Attitudes toward homosexuality among U.S. residents of Mexican descent.

Many authors have noted that homosexuality is stigmatized among U.S. residents of Mexican descent and, more generally, in Latino communities in the United States (Ayala & Dfaz, 2001; Diaz, 1998; Diaz, Ayala, Bein, Henne, & Marin, 2001; Flaskerud, Uman, Lara, Romero, & Otherset, 1996; Gonzalez & Espin, 1996; Marin, 2003; Morales, 1990). However, relatively little empirical research has directly examined attitudes toward homosexuality among Latinos or Hispanics. Of the research that has been published in this area, most studies have focused mainly on comparing the direction and intensity of those attitudes to other groups (e.g., Latinos' attitudes versus those of non-Hispanic Whites or African Americans).

For example, Crawford and Robinson (1990) found that Latinos in an ethnically-diverse convenience sample of male high school students were significantly less antigay than their non-Hispanic White counterparts. Bonilla and Porter (1990), using data from the General Social Survey, found that Hispanics did not differ from Whites but were more tolerant than Blacks in their moral judgments about homosexual behavior (although a majority of all three groups judged homosexual behavior to be "always wrong"). Compared to the other groups, however, Hispanics were less supportive of free speech rights and civil liberties for homosexuals. Bonilla and Porter found no differences in attitudes between Mexican-Americans and other Hispanics.

Other studies have failed to find substantial intergroup differences in attitudes. Using a 9-item scale that tapped opinions about civil rights and civil liberties for gay men and lesbians (e.g., equal rights in employment, adoption, and marriage), Sherrod and Nardi (1998) found that gender was more important than ethnicity in predicting attitudes: Latino and non-Hispanic White males expressed more antigay attitudes compared to Latino and non-Hispanic White females and all African Americans. Moreover, the magnitude of the intergroup differences in this sizable convenience sample (N = 3,542) was small--less than 2 points on a 27-point scale--suggesting that they may have had little substantive significance (see also Alcalay, Sniderman, Mitchell, & Griffin, 1989-1990).

The psychological components of attitudes toward homosexuality and toward gay people among adults of Mexican ancestry in the United States remain largely unexamined. Only the previously mentioned study by Sherrod and Nardi (1998) examined the correlates of such attitudes in depth and, like most other published research in this area, it did not differentiate among cultural subgroups of Latinos. In that study, higher levels of sexual prejudice in Latinos were associated with having few lesbian or gay close friends and with describing one's own political ideology as conservative. In addition, Latinos' anti-gay attitudes were associated with agreeing that religious beliefs are always important in guiding their daily decisions.

Sherrod and Nardi's (1998) exploratory study provides a useful starting point for a social psychological analysis of Latinos' attitudes toward homosexuality, although it has important limitations, It did not use a validated attitude measure with known psychometric properties and did not differentiate attitudes toward lesbians from attitudes toward gay men. Furthermore, the study did not examine how cultural variables might affect attitudes within the portion of the sample identified as Hispanic. Indeed, the reported analyses did not differentiate among cultural groupings of Latinos. Finally, the questionnaire apparently was administered only in English, thereby excluding respondents whose preferred reading language was Spanish.

Empirical research is needed that describes the direction, intensity, and correlates of attitudes toward lesbians and gay men within specific U.S. cultural groups of Latinos. Moreover, such research should address the cultural context of these attitudes by examining how individual Latinos' attitudes toward homosexuality are related to their personal ethnic identity and feelings about Latino and U.S. culture. This study reports such data from a convenience sample of Mexican-American adults in northern California. Attitudes toward homosexuality were operationalized using a bilingual version of a previously validated measure, the Attitudes Toward Lesbians and Gay Men (ATLG) scale (Herek, 1994), which yields separate scores for attitudes toward male and female homosexuality.

The study is based on the premise that the social construction of sexuality affects not only how individuals understand and structure their own sexual feelings and behaviors, but also how they think about various types of sexual conduct, the individuals who practice it, and the social categories to which the latter are assigned. Ethnic minority individuals' personal understandings of sexuality are shaped both by their own culture and by the dominant culture. When these two constructions of sexuality diverge, the relative influence of each is affected by the extent to which an individual's personal identity and daily life experiences are embedded mainly within one culture or the other.

Given the different ways in which homosexuality is regarded in U.S. and Mexican culture (Diaz, 1998; Gonzalez & Espin, 1996; Manalansan, 1996), it is possible that U.S. residents of Mexican descent will differ in their views depending on their primary cultural identification. Compared to those whose lives are more embedded in Mexican culture, for example, those who strongly identify with U.S. culture may have a greater internalization of the heterosexual-homosexual dichotomy that predominates in the United States. This may result in a more clearly delineated cognitive category for the construct of homosexual. Having such an internalized dichotomy might foster more favorable attitudes toward gay people because they are perceived as a cultural outgroup that, like one's own ethnic group, experiences discrimination. Alternatively, gay people might be perceived as an outgroup that is distinct not only from the larger society, but from Latino culture as well. In either case, the relationship between attitudes toward homosexuality and cultural identity warrants examination.

Based on these considerations, this study's goals were (a) to describe the direction and intensity of attitudes toward homosexuality in a community-based convenience sample of U.S. residents of Mexican descent; (b) to examine the associations between those attitudes and theoretically relevant demographic, social, and psychological variables; and (c) to explore how various facets of cultural identity might affect those associations. For the first goal, we used the ATLG scale. Given the relative absence of data on the attitudes of U.S. residents of Mexican descent (or, more broadly, on the attitudes of Latino Americans in general), hypotheses relevant to the second goal were derived from previous research describing heterosexuals' attitudes toward lesbians and gay men, which has been conducted mainly with non-Hispanic samples (e.g., Herek, 1984, 1994, 2000a; Herek & Capitanio, 1995, 1996; Kite & Whitley, 1996, 1998). We tested six hypotheses:

H1. Men of Mexican descent will express more negative attitudes toward homosexuality than women of Mexican descent, and this difference will be more pronounced in attitudes toward gay men than in attitudes toward lesbians. This hypothesis is based on previous findings of a reliable gender difference in attitudes toward gay men and lesbians in samples of non-Hispanic Whites (Kite & Whitley, 1998) and Hispanics (Sherrod & Nardi, 1998).

H2. U.S. residents of Mexican descent who are less educated, older, and married will express more negative attitudes toward homosexuality compared to those who are highly educated, younger, and single. These demographic differences have been reliably observed in survey research with national probability samples (Glenn & Weaver, 1979; Herek & Capitanio, 1995; Herek & Glunt, 1993; Irwin & Thompson, 1977; Schneider & Lewis, 1984). In addition, Sherrod and Nardi (1998) reported a statistically significant correlation between attitudes and marital status among Hispanic women in their sample.

H3. U.S. residents of Mexican descent will express more negative attitudes toward homosexuality to the extent that they are highly religious and belong to denominations with strongly negative views of homosexuality. Religiosity, as measured by frequency of attendance at religious services, is a reliable predictor of non-Hispanic White heterosexuals' attitudes toward lesbians and gay men (Herek, 1994; see also Sherrod & Nardi, 1998). Given the central role played by religious institutions in Mexico and in the Mexican-American community (Marin & Marin, 1991), we hypothesized that this relationship would hold in the present sample as well. Furthermore, we expected respondents who belonged to more fundamentalist religious denominations (e.g., Baptists, Evangelicals) to exhibit the most negative attitudes.

H4. U.S. residents of Mexican descent will express more negative attitudes toward homosexuality to the extent that they are politically conservative. In recent years, the issue of gay rights in the United States has become increasingly politically charged, with liberals generally supporting the passage of antidiscrimination statutes, whereas conservatives have denounced gay men and lesbians as immoral (Herek, 1994; Herman, 1997). We hypothesized that the same political dynamics would be present among U.S. residents of Mexican descent (see also Sherrod & Nardi, 1998).

H5. U.S. residents of Mexican descent will express less negative attitudes toward homosexuality to the extent that they have had personal contact with gay people. Empirical research has consistently shown that such contact is correlated with tolerant attitudes and, indeed, is one of the best predictors of such attitudes (Herek & Capitanio, 1996; Herek & Glunt, 1993; Schneider & Lewis, 1984). This pattern also was observed among Latinos in Sherrod and Nardi's (1998) sample.

H6. U.S. residents of Mexican descent will express more negative attitudes toward homosexuality to the extent that they endorse traditional values about gender and the proper roles of men and women. This pattern has been reliably observed among non-Hispanic samples (e.g., Kite & Whitley, 1998).

To address the study's third goal---exploring how cultural identity might affect these associations--we tested each of the preceding hypotheses within cultural subgroups defined by language preference (English vs. Spanish) and ethnic identification (Mexican vs. Mexican-American or Chicano/a). In addition, we examined the extent to which attitudes toward homosexuality were correlated with social psychological measures of personal identification with Latinos as a group, self-rated importance of Mexican and U.S. customs and celebrations, and preferences for Mexican versus U.S. cultural contexts. Because differences between males and females have been observed consistently in research with non-Hispanic samples (Herek, 2000b; Kite & Whitley, 1998), we also conducted analyses to assess whether the correlations between attitudes toward homosexuality and the variables listed above differed by respondent gender.

Finally, the extent to which attitudes toward homosexuality are related to perceptions of the ethnicity of gay people was assessed. The attitudes of some U.S. residents of Mexican descent may be premised on the assumption that they differ from gay people not only in sexual orientation, but also in ethnic identification. This assumption of dual differences may result in more negative attitudes than would be the case if gay people were perceived as different solely on the dimension of sexuality. This type of pattern has been observed among African Americans, who tend to express more strongly antigay attitudes if they equate being gay with being White rather than Black (Herek & Capitanio, 1995). Thus, we hypothesized that people of Mexican descent whose cognitive category of homosexual overlaps with their category of Mexican Americans or Latinos will manifest more favorable attitudes toward gay people generally than those who perceive homosexuality as a phenomenon manifested by ethnic outgroups (i.e., non-Latinos).

METHOD

The data were collected in a larger study of AIDS education for people of Mexican descent. All questionnaire items reported here were completed by participants in a pretest before the AIDS study.

Participants and Recruitment

Recruitment focused on adults of Mexican descent who were residents of Yolo and Sacramento counties in northern California. Prior to the study, the research team conducted focus groups with local Mexican American leaders, moderated by the second author. These groups introduced the project staff to community leaders and explained the goals of the larger study. Participants discussed their concerns related to AIDS, suggested strategies for recruiting participants, and volunteered resources for making the local community aware of the study. Based on these consultations, recruitment efforts subsequently focused on five sources: (a) community festivals, at which the research team staffed booths and tables; (b) organizations, at which a member of the research team was introduced by a community leader who had participated in the initial focus groups; (c) Latino dance clubs; (d) widely-circulated fliers and advertisements in community newsletters and Latino newspapers, which included a brief description of the study and a toll-free telephone number for potential participants to call; and (e) personal networks of socially influential individuals whom the research team identified in the course of community outreach.

At the time of recruitment, potential participants were informed of the study's eligibility requirements: that they must be at least 18 years of age and able to read and write in English or Spanish. Nevertheless, four participants younger than 18 years old completed the questionnaire. In addition, some participants had minimal reading skills and required assistance with the questionnaire.

Measures

English- and Spanish-language versions of the questionnaire were prepared in consultation with a professional translator and an expert in Latin American literature. A Spanish-language version was developed by the second author and then reviewed by the literature expert to ensure that the language was appropriate for Spanish-speaking Mexican Americans. The questionnaires were revised as needed and translated into English by the second author. The English versions were reviewed by the literature expert, refined as needed, and then back-translated to Spanish by a professional translator. The back-translated versions were compared to the originals and modified as necessary to ensure equivalency of the English and Spanish versions. In the final questionnaires, all items were presented in both English and Spanish, with the Spanish version first in bold print followed by the English wording in italics. Three categories of variables were assessed: (a) variables related to homosexuality, (b) variables related to Latino culture, and (c) social and demographic information.

Measures Related to Homosexuality

Attitudes toward male and female homosexuality. Four-item versions of the Attitudes Toward Gay Men (ATG) and Attitudes Toward Lesbians (ATL) subscales of the ATLG were translated into Spanish. Multiple forms of these subscales were originally developed in English and their psychometric properties are well-established (Herek, 1994). For this study, they consisted of four statements about male homosexuality and four parallel statements about female homosexuality, to which respondents indicated their level of agreement or disagreement. Because the extent to which the term gay is understood by non-gay Mexicans has not been documented, we used the term homosexual in the ATLG items to ensure that all respondents understood them.

The English and Spanish versions of the items are presented in Table 1. The items were administered with five Likert response alternatives listed in both Spanish and English: strongly disagree [completamente en desacuerdo], somewhat disagree [en desacuerdo], neither agree nor disagree [ni acuerdo ni desacuerdo], somewhat agree [de acuerdo], and strongly agree [completamente de acuerdo]. Scale scores were computed by assigning numerical values to each response alternative (0 = strongly disagree, 4 = strongly agree) and summing across the items (with scoring reversed as appropriate) so that higher scores indicate higher levels of sexual prejudice (i.e., antigay attitudes; Herek, 2000a).

Personal contact with homosexual men and women. Respondents were asked, "How many men and women do you know--your friends, family, or acquaintances--who are homosexual?" [[??]Cuantos hombres y mujeres homosexuales--conocidos, sus amigos/as o familiares--usted conoce?]. Response options ranged from None to 5 or more.

Ethnic perceptions of homosexual men and women. Respondents were asked, "When you hear someone talking about a man who is a homosexual, which of these groups comes to your mind first?" [Cuando usted escucha a alguien hablar acerca de un hombre que es homosexual,/,cual de los siguientes grupos viene primero a su mente?]. Response options included Anglo [Angloamericano], Black [Negro], Asian [Asiatico], Native American [Indio Estadounidense], Mexican [Mejicano], Chicano [Chicano], Cuban [Cubano], Puerto Rican [Puertorriquefio], Other Latinos [Otros latinos], and Other, please specify [Otro grupo (por favor especifique)]. The question was then repeated to ask about the ethnic group associated with "a woman who is a homosexual" [una mujer que es homosexual].

Measures Related to Cultural Constructs

Ethnic identification. In addition to being asked their country of birth in an open-ended question, respondents were asked to select the cultural group with which they most strongly identified, using a list that included Mexican [Mejicano]; Chicano or Mexican American [Chicano o Mejico Americano]; Caribbean, including Puerto Rican, Dominican, and Cuban [Caribeno: puertorriqueno, dominicano, cubano]; Latin American Indian [Indio Latinoamericano]; Native American [Indio Americano Estadounidense]; Central American, including Guatemalan, Panamanian, Salvadorean [Centroamericano: guatemalteco, panameno, salvadoreno]; South American [Suramericano]; European or European American [Europeo o Europeo-Americano]; African or African American [Africano o Africo Americano]; and Asian/Pacific Islander [Oriental o Asiatico].

Language preference. Respondents completed the 4-item version of the Marin measure of acculturation, which assesses language preferences in reading, speaking, and thinking (Marin & Marin, 1991; Marin, Sabogal, Marin, & Otero-Sabogal, 1987). Five response alternatives were provided: Only Spanish [$610 espanol], Spanish more than English [Mas espanol que ingles], Both equally [Ambos por igual], English more than Spanish [Mas ingles que espanol], and Only English [Solo ingles]. Responses to the items were summed to yield a scale score ([alpha] = .94), with higher scores indicating higher levels of U.S. acculturation, based on language use.

Four additional scales were constructed to assess aspects of cultural identity:

Personal identification with Latinos as a group. A scale of identification with Latinos as a group ([alpha] = .68) included three items: (a) "I believe that being Latino/a is a good and positive experience" [Yo creo que ser latino/a es una experiencia buena y positiva]; (b) "I feel a strong attachment to Latinos" [Yo me siento muy apegado a la gente latina]; and (c) "I feel excited and joyful in Latino surroundings" [Yo me siento entusiasmado y feliz cuando estoy en ambientes latinos].

Preference for Mexican and U.S. customs and celebrations. The questionnaire included nine items that assessed how much importance respondents attached to Mexican and U.S. customs and celebrations. All items were rated on a 5-point scale ranging from 0 (not at all important) to 4 (very important). Principal components analyses with oblique rotation revealed that eight of the items loaded on two distinct factors, one reflecting Mexican customs and the other U.S. customs. Based on these factors, we constructed two 4-item scales (see Appendix). The ninth item--"dance to Tex-Mex music (e.g., 'banda,' 'quebraditas')" [bailar musica Mejico-tejana (ej. banda, quebraditas)]--did not load on either factor and was dropped from the analysis. Reliability coefficients were acceptably high for both the Preferences for Mexican Customs and Celebrations scale ([alpha] = .78; hereafter referred to as the Mexican Customs Scale) and the Preferences for U.S. Customs and Celebrations scale ([alpha] = .68; hereafter the U.S. Customs Scale). Scores on the two scales were negatively correlated (r = -23). Respondents rated U.S. customs as substantially more important than Mexican customs (means = 9.29 and 3.38, respectively; standard deviations = 3.66 and 3.43, respectively).

Preferences for Mexican vs. U.S. cultural contexts. Finally, the questionnaire included 4 items that assessed preferences for socializing, movies, food, and self-image (see Appendix). Each item offered an ordinal list of response alternatives that ranged from a Mexican context to a U.S./Anglo context. The items were scored on ordinal scales such that higher scores indicated greater preference for U.S./Anglo activities. Reliability for the scale was acceptable ([alpha] = .68). The scale was weakly correlated with scores on the 4-item U.S. Customs scale described above (r =. 12, p < .05) and moderately correlated with the 4-item Mexican Customs scale (r = .40, p < .001).

To understand the meaning of responses to the cultural measures, scores on the Latino Group Identification, Mexican Customs, U.S. Customs, and Preferences for Cultural Contexts scales were subjected to a 2 (Ethnic identification: Mexican vs. Chicano/a or Mexican-American) x 2 (Language preference) MANOVA. Language preference was based on an item from the Marin acculturation scale assessing preferred language for reading and speaking, with responses dichotomized as Spanish versus English/both equally (this dichotomy is hereafter characterized as Spanish vs. English).

Examination of the univariate ANOVAs for the Mexican Customs scale revealed a significant main effect for ethnic identification, F (1,349) = 19.75 (p < .001, [[eta].sup.2] = .054) which was qualified by a significant Ethnic Identification x Language Preference interaction, F (1, 349) = 7.957 (p < .01, [[eta].sup.2] = .022). Respondents who identified as Mexican attached less importance to Mexican customs and celebrations (i.e., had lower scores) than did respondents who identified as Chicano/a or Mexican American, but the difference was significant only for respondents whose language preference was Spanish (M = 2.48 for Spanish-preferring Mexicans vs. 6.70 for Spanish-preferring Mexican Americans; SD = 2.9 and 4.2, respectively). Among respondents whose language preference was English or both English and Spanish, the pattern was similar but the difference was not statistically significant (M = 3.11 for English-preferring Mexicans vs. 4.05 for English-preferring Mexican Americans; SD = 3.3 and 3.4, respectively).

Scores on the U.S. Customs scale were significantly higher (i.e., more importance attached to U.S. customs and celebrations) for respondents whose language preference was English (M = 9.51, SD = 3.6) than for those who preferred Spanish (M = 8.95, SD = 3.7), F (1,349) = 3.80 (p = .05, [[eta].sup.2] = .011). For the Cultural Preferences scale, significant main effects were obtained for ethnic identification, F (1,349) = 35.98 (p < .001, [[eta].sup.2] = .093) and language preference, F (1,349) = 14.58 (p < .001, [[eta].sup.2] = .040). Scores were significantly higher (indicating greater preference for U.S. cultural contexts) for respondents whose language preference was English (M = 7.50 [SD = 2.2] vs. 5.28 [SD = 1.5] for respondents who preferred Spanish) and for those who identified as Mexican American or Chicano/a (M = 8.22 [SD = 2.0] vs. 5.65 [SD = 1.8] for those who identified as Mexican). No significant differences were observed for Latino Group Identification scores.

Gender Attitudes

A pool of 11 statements expressing opinions about gender roles and equality relevant to Latino culture was included in the questionnaire, intermixed with the ATL and ATG items. A principal components analysis using oblique rotation revealed that seven of the items were highly intercorrelated and could be combined meaningfully into a scale of traditional gender-related attitudes ([alpha] = .80). The items are listed in the Appendix.

Social and Demographic Measures

The questionnaire included items to ascertain respondent sex, age, educational level (number of years of formal schooling and highest degree or diploma), marital status, and parental status. Questions were also included to assess religious denomination, frequency of attendance at religious services, political party, and voting behavior in the most recent presidential election. Because of concerns about negative reactions from some participants, we did not assess sexual orientation. Consequently, a small number of gay, lesbian, or bisexual individuals may have completed the questionnaire. If so, this may reduce the magnitude of correlations between ATG and ATL scores and other variables.

Procedure

All members of the research team who interacted directly with participants were Latinos; they remained unaware of the study's hypotheses during data collection. Questionnaires were completed by respondents in small groups. A member of the research team explained the purpose of the study to each participant and gave him or her $15 in cash, a form explaining the rights of participants, and a blank questionnaire. All participants watched a brief video with instructions in both English and Spanish for completing the questionnaire. The research team also provided additional assistance to participants who needed help reading the questionnaire items.

RESULTS

Sample Characteristics

A total of 616 respondents completed questionnaires. Slightly more than half (51%) were born in Mexico, with another 42% born in the United States (excluding Puerto Rico). Roughly 1% were born elsewhere, and 6% did not answer the question about birthplace. Almost half of the respondents (46%) self-identified as Mexican, with another 28% identifying as Chicano/a or Mexican American. Fewer than 4% endorsed any of the remaining ethnic identities (including "other"); 23% did not answer the ethnic identification question. (Nonresponse to this item was not significantly associated with educational level or preferred reading language.) Most respondents who identified as Mexican (86%) were born in Mexico, whereas nearly all respondents who identified as Chicano/a or Mexican American (91%) were born in the United States.

The sample was predominantly female (66%). The mean age was 30 years (range = 15-89 years) and the median educational level was 12 years, with 63% having a high school diploma or equivalent and 11% having a bachelor's degree. Nearly half of the respondents (48%) were single, with another 41% married, 9% separated or divorced, and 2% widowed. Half had children (median number of children = 3). Most respondents (75%) were Catholic, with another 5% belonging to a fundamentalist or conservative denomination (Baptist, Mormon, Evangelical, Pentecostal), 7% belonging to another denomination (e.g., Episcopalian), and 11% having no denominational affiliation. Most participants (53%) were currently employed. The sample was not politically active, with fewer than one fourth reporting that they had voted in the previous national election.

ATG and ATL Scales

Mean scores for responses to the individual ATG and ATL items are reported in the first data column of Table 1. The collapsed percentages of respondents who agreed (strongly or somewhat), disagreed (strongly or somewhat), or neither agreed nor disagreed (i.e., gave the midpoint response) with each item are presented in the remaining columns. As shown in Table 1, the proportions of respondents expressing negative attitudes on each item varied from roughly 27% to 36% for the ATG items and from 29% to 43% for the ATL items. The proportions expressing positive attitudes ranged from roughly 37% to 45% for the ATG and from 27% to 47% for the ATL. Expressions of positive attitudes were more frequent than expressions of negative attitudes on 7 of the 8 items (i.e., all but the ATL PERVERSION item).

The four items were summed, with responses to the NATURAL item reversed, to yield ATG and ATL scores. Internal consistency of the scales was assessed with coefficient alpha. Because all respondents received a bilingual questionnaire, separate reliability coefficients for the English- and Spanish-language versions of the scales could not be computed. However, it is reasonable to assume that respondents who reported a Spanish-language reading and speaking preference used the Spanish-language version of the ATG and ATL. Accordingly, separate alpha coefficients were calculated for respondents in each reading category. Because respondents who preferred English or both English and Spanish equally tended to have higher educational levels, and because educational level is likely to be associated with reading facility which, in turn, may affect response reliability, respondents were also categorized according to educational background.

Scores for the 4-item ATG and ATL had somewhat low internal consistency ([alpha] = .60 and .64, respectively). Further examination revealed that reliability was particularly low (< .30) for respondents whose language preference was Spanish and who had not graduated from high school. Examination of the inter-item correlations revealed that the low reliability for this portion of the sample was due mainly to responses to the reversed NATURAL item. With that item omitted, alpha coefficients increased for the entire sample (overall ATG [alpha] = .70, ATL [alpha] = .71) and for Spanish-speaking respondents with low levels of formal education (see Table 2). Accordingly, 3-item versions of the two scales--comprising the WRONG, DISGUST, and PERVERSION items--are used in subsequent analyses. Nevertheless, because of the relatively low alpha coefficients for Spanish-speaking respondents with low levels of formal education, caution must be exercised in interpreting scale scores for that group. A similar pattern was observed when respondents were categorized by educational level and birthplace (Mexico vs. U.S.), with [alpha] [greater than or equal to] .70 for all groups except Mexican-born respondents with less than a high school diploma. For that group, [alpha] = .56 for both the ATG and ATL.

Hypothesis Tests

We conducted a series of MANOVAs (with ATG and ATL scores as dependent variables) to test the six hypotheses described above. When independent variables included three or more categories, post-hoc comparisons were made using the Student Newman Keuls statistic. The results are reported in Table 3.

H1. Sex Differences

Consistent with previous findings from non-Hispanic samples, male respondents expressed significantly more negative attitudes toward gay men than did female respondents. MANOVA yielded a significant main effect for ATG scores for sex, F (1,561) = 5.25 (p < .05, [[eta].sup.2] = .009). In this and subsequent analyses, univariate effects are reported only if MANOVA yielded a significant (p < .05) multivariate effect.

The difference between males' and females' ATL scores was not statistically significant. This overall pattern is similar to one observed with samples of non-Hispanic Whites and African Americans (Herek, 2002; Herek & Capitanio, 1995). For the latter groups, however, the lack of significant sex differences in attitudes toward lesbians results mainly because male respondents tend to express more favorable attitudes toward lesbians than toward gay men, while female respondents generally express similar attitudes toward both gay men and lesbians. In the present sample, by contrast, males' attitudes did not differ significantly across attitude targets, whereas females' attitudes toward lesbians were significantly less favorable than their attitudes toward gay men, F (1,366) = 5.72 (17 < .001, [[eta].sup.2] = .041; see Table 3). In other words, the lack of a significant difference between males' and females' ATL scores in the present sample resulted mainly from females' relatively negative attitudes toward lesbians rather than males' relatively favorable attitudes toward them.

H2. Education, Age, and Marital Status

As hypothesized, ATG and ATL scores varied as a function of respondents' age and education. The correlations of age with ATL and ATG scores were modest, indicating less than 1% of shared variance (see Table 4). The correlations with number of years of formal schooling were nearly twice as large but still modest (approximately 2% of variance shared). To examine the hypothesis regarding educational level, respondents were divided into three categories: (a) did not complete high school, (b) high school diploma or equivalent, and (c) any post-secondary education. MANOVA revealed significant main effects for education for both the ATG, F (2, 555) = 6.63 (p = .001, [[eta].sup.2] = .023), and the ATL, F (2, 555) = 10.35 (p < .001, [[eta].sup.2] = .036). Post-hoc comparisons revealed that respondents with any post-secondary education scored significantly lower (indicating less prejudice) than others on both scales (Table 3). For the hypothesis regarding marital status, unmarried respondents scored somewhat lower than others on both scales, but the difference was not statistically significant.

H3. Religiosity

As hypothesized, highly religious respondents expressed significantly more negative attitudes toward gay men and lesbians than their less religious counterparts. Attendance at religious services had a small but statistically significant positive correlation with ATG and ATL scores (see Table 3). With the attendance variable divided into six categories for MANOVA, a main effect was observed for the ATG, F (5, 548) = 5.37 (p < .001, [[eta].sup.2]= =.047), and the ATL, F (5, 548) = 7.91 (p < .001, [[eta].sup.2] =.067). Post-hoc comparisons revealed that respondents who attended religious services more than once per week expressed significantly more negative attitudes toward lesbians and gay men than all other respondents. In addition, respondents who attended religious services on a weekly basis scored higher than respondents who attended sporadically (Table 3).

The relationship between religion and attitudes also was evident in a main effect for denomination, which was significant for the ATG, F (3,556) = 6.07 (p < .001, [[eta].sup.2] = .032), and the ATL, F (3, 556) = 8.46 (p < .001, [[eta].sup.2] = .044). Post-hoc comparisons revealed that respondents who belonged to a conservative Christian denomination (Baptist, Mormon, Evangelical, Pentecostal) scored significantly higher on the ATG and ATL than respondents who were Catholic, members of another Protestant denomination (e.g., Episcopalian), or respondents who had no religious denomination.

H4. Political Ideology

As shown in Table 3, politically conservative respondents (i.e., those who identified as Republican or voted for the Republican candidate in the previous presidential election) scored significantly higher on both scales. For the ATG, F (1,569) = 11.14 (p = .001,[[eta].sup.2] = .019). For the ATL, F (1, 569) = 12.66 (p < .001, [[eta].sup.2] =.022).

H5. Personal Contact

ATG and ATL scores were negatively correlated with the number of lesbians and gay men that respondents reported knowing personally (see Table 4). Consistent with that finding, MANOVA yielded a significant main effect for number of contacts for the ATG, F (5, 436) = 6.03 (p < .001, [[eta].sup.2] = .065), and for the ATL, F (5,436) = 6.30 (p < .001, [[eta].sup.2] = .067). Post-hoc comparisons showed that the contact effect was most pronounced for respondents who knew a large number of gay people: those who reported knowing five or more homosexual men or women (n = 93) had significantly lower ATG and ATL scores than all other respondents.

H6. Traditional Gender Attitudes

As expected, respondents who endorsed traditional beliefs about gender roles were more likely to express negative attitudes toward lesbians and gay men. The correlation between traditional gender attitudes and both ATG and ATL scores was r = .32 (see Table 4).

Cultural Patterns in Attitude Correlates

ATL scores differed significantly according to ethnic identification, with significantly more negative attitudes toward lesbians expressed by respondents who identified as Mexican (M = 6.13, SD = 3.6) than by respondents who identified as Mexican American or Chicano/a (M = 5.35, SD = 4.3), F (1,425) = 4.14 (p < .05, [[eta].sup.2] =.01). The groups did not differ in ATG scores. Neither ATG nor ATL scores differed significantly according to respondents' language preference.

ATG and ATL scores were not correlated with the Latino Group Identification, Mexican Customs, U.S. Customs, or Preferences for Cultural Contexts scales (see Table 4). The only statistically significant correlations observed among the cultural variables were between ATG and ATL scores and scores on the Marin acculturation scale. To the extent that respondents preferred English, they tended to express less negative attitudes toward lesbians and gay men. However, this association was fairly weak (r = -.12).

As shown in Table 3, to the extent that respondents associated a particular ethnicity with gay people, it was Anglo (50% reported this association for gay men, 58% for lesbians). The second most common association was with Mexicans or Chicanos (24% for gay men, 18% for lesbians). These perceptions were not reliably associated with ATG or ATL scores.

Correlates of ATG and ATL Scores Within Cultural Groups

The analyses reported to this point have examined the sample as a whole without regard to differences in cultural variables. However, subgroups of the sample might differ in how ATG and ATL scores correlate with the variables described in Table 3. Examination of these patterns with ANOVA procedures is problematic because of unacceptably small cell sizes. To avoid this problem, we conducted a series of ordinary least squares (OLS) regression analyses, creating separate equations for ATG and ATL scores. On the first step, a variable described above in the hypothesis tests (e.g., education) was entered along with dummy-coded variables representing the respondent's preferred language (1 = English or both English and Spanish, based on an item from the Marin scale), ethnic identification (1 = Mexican American), and sex (1 = female). On the second step, interaction terms were entered for the main predictor variable multiplied by the dummy variables for language, identification, and gender (e.g., education x language, education x identification, education x gender). This analysis was repeated for each variable previously examined in the hypothesis tests.

No interaction terms were significant for the analyses of age, marital status, political ideology, ethnic images of gay men or lesbians, or religious denomination. That is, the relationships between these variables and ATL and ATG scores did not differ according to cultural subgroups. Differences were observed, however, for religious attendance, personal contact, education, number of children, and traditional gender attitudes.

For religious attendance, the Sex x Attendance interaction term was significant and positive for both ATG (b = 1.02, t = 3.79, p < .001) and ATL scores (b = 1.03, t = 3.80, p < .001), indicating that more frequent attendance was a significant predictor of negative attitudes for women but not for men. Moreover, the Language x Attendance interaction was significant for ATG scores (b = 0.85, t = 2.82, p < .01), indicating that more frequent attendance predicted more negative attitudes toward gay men for respondents whose preferred language was English.

The language preference interaction was also significant for the analyses of personal contact for both the ATG (b = -.94, t = -3.93, p < .001) and ATL (b = -1.03, t = -4.18, p < .001). The negative sign of these regression coefficients indicates that personal contact predicted lower ATG and ATL scores (more favorable attitudes) for English-preferring respondents.

Significant interactions were observed between the ethnic identification variable and education (for ATG, b = -0.32, t = -3.03, p < .01; for ATL, b = -0.29, t = -2.67, p < .01), number of children (for ATG, b = 0.55, t = 2.84, p < .01), and traditional gender attitudes (for ATG, b = 0.23, t = 3.07, p < .01; for ATL, b = .21, t = 2.70, p < .01). These interactions indicate that higher levels of formal education predicted less negative attitudes toward gay men and lesbians among respondents who identified as Mexican American or Chicano/a. By contrast, parenting a larger number of children was associated with more negative attitudes toward gay men in this group. Finally, whereas traditional gender attitudes were predictive of greater hostility toward lesbians and gay men for the sample as a whole, this association was even stronger among respondents who identified as Mexican American or Chicano/a.

Predictors of Attitudes

The analyses presented thus far have examined correlates of respondents' attitudes toward male and female homosexuality and identified which variables have differential associations with ATG and ATL scores depending on the respondent's language preference, ethnic identification, or sex. In a final analysis, we sought to identify a core group of key variables that best predicted ATG and ATL scores.

For this purpose, separate OLS regression equations were constructed for ATG scores and ATL scores. Because including all potential predictor variables and their respective interaction terms (by gender, language preference, and ethnic identification) would produce equations with an unwieldy number of variables and substantially-reduced statistical power, preliminary analyses were conducted to eliminate unnecessary variables. Three sets of equations were computed, one that included all of the interaction terms based on respondent gender, another with all of the interaction terms based on language preference, and a third with all of the ethnic identification interaction terms. On the first step, each equation included the cultural variables (language preference, ethnic identification, Latino Group Identification scale, Mexican Customs and U.S. Customs scales, Preferences for Cultural Contexts scale). The second step included the social, demographic, and attitudinal variables (gender, education, age, marital status, number of children, traditional gender attitudes, religious denomination, religious attendance, political ideology, contact with gay people, perceptions of lesbian and gay ethnicity). The third step included the relevant multiplicative interaction terms (e.g., the dummy variable for reading language preference multiplied by each of the variables entered on Step 2).

For these equations, responses to the single item on the Marin scale were used on the first step, rather than the total 4-item score. This procedure was adopted for the sake of consistency because the single item was used as a dummy variable for computing interaction terms. When the analyses were repeated with the total 4-item acculturation score, the results were essentially identical.

We examined the resulting coefficients for these equations and dropped any nonsignificant interaction terms from further analysis. Two new combined equations were constructed (once again, one for ATG scores and the other for ATL scores) with the cultural variables entered on the first step; the social, demographic, and attitudinal variables entered on the second step; and all statistically significant interaction terms from the previous three sets of equations combined on the third step. Once again, interaction terms that did not achieve statistical significance were dropped from further analysis, as were variables from Step 2 that did not achieve statistical significance and did not have an associated interaction term that achieved statistical significance. This resulted in final regression equations that explained 31% of the variance in ATG scores and 29% of the variance in ATL scores. The variables included in the equations are listed in Table 5.

As shown in Table 5, traditional gender beliefs emerged as a significant predictor in the final equations for both ATL and ATG scores. The significant regression coefficients for ethnic identification and language preference are qualified by significant interaction terms. Consistent with the individual regression analyses reported in the previous section, the attitudes of respondents who identified as Mexican American or Chicano/a were predicted by formal education and, for ATG scores, by number of children. The attitudes of respondents with a preference for English were predicted by personal contact. For female respondents, more years of formal schooling predicted more favorable attitudes toward lesbians. Although Latino group identification was a statistically significant predictor of ATG scores, it explained a negligible amount of variance (< 0.1%).

DISCUSSION

The data offer important insights into the social psychology of attitudes toward male and female homosexuality among U.S. residents of Mexican descent. On one hand, the results reveal commonalities between the Latinos in this sample and non-Hispanic respondents in earlier research, as well as similarities between Latinos who identify with Mexican and U.S. cultures. Consistent with previous studies, attitudes were correlated in the expected directions with age, education, number of children, attendance at religious services, membership in a fundamentalist denomination, political conservatism, personal contact with gay people, and traditional gender attitudes. Of these variables, support for traditional gender roles among Latinos was perhaps the most robust predictor of attitudes toward homosexuality. It accounted for more than 5% of the variance in ATL scores and 3% of the variance in ATG scores, even when the direct and moderated effects of all other variables were statistically controlled.

Whereas the relationships between these variables and attitudes toward homosexuality appeared to be similar across the entire sample, cultural differences were evident in the direction and intensity of attitudes as well as their underlying correlations with demographic, social, and psychological variables. Attitudes toward female homosexuality were significantly more favorable among respondents who identified as Mexican American or Chicano/a. It is also noteworthy that a plurality of the sample generally disagreed with the questionnaire statements expressing negative attitudes toward homosexuality. Because this was a convenience sample, generalizations from it to the population of U.S. residents of Mexican descent are not appropriate. However, it will be interesting and valuable for future research to assess the extent to which attitudes toward homosexuality in Latino samples can be characterized as generally tolerant or intolerant.

Previous research with non-Latino samples has consistently found that attitudes toward homosexual men are significantly more negative among heterosexual males than among heterosexual females, and that males' attitudes toward gay men tend to be more negative than their attitudes toward lesbians, whereas females' evaluations of lesbians and gay men do not differ substantially (Herek, 2002). This study only partially replicated these patterns. Men's ATL and ATG scores did not significantly differ, whereas women expressed more negative attitudes toward lesbians than toward gay men. However, results of the final regression analysis (Table 5) suggested that attitudes toward lesbians are more favorable among highly-educated women than among women with less formal schooling. Perhaps women in our sample with less education had relatively little contact with lesbians and this affected their attitudes. Unfortunately, the questionnaire item about personal contact did not differentiate between contact with gay men and contact with lesbians. Another possible explanation for the pattern is that gender-related expectations in Latino cultures allow greater latitude in sexual behavior for men than for women. Indeed, the belief that men should be permitted sexual transgressions while women should remain morally superior is encapsulated in several items in the Traditional Gender Attitudes scale (see Appendix). The interaction observed in this study between respondent gender and gender of the attitude target (i.e., male versus female homosexuality) suggests promising areas for future research.

Moreover, previously reported associations between attitudes toward homosexuality and various demographic, social, and psychological variables appeared most likely to be replicated among the respondents who, in cultural terms, more closely resembled participants in previous studies: that is, those who spoke and read English and identified as members of U.S. culture rather than Mexican culture. Education and number of children predicted attitudes toward both male and female homosexuality mainly for respondents who identified as Mexican American or Chicano/a. In addition, support for traditional gender roles was an even stronger predictor of attitudes in this group than for the rest of the sample. Furthermore, personal contact with gay people and religious attendance were significant predictors of attitudes among respondents whose language preference was English or both English and Spanish.

It might be expected that these patterns of differential associations would be explained by respondents' cultural preferences and identification. Yet, measures of those variables were not substantially correlated with attitudes toward homosexuality, and the significant effects described above emerged even when the cultural variables were statistically controlled. Nor did respondents' perceptions of the ethnicity of homosexual men and women affect their attitudes. The study's lack of significant findings in this area means that further research is needed to identify the underlying social psychological processes that account for the observed associations between attitudes toward homosexuality and language and ethnic identification. Such research should also attempt to identify additional correlates of attitudes among Latinos who prefer Spanish and who identify with their ancestral country more than the United States.

The Spanish-language versions of the ATG and ATL developed for this study will be useful tools for studying attitudes toward homosexuality among individuals of Mexican descent. At the same time, the tradeoffs inevitably associated with adapting an existing measure for a cultural group different than the one for which it was initially developed (e.g., Marin & Marin, 1991) were evident. Contrary to expectation, the one item in the original English scales that expresses favorable attitudes and is reverse-scored--"Male [or female] homosexuality is a natural expression of sexuality in men [or women]"--was problematic for some respondents. Especially among Spanish-speaking respondents with relatively little formal schooling, inclusion of the NATURAL item reduced reliability for the ATG and ATL. There are at least two explanations for this pattern. First, it may reflect differing connotations for the term "natural" in Spanish and American English. For example, whereas some respondents may have understood the Spanish natural to mean "innate" or "normal"--similar to the connotation of the English "natural"--others may have taken it to mean "primitive" or "uncivilized." The latter interpretation might have been more common among Spanish-preferring respondents with lower levels of education, which could account for the pattern of correlations between the NATURAL item and the other ATG and ATL items in this group. Alternatively, the overall pattern may reflect an acquiescent response style. A tendency to agree generally with assertions of opinion, even when they are contradictory, is especially likely to be observed in respondents with minimal education and those who lack extensive experience completing self-administered questionnaires (e.g., Krosnick, 1999). In the current sample, the Spanish-preferring respondents without a high school diploma most closely fit this profile. Further research is needed to illuminate why the NATURAL item may be problematic for the ATG and ATL with Spanish-speaking Latino samples.

This study has several limitations. As already mentioned, although respondents were recruited through a variety of community venues, the generalizability of the results from this convenience sample to a larger population cannot be known. In addition, because sexual orientation was not assessed, the sample might have included lesbian, gay, or bisexual individuals, which could have weakened the relationships among some of the variables. Moreover, several measures of cultural constructs were developed specifically for this study and will benefit from further validation. Finally, although the use of bilingual questionnaires may have helped respondents with some degree of fluency in both Spanish and English, this procedure made strict comparisons between Spanish- and English-language versions of the items impossible.

In future research, it will be important to assess the extent to which bilingual instruments are needed and to identify the samples for which they are most appropriate. Future studies might use brief assessments of English and Spanish proficiency to assign participants to language categories. Alternatively, bilingual respondents could be randomly assigned to complete monolingual Spanish or English versions of the scales. This procedure might be especially useful for determining the scales' properties with Latino respondents in the United States who use "Spanglish," a combination of English and Spanish. The properties of the Spanish-language ATG and ATL will also be better understood when they are administered to samples of Spanish-speaking Latinos outside the continental United States.

The results suggest that some of the same strategies previously discussed for reducing sexual prejudice among non-Latinos may be applicable with U.S. residents of Mexican descent. For example, the finding that attitudes toward homosexuality were closely related to beliefs about appropriate roles for Latin men and women is consistent with Diaz' (1998) observation that homosexuality is defined in Latino culture mainly in terms of gender. It suggests that sexual prejudice might be reduced by modifying traditional gender role attitudes to allow for greater flexibility for men and women. Although Diaz framed his analysis mainly in terms of male gender expectations, it is possible that greater flexibility in female gender roles will result in more favorable attitudes toward lesbians. Alternatively, negative attitudes toward gay men and lesbians might be changed if cultural definitions of homosexuality shift from being based mainly on gender to, for example, a focus on the minority group status of gay men and lesbians (e.g., Herek, 2000b). Other attitude-change strategies may work differentially with subgroups in Latino communities. Among English-speaking Mexican Americans, for example, direct personal contact with lesbians and gay men may lead to more favorable attitudes toward gay people as a group.

Replication of this study with other samples of people of Mexican descent and with other Latino groups is needed. Also needed is further study of how experiences such as those associated with interpersonal contact and education affect sexual prejudice, and why this pattern occurs mainly in Latinos who are more embedded in U.S. culture. Further study with different and larger samples may help to explain how cultural factors moderate the effects of education and contact. Finally, future research should continue to evaluate the utility of the ATG and ATL scales with Latino samples. This study suggests that the items do tap a set of attitudes similar to those observed with non-Hispanic samples in the United States and other countries.

APPENDIX: ADDITIONAL SCALES

Preference for Mexican Customs and Celebrations

Utilizando las siguientes alternativas indique cumin importante es para usted ... [Using the following options, please tell me how important it is for you to ...]

1. Celebrar dfas feriados mejicanos tales como el 5 demayo, el dla de reyes, ect.? [Celebrate Mexican holidays such as "Cinco de Mayo," "Three Kings Day," etc.?]

2. Seguir las costumbres y tradiciones mejicanas? [Follow Mexican customs and traditions?]

3. Conocer sobre la historia de Mrjico? [Know about the history of Mexico?]

4. Bailar musica mejicana (rancheras, boleros)? [Dance to Mexican music, "rancheras," "boleros"?[

Preference for U.S. Customs and Celebrations

Utilizando las siguientes alternativas indique cuan importante es para usted ... [Using the following options, please tell me how important it is for you to ...]

1. Celebrar dias de fiesta estadounidenses tales como el cuatro de julio? [Celebrate U.S. holidays, such as the Fourth of July?]

2. Seguir las costumbres y tradiciones estadounidenses? [Follow U.S. customs and traditions?]

3. Conocer sobre la historia de los Estados Unidos? [Know about the history of the United States?]

4. Bailar musica estadounidense (ej. rock, jazz)? [Dance to U.S.A. music (e.g., rock, jazz big bands)?]

Preference for Mexican vs. U.S. Cultural Contexts

1. Si usted llega a un lugar nuevo de trabajo y tres companeros de quienes solo conoce el origen etnico, le invitan a almorzar, [??]Mexicon quien preferira irse? [If you are new on your job and three coworkers ask you out for lunch, and you know only their ethnic backgrounds, with whom would you prefer to go?]

(a) Companero\a mejicano\a [Mexican coworker]

(b) Companero\a chicano\a [Chicano/a coworker]

(c) Companero\a americana (Blanco/a) [White American coworker]

2. Si usted tiene la posibilidad de ver una misma pelicula en una de cuatro versiones distintas, ?en que version preferiria verla? [If you have the choice of watching a movie that has four different versions, which version would you prefer to see?]

(a) version en espanol [Spanish version]

(b) version en espanol con subtitulos en ingles [Spanish version with English subtitles]

(c) version en ingles con subtitulos en espanol [English version with Spanish subtitles]

(d) version en ingles [English version]

3. Si unos visitantes le indican que desean comer el tipo de comida que usted come mas frecuentemente, ?que tipo de comida usted les serviria? [If some visitors ask to have the kind of meal that you most often eat, what kind of meal would you serve to them?]

(a) comida mejicana (ej. chile relleno, tacos, polio en mole, ect.) [Mexican food]

(b) comida Tex-Mex o Cal-Mex (ej. burritos, fajitas, ensalada de taco, ect.) [Tex-Mex or Cal-Mex food]

(c) comida americana (ej. filetes de came, emparedados, polio frito, ect.) [American food]

4. Si usted va a conocer un grupo donde nadie le conoce de antemano, ?que imagen usted le querni dar a su audiencia? [If you are going to meet a group of people who do not know you, what image would you like to give to them?]

(a) la imagen de una persona mejicana [the image of a Mexican person]

(b) la imagen de una persona chicana [the image of a Chicano/a person]

(c) la imagen de una persona americana de los Estados Unidos [the image of an American from the United States]

Traditional Gender Attitudes

1. Las mujeres latinas deben obedecer a sus esposos. [Latin women should obey their husbands.]

2. El hombre debe estar a cargo de las finanzas de la familia. [Men should be in charge of the family's finances.]

3. El hombre latino necesita mas libertad que la mujer latina. [Latin men need more freedom than Latin women.]

4. La mujer debe perdonar si el esposo tiene relaciones fuera del matrimonio. [A woman should be understanding if her husband has extramarital affairs.]

5. Es natural que el hombre sea mas violento que la mujer. [It is natural for men to be more violent than women.]

6. La mujer latina debe ser mas sacrificada que el hombre. [Latin women should make more sacrifices than men.]

7. En las familias latinas el rol del hombre es proteger a la familia. [In Latin families, it is the man's job to protect the family.]

Note. Preparation of this paper was supported by grants to Gregory M. Herek from the National Institute of Mental Health (R01 MH43823 and K02 MH01455). We gratefully acknowledge the invaluable assistance of Adriana Bonilla, Janet Casaverde, Waded Cruzado, Gerardo Medina, Felipa Ortiz, and Modesto Ortiz. We also thank Mary Ellen Chaney and Eric Glunt, as well as the Sacramento- and Davis-area individuals and organizations that provided facilities and assistance for data collection.

Manuscript accepted September 1, 2005

REFERENCES

Alcalay, R., Sniderman, P. M., Mitchell, J., & Griffin, R. (1989-1990). Ethnic differences in knowledge of AIDS transmission and attitudes towards gays and people with AIDS. International Quarterly of Community Health Education, 10, 213-222.

Ayala, G., & Diaz, R. (2001). Racism, poverty and other truths about sex: Race, class and HIV risk among Latino gay men. Revista Interamericana De Psicologia, 35, 59-77.

Bonilla, L., & Porter, J. (1990). A comparison of Latino, Black, and non-Hispanic White attitudes toward homosexuality. Hispanic Journal of Behavioral Sciences, 12, 437-452.

Crawford, I., & Robinson, W. L. (1990). Adolescents and AIDS: Knowledge and attitudes of African-American, Latino, and Caucasian Midwestern U.S. high school seniors. Journal of Psychology and Human Sexuality, 3(2), 25-33.

Diaz, R. M. (1998). Latino gay men and HIV: Culture, sexuality, and risk behavior. New York: Routledge.

Diaz, R. M., Ayala, G., Bein, E., Henne, J., & Marin, B. V. (2001). The impact of homophobia, poverty, and racism on the mental health of gay and bisexual Latino men: Findings from 3 U.S. cities. American Journal of Public Health, 91,927-932.

Flaskerud, J. H., Uman, G., Lara, R., Romero, L., & Otherset, A. (1996). Sexual practices, attitudes, and knowledge related to HIV transmission in low-income Los Angeles Hispanic women. The Journal of Sex Research, 33, 343-353.

Glenn, N. D., & Weaver, C. N. (1979). Attitudes toward premarital, extramarital, and homosexual relations in the U.S. in the 1970s. The Journal of Sex Research, 15, 108-118.

Gonzalez, F. J., & Espin, O. M. (1996). Latino men, Latina women, and homosexuality. In R. P. Cabaj & T. S. Stein (Eds.), Textbook of homosexuality and mental health (pp. 583-601). Washington, DC: American Psychiatric Press.

Herek, G. M. (1984). Beyond "homophobia": A social psychological perspective on attitudes toward lesbians and gay men. Journal of Homosexuality, 10(1-2), 1-21.

Herek, G. M. (1994). Assessing heterosexuals' attitudes toward lesbians and gay men: A review of empirical research with the ATLG scale. In B. Greene & G. M. Herek (Eds.), Lesbian and gay psychology: Theory, research, and clinical applications (pp. 206-228). Thousand Oaks, CA: Sage.

Herek, G. M. (2000a). The psychology of sexual prejudice. Current Directions in Psychological Science, 9, 19-22.

Herek, G. M. (2000b). Sexual prejudice and gender: Do heterosexuals' attitudes toward lesbians and gay men differ? Journal of Social Issues, 56, 251-266.

Herek, G. M. (2002). Gender gaps in public opinion about lesbians and gay men. Public Opinion Quarterly, 66, 40-66.

Herek, G. M., & Capitanio, J. P. (1995). Black heterosexuals' attitudes toward lesbians and gay men in the United States. The Journal of Sex Research, 32, 95-105.

Herek, G. M., & Capitanio, J. P. (1996). "Some of my best friends": Intergroup contact, concealable stigma, and heterosexuals' attitudes toward gay men and lesbians. Personality and Social Psychology Bulletin, 22, 412-424.

Herek, G. M., & Glunt, E. K. (1993). Interpersonal contact and heterosexuals' attitudes toward gay men: Results from a national survey. The Journal of Sex Research, 30, 239-244.

Herman, D. (1997). The antigay agenda: Orthodox vision and the Christian Right. Chicago: University of Chicago Press.

Irwin, P., & Thompson, N. L. (1977). Acceptance of the rights of homosexuals: A social profile. Journal of Homosexuality, 3(2), 107-121.

Kite, M. E., & Whitley, B. E., Jr. (1996). Sex differences in attitudes toward homosexual persons, behaviors, and civil rights: A meta-analysis. Personality and Social Psychology Bulletin, 22, 336-353.

Kite, M. E., & Whitley, B. E., Jr. (1998). Do heterosexual women and men differ in their attitudes toward homosexuality? A conceptual and methodological analysis. In G. M. Herek (Ed.), Stigma and sexual orientation: Understanding prejudice against lesbians, gay men, and bisexuals (pp. 39-61). Thousand Oaks, CA: Sage.

Krosnick, J. A. (1999). Maximizing questionnaire quality. In J. P. Robinson, P. R. Shaver, & L. S. Wrightsman (Eds.), Measures of political attitudes (Vol. 2, pp. 37-58). San Diego: Academic Press.

Manalansan, M. F. (1996). Double minorities: Latino, Black, and Asian men who have sex with men. In R. C. Savin-Williams & K. M. Cohen (Eds.), The lives of lesbians, gays, and bisexuals: Children to adults (pp. 393-415). Ft Worth, TX: Harcourt Brace College Publishers.

Marin, B. V. (2003). HIV prevention in the Hispanic community: Sex, culture, and empowerment. Journal of Transcultural Nursing, 14, 186-192.

Marin, G., & Marin, B. V. (1991). Research with Hispanic populations. Newbury Park, CA: Sage.

Marin, G., Sabogal, E, Marin, B. V. O., & Otero-Sabogal, R. (1987). Development of a short acculturation scale for Hispanics. Hispanic Journal of Behavioral Sciences, 9, 183-205.

Morales, E. S. (1990). HIV infection and Hispanic gay and bisexual men. Hispanic Journal of Behavioral Sciences, 12, 212-222.

Schneider, W., & Lewis, I. A. (1984, Feb.). The straight story on homosexuality and gay rights. Public Opinion, pp. 16-20, 59-60.

Sherrod, D., & Nardi, P. M. (1998). Homophobia in the courtroom: An assessment of biases against gay men and lesbians in a multiethnic sample of potential jurors. In G. M. Herek (Ed.), Stigma and sexual orientation: Understanding prejudice against lesbians, gay men, and bisexuals (pp. 24-38). Thousand Oaks, CA: Sage.

Gregory M. Herek

University of California at Davis

Milagritos Gonzalez-Rivera

University of Puerto Rico, Mayaguez

Address correspondence to Gregory M. Herek, Department of Psychology, University of California, 1 Shields Avenue, Davis, CA, 95616-8686.

Table 1. Response Means and Frequencies for ATL and ATG Items

                                                 Mean
Attitudes Toward Gay Men (ATG)                   (SD)    % Agree

(ATG WRONG): Sex between two men is just
  plain wrong. [Las relaciones sexuales           1.85
  entre dos hombres simplemente estan mal.]      (1.6)     35.9

(ATG DISGUST): I think that male homosexuals
  are disgusting. [Yo pienso que los hombres      1.89
  homosexuales son repugnantes.]                 (1.4)     30.7

(ATG NATURAL): * Male homosexuality is a
  natural expression of sexuality in men. [La
  homosexualidad masculina es una expresion       2.21
  natural de la sexualidad del hombre.]          (1.4)     39.5

(ATG PERVERSION): Male homosexuality is a
  perversion. [La homosexualidad masculina        1.74
  es una perversion.]                            (1.4)     26.9

Attitudes Toward Lesbians (ATL)

(ATL WRONG): Sex between two women is just
  plain wrong. [Las relaciones sexuales entre     1.75
  dos mujeres simplemente estan mal.]            (1.6)     32.1

(ATL DISGUST): I think that lesbians are
  disgusting. [Yo pienso que las lesbianas        1.82
  son repugnantes.]                              (1.6)     28.5

(ATL NATURAL): * Female homosexuality is a
  natural expression of sexuality in women.
  [La homosexualidad femenina es una expresion    2.06
  natural de la sexualidad de la mujer.]         (1.3)     33.5

(ATL PERVERSION): Female homosexuality is a
  perversion. [La homosexualidad femenina es      2.24
  una perversion.]                               (1.5)     42.8

                                                    %          %
Attitudes Toward Gay Men (ATG)                   Disagree   Neither

(ATG WRONG): Sex between two men is just
  plain wrong. [Las relaciones sexuales
  entre dos hombres simplemente estan mal.]        44.8       19.3

(ATG DISGUST): I think that male homosexuals
  are disgusting. [Yo pienso que los hombres
  homosexuales son repugnantes.]                   36.7       32.7

(ATG NATURAL): * Male homosexuality is a
  natural expression of sexuality in men. [La
  homosexualidad masculina es una expresion
  natural de la sexualidad del hombre.]            27.3       33.2

(ATG PERVERSION): Male homosexuality is a
  perversion. [La homosexualidad masculina
  es una perversion.]                              42.0       31.1

Attitudes Toward Lesbians (ATL)

(ATL WRONG): Sex between two women is just
  plain wrong. [Las relaciones sexuales entre
  dos mujeres simplemente estan mal.]              47.1       20.8

(ATL DISGUST): I think that lesbians are
  disgusting. [Yo pienso que las lesbianas
  son repugnantes.]                                37.2       34.3

(ATL NATURAL): * Female homosexuality is a
  natural expression of sexuality in women.
  [La homosexualidad femenina es una expresion
  natural de la sexualidad de la mujer.]           30.1       36.4

(ATL PERVERSION): Female homosexuality is a
  perversion. [La homosexualidad femenina es
  una perversion.]                                 27.1       30.1

* The NATURAL items were dropped from the final ATL and ATG
subscales.

Note. Item responses can range from 0 to 4, with higher values
indicating more agreement with the statement.

Table 2. Reliability Coefficients for ATL and ATG (3-Item
Versions) By Respondent Educational Level and
Language Reading Preference

                              Highest Educational Level

Reading Preference       Less Than 12 Years   12 Years or More

Mainly Spanish
  ATG                           .45                 .58
                             (n = 160)            (n = 41)
  ATL                           .54                 .71
                             (n = 165)            (n = 41)
Mainly English or both
  English and Spanish
  ATG                           .82                 .78
                              (n = 60)           (n = 288)
  ATL                           .79                 .77
                              (n = 61)           (n = 283)

Note. Table reports coefficient alpha for each cell, based on 3-item
versions of the ATL and ATG. Variations in sample size between ATL
and ATG scores reflect missing data.

Table 3. Mean ATG and ATL Scores By Theoretically
Relevant Categorical Variables

                                         ATG               ATL

Sex
  Male                             5.93 (3.7) (a)    6.01 (3.4)
  Female                           5.22 (3.4) (b)    5.69 (3.5)
Educational Level
  Less than High School            5.91 (3.5) (a)    6.48 (3.3) (a)
  High School Diploma              6.11 (3.5) (a)    6.32 (3.4) (a)
  Any Post-Secondary Education     4.91 (3.4) (b)    5.10 (3.6) (b)
Marital status
  Single                           5.23 (3.5)        5.64 (3.5)
  Married/Widowed                  5.84 (3.6)        6.11 (3.5)
  Separated/Divorced               6.34 (3.2)        6.34 (3.5)
Religious attendance
  Never                            5.39 (3.4) (b)    6.31 (3.6) (bc)
  Once or twice                    5.25 (3.6) (b)    4.93 (3.5) (c)
  Several times                    4.80 (3.4) (c)    5.20 (3.4) (c)
  Once or twice a month            5.07 (3.3) (c)    5.50 (2.8) (bc)
  About once a week                6.61 (3.4) (ab)   6.81 (3.6) (b)
  More than weekly                 6.95(4. 1) (a)    7.97 (3.4) (a)
Religious affiliation
  No affiliation                   4.32 (3.5) (a)    4.52 (3.4) (a)
  Catholic                         5.50 (3.4) (a)    5.87 (3.4) (a)
  Conservative religious
    affiliation                    7.27 (4.1) (b)    7.93 (3.3) (b)
  Other denomination               5.81 (4.0) (a)    5.32 (3.8) (a)
Political ideology
  Conservative                     7.08 (3.6) (a)    7.47 (3.4) (a)
  Not conservative                 5.36 (3.5) (b)    5.66 (3.5) (b)
Number of homosexuals known
  None                             6.18 (3.9) (a)    6.55 (3.7) (a)
  One                              6.36 (3.6) (a)    6.27 (3.6) (a)
  Two                              5.55 (3.2) (a)    6.02 (3.4) (a)
  Three                            6.08 (3.8) (a)    5.94 (3.8) (a)
  Four                             5.46 (3.4) (a)    5.58 (3.5) (a)
  Five or more                     3.73 (3.7) (b)    3.88 (4.2) (b)
Ethnic association with gay men
  Mexican                          5.72 (3.8)        6.24 (3.7)
  Other Latino                     5.55 (3.6)        6.43 (3.3)
  Non-Hispanic White               5.71 (3.8)        5.74 (4.0)
  Other                            5.95 (4.2)        5.75 (4.1)
Ethnic association with lesbians
  Mexican                          5.46 (3.9)        6.07 (4.0)
  Other Latino                     5.45 (3.3)        6.29 (3.5)
  Non-Hispanic White               5.75 (3.8)        5.77 (3.9)
  Other                            6.06 (4.2)        6.10 (4.1)

Note. ATG = Attitudes Toward Gay Men (3 items); ATL = Attitudes
Toward Lesbians (3 items). Table reports mean scale scores and (in
parentheses) standard deviations. The possible range for ATG and ATL
scale scores is 0-12, with higher scores indicating more negative
attitudes (i.e., higher levels of sexual prejudice). Within each
variable, ATG or ATL scores with different superscripts are
significantly different using MANOVA with p < .05. For variables with
3 or more categories, differences were tested with the Student Newman
Keuls procedure (p < .05).

Table 4. Zero-Order Correlations of ATL and ATG Scores
with Other Theoretically Relevant Variables

                                                ATG        ATL

Age                                            .08 (a)    .09 (a)
Education (number of years)                   -.15 (c)   -.16 (c)
Number of children                             .10 (b)    .13 (c)
Traditional gender attitudes                   .32 (c)    .32 (c)
Number of gay men/lesbians personally known   -.22 (c)   -.23 (c)
Religious Attendance (past year)               .15 (c)    .17 (c)
Latino group identification                   -.03        .04
Acculturation scale                           -.12 (b)   -.12 (b)
Importance of Mexican cultural activities      .05        .04
Importance of U.S. cultural activities         .00        .07
Cultural preferences scale                    -.03       -.07

Note. ATG = Attitudes Toward Gay Men (3 items); ATL = Attitudes
Toward Lesbians (3 items). All statistical tests are one-tailed.
Higher ATG and ATL scores indicate more negative attitudes.

(a) p < .05 (b) p < .01 (c) p < .001

Table 5. Predictors of ATG and ATL Scores (ULS Regression Anaiysis)

                                                     ATG

Variable                                   b        [beta]   [R.sup.2]

Traditional gender beliefs              0.181 (c)    .271      .030
Ethnic identification                   3.231 (a)    .417      .003
Language preference                     2.002 (a)    .240      .012
Latino group identification            -0.225 (a)   -.116      .000
Education (# years)                     0.129       ...        ...
Gender                                  1.526       ...        ...
# Gay/lesbian friends                   0.158       ...        ...
Number of children                     -0.184       ...        ...
Mexican cultural importance             0.120       ...        ...
U.S. cultural importance                0.058       ...        ...
Cultural preferences scale             -0.030       ...        ...
Religious attendance                    0.344       ...        ...
Catholic denomination                  -0.230       ...        ...
Conservative Christian denomination     0.518       ...        ...

Language x # Gay friends               -0.790 (c)   -.409      .019
Ethnic identification x Education      -0.209 (a)   -.392      .013
Ethnic identification x Number of
  children                              0.575 (b)    .207      .005
Sex x Education                        -0.167       ...        ...
Sex x Religious Attendance               .160       ...        ...
Language x Catholic Denomination        -.353       ...        ...
Language x Conservative Christian
  Denomination                          1.024

Adjusted [R.sup.2] (Total Equation)                  .310

                                       ATG

Variable                                   b        [beta]   [R.sup.2]

Traditional gender beliefs              0.132 (b)   .193       .055
Ethnic identification                   3.622 (a)   .459       .010
Language preference                     2.032 (a)   .240       .006
Latino group identification            -0.007       ...        ...
Education (# years)                     0.101       ...        ...
Gender                                  1.333       ...        ...
# Gay/lesbian friends                   0.126       ...        ...
Number of children                     -0.029       ...        ...
Mexican cultural importance             0.043       ...        ...
U.S. cultural importance                0.109       ...        ...
Cultural preferences scale             -0.123       ...        ...
Religious attendance                    0.075       ...        ...
Catholic denomination                  -0.483       ...        ...
Conservative Christian denomination     0.950

Language x # Gay friends               -0.653 (b)   -.332      .029
Ethnic identification x Education      -0.250 (a)   -.459      .010
Ethnic identification x Number of
  children                              0.286       ...        ...
Sex x Education                        -0.195 (a)   -.337      .008
Sex x Religious Attendance              0.477       ...        ...
Language x Catholic Denomination       -0.191       ...        ...
Language x Conservative Christian
  Denomination                          1.488       ...        ...

Adjusted [R.sup.2] (Total Equation)                 .292

Note. Table reports unstandardized (b) and standardized ([beta]) OLS
regression coefficients. Percentage of explained variance ([R.sup.2])
was computed for main effects variables on Step 1 and for interactions
on Step 2. Nonsignifcant [beta]s and [R.sup.2] are not shown. For
ethnic identification, 1 = Chicano/Mexican American. For language
preference, 1 = English or English and Spanish equally. For sex, 1 =
Female. For Catholic denomination, 1 = Catholic, 0 = Conservative
Christian, -1 = No religion or other religion. For Conservative
Christian denomination, 1 Conservative Christian, 0 = Catholic, -1 =
No religion or other religion.

(a) p < .05 (b) p < .01 (c) p < .001