Sei sulla pagina 1di 18

P , y h l o g y of Women @iurterly, 23 (lYYS), 519-536.

Printed in the United States of America

THE AMBIVALENCE TOWARD MEN INVENTORY


Differentiating Hostile and Benevolent Beliefs About Men

Peter Glick Lawrence University

Susan T. Fiske Universify of Massachusetts at Amhersf

We present a measure, the Ambivalence toward Men Inventory (AMI), that differentiates between womens hostile and benevolent prejudices and stereotypes about men. The Hostility toward Men (HM) and Benevolence toward Men (EM) subscales of the AM1 tap conventional attitudes toward men that have opposing valences. Each subscale assesses subfactors concerning mens power, gender differentiation, and heterosexuality. Three studies with predominately White, male and female participants (two with undergraduates and one with a community sample) establish the factor structure, reliability, convergent validity, and predictive validity of the AMI. The AM1 was strongly related to its sister scale, the Ambivalent Sexism Inventory (Click & Fiske, 1996) and to two established scales of attitudes toward men (Downs & Engleson, 1982; lazzo, 1983). Only the AMI, however, successfully distinguished between subjectively positive and subjectively negative beliefs about men. A copy of the 20-item AM1 i s provided as a tool for further exploration of womens ambivalence toward men.

We thank Christine McCroarty for her help in refining the AM1 and ninning some of the stndies. Address correspondence and reprint reqiirsts to: Peter Glick, Psycholo9 Department, Lawrence University, Appletori, WI 54912-0599 (E-mail: peter.s.glick@lawrence.edu) or to Susan T. Fiske, Psychology Department, University of Massachusetts at Amherst, Amherst MA 01003 (E-mail: fiske@psych .urnass. edu).
Published by Cambridge University Press 0361-6843/99 $9.50

519

520

CLICK AND FISKE

What a change of temper a fixed income will bring about. . . I need not hate any man; I r e cltnnot hint me. I need not flatter any man; he has nothing to givc me. -Virginia Woolf, A Rnorri of Ones Otcn

The relationship between women and men as social groups is unique. No other groups have endured so long a relationship of status inequality coupled with such close physical and psychological intimacy. Cross-culturally and historically men have possessed greater structural power (control over important social, political, economic institutions), while at the same time women have been viewed as possessing significant dyadic power (power due to mens dependence on women as wives, mothers, and romantic partners; Cuttentag & Secord, 1983). These aspects of male-female relations, we have argued, create ambivalent attitudes on the part of men toward women (click & Fiske, 1996,1997). In this article, we ask how women are likely to respond to these same social facts. As Virginia Woolf (1929/1957) implies in the epigraph, womens dependence on men may foster ambivalent attitudes; they may resent mens power, but their dependency may also provide incentive to flatter men a s well. We discuss the sources of this proposed ambivalence and present a rneasure, with evidence of its reliability and predictive validity, for assessing womens ambivalence toward men.
W H Y WOMEN ARE AMBIVALENT TOWARD MEN

Theories of intergroup relations emphasize power differences and the drive to establish group identity as determinants of intergroup attitudes (e.g., social identity theory; Tajfel, 1981). Additionally, although general theories of group relations do not typically address sexuality, the importance of heterosexuality to the attitudes of each sex toward the other is quite obvious (Deaux, 1995),with whole fields of study devoted to heterosexual romantic: love and to hostile forms of heterosexuality, such as sexual harassment and sexual violence. Click and Fiske (1996) provide evidence that mens structural power combined with their dependency on women creates ambivalence toward women along these dimensions, as measured by the Ambivalent Sexism Inventory (ASI). The AS1 contains the Hostile Sexism (HS) and Benevolent Sexism (BS) subscales, each of which covers issues of power, gender differentiation, and heterosexuality. Hostile sexism justifies mens power by derogating women, whereas benevolent sexist attitudes recognize mens dependence on women by idealizing them as pure creatures who ought to be protected by men.\Vomen, we argue, also respond to the simultaneous facts of male structural power and female dyadic power by holding both hostile and benevolent (i.e., ambivalent) beliefs about men that encompass these same dimensions. On the hostile side, subordinate group members may resent the power and higher status accorded to the dominant group. In terms of gender relations, we label this resentrrumt of paterrudism. Social identity theory (Tajfel, 1981) suggests that individuals may cope with negative social identities through hostile competition with other groups (including dominants). Similarly, Allport (1954) noted that one potential response to being a victim of prejudice is increased prejudice toward other groups, most often expressed toward dominants in subtle, covert ways. This resistance may also be reflected in negative stereotypes of the dominant

Ambiva/ence TowardMen

521

group that compensate for the negative identity thrust on the subordinate group hy characterizing dominants as inferior in ways tlyat are safe to criticize (e.g., that Inen are babies when sick) and by attributing to dominants the negative traits associated with power (e.g., arrogance). Indeed, conventional stereotypes about men have been found to have significant negative content (Eagly & Mladinic, 1993).For relations between the sexes, we term this compensato y gender diflerentiation, which allows women (despite a lower status) to differentiate themselves positively from men (cf. Crocker, Thompson, McGraw, & Ingerman, 1987). Finally, heterosexual hostility may occur on the part of women toward men as a result of resentment of male sexual aggressiveness and because of paternalism in close heterosexual interpersonal relationships (e.g., men dominating conversations). Male sexual aggression thrives in cultures and subcultures that promote gender inequality (Bohner & Schwarz, 1996) and the threat of sexual violence has been popularly characterized as a means hy which men control women to maintain inequuality (Brownmiller, 1975).Womens awareness of this threat has been shown to have profound effects on their attitudes and behavior (Bohner & Schwarz, 1996) including, we suggest, considerable resentment toward men. Women, however, are not likely to be purely hostile in their attitudes toward men. Just as most men are dependent on women, so too are most women dependent on men. Traditionally, it was through men that women had the opportunity to gain social status and economic security. Further, dependence on the other gender for sexual reproduction and (for heterosexuals) romantic relationships works both ways. These factors are likely to motivate women to hold subjectively positive attitudes toward men and may foster an attitude of maternali.sm. Like protective paternalism (Glick & Fiske, 1996), maternalism presumes a weakness in the other gender, but regards this as something that justifies protectiveness or nurturance (e.g., that a woman rnust take care of her man at home because he is incapable of doing so himself). Although maternalism casts women as being more competent and powerful (in a limited domain), it justifies womens service to men. Benevolence toward men may also stem from admiration for mens higher status. Stable status dfferences between groups lead to consensual stereotypes that assign favored traits to members of the high-status group as a way of explaining the status difference. Members of low-status groups may share the belief that they are less ambitious, intelligent, and competent than members of the dominant group and evince considerable respect for the power and the abilities of the dominant group (Jost & Banaji, 1994). In the case of womens attitudes toward men, WP label this complementary gendm differentiation, Additionally, bonds of affection between women and men are energized by heterosexual attraction. As Berscheid and Peplau (1983) note, people often identify romantic relationships with other-gender partners as one of the most important ingredients for happiness in life. Just as many men endorse the sentiment that a man is incomplete without the love of a woman, so too may many women beIieve that a womans personal fulfillment requires a romantic relationship with a man. In short, there are ample reasons why women are likely to hold ambivalent attitudes about men as a group, resenting mens structural power, sexist attitudes that portray men as superior, and the manner in which inen assert control within intimate heterosexual relationships, but at the same time experiencing subjectively positive feelings of affectionate protectiveness, admiration, and connection with

522

CLICK A N D

FISKE

men, with whom they are often intimate and interdependent. This ambivalence may be especially likely to occur among women who hold more traditional attitudes about men, endorsing both subjectivelypositive (e.g., competent, ambitious, strong) and negative (e.g., domineering, arrogant, hostile) stereotypes about men (Eagly & Mladinic, 1993). These women may resent mens power even as they subscribe to beliefs that support it, perhaps because that is how they have learned to act and because it is perceived as the most effective way of getting along in a world where men exert greater control (see Jackman, 1994).
OVERVIEW

To test for this predxted ambivalence of (at least some) women toward men, we created the Ambivalence toward Men Inventory (AMI), a 20-item self-report inventory consisting of two subscales: Hostility toward Men (HM) a d Benevolence toward Men (BM),each with three subfactors that address issues of male structural power (paternalism and its counterpart, maternalism), gender differentiation, and sexuality (see Appendix for the AM1 and scoring instructions).We conducted three studies to develop the AM1 and establish its reliability and validity. The results are organized according to the theoretical and empirical issues addressed, rather than in separate sections for each study.
METHOD

Participants and Procedures


Study 1 A total of 480 undergraduate participants (147 men and 333 women) were recruited at three different campuses: the University of Massachusetts-Amherst (54 men and 140women), Lawrence University (66 men and 121 women), and Smith College (27 men and 72 women). Participants were compensated either through extracredit points or a chance to win a cash lottery. All participants completed a 133item survey entitled Women, Men, and Their Relationships that formed the initial pool of items from which the final version of the AM1 was developed. Respondents on all three campuses were roughly similar in terms of age (over 90% were between 17 and 24 years old) and ethnicity (86.5%were White, 4% Asian, 1.4% Hispanic, 1% Native American, 1.2% African American; 5.9% failed to respond). Study 2 All 208 participants (74 men and 134 women) in Study 2 were undergraduates
recruited for extra credit in psychology classes at the University o f MassachusettsAmherst. These students were similar in age and ethnicity to those in Study 1. Participants completed a short (32-item) version of the AMI; the AS1 (Glick I% Fiske, 1996), which has a similar theoretical derivation to the AMI, but taps ambivalence toward women with Hostile Sexism and Benevolent Sexism subscdes; two additional questionnaires, both called the Attitudes Toward Men Scale, one by Downs and Engleson (1982) and the other by Iazzo (198-3).All scales were

Ambivalence Toward Men

523

similar in format to the AM1 (a series of statements to which respondents agreed or disagreed on a 6-point scale). Participants also generated 10 traits they associate with men as a group and rated the negativity or positivity of each trait on a -2 to +2 scale as a predictive validity check.

Study 3 We recruited a sample of 266 (mostly) older adults (102 men and 164 women). Lawrence University undergraduates volunteered to send copies of the AM1 and the predictive validity task used in Study 2 to adult relatives and friends (volunteers mailed between 2 and 8 questionnaires). Additional respondents were recruited from classes at a community college, University of Wisconsin-Fox Valley, which primarily serves adults seeking continuing education. For both sets of participants a cash lottery was used as incentive. The return rate was close to 60%. Ages ranged from 16 to 86 (median=44 for women, 48 for men). Participants held a wide variety of occupations, including secretary, clerk, miner, truck driver, police officer, firefighter, teacher, lawyer, nurse, physician, minister, administrator, as well as several homemakers. Ethnically, however, the sample was not diverse: 95% of the respondents were White, 3% Asian, 1.4% African American, and fewer than 1% Hispanic.

Developing the AM1 The AM1 item pool included 133 statements that respondents in Study 1 rated on a 0 (strongly disagree) to 5 (strongly agree) scale with no neutral midpoint (forcing respondents to agree or disagree at least slightly). In the subsequent studies, the AM1 was reduced to a 32-item form, from which the 20 best performing items were chosen for the final version. Many specific items within the theoretical dimensions were inspired by a small group of (20- to 30-year-old) women who informally discussed and recorded their attitudes tovyard men (without the authors being present).

Hostility Toward Men


HM items tapped Resentment of Paternalism (e.g., Men will always fight to have greater control in society than women), Compensatory Gender Differentiation (e.g., Men would be lost in this world if women werent there to guide them), and Heterosexual Hostility (e.g., A man who is sexually attracted to a woman typically has no morals about doing whatever it takes to get her in b e d ) . Each category was represented by about 20 items.

Benevolence Toward Men


BM items tapped Matemalism (e.g., Women ought to take care of their men at home, because men would fall apart if they had to fend for themselves), Complementary Gender Differentiation (e.g., Men are less likely to Fall apart in emergencies than women are), and Heterosexual Intimacy (e.g., Every woman needs a male partner who will cherish her). Each category was represented by approximately 20 items.

524
RESULTS

CLICK AND FISKE

Data from three studies that explore the factor structure, reliability, convergent and discriminant validity, and predictive validity of the AM1 are reported. For all analyses, gender differences were examined and are reported if statistically significant.
Factor Analyses

Study 1 was used to select the best performing items for a shorter version ofthe

AM1 and to establish its factor structure (with replications provided by Studies 2 and 3). Exploratory factor analyses (using principal-components factor analysis
with a Varimax rotation) yielded similar results for participants from the different colleges, allowing us to pool the data. The 32 items that most cleanly loaded on the separate factors yielded by this analysis were used in subsequent studies. Based on the data of all three studies 20 items were chosen for the final version of the AM1 upon which all analyses reported in this article are based. Rather than present the exploratory analyses, we will focus on the factor structure of the 20-item AM1 as revealed by confirmatory factor analysis using LISREL 8.0 (Joreskog & Sorbom, 1993) across three replications (our three separate samples). The advantage of using confirmatory factor analysis is that it allows the researcher to construct and test the fit of specific theoretical models, which can include second-order factors (as does our preferred model). The preferred model (HM and BM as correlated superordinate factors, each with three subfactors, described previously) was tested against more parsimonious alternatives: (a) HM and BM as distinct, though correlated, factors, but with no HM or BM subfactors; arid (1)) a one-factor rriodel with all items loading on an overall attitude toward men factor. The fit of each of these models was tested on the 20-item AM1 for each of the three studies using the Goodness of Fit Index (GFI), the more conservative Adjusted Goodness of Fit Index (AGFI), the Parsimony Goodness of Fit Index (PGFI; which adjusts for the complexity of the model and rewards parsimony), and chi-square (see Jiireskog & Siirbom, 1993, for explanations of these fit indices). The GFI, AGFI, and PGFI are standardized so that the closer the index is to 1.0, the better the fit. Chi-square is a measure of lack of fit (i.e.,a nonsignificant chi-square is desired) and is not standardized. Even though sarriple size affects whether chi-square will be significant (such that large sample sizes may lead to statistical significance even for models that fit tolerably well), it is useful for model comparisons. The decrease in chi-square for each successively more complex model (decrease from one to two factors; two factors to full model) tests whether the more complex models fit the data significantly better (Bollen, 1989).

Goodness of Fit As indicated in Table 1, in all three studies the preferred rriodel fit significantly better than the simpler two-factor and one-factor models (as indexed by a significant decrease in chi-square and by the increase in the GFI, AGFI, and PGFI). These analyses provide strong support for our theoretical model, which suggests that

Ambivalence TowardMen

525 Table 1

Fit of Confirmatory Factor Models Across 3 Studies

Fit statistics
One-Factor Model GFI AGFI PGFI (170) Two-Factor Model (HM and BM, no subfactors) GFI AGFI PGFI (169) Decrease in x2 (vs. one factor)" Full Model (HM and BM, each with 3 subfactors) GFI AGFI PGFI (163) (vs. two factor)b Decrease in

3
.66 .58 53 642.20"

x2

.64 .55 .52 1522.32'

.67 .59 54 746.85'

x2

.86 .83 .69 702.35" 819.97"

.78 .73 .62 506.97" 239.88"

.79 .73 .63 416.49' 225.71'

x2

x2

.93 .91 .72 375.58" 326.77" 489

.84
.80

.85
.80

.65 383.49 123.48' 205

.66 310.40" 106.09" 168

Note: GFI = Goodness of Fit Index, AGFI =Adjusted Goodness of Fit Index, PCFI = Parsimony Goodness of Fit. "Distributed as x2 with 1 d& "Distributed as with 6 df. * p < .01.

x2

attitudes toward men have separate hostile and benevolent components, each of which consists of attitudes revolving around issues of male power, gender differentiation, and heterosexuality. A multiple groups analysis testing the comparative fit of the model for women and men yielded a GFI of .92 in Study 1 (the only sample with sufficient numbers of men and women to perform this comparison) inlcating no significant gender differences in the fit of the preferred model.

Factor Correlations Table 2 reveals the relationships between the factors and the subfactors. HM and BM correlated in the .4to .5 range (in the simpler two-factor model the T S ranged from .35 to .47).That hostility and benevolence toward men tend to occur together
indicates that there is ambivalence toward men (i.e.,there are women who endorse both hostile and benevolent beliefs about men). This result mirrors the positive correlation that is typically found between hostile and benevolent sexism (with the exception of adult men, for whom HS and BS are uncorrelated; Glick & Fiske,

526 Table 2

CLICK A N D FSKE

Relationships Among AM1 Factors Across 3 Studies

st l l dy
AM1 factor
Superordinate Factors: HM arid BM Loadings of Subfactors on HM Resentment of Paternalism Compensatory Gender Differentiation Heterosexual Hostility Loadings of Subfactors on BM Maternalism Complementary Gender Differentiation Heterosexual Intimacy N
1 2
3

.45

.54 .98 .62 .89 .99 .78 .72 205

.47 .96 .76 .63


.80 .65 .81 168

.93 .75
.80
.88

.79 .79 485

Note: AM1 = Amhivalmce toward Men Inventory, HM = Hostility toward Men, BM = Bcnevolerrce toward Men. Correlations between the HM subfactors arid tlic superordinate BM factor arid I)etweeii the BM subfactors and the superordinate HM &tor ranged froin .30 to 5 5 .

1996). Like HS and BS, HM and BM may represent conventional, stereotypical views of male-female relations (a possibility explored below). Loadings of the subfactors on HM and BM are presented in Table 2 .

Item Loadings
Factor loadings of individual AM1 items (with gist wording) for the preferred model are presented in Table 3. Bear in mind that in this confirmatory model items were allowed to load on only one subfactor (i.e., loadings were set to 0 on all other factors and are therefore not reported). Table 3 reveals that the factor loadings for individual items were generally very strong and stable across the three studies.
Properties of AM1 Raw Scores

Reliability
Researchers interested in the AM1 will most likely rely on raw score averages for the scales (rather than factor scores). The AM1 yielded acceptable reliability coefficients across the three samples; alphas ranged from .83to .87 for total AM1 score, .81 to .86 for HM scores, and .79 to .83for BM scores. Given the complex, multidimensional factor structure of each subscale, their internal consistency reliability is impressive.

Relationship Between H M and B M Suhscdes


The correlations between the raw score averages on each of the AM1 subscales was, like the factor score correlations, positive and moderate across all three studies (ranging from .34 to .39 for women, .33 to ,135 for men). The correlations for male

Ambivafence Toward Men

527

Table 3
Factor Loadings of AM1 Items Across 3 Studies
Study

AM1 scales and iteins


Hostility toward Men Resentment of Paternalism Men will always fight for greater control in society Even sensitive men want traditional relationships Men pay lip service to equality, but cant handle it Compensatory Gender Differentiation Most men are really like children Men would be lost without women to guide them Men act like babies when they are sick Heterosexual Hostility When in positions of power, men sexually harass women Men have no morals in what they will do to get sex When men help women it is to prove they are better Men usually try to dominate conversations with women Benevolence toward Men Maternalism Even if both work, woman should take care of man at home Men are mainly useful to provide financial security for women Woman should take care of man at home, else hed fall apart Complementary Gender Differentiation Men are more willing to risk self to protect. others Men are more willing to take risks than women Men are less likely to fall apart in emergencies Heterosexual Intimacy Every woman needs a male partner who will cherish her Woman is never fulfilled without romantic relationship Every woman ought to have a man she adores Women are incomplete without men N

.61 .71 .68

.64 .56 .65

.56 .65 .54 .77

.70
.64 .68 .67 .58 .67 .63

.78

.so
.65 .53 .60 .66 .54

.59
.73 .61 .67 .62 .56

.69 .69 .68 .61 .71 .72

.80

.73 .55 .39

.56 .60
.80 .76 .68

.77 .70 .55

.83 .69 .71 .66 .75 .73 .75 .69 .63 .63 .70 .65 168 489 205

and female respondents were significantly different in Study 2, z = 2.54, p < .01 ( r for women = .38, r for men = .6,5),but not in the other two studies, zs I 1.32, n s . (with rs ranging from .34 to .49).

Gender Dzfierences in H M and B M Scores Theories of intergroup relations suggest that female, in contrast to male, respondents would exhibit greater hostility and less benevolence toward men (i.e., women

528

GLlCK A N D

FISKE

Table 4
AM1 Means for Men and Women Across 3 Studies
Study AMI score
1

Female respondents
AM1 M SD Hostility toward Men M SD Benevolence toward Men

1.94 .74 2.38 .91 1.50 .91 343

1.95 .78 2.36 .95 1.53 .90 134

1.79 .72 2.23 .84 1.34 .82 164

SD
N

Male respondents
AM1 M SD Hostility toward Men M SD Benevolence toward Men M SD N

1.88 .78
1.84 .89

2.04 .93 2.05 .82 2.03 1.22 74

1.68 .77 1.74 .87 1.62 1.01 102

1.93 .95 146

Note: Scalc. scores are averages of item scores, with possible range of 0 (disagree strongly) to 5 (agree strongly). Total AM1 score = avcragr of all items; H M and BM siihscalr scores = averages of items within each scale

demonstrating bias against men as an outgroup and men showing an ingroup bias). Female respondents consistently scored higher on HM and lower on BM than did male respondents within all three samples (see Table 4 for means). A 2 (gender of respondent) x 2 (AM1 subscale) analysis of variance, with AM1 subscale (HM, BM) as a repeated-measures factor, was computed within each sample. All three samples showed significant main effects for AM1 subscale, with HM scores consistently higher than BM scores, all F s (1, 162 or greater) > 34.90, p c .001. There were no gender differences in overall AM1 scores, as indicated by the lack of a significant gender of respondent main effect, all F s (1, 1fi2 or greater) < .80, n.s. In all three studies, however, there were significant AM1 subscale x Gender of Respondent interactions, all Fs (1, 162 or greater) > 26.15, p < ,001. Post hoc coinparisons (using Tukeys test) showed that feaiale, as compared to male, respon-

Ambivalence TowardMen

529

dents were significantly more hostile and less benevolent in their attitudes toward men in all three samples, all q s (4,162 or greater) 2 3.89, p < .05. As these mean differences are consistent with theories of bias against outgroups and in favor of ones ingroup, they can be taken as preliminary evidence for the AMIs validity. A second set of contrasts showed that, in all three samples, women scored significantly higher on HM than on BM (all 4s 2 13.61,p < .Ol), whereas men scored equally on each scale (all q s 21.82, n.s.).
Convergent and Discriminant Validity of the AM1

Study 2 was aimed at establishing the convergent and discriminant validity of the AM1 by exploring its relationship with other scales of attitudes toward men and with the hostile and benevolent sexism subscales of the ASI.

The AM1 and Other Scales o f Attitudes Toward Men Although there are many fewer scales of attitudes toward men than of attitudes toward women, several have been published (see Beere, 1990, for a review of the various scales). Two of the more comprehensive and psychometrically sound scales are those of Downs arid Engleson (1982) and Iazzo (1983), both labeled the Attitudes Toward Men Scale. Both scales tap a wide variety of domains (e.g., work, sexuality, parenting), Although the scales bear some similarity to the AMI, neither attempts to distinguish simultaneously between traditional attitudes abont men that are subjectively positive and those that are subjectively negative. The Downs and Engleson scale explicitly distinguishes between traditional and nontraditional (or egalitarian) attitudes about men and their roles, but does not address the subjective positivity or negativity of these attitudes. Examples of items incluck Men find wartime combat enjoyable because it gives them a sense of power, It is ridiculous for men to stay at home and keep house, and Men prefer sex to other activities. Some of the items on this scale seem to represent subjectively positive attitudes (e.g., Men are more sure of what they can do than women are), whereas others are clearly negative (e.g., Men feel threatened by intelligent, educated women). In contrast, Iazzo casts his scale more in terms of favorable versus unfavorable attitudes about men, though the items on the scale may be seen as tapping traditional versus nontrahtional beliefs. Examples include Men often view women as one of their possessions, A mans job is the most important thing in his life, and Men only want one thing from a woman. As we scored this scale, high scores indicate what Iazzo claims to be negative attitudes toward men. This is opposite to Iazzos scoring, hut makes aliigli score (we think) more traditional, and thereby simplifies comparisons with the other scales. To the extent that both Attitudes Toward Men scales measure traditional views of men, convergence with the AM1 was expected. Because the Downs and Engleson scale does not distinguish between subjectively positive and subjectively negative traditional attitudes, but seems to assess both, the HM and BM subscales ought to correlate equally well with this scale. As Table 5 reveals, this was the case for both men and women. The total AM1 score correlated very strongly with the Downs and Engleson scale (listed as AMSl in the table) and each AM1 subscale (HM and HM) contributed about equally to this relationship (as indicated by the partial

530

GLICK A N D FISKE

Table 5
Relationship of the Ambivalence toward Men Inventory (AMI) and its Subscales to Other Measures of Attitudes toward Men and Women

Measures of AttitiAda Towerd Men and Women AMI score AMSl


AMS2 AS1

HS

BS

Femnb respondents (n = 114)


AM1 overall HM BM HM controlling for BM BM controlling for HM .85" .72"
32"

.70" .61"
68"

.46"
.64"

.34" .45"

.76" .56" .79" .15 .67"

.71" .49" .74" .09 .63"

.61" .47" .61" .14


.4S"

Male respondents (n = 65)


AM1 overall HM BM HM controlling for BM BM controlling for HM
.68" 56"

.70" .67"
.48"

.69" .36"
.80* .ll

57" .43" .48"

.60" .34"

.76"

.49" .17 .65" -. 13 .64"

.73"
.48"

74" .32" .67"

Note: AM1 = Amhivalence toward Men Inventory; H M = Hostility toward Men; BM = Benevolence toward Men: AMSl =Attitudes toward Men Scale by llowns and Engleson (1982); AMS2 = Attitudes toward Mrri Scale by lazzo (1983); AS1 = Arnbivaleiit Sexism Inventory; HS = Hostile Sexism; BS = Benevolent Sexism. The sample sizes are lower as not all respondents completely filled out the attitudes toward men scales. * p < .01.

correlations, which control for the relationship between the HM and BM subscales). If Iazzo is correct that his scale (listed as AMS2 in the table) measures positive versus negative attitudes toward men, then HM and BM should show opposing correlations with this scale (positive for HM and negative for BM). This was not the case. Rather, the results were quite similar to those with the Downs and Engleson scale, suggesting (presuming the AM1 is valid, an assumption for which evidence will be presented) that the Iazzo scale also measures traditional versus nontraditional attitudes about men, rather than distinguishing positive from negative attitudes (additional evidence for this interpretation is provided by the predictive validity data that follows)

The AM1 and the AS1 Because of the common theoretical backgroiirid of the AM1 and the ASI, we expected some strong convergences between the scales. Theoretically, the attitudes tapped by both the AM1 and the AS1 are a result of men's structural power and women's dyadic power. Further, the AM1 and AS1 subscales are all geared to tap into issues of power, gender differentiation, and heterosexuality. BS and BM, in particular, ought to be strongly related to each other, for both sets of attitudes presume that men and women are complementary halves of a whole (e.g., that

Ambivalence Toward Men

531

men and women need each other to be complete), with the distinction that BM examines these beliefs through statements in which men, rather than women, are the attitude object. HS and BM should also converge because hostile sexists (who view women as inferior to men) would, almost by definition, be likely to have benevolent attitudes toward men. In contrast, we predicted that the HM subscale should not show strong convergence with the HS and BS subscales (once its correlation with BM is partialed out) because, unlike BM and the AS1 subscales, the attitudes tapped by the HM scale do not explicitly serve asjustijications of traditional gender relations (rather, HM involves an acknowledgment of such traditional arrangements along with a resentment of the power this gives men). Thus, the HM scale should demonstrate discriminant validity when compared to the ASI. Correlations between the AM1 and AS1 are shown in Table 5. As expected, BM was strongly positively correlated with both HS and BS. This held true for both male and female respondents. Thus, those who are sexist toward women are also likely to be benevolent toward men. In contrast, the HM subscale (once its relationship with BM was partialed out) tended not to correlate with the AS1 subscales. The one exception to this rule was a weak but significant correlation between HM and BS among male respondents; this is consistent with earlier research which demonstrates that benevolent sexism in men is correlated with recognition that women are discriminated against (Click 87 Fiske, 1996) and more recent data that benevolent sexism in men predicted suspicion about the motives of a man observed to offer unsolicited help to a woman (Wegmann 87 Bushman, 1997).If one considers who benevolent sexist men seek to protect women against, the answer may often be other men. Thus, benevolent sexist inen may be suspicious of other mens motives, resulting in higher HM scores.
Predictive Validity of the AM1

A more stringent test of the validity of the AM1 was built into Studies 2 and 3, in which participants spontaneously generated traits they associate with men and then rated whether each trait was a negative or a positive attribute (on a -2 to +2 scale). If the AM1 subscales differentiate between hostile and benevolent attitudes toward men, then (after controlling for the positive relationship between the AM1 subscales) HM ought to correlate negatively and BM positively with the valence (negative/ positive) ratings of the traits respondents spontaneously associated with men. In contrast, the total AM1 score ought not to correlate with the valence ratings as it is conrposed in equal measure of hostile and benevolent beliefs about men. This is precisely the pattern of results obtained for both men and women (see Table 6). Both for female undergraduates and women in the more diverse community sample, HM was significantly negatively correlated to the valence ratings and BM positively correlated to these ratings. Although the total AM1 score was significantly negatively correlated with valence ratings among female undergraduates, this correlation was of trivial magnitude; for women in the community sample, the correlation was nonsignificant. Male respondents generally showed a similar pattern of results, although the negative correlation between HM and the valence ratings was trivially small and nonsignificant in the community sample. In general, the consistency of

532

CLICK AND FISKE

Table 6
Correlation of Hostility toward Men (HM) and Benevolence toward Men (BM) to the Valence of Traits Associated with Men Across 2 Studies

A M I score

Undergraduate sample (Study 2 )

Comrtiunity sample (StILd, 3 )

Fernale respondent,>

AM1 overall IIM controlling for BM BM controlling for HM


N

-. 19"
-.44*"

.25"" 114

-.07 -.33"* .26" 75

Mule respondents

AM1 overall HM controlling for BM BM controlling for HM N

.10
-.26"

.31""

65

.18 -.10 .33*" 47

Note: AM1 = Anrl~ivnlrnce toward Men Iriveirtory, I-IM = Ilostility toward Men, BM = Benevolence toward Men. Tlrc sarnplc sizes are lower hecaiise wmre participants in Studies 2 and 3 did not frilly complete the rrieasurca. 7 1 < .05; **p < .01.

the pattern of correlations across the sexes and across two different kinds of samples supports the predictive validity of the AMI. In contrast, with the exception of a weak, but significant negative correlation for women on the Iazzo scale ( r = -.20, p < .05), neither of the Attitudes Toward Men scales predicted the valence of the trait ratings (strongest r = .17, n.s.). As we argued previously, it makes sense that hostile sexists, who view women as inferior, would have favorable views of men as a group. This was confirmed by the positive correlation between HS and the valence ratings, for both female ( r = 21, p < .05) and male (r = 2 6 , p < .05) respondents.
DISCUSSION

These three studies suggest that some women experience an ambivalence toward men that is structured along the predicted dimensions. The factor structure of the AM1 proved to be highly stahle across three samples, supporting the theoretical contentions on wliich both the AS1 and the AM1 are built and confirming that attitudes of each gender toward the other can be understood in terms of the dimensions of paternalistic power, gender differentiation, and heterosexuality. In all three studies, the preferred factor model ( H M and BM, each with three subfactors) fit the data well, and significantly better than more parsimonious alternative models. The AM1 also demonstrated reliability and predictive validity (across two samples), with the HM subscale predicting the ascription of negative traits and

Ambivalence Toward Men

533

the BM subscale the ascription of positive traits to men as a group. Although the AM1 strongly converged with other scales of attitudes toward men and with the ASI, none of the other scales showed the AMIs ability to predict both positive and negative evaluations of men as a group. The Nature of Womens Ambivalence Toward Men That HM and BM correlate positively with each other suggests that some women simultaneously hold beliefs that actively support and justify male dominance (BM) at the same time that they resent the consequences of this dominance (HM). This may seem to be a dissonant set of beliefs, but this pattern of ambivalence is highly consistent with Jackmans (1994) work on the nature of paternalism and its effects on subordinate groups. Subordinate group members are often coerced and enticed to accept justifications of dominants higher status (see also Jost & Banaji, 1994), even as resentment of this status festers. Jackman (1994) shows that dominants rig the social and economic reward structure to give subordinates considerable incentive to buy into the status quo by providing them enough of a stake in the system that they have something to lose if they rebel and by punishing them if they do. Hostile and benevolent sexism may reflect these forms of control. Glick, Diebold, Bailey-Werner, and Zhu (1997) demonstrated that, for men, benevolent sexism predicts more favorable evaluations of women who adopt traditional roles. In contrast, hostile sexism predicts less favorable evaluations of women who take on nonconventional roles. Given mens ability to reward women who adopt traditional roles and punish those who do not, it is not surprising that many women would adopt benevolent beliefs about men that justify male power. The women who adopt such beliefs (and, presumably, the more traditional roles that go along with them) are, at the same time, increasing their dependence on men, providmg strong reasons both to resent and to respect mens power. In short, we speculate that the greater the dependence a woman has on men, the more she is likely to experience both benevolence and hostility toward men; the former because of her recognition of her investment in men and the latter because ofresentment over her dependence. A less interesting possibility as to why HM and BM are correlated is a possible methodological artifact: acquiescence bias. Despite several attempts, we were unsuccessful in structuring reverse-worded items that would yield acceptable loadings on the AM1 factors. Thus, it is possible that HM and BM correlated because of respondents who tended to endorse all items. We argue against this explanation for two reasons. First, the HM-BM correlation mirrors the HS-BS correlation found for the ASI, a correlation that was unaffected hy the addition of reversal items (Glick & Fiske, 1996), indicating that acquiescence bias is not a problem with the AMIs sister-scale (a scale that is itself highly correlated with the AM1 and that taps related issues). We believe that the difficulty in constructing reversal items is a semantic issue (the difficulty of framing negative versions of the statements). In support of this, we have found that only the reversed items (not the nonreversed items) on the AS1 do not perform well when translations are attempted into other languages. Second, acquiescence bias cannot explain the correlation between both HM and BM and the Downs and Engleson (1982) scale, which measures traditional attitudes toward men, as this scale contains many reversal

534

CLICK A N D

FISKE

items. That both HM and BM correlate with this scale adds support to our contention that HM and BM are correlated because they both measure traditional attitudes. Regardless of the explanation for its existence, partial correlations should be used to control for the HM-BM correlation when using the subscales as predictors. Although the AM1 was developed with womens attitudes toward men in mind, the factor structure of the scale was similar for n d e and female respondents. Additionally, as with female respondents, the AM1 demonstrated some degree of predictive validity for male respondents (though not as consistently). For both sexes, the endorsement rate of the scales was generally low, though (as predicted) women, as compared to men, were more hostile and less benevolent toward men. The low endorsement rates may reflect politically liberal samples (college students, their relatives). Nevertheless there was sufficient variance in responses for the scales to demonstrate predictive validity.

The AM1 and Other Scales


We expected and obtained evidence for convergence between the AM1 and existing scales of attitudes toward men. The AM1 was strongly correlated with the Downs and Engleson (1982)and the Iazzo (1983)Attitudes Toward Men scales, but neither of these scales can yield separate hostility and benevolence scores for attitudes toward men. Although the HS scale of the AS1 was also predictive of more positive attitudes about men, the AM1 stands alone in its ability to differentiate between subjectively positive and negative attitudes about men (and, consequently, to measure ambivalence). The ability to make this distinction is crucial for understanding the emotional and evaluative aspects of womens attitudes about men, not only better to understand intergroup attitudes, but interpersonal relationships as well. Women high in both sets of attitudes may be prone to more volatile reactions toward men, just as ambivalent sexists have more polarized views about women (Click et al., 1997).
CONCLUSIONS

We believe that the AM1 will prove useful for assessing the ambivalent attitudes that women have toward men, revolving around issues of paternalistic power, gender differentiation, and heterosexuality. As such, it complements the ASI, which prilriarily assesses mens ambivalence toward women on these same dimensions. Given the frequency with which men and women interact and the intimacy that male-female relationships ,often achieve, it is crucial to understand how each gender perceives and feels about the other. Greater attention has been given to how men conceive of and evaluate women than how women evaluate men. We offer the AM1 as a tool for exploring the relatively neglected topic of how women think and feel about those who have, since the beginning of recorded human history, been their oppressors as well as their protectors, their strongest foes and most intimate partners.
initid slibnrissiun: October 21, 1997 Initial nccqtance: December 15, 1998 Final acceptance: March I , 1998

Ambiva/ence Toward Men


NOTE

535

1.Iazzo does not present any direct evidence that his scale ineasiires a positive-negative dimension. Rather, he supports this claim by showing that feminists, rape victims, battered wives, and lesbians had increasingly higher (according to onr coding scheme), more negative scores than &d a control group of women. The meaning of these differences, however, is not clear. Furthermore, the Iazzo scale was not strongly related to the valence (negativeipositive)of the traits respondents in our Study 2 associated with men.
REFERENCES

Allport, G. W. (1954). The nature of prejudice. Reading, MA: Addison-Wesley. f test.s and measures. New York: Greenwood Beere, C. A. (1990). Gender roles: A handbook o Press. Berscheid, E., & Peplau, L. A. (1983). The emerging science of relationships. In H. H. Kelley, E. Berscheid, A. Christiansen, J. H. Harvey, T. L. Huston, 6. Levinger, E. McClintock, L. A. Peplau, & A. Peterson (Eds.), Close relationships. New York: W. H. Freeman. Bohner, G., & Schwarz, N. (1996). The threat of rape: Its psychological impact on nonvictimized women. In D. M . Buss & N. M. Malamuth (Eds.), Sex, power, conflict. Evolutionanj and feminist perspectives. New York: Oxford University Press. Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley. Brownmiller, S. (1975). Against our will. New York: Simon and Schuster. Crocker, J., Thompson, L. L., McGraw, K. M., & Ingerman, C. (1987). Downward comparison, o f Personality prejudice, and evaluations of others: Effects of self-esteem and threat. Jm~rnal and Social Psychology, 52, 907-917. Deaux, K. (1995). How basic can you be? The evolntion of research on gender stereotypes.Journal of Soeinl 1.s.v1e.s9 51, 11-20, Downs, A. C., & Engleson, S. A. (1982). The Attitudes Toward Men Scale (AMS): An analysis of the role and status of men and masculinity. Catalog o f Selected Doccrmnts in Psychology, 12(4), 45, (Ms. No. 2503). Eagly, A. H., & Mladinic, A. (1993). Are people prejudiced against women? Some answers from research on attitudes, gender stereotypes, and jiidgments of competence. In W. Stroebe & M. Hewstone (Eds.),European rcview of social psychology (Vol. ,5, pp. 1-35). New York: Wiley. Glick, P., Diebold, J., Bailey-Werner, B., & Zhu, L. (1997). The two faces of Adam: Ambivalent sexism and polarized attitudes toward women. Pcrsonczlity and Social Psychology Bulletin, 12, 1323-1334. Glick, P., & Fiske, S. T. (1996). The Ambivalent Sexism Inventory: Differentiating hostile and benevolent sexisin. ]mima1 o f Personality and Social Psychology, 70, 491-512. Glick, P., & Fiske, S. T. (1997). Hostile and benevolent sexism: Measuring ambivalent sexist attitudes toward women. Ps!ychology o f Women Quarterly, 21, 119-136. Guttentag, M., & Secord, P. (1983). Too many women?-Beverly Hills, CA: Sage. Iazzo, A. N. (1983).The construction and validation of Attitudes Toward Men Scale. Psychological Record, 33, 371-378. Jackman, M. R. (1994). The d u e t glove: Paternalism and conflict in gender, cla.ss, and raw relations. Berkeley, CA: University of California Press. Joreskog, K., & Sorbom, D. (1993). LZSREL 8: Structural eyuation modeling with the SIMPLIS command language. Hillsdale, NJ: Erlbaum. Jost, J., & Banaji, M. (1994). The role of stereotyping in system justification and the production f Social Psychology, 33, 1-27. of false consciousness. British Journal o Tajfei, H. (Ed.).(1981). Social identity arid intergroup relations. London: Cambridge University Press. help. Wegrnnnn, S., & Birshrnan, N. (1997). Let me help you: The effect ofgender on nr.~trmptiue Unpublished honors thesis, Lawrence University, Appleton, WI. Woolf; V. (I929/1957). A room ofone,u oz~n. San Diego: Harcoiirt Brace Jovanovich.

536
APPENDIX: Relationships Between M e n and Women

CLICK AND FISKE

Helow are a series of statements concerningmen and women and their relationships in contemporary society Please indicate the degree to which you agree or disagree with each statement usiiig the scale helow:
0 disagree strongly

1
disagree somewhat

2
disagree slightly

3
agree slightly

4 aggrcr somcwliat

5
agree strongly

1.Even if both meinbers of a couple work, the woinan ought to be more attentive to taking care of her inan at tionir:. 2. A man who is sexually attracted to a woman typically has no morals about doing
wliatrvrr it takes to get her in bed.

3. Men are less likely to fall apart in emergencies than women are. 4. Wien n1e11 act to helpwornen, they are often trying to prove they are hetter
than women.

5. Every woinan needs a male partner who will cherish her. 6. Men would be lost in this world if women werent there to guide them. 7 . A woinan will never be truly fulfilled in life if she doesnt have a committed,
long-term relationship with a rnan.

8. Mcri act like babies when they are sick. 9. Men will always figlit to have greater control in society than woinen. B(M) 10. Men are mainly useful to providr financial security for woinen. H(1) _ _ 1 1 . Even men who claim to be sensitive to womens rights really want a traditional relatioilship at home, with the woman performing rrtost of the housekeeping and child care:. B(S) -12. Every woman ought to have a rnan she adores. B(C) 1:3. Men are inore willing to put themselves in danger to protect others. H(S) _ _ 14. M e n usiially try to dominate conversations \vIien talking to women. H(P) -15. Most men pay lip service to equality for woinen, but cant handle having a
woman as an equal

B(S) -16.Wornen are incomplete without men. H(Gj ___ 17. When it comes down to it, most man are really like children. R(Gj 18. Men are more willing to take risks than women. II(S) _ _ 19. Most men sexually harass woinen, even if only in subtle ways, once they arr in a position of power over them. 20. Women ought to take care of their nien at lioine, because men would fall apart R(M) ~

if they had to fend for themselves.


Copyright 199fi by Peter Click and Sirvan T. Fiske. Use oftliis scale requires permission olone of the aiithnrs. Note: The NM siibscales are indicatcd by the following notation: II(P)Rrsrntrrlent of Paternalism, M(G) = Colripensatory Grnder Differentiation, 11(S) = Hetereosexual Hostility. The BM subscales are indicated by the tidlowing ~iotatiori:B( M)= Maternalisin, B(G) = Corr~plei~~entary Gender Differentiation, B(S) = 11cteri)sr~x~Id
Irrtilrlilcy.
Scoring [irstniclions:

Hostility tow;ird Men (IIM) wore = average of the following itrrria: 2,4,6,8,Y,11,14,15,1i,lY. Bmr\oleiic.r tow.ird Me11 (BM) scorc = avr.rage the lollowi~~g items: 1.:3.5.9,10,12.13.1lj,18,20. Overall A M 1 score = average of dl items. For correlational research, purer measures of H M and BM caii hc obtained by paltiding each of tllcw subscalcs from the other to control for the correlation between the two subscalcs. A n ovrrall AM1 scow ear1 he ohiained merely by iwvmiging scores m i ;dl items (01- averaging thr HM a r i d BM suhscal

Potrebbero piacerti anche