with air-conditioning experienced lower death rates during hot weather than persons in households without air-conditioning. This question was examined for both central and room air-conditioning. The analysis was based on a state-by-state approach, that cross-tabulated deaths by air-conditioning status (yes or no) and average temperature during the month of death (less than 21.2 degrees C (less than 70 degrees F) or greater than or equal to 21.2 degrees C (greater than or equal to 70 degrees F)). The Mantel-Haenszel and sign tests were used to summarize the data. For central air-conditioning versus no air-conditioning, statistically significant benefits (p less than 0.05, Mantel-Haenszel test) were observed for the overall total, for females, for persons not in the labor force, and for persons living in fewer than six rooms. These groups had more exposure to air-conditioning. The relative risk for the total group was 0.58, implying that in hot weather, the death rate for persons who had central air-conditioning was 42 percent lower than the rate for persons who did not have air-conditioning, after confounding variables had been controlled for. For room air-conditioning versus no air-conditioning, the odds ratio for the total group was 0.96, which was not significantly different from 1.0, suggesting that no real benefit was derived from room air-conditioning. Some reasons for the lack of a demonstrable benefit for room air-conditioning are given.