In Reply--We concur with Dr Kirkland that subgroups of exposure are of less interest than. total exposure to ETS and, as noted in our report, some exposures are more likely to be accurately recalled and reported than others. However, to estimate the risk of exposure to ETS, it is necessary to ask specific questions about exposures in a variety of settings during different time periods and then to summarize those data as we did in Table 5 for all childhood exposures combined and in Tables 7 and 8 for all adulthood exposures combined. The study was desig1led to collect data on a finite number of subgroups and to combine the subgroup data to categorize study subjects into levels of exposure ranging from little or none to lengthy. The unit of exposure was "years," which does not presume a level of precision that is unachievable (e.g., hours per day per year). To inquire only about a spouse's smoking habit or about exposure in the workplace would provide a glimpse of one part of the total picture in which individuals are actually exposed to greater and lesser degrees at home, in the workplace, in carpools, during bridge games, and so forth. Estimates of risk in those various settings were provided in the report, but were certainly not the main focus. As noted on page 1758, "[a] positive dose response between ETS exposure during adult life and lung cancer risk was found when individual sources of exposure, such as household, occupational, and social settings, were examined separately, and this pattern of risk was clearest when these exposure sources were considered jointly. The point estimates are somewhat higher for exposures in occupational and social settings than within households, but these differences are not statistically significant." While exposures in occupational and social settings may be equally or more important than household exposures because the concentration of ETS, a function of the number of active smokers, may be much higher and the opportunity to reduce the concentration by opening a window may not exist, our data suggest great consistency because the confidence intervals around these estimates are all virtually super-imposable and all point estimates are consistent with a small to moderate effect. We likewise concur with Dr Kirkland as to the main conclusion: long-term exposure to ETS increases the risk of lung cancer in women who have never used tobacco.