Sexual-moral attitudes, love styles, and mate selection.

The National Health and Social Life Survey (NHSLS; Laumann, Gagnon, Michael, & Michaels, 1994; Michael, Gagnon, Laumann, & Kolata, 1994) was a landmark research study, known both for its pioneering use of state-of-the-art sampling and interviewing techniques in a national sexuality study and for its substantive findings on the sexual attitudes and behaviors of Americans ages 18 to 59. The present study focuses on one aspect of the NHSLS, namely its measurement and characterization of sexual-moral attitudes. By assessing respondents' attitudes on issues such as homosexuality, pornography, and abortion, the NHSLS researchers were able to develop profiles of different subgroups of Americans who exhibited varying patterns of liberal and conservative views.

Although the National Health and Social Life Survey sexual-moral attitudes framework has now been around for several years, there appears to be little research following up on it (Carpenter, 1998; Laumann & Michael, 2001). We thus propose to extend the study of sexual-moral attitudes by relating them, conceptually and empirically, to two other constructs: love styles (Hendrick & Hendrick, 1986) and attraction criteria (Simpson & Gangestad, 1992). All three constructs are described in greater detail below.

An important reason for attempting to relate sexual-moral attitudes to love styles and attraction criteria is to seek evidence of sexual-moral attitudes' (and the other variables') construct validity via the establishment of a nomological network (Cronbach & Meehl, 1955; Trochim, 2002). A nomological network seeks to relate theoretical constructs to each other, theoretical constructs to observable measures, and observable measures to each other. Cronbach and Meehl (1955) note that "'Learning more about' a theoretical construct is a matter of elaborating the nomological network in which it occurs ..." (p. 290) and that "... 'operations' which are qualitatively very different 'overlap' or 'measure the same thing' if their positions in the nomological net tie them to the same construct variable" (pp. 290-291). In short, nomological analysis can inform us about a construct by showing other constructs to which it is conceptually and empirically related.

SEXUAL-MORAL ATTITUDES

The National Health and Social Life Survey administered nine items covering the areas of love and sex, pornography, religion, abortion, and premarital, extramarital, and homosexual sex. It then used cluster analysis to cleave the American public into distinct attitudinal clusters. Three higher order clusters were termed the traditional, recreational, and relational groups. Respondents in the traditional category felt their religious beliefs were the guiding force behind their sexual behavior and were the most conservative on all sexual attitudes. Individuals in the recreational category believed sex does not have to have anything to do with love and were the most liberal in their sexual attitudes. Relational respondents believed sex should be part of a loving relationship but not necessarily in marriage.

The NHSLS researchers further subdivided the three higher order categories into a larger number of lower order ones based on the cluster analysis. They divided the traditional category into conservative and pro-choice; these groups differed on a few of the issues, but most markedly on abortion. Attitudes toward abortion were also the basis for a division in the recreational category into pro-life and libertarian subgroups; with limited exceptions, the latter group had the most liberal responses of any group on all items. The relational category broke down into three subgroups called religious, conventional, and contemporary religious. These three subgroups were a "mixed bag," virtually never all agreeing with each other across the set of issues. Only in their unanimous rejection of the proposition that premarital sex is always wrong did the three relational groups resemble each other.

In the present study, we also conducted a cluster analysis on the sexual-moral attitude items to remain consistent with the methodology of the National Health and Social Life Survey. We then compared the resulting cluster groups' means on various love-style and attraction indices. Because of obvious differences in sample composition between our participants (students at one university) and those surveyed in the NHSLS (a nationally representative sample of 18-59year-olds), we focussed on the more general, higher order characteristics of the clusters (i.e., liberal-conservative) as opposed to the more detailed, lower order profiles. Also, given how little the NHSLS relational subgroups had in common, we also deemphasized the relational category.

LOVE STYLES

One could reasonably expect there to be some consistency between love styles and sexual attitudes, since beliefs about sexuality could very well determine the choice of a romantic partner and the course of such a relationship (or perhaps experiences in the relationship could affect attitudes regarding sexuality). Lee (1973) developed a typology of love that includes six styles: eros, indus, pragma, mania, agape, and storge.

Eros can be described as a passionate type of love; this style "involves powerful attraction to the partner, both physically and emotionally" (Montgomery & Sorell, 1997, p. 55). Lee (1973) describes the erotic lover as one who is "open and honest" and "seeks a deep, pervasive rapport with the partner" (Hendrick & Hendrick, 1987, p. 284). Since Hendrick and Hendrick (1987) predict that permissiveness is negatively related to disclosure (pervasive rapport), then it is reasonable to assume that someone who endorses the eros love style would be more likely to fall into the more conservative sexual groups.

Ludus is described as the "game-playing" type of love, characterized by being noncommital and distant (Hendrick & Hendrick, 1989, p. 785; Montgomery & Sorell, 1997). Since individuals who endorse ludus have a more casual attitude toward relationships, they might be likely to endorse liberal (recreational) sexual attitudes more so than conservative (traditional) ones.

Pragma is characterized as the practical type of love, in which an individual chooses a romantic partner based on considerations such as how well the partner will reflect on the chooser's career or family. A pragmatic orientation such as this appears to reflect a careful, deliberative process in choosing a partner (and perhaps also in one's sexual behaviors), thus seeming to put it in opposition to a liberal (recreational) orientation to sex.

The mania type of love can be characterized as obsessive in that it is possessive and dependent (Montgomery & Sorell, 1997). Someone holding mania-type feelings may be unlikely to seek a partner with open (i.e., recreational) sexual attitudes, as the latter might not see love as necessary for sex and therefore might not be committed to the manic person.

Agapic individuals tend to put their partner's needs and wishes above their own. Prior research has found agape to be negatively related to permissiveness and permissiveness positively related to ludus (Hendrick & Hendrick, 1987). We thus predicted that agape would be associated with conservative sexual-moral attitudes in the present study.

Storge love tends to be based on friendship (Hendrick & Hendrick, 1989). It stems from the belief that romantic love grows out of a deep friendship and sometimes the parties involved do not even notice that the line has been crossed from one to the other. We did not have strong hypotheses about how storge should relate to sexual-moral attitudes or attraction criteria; thus, we examined it in an exploratory sense.

ATTRACTION CRITERIA

Attraction is intertwined with sexual attitudes and love styles and consists of a variety of attributes that one person may seek in another person. Similarity of attitudes, values, and beliefs between partners has been found to be important in romantic partner choices (Hester, 1996; Morell, Twillman, & Sullaway, 1989; Sprecher & Duck, 1994). However, there are more specific attitudes, values, and beliefs that can contribute to overall similarity or just be important in their own right.

Simpson and Gangestand (1992) used factor analysis to divide romantic partner attributes into the two dimensions of personal/parenting (with items including "responsibility," "kindness and understanding") and attractiveness/social visibility (with items including "physical attractiveness" and "sex appeal"). These dimensions are grounded in Simpson and Gangestad's model of sociosexuality, which refers to "individual differences in willingness to engage in sexual relations without closeness, commitment, and other indicators of emotional bonding" (p. 35). Simpson and Gangestad found that sexually unrestricted individuals desire, choose, and acquire romantic partners who are more physically and sexually attractive and who have higher social visibility. Choosing partners based on attractiveness and visibility characteristics would thus appear to fit conceptually with a liberal-recreational orientation in the attitude domain and a ludic outlook regarding love styles. In contrast, Simpson and Gangestad have found restricted individuals to seek romantic partners who are more kind and affectionate, responsible, and loyal and faithful, in keeping with more conservative attitudes and committed love styles.

GENDER

Various studies show aspects of sexual attitudes, love styles, and attraction to be related to gender. Hendrick and Hendrick (1995) found men to be more sexually permissive than women. Further, men reported endorsing a ludus love style more than did women, and women endorsed storge more than did men. According to Longmore (1998), "The male peer group helps to elaborate a system of norms that devalue love and long-term commitment but encourage a kind of game mentality--males attempt to 'get over' not only on the females involved but in a social competition with other males" (p. 45). Michael et al. (1994) found that men ages 18 to 24 were more likely to fall into the recreational group than were those in the other age groups. Some of the above findings as well as hypotheses for the present study that are shown below appear congruent with either social learning/script theory or psychoevolutionary theory. Because we cannot empirically distinguish these two theories, we do not address them further.

HYPOTHESES

The hypotheses of this study are summarized in Table 1. As we have suggested, sexual-moral attitudes, love styles, attraction criteria, and gender form a conceptual nexus or set of interconnections. For ease of presentation, Table 1 shows both a Nexus I, focusing on liberal or permissive attitudes and behavior, and a Nexus II, focusing on conservative attitudes and behavior. We recognize, however, that each nexus may largely be the opposite side of the same underlying dimension.

As can be seen by reading across Nexus I, we expected respondents who were more liberal (recreational) in their sexual attitudes to be more likely to endorse a ludus love style than would be their more conservative counterparts. Further, we hypothesized that liberal respondents would seek a partner with attractiveness and social visibility attributes as opposed to personal and parenting qualities, and that respondents who endorsed a ludus love style would also seek a partner based on attractiveness and visibility. Regarding gender, we expected men would be proportionally better represented than women in the liberal attitude clusters and on the ludus love style. Also, we thought men would be more likely to place importance on physical attractiveness and sex appeal in potential romantic partners.

Tracking along Nexus II, conservative respondents (loosely corresponding to the National Health and Social Life Survey's traditional and relational categories) would be more likely to endorse mania, eros, pragma, and agape styles of love than would be their more liberal counterparts. Further, conservative respondents would be more likely to seek a partner with personal and parenting attributes. Finally, women would be more likely than men to endorse mania, eros, pragma, and agape styles of love.

METHOD

Participants

A sample of 330 students (77% females, 23% males) from a large southwestern state university (primarily human development and family studies majors) was surveyed via paper-and-pencil questionnaires. Respondents were asked to volunteer for this study during classes and asked to complete the survey at that time. Forty-one percent of the respondents were ages 20 to 21. In descending order of representation, individuals ages 22 to 23 made up 25% of the sample, and 21% of the sample consisted of those ages 18 to 19. The remaining respondents were anywhere from 24 to 59 years old. The majority of respondents, 85%, were White. The other two most frequently reported races were Hispanic (8%) and Black (4%).

Measures

Demographics. Variables measured included gender, age, race and ethnicity, religion, attendance at religious services, college major, type of class in which respondent was filling out the survey, current relationship status, and length of current relationship (if applicable).

College students would likely represent different relationship statuses (e.g., married, engaged, dating, not currently in a relationship), and status could potentially affect responses to some of the relationship-based items. To control for this, we asked respondents to whom they would be referring in completing the relationship items (current partner, indicated by 62% of participants; most recent former partner, 32%; and hypothetical partner, 6%).

Love Attitudes Scale (love styles). This questionnaire was replicated from Hendrick, Hendrick, and Dicke's (1998) short form of the Hendrick and Hendrick (1986) Love Attitudes Scale. Hendrick and Hendrick created a well-standardized measure of Lee's (1973) love styles (eros, ludus, pragma, mania, storge, and agape). The short form has four items per subscale (reduced from the original seven). In keeping with Hendrick et al. (1998), low scores indicate greater endorsement of the construct. Means, coefficient alpha ([alpha]) internal consistency reliability estimates, and a sample item for each of the six subscales in the present study are as follows: eros (M = 1.95, [alpha] = .80, "I feel that my partner and I were meant for each other"); ludus (M = 3.93, [alpha] = .72, "I believe that what my partner doesn't know about me won't hurt him/her"); storge (M = 2.61, [alpha] = .84, "Our love is the best kind because it grew out of a long friendship"); pragma (M = 2.90, [alpha] = .77, "A main consideration in choosing my partner was how he/she would reflect on my family"); mania (M = 3.23, [alpha] = .68, "When my partner doesn't pay attention to me, I feel sick all over"); and agape (M = 2.58, [alpha] = .87, "I cannot be happy unless I place my partner's happiness before my own"). Means and alphas in the present study were virtually identical to Hendrick et al.'s (1998) values.

Two aspects of the love-style measurement must be noted. First, the pragma scale contained an item about choosing a partner based on whether he or she would be a good parent, which could overlap with the personal/parenting scale; this item was thus deleted when examining the correlation between the two indices. Second, the Hendrick et al. (1998) Love Attitudes Scale items are usually scored on a scale from 1 to 5. However, the present study used Scantron (computer scored) sheets that had response options of 0 to 9. Due to concern that respondents who wanted to give the lowest score would fill in 0 instead of 1, we changed the love-style options on the survey to 0 to 4. However, to make scores comparable to published norms, we added a point to every response in the final scale computation. Similar adjustments were made where necessary to some of the other instruments.

Romantic Partner Attribute Index. Students filled out this 15-item instrument from Simpson and Gangestad (1992). Each item consisted of a brief phrase (e.g., "kindness and understanding," "physical attractiveness") as described above. Participants rated each attribute from 0 (not at all important) to 8 (extremely important) according to the influence each attribute has on their selection of a romantic partner (a point was later added to each item to restore the original 1-9 metric). As noted earlier, Simpson and Gangestad delineated two factors in their study representing personal/parenting qualities and attractiveness/social visibility. Means and coefficient alphas for the two factors in the present study were as follows: personal/parenting, M = 7.98, [alpha] = .88; attractiveness/social visibility, M = 6.29, [alpha] = .73.

Sexual-moral attitudes. Participants answered nine sexual-moral attitude items based on those from the National Health and Social Life Survey (Laumann et al., 1994; Michael et al., 1994). The items used in the present study were highly similar to the original NHSLS ones, just shortened in some instances. For example, one original NHSLS item read "What is your opinion about a married person having sexual relations with someone other than the marriage partner?", which we shortened to "If a married person has sexual relations with someone other than the marriage partner, it is ..." (followed by choices such as "always wrong"). Our modified NHSLS wordings are shown in Table B of the supplemental results webpage, described below.

The items pertained to respondents' views on whether premarital sex in general, premarital sex among teenagers, extramarital sex, and same-gender sex are wrong; whether there should be laws restricting adults' access to pornography; whether respondents would only have sex with someone if in love with the person; whether religious beliefs have guided respondents' sexual behavior; and whether a woman should be able to obtain an abortion for any reason (one item) or only if she was raped (a separate item). Depending on the item, respondents answered on different response scales: always wrong to not wrong at all; strongly agree to strongly disagree; and "yes" or "no" (with "don't know" and "not applicable" options). Given the cluster-analytic technique used with the sexual-moral attitudes, the items were all treated as separate, individual entities and not aggregated into indices (hence the absence of any alpha reliability estimates).

The two items related to abortion had numerical response scales on the questionnaire that erroneously were different from those of surrounding items. This apparently led many respondents to give out-of-range answers on the abortion items. For this reason, the two abortion items were excluded from all analyses.

RESULTS

Only essential tables are included in the present article. Additional tables, noted below, are available at a special project website at http://www.hs.ttu.edu/research/reifman/sexual-attitudes.htm.

Creation of Clusters

The first step involved cluster analyses (Rapkin & Luke, 1993) of the National Health and Social Life Survey sexual-moral attitude items (as noted above, only seven of the items were included as the abortion-related items were dropped). This procedure involved empirically dividing respondents into clusters so that within a cluster, the members would have a relatively uniform pattern of responses on the items. Rapkin and Luke suggested several criteria for determining the number of clusters to retain, including interpretability of clusters, reproducibility of cluster composition by different algorithms, and whether there are an adequate number of cases per cluster. We used two different forms of cluster analysis: iterative and hierarchical (Table A on the project website provides further detail on the clustering procedures).

We ultimately formed four clusters. The iterative solution yielded four interpretable clusters (two generally liberal leaning and two conservative leaning) with at least 45 respondents per cluster. Hierarchical analyses suggested a three-cluster solution, in part by combining what in the iterative solution had been two different liberal clusters into one and what had been two conservative clusters into one. As described below, substantive differences between the two conservative clusters and between the two liberal clusters from the iterative solution were apparent, so we retained the four clusters produced by the iterative analysis.

Among the liberal clusters, one was named "liberal across the board" (n = 45). Its members were very tolerant of different sexual behaviors, with virtually no one saying premarital sex, teenage premarital sex, or same-gender sex is always wrong. Even though 67% of this group condemned extramarital sex, that figure was lower than for any other group. These across-the-board liberals also opposed antipornography laws, would not require being in love with someone to have sex with the person, and did not see religious beliefs as guiding their sexual behavior. (Table B on the project website shows how members of this and the other clusters responded to the attitude items.)

The other liberal cluster was named "liberal except must be in love for sex" (n = 52). Ninety percent of this group endorsed the "no sex unless in love" item, whereas only 16% of the liberal across the board group did. The liberal except must be in love for sex group was also modestly more conservative than the liberal across the board group on teen premarital sex, antipornography laws, and religion guiding sexual behavior.

Among the conservative clusters, one was called "traditional conservative" (n = 128). Although we might have called this cluster "conservative across the board" in parallel fashion to one of our liberal groups, we opted for traditional conservative because the National Health and Social Life Survey investigators had used this name for a highly similar cluster. Among all of our clusters, the traditional conservative group stood out in its broad (66%) endorsement of laws against the sale of pornography to adults; no other group had higher than 37% endorsement.

The other conservative cluster was named "conservative except for premarital and teens" (n = 87). Members of this group differed from the traditional conservative group (and even resembled those in the liberal groups) in their virtually complete rejection of the idea that premarital sex and teen premarital sex were always wrong.

Although comparisons between the National Health and Social Life Survey and the present study are tentative at best due to obvious geographical and other demographic differences, it seems clear that our sample was more conservative than similarly aged participants in the NHSLS. The traditional conservative cluster alone constituted 41% of our sample (128 of 312 participants; 18 of the original 330 cases were lost due to not fitting on any of the clusters). Among 18- to 24-year-olds in the NHSLS, the higher order traditional category (which includes the more specific traditional conservative cluster) comprised approximately 17% of the men and 23% of the women (Michael et al., 1994; age breakdowns for the lower order categories were not shown).

Cluster by Gender Analyses on Love Styles

We conducted 4 (Attitude Cluster) X 2 (Gender) analyses of variance (ANOVAs), first with love styles and then with attraction criteria as dependent variables.

For the love styles, we found cluster main effects on ludus, F(3,302) = 5.84, p = .001; storge, F(3, 302) = 3.40, p < .05; and pragma, F(3, 302) = 9.86, p < .001. There were five main effects of gender: eros, F(1,302) = 7.37, p < .01; ludus, F(1,302) = 8.74, p < .01; storge, F(1 , 302) = 8.06, p < .01; pragma, F(1, 302) = 6.19, p < .05; and agape, F(1,302) = 4.49, p < .05. We found no significant Gender X Cluster interactions on the love styles; all cluster-based differences thus held equally well for both genders. The love style means, by cluster and gender, are shown in Table C on the project website.

To control for the large number of statistical tests entailed in comparing each pair of clusters "head to head" on six love styles, we used conservative Scheffe tests. A chart showing the significant (p < .05) Scheffe tests appears in Table 2. The liberal across the board cluster exhibited several differences from the other clusters. It was more ludic than the liberal except must be in love for sex cluster and less pragmatic than the conservative except for premarital and teens cluster. The liberal across the board and traditional conservative clusters differed on four love styles. The liberal across the board cluster was more ludic, whereas the traditional conservatives were more storgic, pragmatic, and erotic (the difference on eros should be taken with caution, as the ANOVA main effect for cluster on eros was not significant). Lastly, the conservative except for premarital and teens cluster was more ludic than the traditional conservatives. Several of the findings--the liberal across the board group tending to be high on ludus and the conservative groups tending to be highly pragmatic--were consistent with our nexus-based hypotheses.

Significant gender differences emerged in the direction of women exhibiting more eros, storge, and pragma than men, but men showing more ludus and agape than women. The findings of women's more erotic and pragmatic attitudes were hypothesized, as was that of men's higher ludus. Men's higher agape, though counter to our theoretical framework, has been found in the past (Hendrick et al., 1998).

Cluster by Gender Analyses on Attraction

Significant cluster main effects on attraction criteria emerged for personal/parenting, F(3, 303) = 3.83, p = .01, and attractiveness/social visibility, F(3, 303) = 7.12, p < .001. The gender main effects on attraction criteria involved personal/parenting, F(1,303) = 32.08, p < .001, and attractiveness/social visibility, F(1, 303) = 8.71, p < .01. There was no Cluster X Gender interaction for attractiveness/visibility, but there was for personal/parenting, F(3, 303) = 3.40, p < .05. Table D on the project website shows the means by cluster and gender on attraction criteria.

Scheffe tests revealed significant differences (p < .05) between clusters on the personal/parenting and attractiveness/social visibility qualities (Table 3). Members of the traditional conservative cluster desired personal/parenting qualities more than did respondents of the two liberal groups. Further, respondents in the conservative except for premarital and teens cluster desired personal/parenting qualities more than did liberal across the board respondents. The conservative except for premarital and teens cluster also exhibited a greater preference for attractiveness/social visibility than did the traditional conservative and liberal except must be in love for sex groups. That conservatism tended to be associated with desire for personal/parenting characteristics was consistent with our nexus-based hypotheses.

Females scored significantly higher on both personal/parenting and attractiveness/social visibility than did males. The former result was hypothesized, but the latter was not. The nature of the obtained Cluster X Gender interaction was that in the traditional conservative and liberal except must be in love for sex clusters, females scored noticeably higher on personal/parenting than their male counterparts (more than a 1-point difference), whereas the gender differences were much smaller in the other clusters.

Examining the Role of Target

To examine whether the major relationships between cluster membership and the dependent variables varied based on whom the respondent was answering the questions in reference to (current, past, or hypothetical partner), we conducted cluster by target ANOVAs. Out of eight possible Cluster X Target interactions (on eros, indus, storge, pragma, mania, agape, personal/parenting, and attractiveness/social visibility), only one, for the love style of mania, was significant, F(6, 297) = 2.33, p < .05. Two further points are worth noting. First, mania was not involved in any of the cluster differences reported earlier. Second, some of the cells in the Cluster X Target analysis were very small, due to the small number of respondents overall who completed the questionnaire in reference to hypothetical partners. In cells containing at least 10 respondents, mania means were fairly similar, ranging from 2.76 to 3.48; some of the smaller cells had means well above or well below this range. Given the largely nonsignificant nature of the interactions and these other consideratons, it appears that the relationships between cluster membership and the dependent variables did not vary in any substantial way by target. Although not using exactly the same research design as in the present study, other studies using the Love Attitudes Scale have found individuals with different love or relationship statuses to differ on some subscales but not on others (Erwin, 1999). This mixed picture appears roughly consistent with our findings related to target.

Correlations Among Continuous Variables

With the exception of the clusters (which, as a set, formed a nominal variable), the other variables were continuous (i.e., at least ordinal), warranting correlational analyses. Most of the significant correlations were consistent with the hypotheses. It must be kept in mind when interpreting the correlations that for the love styles, a lower score indicates a stronger attitude. Greater endorsement of eros was associated with a greater desire for personal/parenting attributes in a partner (r = -.18, p < .01, 2-tailed). Ludus and personal/parenting were inversely related (r = .23, p < .01), as were agape and attractiveness/social visibility (r = .14, p < .05). A full correlation matrix is available as Table E on the project website.

Chi-Square Analysis of Cluster Membership and Gender

It was expected that men relative to women would be proportionally better represented in the liberal and recreational categories, and women better represented in the conservative ones. However, a chi-square test indicated that the gender composition of the clusters did not differ significantly from what would be expected if the overall sample base rate (77% female, 23% male) applied to every cluster, [chi square] (df= 3, n = 312) = 4.97, p = .17.

DISCUSSION

We see the major contribution of the present study to be its conceptual and empirical integration of three lines of research--sexual-moral attitudes, love styles, and attraction criteria--that heretofore had been relatively independent. In terms of construct validity, or what the National Health and Social Life Survey sexual-moral attitude items are really measuring, we can now see that they appear to encompass attitudes about respondents' own interpersonal relationships (i.e., love styles and attraction criteria) and not just personal standards and public policy prescriptions. Our understanding of the love-style and attraction instruments is similarly enriched, as they have now been shown to relate to each other as well as to the sexual-moral attitudes. A further benefit of the present study comes in the form of parsimony. The idea of a liberal and conservative nexus appears to provide a relatively simple, empirically supported framework for describing people's attitudes toward sex, love, and relationships.

Gender differences were also a part of the nexus system. Many of the hypotheses were supported, both in terms of the love styles (with women showing more erotic and pragmatic attitudes than men and men exhibiting higher ludus than women) and attraction criteria (with women expressing a greater preference for personal/parenting qualities than men). Results were not entirely supportive, however. One surprising gender difference was that women sought attractiveness/social visibility attributes more in a potential partner than did men. In trying to interpret this finding, we took special note of the fact that physical attractiveness and financial resources were included on the same scale. When we looked separately at each item (physical attractiveness, sex appeal, social status, and financial resources), the only significant gender differences were that women scored higher on financial resources and social status, thus leading to women's overall higher scores on this index. Even the idea that men would score higher than women on just the physical attractiveness and sex appeal items rather than the full subscale was not supported due to the null findings with these items. That some predictions derivable from psychoevolutionary theory did not hold up may not be surprising. Eagly and Wood (1999) have shown that cultures' levels of societal gender equality may affect the degree of gender differences on mate-selection criteria.

Of the four clusters in the present study, two (liberal across the board and traditional conservative) represented coherent, comprehensive ideological profiles across the full set of issues. The other two clusters represented situations in which their members largely adhered to a political ideology, but deviated when it came to a particular personal behavior. For example, one liberal subgroup exhibited conservatism on the matter of having to be in love to have sex with someone. Similarly, one conservative subgroup was liberal on the issues of premarital and teen sex. It should be noted, however, that college-age individuals tend to have less crystallized attitudes than their older counterparts (Sears, 1986), which could affect the coherence and long-term stability of the clusters.

This study has various strengths and weaknesses. Among the strengths were the use of a novel combination of existing measures and their use in a type of sample (college students) to which the NHSLS sexual-moral attitude items had not been applied (to our knowledge). Further, our relatively large sample size allowed numerous subgroup comparisons (e.g., cluster by gender). Among the study's weaknesses was the lack of a random sample similar to the one used by Michael et al. (1994). Also, the college population from which the present sample was drawn was biased compared to all the adults in the United States (or even all college students) in that there was a narrower age range, the respondents were students taking classes in one department of the university, and the gender breakdown was not balanced. There might also be specific political attitudes that were not well represented within this sample due to the region of the country from which it was drawn (as implied by the earlier finding that the present sample appeared to be more conservative than similarly aged NHSLS respondents). Another potential limitation is the lack of items on actual sexual behaviors; these were omitted to keep the instrument as short as possible and due to concerns regarding sensitivity and privacy. Lastly, direction of causation between attitudes, love styles, and attraction cannot be stated from correlational data.

Out of the above limitations, however, can flow ideas for further research. First, replications can be attempted in other parts of the U.S. and internationally, where local norms toward sexuality and morality may differ from those in the present study. Studying how residents of different communities cluster on the National Health and Social Life Survey sexual-moral attitude items would likely shed more light on the connection between local norms and attitude structures. Second, future studies can incorporate self-reported sexual behavior to see how it correlates with the constructs in this study. Third, although strict causation cannot be studied, longitudinal panel surveys can be conducted to examine whether there are patterns of temporal precedence in the relations of sexual-moral attitudes, love styles, and attraction criteria (i.e., whether early levels of one of these constructs predict later development of others). Many of the attitudes, beliefs, and preferences studied here likely take form before college age in many individuals, so the study of younger participants might also be fruitful.

In conclusion, we have documented connections between sexual-moral attitudes, love styles, and attraction criteria. With future research to tackle the conceptual and methodological complexities that remain, we can look forward to greater understanding in this area.

Table 1. Summary of Hypothesized Relations Among the Three Constructs
and Gender

            Sexual-moral
Nexus    attitude clusters    Love styles (a)

I         Liberal             Ludic
II        Conservative        Manic, erotic, pragmatic, agapic

Nexus    Attraction criteria               Predominant gender

I         Attractive, social visibility      Male
II        Personal, parenting                Female

(a) No firm hypotheses were developed for storge; it was thus examined
in an exploratory sense.

Table 2. Significant Scheffe Differences by Attitude Cluster on Love
Styles

                   Liberal except                   Conservative except
                     must be in    Traditional        for premarital
                    love for sex   Conservative         and teens

Liberal across     Ludus (LAB       Ludus (LAB       Pragma (CEPT more
  the board          more ludic)    more ludic),     pragmatic)
                                    storge (TC
                                    more storgic),
                                    pragma (TC more
                                    pragmatic),
                                    eros (a) (TC
                                    more erotic)

Conservative                        Ludus (CEPT
  except for                        more ludic)
   premarital
  and teens

Note. If one or more dependent variables are listed in a cell at the
intersection of two attitude-cluster groups, this indicates the two
groups differed (p < .05) on these dependent variables. The direction of
the difference is given in parentheses: LAB = liberal across the board;
TC = traditional conservative; CEPT = conservative except for
premarital and teens.

(a) The cluster main effect from the analysis of variance was not
significant for eros.

Table 3. Significant Scheffe Differences by Attitude Cluster on
Attraction Criteria

                           Liberal across the    Liberal except must
                                board            be in love for sex

Traditional conservative   Personal/parenting    Personal/parenting
                             (TC higher)          (TC higher)

 Conservative except for    Personal/parenting   Attractiveness/social
  premarital and teens       (CEPT higher)        visibility (CEPT
                                                   higher)

                             Traditional conservative
Traditional conservative

 Conservative except for      Attractiveness/social
  premarital and teens         visibility (CEPT higher)

Note. If one or more dependent variables are listed in a cell at the
intersection of two attitude-cluster groups, this indicates the two
groups differed (p < .05) on these dependent variables. The direction
of the difference is given in parentheses: TC = traditional
conservative; CEPT = conservative except for premarital and teens.

REFERENCES

Carpenter, L. (1998). From girls into women: Scripts for sexuality and romance in Seventeen magazine, 1974-1994. The Journal of Sex Research, 35, 158-168.

Cronbach, L., & Meehl, P. (1955). Construct validity in psychological tests. Psychological Bulletin, 52, 281-302.

Eagly, A. H., & Wood, W. (1999). The origins of sex differences in human behavior: Evolved dispositions versus social roles. American Psychologist, 54, 408-423.

Erwin, P. G. (1999). Love attitudes and romantic involvement: A replication and extension. Perceptual and Motor Skills, 88, 317-318.

Hendrick, C., & Hendrick, S. (1986). A theory and method of love. Journal of Personality and Social Psychology, 50, 392-402.

Hendrick, C., & Hendrick, S. (1989). Research on love: Does it measure up? Journal of Personality and Social Psychology, 56, 784-794.

Hendrick, C., Hendrick, S., & Dicke, A. (1998). The Love Attitudes Scale: Short form. Journal of Social and Personal Relationships, 15, 147-159.

Hendrick, S., & Hendrick, C. (1987). Love and sexual attitudes, self-disclosure, and sensation seeking. Journal of Social and Personal Relationships, 4, 281-297.

Hendrick, S., & Hendrick, C. (1995). Gender differences and similarities in sex and love. Personal Relationships, 2, 55-65.

Hester, C. (1996). The relationship of personality, gender, and age to adjective check list profiles of the ideal romantic partner. Journal of Psychological Type, 36, 28-35.

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

Laumann, E. O., & Michael, R. T. (Eds.). (2001). Sex, love, and health in America: Private choices and public policies. Chicago: University of Chicago Press.

Lee, J. (1973). The colors of love: An exploration of the ways of loving. Don Mills, Ontario, Canada: New Press.

Longmore, M. (1998). Symbolic interactionism and the study of sexuality. The Journal of Sex Research, 35, 44-57.

Michael, R. T., Gagnon, J. H., Laumann, E. O., & Kolata, G. (1994). Sex in America: A definitive survey. Boston: Little, Brown.

Montgomery, M., & Sorell, G. (1997). Differences in love attitudes across family life stages. Family Relations, 46, 55-61.

Morell, M., Twillman, R., & Sullaway, M. (1989). Would a Type A date another Type A? Influence of behavior type and personal attitudes in the selection of dating partners. Journal of Applied Social Psychology, 19, 918-931.

Rapkin, B., & Luke, D. (1993). Cluster analysis in community research: Epistemology and practice. American Journal of Community Psychology, 21,247-277.

Sears, D. O. (1986). College sophomores in the laboratory: Influences of a narrow data base on social psychology's view of human nature. Journal of Personality and Social Psychology, 51, 515-530.

Simpson, J., & Gangestad, S. (1992). Sociosexuality and romantic partner choice. Journal of Personality, 60, 31-49.

Sprecher, S., & Duck, S. (1994). Sweet talk: The importance of perceived communication for romantic and friendship attraction experienced during a getacquainted date. Personality and Social Psychology Bulletin, 20, 391-400.

Trochim, W. M. K. (2002). The nomological network. Retrieved November 27, 2002, from http://lrochim.human.cornell.edu/kb/nomonet.htm

Manuscript accepted June 2, 2003

This article is based on Rachel Saul Lacey's master's thesis in Human Development and Family Studies at Texas Tech University. An earlier version of this article was presented at the 2000 meeting of the Midwestern Psychological Association, Chicago, Illinois.

Address correspondence to Alan Reifman, Department of Human Development and Family Studies, College of Human Sciences, Texas Tech University, Lubbock, TX 79409-1162; e-mail: [e-mail].

Rachel Saul Lacey, Alan Reifman, Jean Pearson Scott, Steven M. Harris, and Jacki Fitzpatrick

Texas Tech University