We appreciate the comments of Mr. Langer and Dr. Xia and colleagues. They highlight important issues on placement and phrasing of sexual behavior questions in population surveys and contribute to the limited body of population-level data reflecting discordance between sexual identity and sexual behavior. Population-based data presented by Xia and colleagues for comparative purposes are based on a probability sample of men who have sex with men, not a sample of all men (1), and a sample for which both sexual identity and sexual behavior were not ascertained in the same individuals (2).

Our 2003 New York City Community Health Survey (CHS) used a definition of "sex" that has been validated by other research and has been used by other large-scale surveys, such as the National Health and Nutrition Examination Survey (NHANES) (3). Like the NHANES, the CHS provided the definition once to participants and then asked them about sexual partners. Unlike the NHANES, the CHS asked both men and women first about male sex partners and second about female sex partners. As discussed in our paper, we considered the possibility that asking male participants about male before female sexual partners could have affected respondents' answers, artificially increasing our estimates of men who have sex with men. Arguably, however, offering the opposite sex as sexual partner first may introduce a bias toward a socially desirable answer (4). As we state in our article, our data deserve further investigation.

Indeed, the ordering of partner questions was changed for the 2005 CHS, such that male participants were asked first about female sexual partners and then about male sexual partners. Preliminary data from the 2005 CHS suggest a lower estimate of male same-sex behavior compared with that obtained from the 2003 survey and a lower proportion of men who have sex with men who identify themselves as straight. Of note, however, fewer men overall identified themselves as straight and more were unsure of their sexual identity in 2005 than in 2003, although the sexual identity question did not change between the 2 years. The exclusion of these men from constructed identity–behavior groups affected the estimate of straight-identified men who have sex with men. To date, we do not have evidence of "true" prevalence. With additional data from 2005 and 2006, we will be uniquely poised to evaluate the reliability of answers using various ordering of questions.

Although the CHIS and CHS are not directly comparable (questions to ascertain number and sex of partners on the 2 surveys are not identical), we were interested to see the very low prevalence of straight-identified men who have sex with men in California. Even with the change in question order in the 2005 CHS, the prevalence of straight-identified men who have sex with men in New York City in 2005 was still higher than that in CHIS. Observed differences between the 2 surveys may also be related to true differences in the populations studied. We look forward to further investigations to better understand these differences.

Regardless of the "true" proportion of men who have sex with men who identify themselves as straight, our large city and other metropolitan areas have sizable (>10%) male populations who report same-sex behavior. Some straight-identified men who have sex with men will not benefit from public health messages that target gay-identified men. Medical providers must learn to inquire about sexual behavior rather than identity to serve their patients well.