A brief telephone interview to identify lesbian and bisexual women in random digit dialing sampling.

Lesbian health research has relied on non-probability (sometimes called convenience) samples that restrict the generalizability of their results (Committee on Lesbian Health Research Priorities, 1999). Nonprobability samples may be biased because they include highly motivated volunteers. When these volunteers are recruited from sources in the gay community, as is often the case, sampling probability is associated with levels of affiliation in the community. The resultant bias is often unclear: When service organizations (e.g., 12-step groups) are oversampled, samples may overrepresent respondents with poor health outcomes; when activist organizations are oversampled, samples may overrepresent respondents with better than average health outcomes (Harry, 1986; Meyer & Colten, 1999; Sell & Petrulio, 1996). Studies of lesbians typically overrepresent predominantly white, highly educated women, between 25 and 40 years of age. This led the Institute of Medicine's (IOM) Committee on Lesbian Health Research Priorities to call upon researchers to develop more sophisticated methodologies and to conduct research that represents the diverse population of lesbian and bisexual women. In particular, the IOM called for research that would "explore the feasibility of using different sampling techniques, by themselves or in combination, for rare or hard-to-identify population subgroups, in order to obtain a probability sample of the lesbian population subgroup" (Committee on Lesbian Health Research Priorities, 1999), p. 12). The purpose of the present study was to explore one such approach to sampling lesbian and bisexual women.

Random digit dialing (RDD) is a valuable and economical technique that has been used in studies of gay and bisexual men and has promise for use in lesbian research (Sudman, 1976; Waksberg, 1978). However, large-scale probability sampling with rare and difficult-to-find populations is costly, making more efficient sampling approaches necessary (Committee on Lesbian Health Research Priorities, 1999). One approach may use a 2-phase study design. At Phase 1, the target population would be screened for sexual orientation, thus identifying lesbian and bisexual women for the sample. Then, at Phase 2, an extensive interview would be administered, covering the study's substantive areas. Either a telephone interview or another methodology (e.g., a mailed self-administered survey) may be used at the second phase. For a Phase 1 screen to be cost effective it must be brief and reliable (Committee on Lesbian Health Research Priorities, 1999). For that purpose, a questionnaire needs to be developed that can be administered over the phone to a large number of women in the briefest possible time.

But is it possible to inquire about sexual matters in a brief telephone survey? Conventional wisdom and survey research methodology suggest that it may not be possible for sex research in general, and for sex studies of women in particular. For example, developers of the National Health and Social Life Survey found it essential that an interviewer develop rapport with the interviewee, putting her at ease by starting the interview with nonthreatening questions and "only after this gentle preparation begin to ask detailed questions about sex partners and sexual practices" (Laumann, Michael, & Michaels, 1994, p. 69). Such an approach would be inefficient for studies that seek community samples of lesbian and bisexual women. If researchers conducted the whole substantive interview before they determined the respondents' sexuality, a very large number of women, most of them heterosexual, would be interviewed even if only a small number of women were required for the study. For example, in the 1996 National Household Survey of Drug Abuse, of 5,888 women who had at least one sexual partner in the year prior to the interview, only 96 (1.8%) reported any same-gender partner (Cochran & Mays, 2000). Clearly, conducting a full interview before one identifies the lesbian sample is not an efficient method for studies that focus on lesbians.

Results of studies that used RDD techniques to assess sexuality and sexual behavior in random samples are encouraging. The National AIDS Behavioral Survey demonstrated a high cooperation rate (70%) among male and female respondents despite the sensitive subject matter of the survey (Binson et al., 1995; Catania, Coates, & Kegeles, 1992; Stall et al., 1992). Meyer and Colten (1999) tested a brief questionnaire to screen sexual orientation in a random sample of New York City gay men and found that it was effective: The measure was valid, and the method yielded a sample that was significantly different from a comparison sample they recruited from a gay community center and similar to a random sample of the general population. We tested Meyer and Colten's brief screen methodology in a sample of women.

METHODS

we used three criteria to identify a neighborhood that we expected to have a high density of lesbian residents, thus allowing us to test the study hypotheses most economically. We defined the neighborhood in terms of a postal ZIP code so that potential respondents could easily identify whether or not they lived in the target area. The criteria we used were: (a) a high proportion of never-married females aged 35 years or older (calculated as ratio of ZIP code area to city wide, U.S. Department of Commerce, 1990), (b) a high proportion of female-headed households who also reported an unmarried female partner in the household (U.S. Department of Commerce, Bureau of the Census, 1990), and (c) a high proportion of female patients from the ZIP code area among all Fenway Community Health Center female patients (Fenway Community Health Center is a major health service provider to gay and lesbian populations in Boston and Cambridge [Mayer et al., 2001]). These criteria led us to select the Jamaica Plain neighborhood in Boston. Women were eligible if they resided in that area, were between the ages of 18 and 59 years, and spoke English well enough to be able to answer the interview questions.

The Survey and Evaluation Research Laboratory at Virginia Commonwealth University conducted the survey. We purchased samples of random digit numbers from Genesys, Inc. Female interviewers used the Computer Assisted Survey Execution System (CASES) to conduct the interviews. Interviews were conducted between September 15, 2000, and November 11, 2000. Potential respondents were given a toll-free phone number to the principal investigator and were invited to call if they wanted to verify that the caller was calling for the stated purpose of conducting research or if they wanted to address any other issue. Respondents were also given a phone number to the Institutional Review Board at Columbia University, which approved the study. Three women used the number to call the principal investigator to verify that the call was related to a research project; none indicated that the sexual questions prompted the call, and none complained about the study's topic.

Phone interviews lasted a mean of 5.6 minutes (SD = 1.5). Respondents who identified as lesbian or bisexual completed an additional specialized section that lasted a mean of 2.5 minutes (SD = 0.7) and inquired about participation in and identification with the gay/lesbian community.

Sample

In total, 1,250 numbers were dialed. Of them, 169 (14%) were nonworking numbers, 165 (13%) were not households (e.g., businesses), 235 (19%) were indeterminable (number never answered by person), and 681 (54%) were households. Of these households, 439 (64%) were successfully screened, 176 (26%) refused or delayed screening, and 66 (10%) could not be screened (e.g., language barriers). Of screened households, 223 (51%) were not eligible. Of 216 eligible households, 202 (94%) women completed the interview. Using the American Association for Public Opinion Research (AAPOR) formula RR3 (AAPOR, 2000), the estimated overall response rate for the survey was 58%. This represents the ratio of completions out of the number of eligible households where the number of eligible households is adjusted to include an estimated number of households whose eligibility was not determined. The estimation of eligibility assumes that the proportion of eligible households among indeterminable households is the same as that found among successfully screened households.

Measures

Sexual orientation. Following (Laumann et al., 1994) we operationalized sexual orientation in three dimensions: behavior, attraction, and identity. To assess behavior, respondents were asked whether their sexual partners for the past year, past 5 years, and since age 18 were male only, female only, or both male and female. To assess attraction, respondents were asked whether they were "sexually attracted to only men, mostly men, both men and women, mostly women, or only women." To assess identity, respondents were asked whether they "consider [themselves] straight or heterosexual, lesbian or gay, bisexual, or something else."

Health status. We used the SF-12, a multipurpose short form generic measure of health status developed by The Health Institute at New England Medical Center (Ware, Kosinski, & Keller, 1994). Questions inquire about both physical and mental health and well-being. Responses were scored according to the authors' instructions to derive summary physical and mental health scores. These scores are standardized to the U.S. population and transformed so that both subscales have a mean of 50 (SD = 10).

Participation in the lesbian/gay community. Respondents who identified as lesbians or bisexuals were asked whether they were affiliated with lesbian/gay organizations, and whether they read or subscribed to lesbian/gay newspapers or magazines. A single subjective item asked respondents how connected to the lesbian community they felt, rating the response on a scale ranging from 0 to 10, with 0 being not at all and 10 being very much. Degree of being out of the closet was evaluated by asking respondents if they were out to none, some, most, or all of their family; gay, lesbian, or bisexual friends; straight friends; coworkers; and health care providers.

Internalized, homophobia. This measure was assessed with a 6-item Liken scale that measured the extent to which lesbians were uneasy about their homosexuality and sought to avoid homosexual feelings (adapted from Martin & Dean, 1987).

Attitudes about sex-related issues. We used two attitude questions from the National Health and Social Life Survey (Laumann et al., 1994) about "a married or coupled person having sexual relations with someone other than her or his partner" and about "sexual relations between two adults of the same sex." Respondents were asked whether they thought this was always wrong, almost always wrong, wrong only sometimes, or not wrong at all.

Hypotheses

We hypothesized that our method would be effective at sampling lesbian/bisexual women in the community, so that (a) few if any women would refuse to participate or avoid questions about sexuality; (b) the prevalence of lesbians and bisexual women (measured by self-identity) would be equal to, or higher than, the 2.6% estimated for major cities in the U.S. (Laumann et al., 1994); and (c) the sample of lesbians/bisexual women would be similar to the sample of heterosexual women in demographic characteristics, health outcomes, and opinions.

Statistical Analysis

The SPSS statistical package (version 10) was used for data analysis. To test associations among variables we used t-tests for comparison of means of normally distributed continuous variables; the Mann-Whitney test for comparison of medians when assumption of normality was not met; and Pearson's chi-square statistic for dichotomized variables.

RESULTS

Table 1 shows demographic characteristics of heterosexual and lesbian/bisexual identified respondents. As compared with the heterosexual women, lesbian/bisexual respondents were significantly younger, more of them were unemployed, and they tended to be more educated; the groups did not differ in race or Latina ethnicity. Lesbian/bisexual women did not differ from heterosexual women in physical health summary score, but had significantly lower scores on the mental health summary scale. (These associations did not change when controlling for age and education in multiple regression analyses.) The groups differed in attitudes regarding extramarital and same-sex sex, with the heterosexual women expressing somewhat more conservative opinions than the lesbian/bisexual women.

Structure of Sexuality

Since age 18, 9 (5%) of the women reported that they had sex with women only, 31 (15%) with both women and men, and 156 (77%) had sex with men only; 3 (1%) women had no sexual contact at all, and 3 (1%) refused or didn't know how to answer this question. Of the 40 women who reported any sex with women since age 18, during the 1-year period prior to the interview, 14 (35%) had sex with men only, 19 (47%) with women only, 5 (12%) with both men and women, and 2 (5%) had no sexual contact at all. In all, in the 1-year period prior to the interview, 153 (76%) women reported heterosexual sexual contact only and 25 (12%) reported same-sex contact, including 19 (9%) who reported sex with women only and 6 (3%) who reported sex with both men and women. 23 (11%) reported no sexual contact at all, and one woman refused to answer this question.

Regardless of current sexual behavior, 67 (33%) of the women reported some nonheterosexual sexual attraction. Half (34) of these women reported that they were attracted mostly, but not only, to men, 12 (6%) to both men and women, 10 (5%) mostly to women, and 11 (5%) reported that they were attracted to women only. Finally, 19 (9%) of the women identified as lesbian or gay, 7 (3%) identified as bisexual, and 2 (1%) reported other homosexual identity; 166 (82%) reported that they considered themselves heterosexual or straight, and 3 (1%) reported other heterosexual identity. Three women (1%) didn't know and 2 (1%) refused to answer the question.

We were particularly interested in patterns of avoidance (don't know or refuse) of the five sensitive sexual questions. Overall, seven women (3%) avoided at least one of the questions on sexual behavior (3 women), sexual attraction (1 woman), and sexual identity (5 women); only 1 avoided (refused) to answer all five questions.

Figure 1 depicts the interaction among sex behavior (since age 18), attraction, and identity for 68 women who reported a nonheterosexual response to at least one of these categories. For only 25 (37%) of the women, nonheterosexual orientation converged. The majority of the women (29, or 43%) reported some attraction to but no sexual behavior with other women, and fewer (13, or 19%) reported same-sex sexual behavior with no lesbian/bisexual identity.

[FIGURE 1 OMITTED]

Affiliation With the Gay Community--Lesbian Versus Bisexual Women

Twenty-six women who identified as lesbian (n = 17) or bisexual (n = 9) were asked about their relationship to the lesbian community. Table 2 shows that lesbians were more likely than bisexuals to be out of the closet and that they tended to report more connection to the gay/lesbian community.

DISCUSSION

We set out to test the feasibility of using a quick telephone interview to screen women in the community and to identify lesbians and bisexual women. We hypothesized that few women, if any, would refuse to participate or would avoid questions about sexuality. Conventional beliefs among survey methodologists suggest that women would be reluctant to disclose information about their sexuality, in particular about their same-sex sexuality, to a stranger over the telephone. Our results refute these beliefs. Our telephone interview lasted less than 6 minutes, not allowing for much rapport between the interviewer and respondent. Nevertheless, 94% of screened eligible households cooperated with the survey, demonstrating that women were willing to respond to questions about their homosexuality. Very few women (3%) avoided the sexual questions and none used a toll-tree number provided by the researchers to complain about the nature of these questions.

To test the feasibility of our method, we selected a Boston neighborhood where we expected a large populace of lesbian or bisexual women. We hypothesized that the prevalence of lesbians and bisexual women (measured by self-identity) would be equal to, or higher than, the 2.6% estimated for major cities in the U.S. (Laumann et al., 1994). Our expectation was confirmed: 33% of the women reported at least some sexual attraction to other women, 20% reported sex with women since age 18, and 14% reported a lesbian sexual identity. That we captured such a high prevalence of nonheterosexual women suggests that our strategy worked: These prevalences are much higher than would be expected in a city-wide survey. It should be stressed, however, that we did not attempt to estimate population prevalence, and therefore did not adjust our estimates for household selection or other common sampling errors.

Finally, we hypothesized that the sample of lesbians/bisexual women would be similar to the sample of heterosexual women in demographic characteristics, health outcomes, and opinions. One of the handicaps of convenience samples of lesbians is that they tend to overrepresent women who are better educated, white, and younger than the populations from which they are supposedly drawn. Our sampling method allows comparison with a heterosexual population in the same neighborhood. Presuming that the true demographic distribution in the neighborhood is similar between these two groups, our method improved on the race/ethnic representation that is typical of lesbian studies; however, lesbians in our sample were significantly younger and a marginal probability suggested they may have been more educated. This is consistent with findings of other random and convenience samples (Committee on Lesbian Health Research Priorities, 1999; Laumann et al., 1994; Morris & Rothblum, 1999) and may be due to at least two factors: (a) more educated and younger women are less reluctant to disclose same-sex orientation, and (b) nonheterosexual sexuality is structured differently among less educated women, so that our method of inquiry was not reliable in measuring these expressions of sexuality. Existing studies of lesbian and bisexual women limit our understanding of the scope of lesbian identities (Bradford, Ryan, & Rothblum, 1994; Kitzinger & Wilkinson, 1995; Morgan, 1992; Skinner & Otis, 1996; Viss & Burn, 1992). Further research on the structure of sexuality and the language used to describe sexuality in women from different cultures and demographic groups will help explore this.

Another handicap of convenience samples is that they overrepresent women who are highly affiliated in the gay/lesbian community. This is an integral part of the sampling technique because when researchers seek volunteers in the community, level of affiliation therein is related to probability of being sampled. An RDD sampling method reduces this bias because respondents are contacted at random. The small number of lesbian/bisexual respondents in our study limits our ability to examine variability among them, but our results show that RDD sampling promises to yield a diverse group of women who varied in how out they were and in their level of affiliation with the lesbian community.

Our results are limited in that we have surveyed one neighborhood selected because it was unique and therefore unrepresentative of most areas. The neighborhood may be more permissive and allow self-disclosure more freely. Lesbian and bisexual women who live in less-permissive neighborhoods may be less likely to disclose personal sexual matter in a telephone interview. If this was the case our method may not be as easily transferred to other less-tolerant populations. We addressed this by testing whether attitudes of women in the neighborhood were tolerant and progressive. We found that heterosexual women were more conservative than the lesbian/bisexual women: more disapproving of homosexual relationships and of extramarital sex. This suggests that heterosexual and lesbian/bisexual women, although from the same neighborhood, do not share values about sexuality, and that the women who disclosed same-sex sexuality volunteered information that could be stigmatizing in their cultural environment. That close to half of the lesbian/bisexual women indicated that they are not completely out to coworkers and family members attests to that. That they nevertheless disclosed their sexual orientation to the telephone interviewer attests to the effectiveness of the methodology.

The development of fast, efficient, and reliable research methodologies is essential for health research on lesbians and bisexual women to progress (Committee on Lesbian Health Research Priorities, 1999). We presented one method that may be used in a two-phase screening, where lesbians are identified in a quick and economical manner, to be followed later by more probing research interviews on the substantive topic of the research. Although further studies need to test our research methodology in a broader population, our results suggest that the brief telephone survey is efficient in eliciting sensitive and potentially stigmatizing information about same-sex sexuality among women.

Table 1. Demographic Characteristics of Randomly Selected Heterosexual
and Lesbian/Bisexual Women in a Select Boston Neighborhood (N = 197)

Characteristic                              Heterosexual (N = 169)

Age [mean (SD)]                                    36 (9.72)
Education--4-year college or more                  69%
Unemployed                                          2%
Nonwhite                                           33%
Latina                                             14%
Physical health score [mean (SD)]                  51 (9.2)
Mental health score [mean (SD)]                    49 (9.5)
Attitude toward sex with someone
  other than spouse or
  partner (a) [mean (SD)]                           1.8 (0.9)
Attitude toward sexual relations
  with same-sex adult (a) [mean (SD)]               3.6 (1.0)

Characteristic                             Lesbian/Bisexual (N = 28)

Age [mean (SD)]                                    32 (9.46)
Education--4-year college or more                  86%
Unemployed                                         11%
Nonwhite                                           25%
Latina                                             11%
Physical health score [mean (SD)]                  53 (7.2)
Mental health score [mean (SD)]                    44 (11.3)
Attitude toward sex with someone
  other than spouse or
  partner (a) [mean (SD)]                           2.5 (0.8)
Attitude toward sexual relations
  with same-sex adult (a) [mean (SD)]               3.9 (0.3)

Characteristic                                     Statistic

Age [mean (SD)]                           t = 2.29, df = 194, p = .02
Education--4-year college or more          [Chi.sup.2] = 3.41, p = .06
Unemployed                                 [Chi.sup.2] = 4.88, p = .03
Nonwhite                                   [Chi.sup.2] = 0.66, p = .41
Latina                                     [Chi.sup.2] = 0.25, p = .62
Physical health score [mean (SD)]        t = -0.84, df = 189, p = .40
Mental health score [mean (SD)]           t = 2.56, df = 189, p = .01
Attitude toward sex with someone
  other than spouse or
  partner (a) [mean (SD)]                t = -3.35, df = 180, p = .001
Attitude toward sexual relations
  with same-sex adult (a) [mean (SD)]    t = -1.93, df = 184, p = .05

Note. Percent respondents and means (SD) indicated. Five women of the
sample's total of 202 women who did not disclose sexual identity were
excluded from these analyses. Table Ns vary due to missing values.

(a) Scale of 1, always wrong, to 4, not wrong at all.
Table 2. Affiliation With the Gay/Lesbian Community for a Random Sample
of Lesbian and Bisexual Women in a Select Boston Neighborhood (N = 28).

                                          All       Lesbians (a)
Characteristic                         (N = 28)       (N = 19)

Belong to lesbian/gay
  organization                          9 (32%)        8 (42%)
Regularly read gay media               14 (50%)       11 (58%)
Out to all
    Family                             15 (54%)       11 (58%)
    LGBT friends                       22 (79%)       18 (95%)
    Straight friends                   20 (71%)       15 (79%)
    Coworkers                          14 (52%)       13 (68%)
    Healthcare providers               18 (67%)       15 (79%)
Connectedness to the
  community [mean (SD)] (b)             7.0 (2.3)      7.8 (1.3)
Internalized homophobia [mean (SD)]     3.7 (0.4)      3.7 (0.4)

                                       Bisexual (a)
Characteristic                          (N = 7)

Belong to lesbian/gay
  organization                          1 (14%)
Regularly read gay media                3 (43%)
Out to all
    Family                              2 (29%)
    LGBT friends                        2 (29%)
    Straight friends                    3 (43%)
    Coworkers                           1 (17%)
    Healthcare providers                3 (50%)
Connectedness to the
  community [mean (SD)] (b)             5.7 (3.1)
Internalized homophobia [mean (SD)]     3.7 (0.3)

                                       Statistic for differences
                                          between lesbian and
                                             bisexual women

Characteristic

Belong to lesbian/gay
  organization                         [Chi.sup.2] = 1.75, p = .19
Regularly read gay media               [Chi.sup.2] = 0.40, p = .53
Out to all
    Family                             [Chi.sup.2] = 1.76, p = .18
    LGBT friends                       [Chi.sup.2] = 12.62, p < .001
    Straight friends                   [Chi.sup.2] = 3.13, p = .08
    Coworkers                          [Chi.sup.2] = 4.96, p = .03
    Healthcare providers               [Chi.sup.2] = 1.89, p =. 17
Connectedness to the
  community [mean (SD)] (b)                      Z = 1.62, p = .11
Internalized homophobia [mean (SD)]              t = 0.13, p = .9

(a) Two women were excluded from these analyses because they identified
as "other, lesbian, or bisexual" and their identity as either lesbian
or bisexual could not be differentiated. (b) Mann-Whitney test for
comparison of medians was used because the assumption of normality of
distribution is not met for the bisexual women. Means are provided for
descriptive purposes.

REFERENCES

American Association for Public Opinion Research. (2000). Standard definitions: Final dispositions of case codes and outcome rates for surveys. (Available on line: http ://www.aapor.org/ethics/stddef.html).

Binson, D., Michaels, S., Stall, R., Coates, T.J., Gagnon, J.H., & Catania, J.A. (1995). Prevalence and social distribution of men who have sex with men: United States and its urban centers. The Journal of Sex Research, 32, 245-254.

Binson, D., Moskowitz, J. M. T., Anderson, K., Paul, J., & Catania, J. A. (1996). Sampling men who have sex with men: Strategies for a telephone survey in urban areas in the United States. Proceedings of the Section on Survey Research Methods, American Statistical Association, 68-72.

Bradford, J., Ryan, C., & Rothblum, E. (1994). National lesbian health care survey: Implications for mental health care. Journal of Consulting and Clinical Psycholog3; 62, 228-242.

Catania, J., Coates, T., & Kegeles, S. (1992). Condom use in multi-ethnic neighborhoods of San Francisco: The population-based AMEN (AIDS in multi-ethnic neighborhood) study. American Journal of Public Health, 82, 284-287.

Cochran, S. D., & Mays, V. M. (2000). Relation between psychiatric syndromes and behaviorally defined sexual orientation in a sample of the US population. American Journal of Epidemiology, 151, 516-523.

Committee on Lesbian Health Research Priorities, I.O.M. (1999). Lesbian health: Current assessment and direction for the future. Washington, DC: National Academy Press.

Harry, J. (1986). Sampling gay men. The Journal of Sex Research, 22, 21-34.

Kitzinger, C., & Wilkinson, S. (1995). Transitions from heterosexuality to lesbianism: The discursive production of lesbian identities. Developmental Psychology. 31, 95-104.

Laumann, E. O., Michael, R. T., Gagnon, J., & Michaels, S. (1994). The social organization of sexuality Sexual practices in the United States. Chicago: The University of Chicago Press.

Martin, J. L., & Dean, L. (1987). Summary measures: Mental health effects of AIDS on at-risk homosexual men. Division of Sociomedical Sciences, Columbia University, Mailman School of Public Health. Unpublished Manuscript.

Mayer, K., Appelbaum, J., Rogers, T., Lo, W., Bradford, J., & Boswell, S. (2001). The evolution of the Fenway Community Health model. American Journal of Public Health, 91,892-895.

Meyer, I. H., & Colten, M. E. (1999). Sampling gay men: Random digit dialing versus sources in the gay community. Journal of Homosexuality 37, 99-110.

Morgan, K. (1992). Lesbians' use of psychotherapy: A matter of attitude? Psychology of Women Quarterly, 16, 127-130.

Morris, J. F., & Rothblum, E. D. (1999). Who fills out a "lesbian" questionnaire? The interrelationship of sexual orientation, years out, disclosure of sexual orientation, sexual experience with women, and participation in the lesbian community. Psychology of Women Quarterly, 23, 537-557.

Sell, R. L., & Petrulio, C. (1996). Sampling homosexuals, bisexuals, gays, and lesbians for public helath research: A review of the literature from 1990 to 1992. Journal of Homosexuality, 30, 31-47.

Skinner, W., & Otis, M. (1996). Drug and alcohol use among lesbian and gay people in a southern U.S. sample: Epidemiological, comparative, and methodological findings from the Trilogy project. Journal of Homosexuality, 30, 59-92.

Stall R., Barrett, D., Bye, L., Catania, J., Frutchey, C., Henne, J., Lemp, G., & Paul, J. (1992). A comparison of younger and older gay men's HIV risk taking behavior: The communication technologies 1989 cross-sectional survey. Journal of Acquired Immune Deficiency Syndromes and Human Retrovirology, 5, 682-687.

Sudman, S. (1976). Applied sampling. New York: Academic Press. U.S. Department of Commerce, Bureau of the Census. Census of population and housing, 1990 [United States]: Summary tape file 4A [Computer file]. Washington, DC: U.S. Dept. of Commerce, Bureau of the Census [producer], 1993. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 1995.

Viss, D. & Burn, S. (1992). Divergent perceptions of lesbians: A comparison of lesbian self-perceptions and heterosexual perceptions. The Journal of Social Psychology, 132, 169-177.

Waksberg, J. (1978). Sampling methods for random digit dialing. Journal of the American Statistical Association, 73, 40-46.

Ware, J., Kosinski, M., & Keller, S. (1994). SF-12 physical and mental health summary scales: A user k manual. Boston, MA, The Health Institute, New England Medical Center.

Ilan H. Meyer and Lindsay Rossano
Columbian University

James M. Ellis and Judith Bradford
Virginia Commonwealth University

The research presented in this paper was funded by a grant from the Lesbian Health Fund of the Gay and Lesbian Medical Association.

Address correspondence to Ilan H. Meyer, Ph.D., Columbia University, Mailman School of Public Health, Department of Sociomedical Sciences, 722 West 168th Street, New York, NY 10032; e-mail: [e-mail].