Beyond traditional gender roles and identity: does reconceptualisation better predict condom-related outcomes for African-American women?

Abstract

African-American women continue to be at high risk for HIV and better prevention efforts are needed. The current paper sought to investigate the relationship between gender roles and condom-related outcomes among African American women. The sample consisted of 398 African-American women, who were administered a survey that contained measures of condom-related outcomes and gender role beliefs. We factor analysed their responses and three domains emerged: caretaking/mindful, interpersonal sensitivity and persistent/active coping. Results indicated that the interpersonal sensitivity domain was a significant predictor of condom use and intention with higher interpersonal sensitivity scores associated with less condom use and intentions. The persistent/active coping domain was a significant predictor of condom negotiation efficacy and condom use with higher scores in this domain associated with more condom negotiation efficacy and use. Results suggest that re-conceptualisations offer a better understanding of underlying traits that may influence condom-related outcomes for this population.

Introduction

HIV/AIDS is a significant health risk for African-American women in the USA. Although African-American women comprise 13% of the female population (US Census 2007), they make up 64% of reported cases of HIV/AIDS among women (Centers for Disease Control and Prevention (CDC) 2008). In 2004, HIV/AIDS was the leading cause of death among African-American women aged 25–34 years (CDC 2008). Heterosexual contact is the primary mode of HIV transmission for African-American women and accounts for 83% of HIV/AIDS diagnoses (CDC 2008). There are numerous social, economic, environmental, cultural and personal factors that place African-American women at risk for HIV. The current study focuses on factors that promote consistent condom use.

Sexual attitudes and behaviours are influenced by culturally determined gender norms (Amaro et al. 2001; Noar and Morokoff 2002). These gender norms convey the attitudes and behaviours that are deemed appropriate for women and men’s social and sexual interactions. Hence, women’s negotiation of condom use is influenced by their gender role beliefs and this relationship is the focus of the present study. In this paper, we identify the underlying constructs for African-American women’s gender role beliefs as these beliefs are likely to increase our understanding of African-American women’s safer sexual behaviour, including condom use.

Defining gender, gender identity and gender roles

What does it mean to be a woman and, more specifically, what does it mean to be an African-American woman? These questions were more easily answered a few decades ago when traditional gender roles stereotypes were ascribed to women (Lipman-Blumen and Tickamyer 1975). While sex refers to biological identity as male or female, gender refers to the social identities attributed to women and men (Andersen and Hill Collins 2001). The transition to the preference of the term gender instead of sex was empowering for women because it reduced limitations placed on them solely based upon their physical attributes and emphasized the social construction of gender roles (Bernard 1971). Not all women are oppressed in the same ways despite holding the common minority status of being female (Andersen and Hill Collins 2001). For instance, bell hooks (1981) asserted that sexism operates independently of and simultaneously with racism to oppress African-American women. Gender, race and class intersect to vary women’s experiences of their gender according to race/ethnicity and socioeconomic status (see Andersen and Hill Collins [2001] and Rothenberg [2007] for a more in-depth discussion of the intersection of race, gender and class). Findings from previous research cannot necessarily be generalised to all African-American women if there is not adequate representation of African-American women in the sample. Similarly, measures that have not been validated with African-Americans may not be appropriate.

Gender identity is conceptualised as an individual’s sense of being male or female regardless of biological sex and can be shaped by social factors, such as family and community norms. Gender roles refer to beliefs regarding divisions of specific tasks and differences in behaviours, abilities and personality that society expects men and women to hold (Littlefield 2003). Gender, gender identity and gender role are strongly influenced by ‘socio-cultural-sociological-psychological’ norms (Bernard 1971). Constructs such as race and class are influenced by and influence the structure of social institutions that include work, families, mass media and education. Similarly, gender oppression and domination are perpetuated by these institutions. In turn, gender, gender identity and gender roles influence social institutions. Researchers have examined gender role and its relation to condom attitudes, condom-related efficacy, condom use intention and condom use behaviour. Some of these studies are described next.

Gender roles and condom use

Common measures of gender roles, the Bem Sex Role Inventory (BSRI: Bem 1974) and the Personal Attributes Questionnaire (PAQ: Spence, Helmreich, and Stapp 1974) classify individuals into four categories: ‘masculine’, ‘feminine’, ‘androgynous’ or ‘undifferentiated’ based on the psychological traits they endorse. Instrumental traits emphasise independence, mastery and assertiveness; they are typically found in masculinity subscales. Expressive traits refer to emotional openness, sensitivity to others’ needs and cooperativeness; they are typically assessed in femininity subscales. Individuals who score high on both the ‘masculine’ and ‘feminine’ scales are categorised as androgynous while individuals who score low on both are classified as undifferentiated (Littlefield 2003).

Using this construction of gender role, researchers have found that ‘feminine’ gender role orientations may contribute to a lack of effective communication about sexual practices, particularly the use of condoms (Fullilove et al. 1990), while traits traditionally viewed as ‘masculine’, such as assertiveness, may lead to increased condom use (e.g. Roberts and Kennedy 2006). Studies also indicate that African-American women do not endorse stereotypical feminine identities on measures such as the BSRI (Littlefield 2003). It seems that given the gender role traits African-American women hold, they should be more comfortable with refusing to engage in unsafe sex. Still, the current incidence of HIV/AIDS infection among African-American women suggests that these gender roles may not apply within the context of their romantic relationships. Binion (1990) found that although African-American women reported androgynous gender identities, they held traditional beliefs about the female role in the family.

Problems with traditional conceptualisations of gender roles for African-American women

The empirical studies on gender role and condom use are limited by their focus on White, middle-class women. The PAQ and the BSRI measures used in this study also have these limitations because they were originally developed with predominantly White female participants reporting on traits that they perceived to be either masculine or feminine. This is a key limitation, given the findings that suggest that African-American women’s gender roles differ from the normative expectation of women in the USA (Collins 2000; Davis 1981; Zinn and Dill 1996). For example, research indicates that the archetypal mainstream female gender role emphasises sensitivity, passivity, dependence, submissiveness and unselfishness (Stockard and Johnson 1980; Thomas 1986). On the other hand, African-American women’s archetypal gender role emphasises strength, independence and nurturance (Davis 1981; Stack 1991; Zinn and Dill 1996). For this reason, the original subscales developed on the PAQ and BSRI may not accurately reflect African-American women’s perceptions of what it means to be masculine, feminine or androgynous.

In general, African-American females are socialised to incorporate traditionally feminine identity traits, such as nurturing, and masculine identity traits, such as assertiveness, into their identity as women (Littlefield 2003). African-American women also emphasise the importance of strength when describing their identity (Shorter-Gooden and Washington 1996). These gender role beliefs reflect their historical experiences in the USA, including their roles during slavery and its aftermath. These gender role beliefs have helped African-American women to succeed within their families and the workforce (Ashcraft and Belgrave 2005).

In the context of sexual interactions, assertiveness or the ability to effectively communicate one’s needs and desires is a salient factor in women’s abilities to negotiate condom use with their male partners. African-American women may feel pressured to adjust their assertiveness to make the men in their lives comfortable with their masculinity (Jones and Shorter-Gooden 2003). Jones and Shorter-Gooden described the process of ‘shifting’ as changing attitudes, behaviour or tone to accommodate others’ expectations related to gender, class or ethnicity. African-American women may feel increased pressure to shift gender-related beliefs and behaviours for the sake of relationship maintenance given the sex-ratio imbalance or the higher rates of mortality and incarceration for men than women (Mays and Cochran 1988, 949; McNair and Prather 2004, 106).

We are interested in examining the cluster of traits, outside the realm of traditional gender roles, that are associated with condom-related outcomes for African-American women. Is it assertiveness, nurturance, independence or another trait? This primary question led to the present study. We speculate that it is possible that there are clusters of traits that may not be related to specific gender domains that may be better predictors of condom use outcomes than the traditional gender domains for African-American women. That is, traditional conceptualisations of gender role may be more appropriate for White women but, perhaps, re-conceptualisations of these constructs will prove to be beneficial for understanding gender roles among African-American women. This research is important for generating knowledge, but also for the development of effective programming.

The purpose of the current study was to explore whether unique patterns emerged from the items on the PAQ and BSRI measures (measures of gender roles) for an African-American sample and whether these items were predictive of condom-related outcomes. We predicted that African-American women will have unique gender-role profiles as indicated by the domains and patterns that emerge from a factor analysis using items from the PAQ and BSRI. We believed that these new domains would offer a better conceptualisation of the kinds of attributes that influence condom-related outcomes for African-American females.

Methods

Participants

Participants included 398 African-American women who identified as unmarried and heterosexual. Women were recruited from an urban city in the south-eastern region of the USA. Participants were invited to participate in an HIV prevention intervention that was advertised as a study on healthy relationships and healthy behaviour. This project involved the implementation and evaluation of an intervention to promote safer sex practices among African-American women with the larger goal of decreasing the incidence of HIV/AIDS within the population. Participants were recruited from churches, community organisations, two predominantly Black universities and one predominately White university. Participants’ ages ranged from 18 to 45, with a mean age of 21.27 years (SD = 5.37). Most participants (84%) were between the ages of 18 and 23 years. Approximately half (51%) of participants reported that they had a ‘main partner’.

Of participants, 77% were recruited from universities and 23% were recruited from churches and community organisations. In terms of level of education, 39% reported their highest level of education as having completed high school/General Education Diploma or less; 61% reported that they attended ‘some college’ or received a college or graduate degree. With regards to employment status, 54% of the sample reported being currently unemployed, 33% reported having part-time employment and 13% report being employed full-time.

Measures

Measures were obtained from the measurement packet of the Sisters Informing Sisters about the Topic of AIDS (SISTAS: Wingood and DiClemente 2000) that were used to assess programme impact (see section on procedures).

Demographic variables

The demographic variables were measured using single items. Participants responded to a single item ‘how old are you?’ to determine their age. Participants were asked, ‘do you currently have a main partner?’ to determine their relationship status. We used age and relationship status as covariates in the study. The link between age and sexual risk taking is a complex one. Research has shown that younger age is associated with risky sexual behaviour (Hallet et al. 2007; Masatu et al. 2009). However, the relationship appears to be curvilinear; increasing age seems to predict increasing condom use up to a point. As people become older, safe sex practices begin to decline as evidenced by literature that shows that older ages predict less safe sex practices (Corneille, Zyzniewski, and Belgrave 2008; Holmes et al. 2008). As a result, we controlled for age in our analyses.

In addition, the literature provides evidence that people in monogamous relationships engage in less preventive sexual behaviours than people in multiple casual relationships (Mehrotra et al. 2009; Misovich, Fisher, and Fisher 1997; Noar, Zimmerman, and Atwood 2004). Individuals with a main partner feel safer during intercourse. They perceive risk of contracting HIV or other STDS from unprotected sex with a casual partner as being much higher than the perceived risk associated with a main sexual partner (Mehrota et al. 2009). As a result, we controlled for relationship status in our analyses.

Gender roles

Gender roles were assessed using a modified version of the gender role scale from the BSRI. This 6-item scale measures gender role beliefs based on subscales for traditionally feminine and traditionally masculine traits. Traits from the traditionally masculine subscale are ‘independent’, ‘assertive’ and ‘aggressive’. Traits from the traditionally feminine subscale are ‘emotional’, ‘warm’ and ‘compassionate’. The response options are ratings from 1 = ‘not at all’ to 5 = ‘very much’. The Cronbach’s alpha for the feminine subscale was .62, the masculine subscale was .28. The reliability of these sub-scales was so low in part because there were only three items in each sub-scale.

Gender roles were also assessed using the PAQ, which consists of 24 descriptive statements that were developed to assess masculine and feminine personality attributes. The PAQ has three subscales, each subscale ‘masculine’, ‘feminine’ and ‘masculine-feminine’ (or ‘androgynous’) has eight items. The items are rated on a 5-point scale that has an adjective at one end with its presumed opposite at the other end (very passive–very active) or with its divergent adjective at the opposite pole (not at all competitive–very competitive). In the current sample, the Cronbach reliability coefficient the feminine subscale was .76, the masculine subscale was .48 and the androgynous scale was .47. For more descriptive information relating to BSRI and PAQ items, please refer to Table 1.

Table 1.

Condom-related measures and outcomes.

Variable	Mean	s.d.	min	max
Condom Attitude	12.25	3.73	7	28
Condom Negotiation	24.35	3.52	9	28
Condom Use Self-Efficacy	20.39	5.07	9	27
Condom Use Intention	12.04	2.13	4	15
Condom Use Consistency (Past 3 Mos.)	0.75	0.39	0	1
BSRI Feminine	11.98	2.10	4	15
BSRI Masculine	11.42	2.99	5	62
PAQ Masculine	29.29	3.77	18	37
PAQ Feminine	31.51	4.27	11	40
PAQ Androgynous	24.44	3.76	15	35

Dependent variables

Several measures were used to assess condom behaviours, attitudes, intentions and efficacy. ‘Condom use consistency’ was determined by asking participants to report the number of times they had engaged in sex in the past three months. Participants were then asked the number of times they used a condom in the past three months. A ratio was computed to determine the frequency of condom use. Additionally, condom use consistency was assessed by asking participants whether they had used a condom the last time they had sex. Response options were ‘yes’ or ‘no’.

The ‘condom use intention’ scale measured women’s intent to use condoms in the future through the use of three dichotomous questions. Items from the condom intention scale include: (1) The next time you have sex, do you plan to use a condom?; (2) In the next three months, do you plan on using a condom if you have sex?; and (3) In the next three months, do you plan on using a female condom if you have sex?

‘Condom attitudes’ were measured by a 4-item scale that assessed women’s attitudes towards condoms. An example item is, ‘Sex with condoms does not feel natural’. Participants responded on a 4-point Likert scale, with lower scores indicating more favourable attitudes towards condoms. The Cronbach reliability coefficient was .86.

‘Condom negotiation efficacy’ was measured with a 7-item scale designed to assess women’s efficacy for negotiating condom use with a primary partner. A sample item includes, ‘Can you insist on condom use if your main partner does not want to use one?’ Participants responded on a 4-point Likert scale, with higher scores indicating higher condom negotiation efficacy. The scale had a Cronbach reliability coefficient of .85 in the current sample.

The ‘condom use efficacy’ scale assessed participants’ efficacy for properly using the male condom. A sample item includes ‘How confident are you that you could put a condom on a hard penis?’ Response options ranged from 1 (not confident) to 3 (very confident). The Cronbach reliability coefficient for the condom use efficacy scale was .92 with the study sample. For more descriptive information relating to dependent variables, please reter to Table 1.

Procedure

Participants were recruited for participation via flyers at universities, churches, clinics and community centres. The flyers invited African American women to participate in a discussion group on healthy relationships. After participants arrived, trained peer facilitators introduced themselves and gave a brief description of the programme. Informed consent was obtained and questions were answered prior to the distribution of the surveys. Participants were informed that their responses were confidential and that they were not required to complete any items that they were uncomfortable answering. Participants were reminded that participation was voluntary. After the consent forms were collected, a research assistant administered the pencil-and-paper surveys. Participants were provided with a US$10 gift card to a local department store. Only baseline data was used in this study.

Data analysis

In order to test the study’s hypotheses, factor analysis, hierarchical multiple regression and binary logistic regression analyses were computed. Factor analysis was employed to determine if the items from the PAQ and BSRI would emerge into meaningful domains. Later reliability analyses indicated stronger factors with the omission of BSRI items. As a result, items from the BSRI were subsequently dropped. We then tested the factors that emerged to investigate whether the new domains influenced sexual outcomes for African-American women. Preliminary analyses were conducted to screen data for violations of the assumptions of multiple regression and also to check the reliability coefficients of our subscales. Age and relationship status were controlled and entered in the first step. Predictor variables were centred. The new domains that emerged from the factor analysis were entered in the second step. Logistic regression analysis was computed for the dichotomous condom use outcome variable.

Results

Preliminary analyses

We predicted that traditional gender role measures would not reliably measure defined gender roles in a female African-American population. Therefore, we computed Cronbach reliability analyses to illustrate this point. We computed the reliability coefficients of subscales from the PAQ. As expected, the reliability coefficients were low. In the current sample, the Cronbach reliability coefficient for the overall measure was .72, the feminine subscale was .76, the masculine subscale was .48 and the androgynous scale was .47. These levels are comparable to those found by Shifren and Bauserman (1996, 841) (α_m = .51 and α_f =.80) though slightly lower than alpha levels reported from Woodie and Fromuth (2009, 318) (α_m = .82 and α_f = .78) and Toller, Suter and Trautman (2004, 85) (α_m = .75 and α_f = .79 α_a = .65). All of these studies used multi-ethnic mixed gender samples.

In addition, we computed the reliability coefficients of the subscales found in the BSRI. As expected, the reliability coefficients were low. The Cronbach’s alpha for the feminine subscale was .62, the masculine subscale was .28 and the reliability of the overall measure was .45. The reliability of the BSRI sub-scales may be so low in part because there were only three items in each sub-scale. In a study using a 30-item shortened version of the BSRI, Littlefield (2003) reported alpha levels of (α_f = .86 and α_m = .77) with a sample of African-American women. Konrad and Harris (2002) and Harris (1996) reported similar alpha levels (α_f = .88, .93 and α_m = .84, .78, respectively) in a 40-item version of the BSRI with a sample of African-American men and women. While these alpha levels are quite a lot higher than those reported in the current study, these alphas were based upon subscales made up of more than three items.

The low reliability of the scales used to measure gender-role orientation in our African-American sample provides some support that traditional conceptualisations of gender roles may not be appropriate for African-American female populations. The next step was to examine whether unique patterns emerged among the items in the gender-role and identity scales and to determine if these new domains were better predictors of condom-related outcomes.

Factor analysis

We were interested in examining different meaningful interrelationships among the items used to measure traditional gender roles. We used factor analysis to determine if the items from the BSRI and PAQ would emerge into meaningful factors that were unique from traditional feminine and masculine domains. A factor analysis was run with a varimax rotation. Three factors emerged from the analysis. The Kaiser-Meyer-Olkin measure of sampling adequacy was .75, above the recommended value of .6, and Bartlett’s test of sphericity was significant, χ² (276) = 1244.63, p = .00. The communalities were all above .3, confirming that each item shared some common variance with other items.

The initial eigenvalues showed that the first factor explained 17% of the variance, the second factor 15% of the variance and a third factor 8% of the variance. Refer to Table 2 for factor loadings. Factor 1 (Cronbach’s α = .84) had items that were related to qualities of nurturance and warmth, so it was named the ‘caretaking/mindful of others’ domain. Factor 2 (Cronbach’s α = .80) had items that were related to dependency and emotional instability and was named the ‘interpersonal sensitivity’ domain. Factor 3 (Cronbach’s α = .71) had items that were related to perseverance and positive coping and was named the ‘persistent/active coping’ domain. It is important to remind the reader that items from the BSRI were dropped from the analyses altogether as the reliability of the factors improved once these items were omitted. Altogether, six out of the original 24 items comprised the two sub-scales.

Table 2.

Factors and item loadings.

Caretaking/Mindful of Others α = .84

Very helpful to others

Very kind

Aware of other’s feelings

Understanding of others

Interpersonal Sensitivity α = .80

Very emotional

Highly needful of other’s approval

Feelings easily hurt

Cries very easily

Strong need for security

Persistent/Active Coping α = .71

Very aggressive

Very active

Very competitive

Never gives up easily

Stands up well under pressure

Hierarchical multiple regression

A hierarchical multiple regression analysis was conducted to predict condom-related outcomes. Using ‘condom attitudes’ as the outcome, age and relationship status were controlled for and entered in the first step of the regression. Items used to predict condom attitudes included scores in the caretaking/mindful of others domain, interpersonal sensitivity domain and persistent/active coping domain and were entered into the second step. The results of the analysis indicated that the model did not account for a significant amount of variance in condom attitudes, F (5,116) = .79, p ≥ .05; R² = .02.

A hierarchical multiple regression analysis was conducted to predict condom negotiation efficacy using the same procedure as described above. The results of the analysis indicated that the model accounted for a significant amount of variance in condom negotiation self-efficacy, F (5,110) = 2.32, p ≤ .05; R² = .10. Scores in the persistent/active coping domain was a significant predictor of condom negotiation self-efficacy, β = .22, t(111) =22.22, p ≤ .05, indicating that higher scores in the persistent/active coping domain predicted higher condom negotiation self-efficacy.

A hierarchical multiple regression analysis was conducted to predict condom use efficacy using the same procedure as described above. The results of the analysis indicated that the model accounted for a significant amount of variance in condom use self-efficacy, F (5,117) = 2.27, p ≥ .05; R² = .09, but none of the variables offered any meaningful explained variance above and beyond the covariates.

A hierarchical multiple regression analysis was conducted to predict condom use intention using the same procedure as described above. The results of the analysis indicated that the model accounted for a significant amount of variance in condom use intention, F (5,137) = 3.10, p ≤ .01; R² = .10. Scores in the interpersonal sensitivity domain were a significant predictor of condom use intention, β = −2.12, t(141) =−2.40, p ≤ .05, indicating that higher scores in the interpersonal sensitivity domain predicted lower condom use intention.

A hierarchical multiple regression analysis was conducted to predict condom use in the past three months using the same procedure as described above. The results of the analysis indicated that the model accounted for a significant amount of variance in the ratio for condom use in the past three months, F (5,115) = 2.31, p ≤ .05; R² = .10. Scores in the interpersonal sensitivity domain were significant predictors of the ratio for condom use in the past three months, β = −.23, t(114) = 2.54, p ≤ .01, indicating that higher scores in the interpersonal sensitivity domain predicted lower condom use in the past three months.

A logistic regression analysis was conducted to predict whether or not condoms had been used during the last sexual encounter (0 = no, 1 = yes). Age and relationship status were controlled for and entered in the first step of the regression. Items used to predict condom use during the last sexual encounter included scores in the caretaking/mindful of others, interpersonal sensitivity and persistent/active coping domains and were entered into the second step. The model was significant χ²(3) = 23.46, p ≤ .01. The Nagelkerke R-square value = .24. The variables correctly predicted 80% of the cases. According to the Wald criterion, the persistent/active coping domain was the only significant predictor of whether or not condoms were used during the last sexual encounter, χ²(1) = 4.86, p ≤ .05. The change in odds associated with a one-unit change in persistent/active coping scores was 1.16, indicating that a one-unit increase in scores in the persistent/active coping domain resulted in a participant being 1.16 times more likely to have used a condom during her last sexual encounter.

Discussion

Researchers have sought to understand sexual behaviour and condom use among African-American women by using traditional gender role variables, such as masculinity and femininity. One of the problems with previous research is the reliance upon measures that have not been developed or validated using African-American women participants. There are many reasons why these traditional measures have not captured the gender role beliefs and orientation of African-American women. First, gender roles and beliefs are developed within a social context and African-American women are confronted with social injustices due to ethnicity and gender. Second, Western European history has influenced the conceptualisations of human differences in dichotomous opposition, such as good/bad, high/low, happy/sad, superior/inferior (Lorde 1984). In many ways, the use of masculine/feminine as descriptors of gender role helps to maintain this simplistic thinking and may counter the thinking of African-Americans and other groups whose thinking may be more holistic (Belgrave and Allison 2006). Third, it could be argued that by categorising and labelling terms such as assertiveness and strength as masculine, systematised oppression is maintained.

We predicted that traditional conceptualisations of gender roles would be inappropriate for an African-American sample. This was confirmed by the poor reliability found for the BSRI and PAQ for our sample population. We predicted that different domains would emerge from a factor analysis of items from the gender identity (PAQ) scale that would fall outside the traditional conceptualisations of female versus male gender role domains. Our predictions were confirmed in that the analysis produced three factors that we named ‘caretaking/being mindful of others’, ‘interpersonal sensitivity’ and ‘persistent/active coping’. These new factors all had reliability coefficients above .70, which are greatly improved above the low reliability indexes of the original factors in the PAQ and BSRI. It is important to note that BSRI items were eventually omitted from our analyses. Reliability of the emergent factors was greatly improved once these items were dropped as we were able to reduce multi-collinearity within our factors. As noted by Spence (1991) in a comparative study of the two measures, the correlation between the BSRI and PAQ was quite high. Because both the BSRI and PAQ had similar properties of discriminant and convergent validity with other measures, the conclusion was that they both measured the same construct.

It was also hypothesised that these re-conceptualisations of gender roles would be associated with condom-related outcomes. Our predictions were confirmed as these new domains offered plausible and conceptually grounded interpretations of the kinds of individual traits and attributes that would affect condom-related outcomes. We found that the interpersonal sensitivity domain was a significant predictor of decreased condom use intention, indicating that higher scores in the interpersonal sensitivity domain predicted lower condom use intention. We found that the interpersonal sensitivity domain was a significant predictor of condom use in the past three months, indicating that higher scores in the interpersonal sensitivity domain predicted lower condom use in the past three months. The interpersonal sensitivity domain is characterised by a ‘neediness of others’ and it is likely that this is the attribute that would place women at higher sexual risk. Women with a high need for approval from others may refrain from being assertive in conversations regarding condom use. This may be an attempt to put the least amount of strain on their social relationships.

Results indicated that the persistent/active coping domain was a significant predictor of condom negotiation efficacy, indicating that higher scores in the active/persistent domain predicted higher condom negotiation self-efficacy. We also found that the persistent/active coping domain was the only significant predictor of condom use during the last sexual encounter, indicating that higher scores in the persistent/active coping domain predicted females being more likely to have used a condom during their last sexual encounter. Women who are high in persistence and active coping are likely to be able to communicate their needs and desires and to be less persuaded by conflicting requests from others; such qualities may improve women’s success at negotiating condom use particularly in instances of partner resistance.

Interestingly, the caretaking/mindful of others domain was not a significant predictor of any of the condom outcomes. This domain is the one that seems most similar to what is traditionally thought of as feminine gender roles. This domain also seems akin to communalism, an orientation toward the inclusion of others found among most people of African descent (Belgrave and Allison 2006). The lack of significance for this trait cluster raises questions about its contributions to understanding condom use and more research is needed.

We believe that the caretaking/mindful of others domain is different from the interpersonal sensitivity domain. Women who score high in the caretaking/mindful of others domain may be more oriented toward the needs of others, whereas women who score high in the interpersonal sensitivity domain may be more oriented toward their own needs and may be described as ‘needy’. Women who score high in the caretaking/mindful of others domain and interpersonal sensitivity domain are both likely to seek emotional intimacy; however, the motivations underlying these two groups may differ as provided in the avoidance-approach framework (Ellliot and Thrash 2002; Gray 1982). Those who are more self-oriented and emotionally needy may be more likely to do whatever they believe is necessary to avoid losing their partner (i.e. having sex without condoms) regardless of associated risks.

High interpersonal sensitivity as a significant sexual risk factor suggests some implications for prevention programming. Programmes that promote empowerment and strength may be especially helpful to these women whose sense of self is linked through approval and dependence on others. These programmes might include opportunities in which women are guided toward reaching small but graduated goals in order to demonstrate competency. Programmes that promote affirmation through sharing stories with other women and the like should also be helpful. African-American women who are high in interpersonal sensitivity may also benefit from programmes that promote affirmation.

Women who score high (or low) on the caretaking/mindfulness of others domain may or may not engage in sexual risk behaviour. While our study found that the caretaking/mindful of others domain did not predict any condom outcomes, more research is needed. Research is needed to replicate this finding and/or to clarify the relationship between caretaking/mindfulness of others and sexual attitudes and behaviours including factors that might modify this relationship such as communication skills and condom attitudes.

In overview, we were able to reconceptualise traditional gender role beliefs and traits so that they produced different clusters that may be better representations of AfricanAmerican women. The domains that emerged from our factor analysis were reliable and meaningful and two of the domains contributed to explaining condom-related outcomes.

Despite support for some hypotheses, the present study has some limitations concerning the measures and sample. First, the items that were used to reconceptualise gender roles came from a traditional gender role measure. Formative research in which African-American women are interviewed regarding gender role beliefs is needed; this research could help to determine if similar or different items and domains emerge. Second, the survey asked women sensitive questions about sex and relationships. Although participants were assured that their responses would be confidential, social desirability bias might be a factor. Another limitation was that information on socioeconomic status (SES) was not collected. Although the sample was considered fairly homogeneous due to most being in college, SES might be a factor that should be addressed in studies with other female populations. For example, lower SES status has been linked to higher rates of HIV infection and higher rates of unsafe sex (Agha 2001, 55; Janssen et al. 2000, 487). Also, factors related to financial resources, such as dependency on partners for financial support and access to contraceptives, may contribute to lack of use of condoms for women with limited financial resources.

Another limitation is that the majority of participants were recruited from local colleges and universities, and the findings may be less applicable to those outside of a college setting. Future studies should investigate whether norms endorsed by African-Americans in college differ from an older and more diverse African American female population. Last, the sample population in the studies was recruited from a southeastern region in the USA and regional differences may influence the desirability of gender role traits (Konrad and Harris 2002, 259). Specifically, Konrad and Harris found that African-American participants from the urban northeast were more traditional in their views of the desirability of feminine attributes for women than African-American participants in the south. The data used in this paper were collected in the southeastern part of the USA and participants’ endorsement of gender roles and identities may differ from other African-American women located in different regions of the USA.

Future research is needed to examine gender role and identity within the context of African-American culture(s). For example, does gender role orientation differ among women who are more or less Africentric, among women with positive or less positive ethnic identity and among those with high and lower levels of religiosity?

Conclusion

In general, the findings from this study indicated that there is a need for new and more dynamic conceptualisations of gender role for African-American women. We found domains that differed and that were psychometrically stronger than those identified in more traditional measures. We also found some association between specific gender role domains and condom use. The interpersonal sensitivity domain was associated with higher sexual risk with regard to less favourable condoms outcomes and the persistence/active coping domain was associated with lower sexual risk via better condom outcomes. This study was exploratory but contributes to the sparse literature on gender role and identity among African-American women. The findings of this study are promising and suggest that reconceptualisation of gender role and identity might be helpful in maximising the effectiveness of prevention programme.