Sex Roles, Vol. 39, Nos. 1/2, 1998
Th e New Old-Fashion ed Girl: Effects of Gender an d Social Desirability on Rep orted 1 Gender-Role Ideology Step hen W. Th eriau lt an d Dian e Holm berg2 Acadia University
In order to assess the effects of social desirability on (1) reported gender-role id eolo gy an d (2) th e relativ e acceptan ce of rape m yth s, a sh ort so cial desirability (SD) scale was in tegrated into a qu estionn aire containin g item s from both attitude sets. Participants were 160 un iversity students. Less than 5% of the sam ple was of other than European eth nicity. A gender by SD interaction was predicted, wherein high SD would predict m ore femin ist beliefs for m en; the relation ship was expected to be less stron g for wom en. Although a significant gender by SD in teraction was fou nd, the pattern of results was un expected. SD did not predict attitudes for m en, an d wom en who were high in SD actu ally reported m ore traditional gender-role attitudes and were m ore ac ceptin g of rap e m yth s th an wom en lo wer in SD. In terpretatio n s an d im plication s of these unexpected findin gs are discussed.
“ Woman, not girl ” . “ Equal pay for work of equal value ” . “ No means no. ” O ve r the last 35 ye ars, arme d with phrase s such as these, represe ntative s of the woman ’s move ment have be en waging a war for the he arts and minds of the populace . The y have fought a numbe r of key battle s ove r those ye ars. O ne broad theme has be en in the area of gende r-role ideology (Kalin & Tilby, 1978; Mille r & Simpson, 1991) . Feminists have atte mpte d to replace a more traditional ide ology (e .g., be lie fs that men and wome n are inhe r1
This study originated in the unde rgraduate Honors thesis of the first author. The study is supported in part by Acadia University, in the form of a Manning Me morial Research Fe llowship, and by the Fundy Me ntal He alth Association, Ke ntville, Nova Scotia. 2 To whom correspondence should be addressed at Departmen t of Psychology, Acadia University, Wolfville, Nova Scotia, B0P 1X0, Canada. E-mail may be se nt to ‘Diane .Holmbe rg@ acadiau.ca ’. 97 0360-0025/98/0700-0097$ 15.00/0
Ó
1998 Plenum Publishing Corporation
98
Th eriau lt an d Holm berg
ently differe nt and that there are rights and role s appropriate to each ge nder) with a more e galitarian or fe minist set of be lie fs (e .g., be liefs that ge nde r is a social construct, and that there are few inhe rent ge nde r diffe rence s). Anothe r important battle has be en fought in the are a of be lie fs about appropri ate se xual inte ractions . For e xample , fe minists have atte mpte d to re duce belief in rape myths, which are prejudicial, stere otype d or false be lie fs about rape , rape victims or rapists (Ande rsen, 1993; Brownmille r, 1975; Dobash and Dobash, 1979) . They have attempted to e ducate people that be liefs such as “ any he althy women can resist a rapist if she really wants to ” or “ all rapists are se x-starve d, insane , or both ” (Burt, 1980) are not only false but also dange rous, le ading to behaviors such as blaming the victim and minimizing the importance of date rape . In at le ast some venue s, feminists have been quite succe ssful in the ir campaign to influe nce pre vailing attitude s. For example , whe ther you de em it progre ss or political corre ctne ss, few acade mics on unive rsity campuse s today would pre sume to ope nly advocate once-common beliefs such as “A woman ’s place is in the home ” or “ Nice girls should know to draw the line before things ge t out of hand ” . Those who do e xpre ss such views are quickly labe le d as sexist or une nlighte ne d. Acade mics as a group thus te nd to be perceived as a re lative ly libe ral, egalitarian lot. Unfortunate ly for inve stigators wishing to conduct e mpirical re se arch on gende r-related issues, such a situation could prove highly proble matic. Constructs such as ge nder-role ideology and rape myth acce ptance are freque ntly studie d using self-report instrume nts (Bee re , King, Bee re & King, 1984; Burt, 1980; Spe nce, Helmre ich & Stapp, 1973) . If re sponde nts are well aware of which response s to such instrume nts would constitute socially-acce ptable vie ws, at le ast in the mind of the acade mic conducting the study, the n social desirability be comes a major thre at to the validity of such measure s. Social desirability has bee n de fined as “ the te nde ncy of individuals to attribute to themselve s, in terms of se lf-descriptive personality statements, those with desirable value s and to re ject those with unde sirable value s” (Edwards, 1957, p. 12) . Such a te nde ncy may come about because the individual is e ngaging in se lf-deception (optimistically playing down the ne gative aspe cts of self and emphasizing the positive aspe cts, probably not fully consciously), or be cause the individual is e ngaging in impression manage ment (consciously trying to make him or he rse lf appe ar in a positive light) (Paulhus, 1984) . In e ithe r case , proble ms can arise whe n a particular se t of re sponse s to a give n measure (patte rn A) is clearly more desirable to a particular audie nce than a diffe rent se t of re sponse s (patte rn B); in such case s, individuals high in social de sirability will te nd to give re sponse s which match patte rn A, regardle ss of the ir true opinions. Such a proble m is ge n-
Old-Fash ion ed Girls
99
erally diagnose d by showing that the measure in que stion corre late s with a standard measure of social de sirability, such as the Marlowe -Crowne social-de sirability scale (Crowne & Marlowe , 1960) . In such case s, a confound is introduce d, thre ate ning the validity of the scale in question; it is no longe r clear whethe r any correlations found be tween this scale and a crite rion variable represent true relationships with the conce pt being measured, or whethe r the y simply re pre sent relationships with social de sirability. Demonstrating freedom from social desirability bias is the re fore an important step in evaluating any measure , particularly in domains whe re the socially-de sirable patte rn of responding is cle arly e vide nt. Fortunate ly for those inte re ste d in studying ge nder-relevant issue s, some commonly-use d se lf-report measure s appe ar to be relative ly free of social de sirability bias. In the are a of ge nder-role ide ology, for e xample , one freque ntly-use d questionnaire , the Sex-Role Ideology Scale (SRIS) has shown non-significant correlations with the Jackson social-de sirability measure (Cota & Xinaris, 1993) ; anothe r measure , the Sex-Role Egalitarianism Scale , shows non-significant corre lations with the Edwards social-de sirability measure (Bee re , e t al., 1984) . Most convincingly, Buhrke (1988) studie d four differe nt measure s of ge nde r-role ide ology, and found that none of the m were significantly corre late d with the Marlowe -Crowne social-de sirability scale. Less work has bee n done with measure s of rape -myth acceptance ; in one study, howe ve r, Spohn ( 1993) found n on -sign ificant corre lations be tween the Rape Myth Acceptance Scale and the Marlowe Crowne social-de sirability scale . Thus, in spite of any message s they may have receive d about the desirability of egalitarian attitude s and be haviors, there was no indication that responde nts in these studie s allowe d the ir tende ncy to re spond in a socially desirable fashion to affect the ir re sponse s to these que stionnaire s. Thre e possibilitie s come re adily to mind to e xplain these results. The first is that the findings do in fact indicate precise ly what the y appe ar to, i.e . re spondents answer the se questionnaire s absolute ly hone stly, with no trace of socially-de sirable re sponding. The second is that the patte rn of re sponse s which form a socially desirable patte rn is not actually clear; if some responde nts fe el it is “ be tter” to respond in a libe ral fashion, while othe rs feel it is “ bette r” to re spond in a more traditional fashion, the two patte rns of response may cancel each othe r out, le aving no overall re lationship with social de sirability. A third possibility is that the scale s are sufficie ntly subtle that responde nts are unable to de te rmine which re sponse s would indicate a libe ral vs. a traditional attitude . This possibility se ems unlike ly, as all of the scales appe ar quite high in face validity. Furthe rmore, a study by Jean and Reynolds (1984) instructe d male and fe male unde rgraduate s to respond to a measure of gende r-role ide ology, the Attitude s towards Women
100
Th eriau lt an d Holm berg
Scale (Spence e t al., 1973) in ways that they conside red to be ‘libe ral’ or ‘conservative ’. Both men and wome n had no difficulty in manipulating the ir score s in either direction. Thus, if individuals were motivate d to appe ar in a particular fashion on the se measure s, it appe ars the y would be able to do so with e ase . We propose a fourth possibility to e xplain the lack of social desirability effects on the se measure s: that the methodology use d in colle cting the data may have maske d true social desirability effects, effects which still might color the results of othe r studie s using these measure s. There are two aspects to our critique . First, some of the se studie s (Be ere e t al., 1984; Buhrke , 1988) do not report social desirability correlations separate ly for men and women. O ne might anticipate that the effe cts of social desirability on gende r-role ideology and rape -myth acce ptanc e would be stronge r for me n than for women, since the social conse quence s of appe aring to be backward or reactionary in the se are as would be greater for unive rsity men than for the ir female classmate s. Additionally, it could be argue d that these are as are of pe rsonal importance to wome n and that the y should have strong, well thought out opinions as a conseque nce, thus re ducing self-de ception or the nee d for impression manage ment. For the se reasons we would e xpe ct that corre lations with social de sirability would be stronge r for me n than for women in the se attitude se ts; studie s which combine results from the two ge nde rs might be masking important inte ractive e ffects. This aspe ct of our critique cannot be the whole story, however, since two studie s (Cota & Xinaris, 1993; Spohn, 1993) do report the gende r breakdown, and the social desirability corre lations remain non-significant for the males (note , howeve r, that the re were twice as many female as male participants in both studie s). The se cond aspe ct of our critique applie s to all of the studie s mentione d. Each re pre sente d a large -scale validation e ffort; as a re sult, e ach study consiste d of participants being give n an assortment of que stionnaire s to re spond to. Whe n give n a batte ry of questionnaire s by a psychologist, participants (particularly young unde rgraduate s with little familiarity with psychological assessment te chnique s) may assume that the y are be ing cleverly analyze d and e xamine d in ways they do not fully compre he nd. Just as participants in “ bogus pipe line ” studie s (Jones & Sigall, 1971) give less socially-de sirable response s when they be lie ve the researcher can ascertain the ir true thoughts and fee lings, so too might participants in the se studie s decide that hone sty is the best policy, since cleve r and consiste nt impression manage me nt across m ultiple que stionnaire s would be too difficult to achie ve. Some variability on the social de sirability measure will remain, due to true diffe re nces in the behaviors be ing que stione d and due to se lf-de-
Old-Fash ion ed Girls
101
ceptive aspe cts of social desirability, but the impression-manage ment aspe ct of social de sirability will be minimize d on all measure s, le aving no significant correlation be twe e n social desirability and the substantive domain. Howe ver, in othe r studie s whe re the ge nder-role ide ology or rape myth acceptance scale is the only one administe re d, atte mpts at impression manage ment may again come to the fore . Two pie ces of research provide sugge stive e vide nce that this line of reasoning may have merit. First, a study by Kashy and DePaulo ( 1996) aske d participants to kee p diary re cords of every lie they told for one week. O ne would certainly e xpe ct a measure of fre quency of lying to be re lated to social de sirability. Contrary to e xpe ctations, howeve r, the re was absolute ly no correlation be twe e n social desirability and numbe r of lie s reporte d. Again in this study, participants were give n a batte ry of personality measure s at the beginning of the study, one of which was the social desirability measure . Face d with this intimidating array of psychological measure s, participan ts m ay ha ve simply de cid e d to te ll the truth to the rese archers across the board, re moving the impression-manage ment aspe cts from their social desirability response s and le aving no re lationship with the lying measure . A study by Albright and Chang (1976) also le nds crede nce to our spe culations. This study was the only one we found which re porte d a relationship be twe e n a ge nde r-role ide ology scale (the Attitude s towards Wome n scale) and social de sirability. In this study, conducte d over 20 years ago, both men and women who were highe r in social desirability espouse d more traditional attitude s on the AWS. This study, showing differe nt results from the othe rs reviewed above , also use s a diffe rent methodology: rathe r than participants being give n all measure s at one time , the y were give n the social-de sirability measure in a take -home form several weeks before be ing administe re d the AWS. Any conne ction be tween the two in participants ’ minds was therefore minimize d. Perhaps whe n participants are give n the opportunity to take questionnaire s one at a time, the y fee l fre e to atte mpt some judicious impression manage ment. In the curre nt study, we se ek to explore the relationship be twee n social desirability and measure s of (a) gende r-role ideology and (b) rape myth acceptance . O ur study uses a some what diffe rent methodology from the previous studie s which e xplore d these links. We inve stigate gende r diffe rence s in the relationship betwe en social de sirability and the se constructs, to e nsure that no inte ractive e ffects are be ing maske d. Additionally, rathe r than presenting participants with a batte ry of measure s, we pre se nt them with a single que stionnaire containing items from a ge nde r-role ideology scale, a rape -myth acce ptance scale , and a social-de sirability scale , all inte rspe rse d.
102
Th eriau lt an d Holm berg
We predict that unde r these circumstance s, when individuals fee l le ss unde r the psychologic al microscope , an e ffect of social desirability will emerge , such that individuals highe r in social desirability will espouse a more e galitarian gende r-role ideology and more ne gative attitude s towards rape myths than will those lower in social de sirability. We also predict a ge nde r by social de sirability interaction, such that the afore mentione d e ffect of social desirability will be stronge r for men than for wome n, for the reasons outline d earlie r in the introduction.
METHOD Participants A total of 164 unive rsity students (124 women and 40 men, mean age = 21.3 years, range = 17 to 50 years) complete d the study’s que stionnaire . A precise breakdown of the sample ’s ethnicity is not available , but le ss than 5% were of other than Europe an ethnicity, a proportion reflective of the composition of our university’s population. The sample was initially recruited for a study on the effects of education (both general unive rsity education and specific education on gender and sexuality issues) on respondents ’ gender-role ideology and acceptance of rape myths. The sample is therefore more diverse than the usual introductory psychology group, including individuals from all four unde rgraduate years, including some individuals who were currently taking or had previously taken a course designe d to addre ss gender and sexuality issues. Materials Participants comple ted a single que stionnaire consisting of items sele cted from thre e separate measure s: the Se x-role Ide ology Scale (SRIS) , the Rape Attitude s and Pe rceptions Q uestionnaire (RAP), and the MarloweCrowne Social Desirability Scale (MCSDS). The SRIS (Kalin & Tilby, 1978) is compose d of 30 ite ms that probe the responde nt’s vie ws on which social roles are appropriate for wome n in a variety of settings: work role s, pare ntal responsibilitie s, pe rsonal re lationships between men and wome n, e tc. The authors ’ initial study re ports split half reliabilitie s ranging from .79 to .84 in stude nt sample s, and a test-re stest reliability of .87. There was also re ported evide nce of concurre nt validity, in that the SRIS was able to distinguish between the members of feminist and traditional crite rion groups. In the format presented in the present study, SRIS score s had a possible range of 12-84. The RAP (Holcomb, Holcomb, Sondag, & Williams, 1991) consists of 20 ite ms which ask the re sponde nt to agre e /disagre e with commonly held
Old-Fash ion ed Girls
103
‘mythical ’ vie ws on sexual assault, such as the idea that women often say no when they mean yes, or that it is impossible to ‘rape ’ a prostitute . The authors re port an alpha value of .74 and split-half reliability of .75. A subseque nt study, using a Likert-type format instead of a dichotomous format, reporte d an alpha of .86 (Holcomb, Sarve la, Sondag & Hatton-Holcom b, 1993) . In the pre sent study, the RAP has a possible range of 8-56. The MCSDS (Crowne & Marlowe , 1960), in its original form, is a 33item scale that pre se nts individuals with a se ries of very unlike ly attitude s and traits and asks them to indicate if each state ment is re flective of the ir own personality. A positive ly scored e xample sugge sts “ I don ’t find it particularly hard to ge t along with loud mouthe d, obnoxious pe ople .” The psychome tric prope rties of the MCSDS have been exte nsive ly inve stigate d (for e xample s, se e E vans, 1979; Fische r & Fick, 1993; Greenwald & Satow, 1970) , and it is still wide ly used after almost 40 years. The short form of the MCSDS use d in the pre se nt study has a possible range of 4-28. We wished to create a shorte ned ve rsion of these measure s, to keep the questionnaire combining ite ms from all three brief, and to avoid arousing evaluation appre hension in our participants. The factor structure of the SRIS has be en inve stigated (Cota & Xinaris, 1993) and found to be unidime nsional, making short forms appropriate . Perhaps because it is a new measure, the RAP has not been so inve stigate d, but the creators of the scale have themselve s used shortene d versions with acceptable re sults (Holcomb, e t al., 1993). Gree nwald and Satow (1970) report that shortened scales of eve n two to four MCSDS items can achie ve acce ptable item-total score reliability. Short forms of all thre e questionnaire s the refore see med to be justifiable . Because validate d short forms of the SRIS and the RAP have not be en publishe d, a pilot study was conducte d to sele ct short forms appropriate to our population. Fifty-seve n re sponde nts (10 males, 47 fe male s, from 18 to 49 ye ars old) were recruite d from summer session classes in psychology and sociology at Acadia Unive rsity and comple te d the full ve rsions of the SRIS and the RAP, as we ll as an alre ady validate d short form of the MCSDS (Fischer & Fick, 1993) . As in the main study, the thre e scales were integrate d into one que stionnaire . Inte rnal consiste ncy estimates for each comple te scale were calculate d; for the SRIS, a = .58, for the RAP, a = .51 and for the MCSDS, a = .44. Each item was e xamine d se parate ly, and ite ms which showed virtually no variability across the group were dele te d. Ite ms were also dele te d if they showed poor item-total corre lations. 3 3
RAP items deleted were # ’s 1,2,3,4,7,8,10,12,15,16,18 (Holcomb et al., 1991) . SRIS items deleted we re # ’s 1,2,3,5,6,7,8,9,13,16,17,1 8,19,20,23,24,28,29 (Kalin & Tilby, 1978) . MCSDS items deleted were # ’s 2,6,12,14,24,28 (Fischer & Fick, 1993) . MCSDS items retained were # ’s 4, 20, 21, 30.
104
Th eriau lt an d Holm berg
This procedure yie lded a 12-ite m SRIS with an a of .69, a 9-ite m RAP with an a of .62 and a 4-ite m MCSDS with an a of .4736. The se 25 re maining items were arrange d in a random orde r, with the provision that not more than two of the MCSDS items should appe ar in seque nce. The full que stionnaire was e ntitle d Attitude s about Gende r Roles and Sexual Assault. Each question was answered on a 7-point scale; ranging from strongly disagre e to strongly agre e. This format was the standard one for the SRIS, and had bee n previously use d with the RAP (Holcomb, Sarvela, Sondag & Hatton-Holcom b, 1993) . The MCSDS was originally designe d with a true -false answe r format; to maintain consiste ncy across items, howe ve r, the same 7-point response scale was used. Likert-type scales have bee n successfully used with the MCSDS in the past, and in fact seem to incre ase the reliability of short forms of the te st (Gree nwald & O ’Conne ll, 1970; Gree nwald & Satow, 1970) . Space s for participants ’ comments were inserted unde r each ite m. An additional page was adde d to allow participants to make more general comme nts.
Procedu re Each participant was informe d that the questionnaire ’s purpose was to measure attitude s about changing gender roles and commonly held be lie fs about sexual assault. Instructions on the que stionnaire e xplaine d the Like rt scale format and how to indicate the appropriate response . Participants who receive d the ir que stionnaire through the mail received additional instructions as to how to return their conse nt forms separate ly from the questionnaire . Every participant provide d their signe d consent to participate . Introductory psychology stude nts were recruited through a student participant pool and comple te d the questionnaire in several group sessions. Group sizes ranged from 6 to 12. Stude nts curre ntly enrolle d in a course on gender and sexuality issues were recruited and comple te d the que stionnaire during a normal class session during the second week of classes; all the stude nts in class on that day agre e d to participate . Stude nts who had previously take n the ge nder and sexuality course , and uppe r-year stude nts who had not take n such a course, were recruite d individually, via the unive rsity’s e-mail and ground mail facilitie s. In the first case , re cruitment was done from class lists provide d by the course ’s instructor; in the latte r case, individuals were randomly se lected from a list of all students provide d by the stude nt union. Response rate s were 67% and 40% , respectivly, for these groups. Afte r comple ting the questionnaire , participants received de brie fing information that e xplaine d the nature of the SD items and the ne cessity of including a measure of social de sirability.
Old-Fash ion ed Girls
105
RESULTS To begin our analyse s, we first calculated the simple correlations between social desirability, as measure d by the MCSDS, and ge nde r-role s and rape myths, as me asured by the SRIS and the RAP respectively. We hypothe sized that using our more subtle measure of social desirability, these correlations would be significant, in contrast to previous re search. Howe ver, we expected the correlations to be relative ly weak, since the simple corre lations mask potential diffe rences in socially-de sirable re sponding by ge nder. The corre lation betwee n scores on the SRI and scores on the MCSDS was in fact significant. Contrary to pre dictions, however, the corre lation was ne gative (r (164) = ¯.25, p = .001) : those individuals who were highe r in social desirability te nde d to re spond in a more traditional manne r, not a more feminist manne r as predicte d. In contrast, the corre lation be tween social desirability and the RAP measure was non-significant (r (164) = .07, p = .36) . To furthe r unde rstand these relationships, we conducte d a MANOVA. The de pende nt variable s were scores on the SRIS and the RAP. The indepe nde nt variable s were ge nde r and social desirability. Gende r had two le ve ls (male and fe male ); social desirability was tre ate d as a continuous variable .4 The MANOVA re vealed a main effect for ge nde r (Wilks lambda = .91; ove rall F (2,160) = 7.51, p = .001) . Consiste nt with previous re se arch (Adams & Betz, 1993; Holcomb et al., 1991; Holcomb e t al., 1993; Kalin & Tilby, 1978) , wome n, as a group, responde d in a more libe ral or fe minist fashion to both the RAP (F(1,161) = 9.77, p < .005; female mean = 13.20, SD = 4.41, male mean = 16.02, SD = 5.56, highe r numbe rs re pre sent gre ate r acce ptance of rape myths; possible score s range from 8-56) and the SRIS (F(1,161) = 10.62, p < .001, fe male mean = 57.83, SD = 8.18, male mean = 54.15, SD = 9.44, highe r numbe rs represent more feminist views, possible scores range from 12-84) than did men. Most re levant to the current inve stigation, howe ver, there was a significant ge nde r by social desirability inte raction (Wilks lambda = .95; overall F(2,160) = 4.58, p < .05) ; as predicted, the patte rn of socially desirable re sponding was differe nt for men and for women, on both the RAP (F(1, 161) = 5.90, p = .02) and the SRI (F(1, 161) = 6.54, p = .01) . To unde rstand the nature of these inte ractions, separate re gression analyse s were run for men and women. For each gende r, two re gressions were run with MCSDS scores as the pre dictor variable : one with SRIS score s as the crite rion variable , and one with RAP scores. As outline d in 4
An analysis treating social desirability as a categorical variable, with two levels defined by a median split, yie lded e xactly the same re sults.
106
Th eriau lt an d Holm berg
the introduction, we had initially anticipate d that high social desirability score s would be associate d with more libe ral vie ws on the RAP and the SRIS for both men and women, but that the re lationship would be stronge r for men. In fact, the re gression analyse s revealed a comple te ly diffe rent patte rn of results. For the men in our sample , social de sirability score s were not at all relate d to scores on e ithe r the SRIS (regression F(1, 39) = .36, p = ns, be ta = ¯.09) or the RAP (re gression F (1, 39) = 1.04, p = ns, beta = ¯.16) . In contrast, for the wome n in our sample , social desirability score s significantly pre dicted scores on both the SRIS (re gression F(1, 121) = 15.59, p < .0001) and the RAP (re gression F(1, 121) = 5.16, p = .02) ; howe ve r, the re lationships were pre cisely in the opposite dire ction as had bee n pre dicte d. For women, high social desirability score s were associate d with lowe r, more traditional score s on the SRIS (beta = ¯.34) and highe r, more rape -tolerant scores on the RAP (beta = .20) .
DISCUSSION In an attempt to clarify this puzzling effect, we conducted post-hoc analyse s, correlating each individual SRIS and RAP ite m with the total MCSDS score. We wishe d to determine if women high in social desirability were responding in a uniformly traditional manne r across all items, or if only certain ite ms accounte d for the se effects. The correlation of each item with the MCSDS score is shown in Table I; items from each scale are arrange d in orde r from those most highly correlated to those le ast highly correlate d with social desirability. The results of these correlational analyse s must be inte rpreted with caution, since a large number of analyse s are being conducte d; they are intended sole ly to sugge st possible inte rpretations of an une xpecte d finding. For the RAP, the majority of items showed no correlation with social de sirability for wome n, which was the re sult we had anticipate d overall. Women both high and low in social desirability are e qually like ly to reje ct all ite ms sugge sting that women provoke rape s, or that que stion the right of women in general to exercise control ove r their sexuality. At least three e xplanations could account for the lack of effect of social desirability on the se ite ms: (a) women could be presenting their hone st, dee ply-fe lt views on such central issues, (b) the socially de sirable re sponse could just happe n to coincide closely with the majority response , making it difficult to distinguish socially de sirable re sponding from true be lie f, or (c) a floor effect may be obscuring a true relationship with social de sirability for purely statistical reasons (five of the six RAP variable s showing non-significant relationships with social desirability have very low means, le ss than 1.5 on a 7-point scale).
Old-Fash ion ed Girls
107
Table I. Correlation of Individual Rape Attitude Perce ption (RAP) and Sex Role Inventory Scale (SRIS) Ite ms with Social Desirability, for Wome na RAP Items Item A prostitute can be raped.
SRIS Items r
b
.26
It is best that men initiate date s.
.21
Some wome n ask to be rape d and may e njoy it.
.08
Rape is often provoked by the victim.
¯ .03
If a woman is he avily intoxi¯ .03 cate d, it is OK to have sex with he r. If a woman says no to having ¯ .03 sex, she means maybe or e ven yes. Women frequently cry rape falsely.
If a woman says no to having sex, she means no.b
p
Items
r
.004 Abortion should be permitted at ¯ .33 the woman’s re que st. .02 Wome n should be paid a salary ¯ .33 by the state for the work they perform as mothers and home make rs. ns More day care centers should be ¯ .25 available to free mothers from the constant caring for their children. ns It is an outdated custom for a ¯ .23 woman to take he r husband ’s name when she marries. ns Wome n should not e xpe ct me n ¯ .21 to offer them seats in buses. ns
.03
ns
¯ .01
ns
A woman should be appreciative of the glance s and looks she rece ives as she walks down the stree t.b Eve ry child should be taught from an e arly age to feel a special honor and respect for Mothe rhood.b Homosexual relationships should be as socially acce ptable as heterosexual relationships. The first duty of a woman with young children is to home and family.b A woman should be no more concerned with her physical appearance on the job than a man. The be st thing a mother can teach her daughte r is what it means to be a girl.b
p .001 .001
.006
.01
.03
¯ .17
.06
.12
ns
¯ .08
ns
¯ .06
ns
¯ .04
ns
.03
ns
a
For RAP items, higher numbers indicate greate r acceptance of rape myths; thus, positive corre lations indicate that those higher in social desirability are more acce pting of rape myths. For SRIS items, higher numbers indicate more feminist vie ws; thus, ne gative correlations indicate that those higher in social desirability hold more traditional vie ws. b These items are reverse-code d.
For the SRIS, approximate ly half of the items showed no relationship with social de sirability. A floor or ceiling e ffect is a le ss like ly explanation for non-significant correlations he re, since the means for the items in question are not extreme (means range from 2.37 to 5.61 on a 7-point scale ). Howe ver, examination of responde nts ’ comments sugge st one plausible e x-
108
Th eriau lt an d Holm berg
planation, lack of consensus on what a socially desirable re sponse might be. In comparison to RAP ite ms, SRIS items are fre quently inte rpre te d in two or more ways. For e xample , the ite m about mothe rs te aching daughte rs what it means to be a girl was some time s interpre te d as representing inculcation of traditional value s, and sometimes inte rpreted in a fe minist light (te aching the next ge neration how to survive as a woman) ; the item about the first duty of a young mothe r be ing to home and family was sometime s reje cted, with the comment “ Why just mothe rs? It should apply to fathe rs, too ” , and some times accepte d, with the comment “ O f course, this applie s to fathe rs, too ” . Corre lations with social desirability may be obscure d on some of these items because of these multiple interpretations; 5 no one cle ar socially desirable alte rnative emerges. For now, this interpre tation re mains spe culative ; analyse s are ongoing comparing the qualitative and quantitative response s to each ite m, but the results are be yond the scope of this pape r. Thus, although the re asons are not comple tely cle ar, a numbe r of items on both scale s show no relationship to social desirability for women, as initially predicte d. O the r items, howeve r, did corre late with social desirability, and in a surprising direction. Examination of Table I reve als that wome n high in social desirability tend to express relative ly conse rvative vie ws on a numbe r of issues relating to the interface of the individual and the state. They endorse some re strictions on abortion, and seem negative towards prostitution. The y are he sitant about advocating salarie s for home makers or e xpansion of day-care alte rnative s. The se conse rvative -sounding young women also seem to e spouse a re lative ly traditional view of the relationship be tween the genders. It is not ne cessarily a bad thing for men to initiate date s, to offer wome n seats on buses, or to offer admiring looks and glance s. It is not necessarily an outdate d custom for a woman to want to take he r husband ’s name when she marrie s. The se young women, high in social desirability, seem to be portraying themselves as relative ly conservative and re lative ly traditional. The re sults of our inve stigation sugge st thre e main le ssons for researche rs inve stigating gende r-relate d social issues. First, participants ’ response s to questionnaire s on such topics may not always be fre e of social desirability bias. The se que stionnaire s appe ar relative ly bias-fre e in validation studie s, where in re sponde nts generally fill out a wide varie ty of measure s. In our study, howe ve r, where in participants are pre sente d with a single que stionnaire containing an unobtrusive social de sirability measure , social desirability e ffects emerge. It may be that individuals are disincline d to attempt impression manage ment whe n faced with a wide array of measures colle cted for rathe r obscure purpose s. Whe n face d with a single measure, howe ve r, especially in the context of a study in which othe r tasks may 5
These multiple interpretations may also account for the relative ly low reliability of our scale s.
Old-Fash ion ed Girls
109
have sensitize d them to ge nde r issues, participants may inde ed attempt to present the mselves in a socially-de sirable fashion. Given that the latte r scenario like ly matche s more studie s than the forme r, researchers have justification for conce rn. Future re se arch should directly compare diffe rent methods of que stionnaire administration to de te rmine the circumstance s unde r which social de sirability is of gre atest concern. The second main le sson for researchers is that the effe cts of social desirability may be gende r-spe cific. This situation should perhaps not be surprising. The scales we are inve stigating show strong ge nde r diffe re nces in our study and in othe rs (Adams & Betz, 1993; Holcomb et al., 1991; Holcomb e t al., 1993; Kalin & Tilby, 1978) ; it is reasonable to assume that men and women would also differ in the ir socially-de sirable re sponding to such strongly gende red issues. The patte rn of ge nde r diffe re nces in sociallydesirable responding, however, was not as hypothe sized. We had anticipate d that men would show stronge r social desirability effects than women; the opposite , in fact, occurred. We can only offer te ntative e xplanation for this une xpe cted finding. Pe rhaps men did not fe el motivate d to engage in e xte nsive impre ssion manage ment in this anonymous surve y colle cted by a male re se archer. Had the re searche r be en fe male , or the situation more public, stronge r attempts at impre ssion manage ment may have occurred. Anothe r possibility is that male responde nts did not feel confide nt that they kne w the “ correct” answe rs to the se e ver-e volving issue s. Relevant to this point, the only ite m on the RAP that showe d any correlation with social desirability, for male s, was the “ no means no” ite m. This one item, the focus of an e xtensive media campaign, was the only one to show e ffe cts: men high in social de sirability were more like ly to agre e that “ no means no” than were men low in social desirability (r = ¯.39, p = .01) . If the message is given cle arly and consiste ntly enough, it may influe nce men’s re sponding. Anothe r source of uncertainty, in the case of the RAP items, is the uniformly low mean item scores, the floor e ffect mentione d e arlie r. It could be that eve ryone , women and men both, are firmly convince d that sexual coercion and aggre ssion are be yond the pale and not to be conside red. This positive interpre tation is in contrast, however, with the large lite rature sugge sting that unive rsity men are not, in fact, ave rse to the use of coe rcion or aggre ssion (se e Ward, Dziuba-Le athe rman, Gerard Staple ton & Yodanis, 1994) . Anothe r, less positive , inte rpre tation is that social de sirability is such a strong force in this area as to flatte n variability in re sponse s to the RAP items. This would sugge st that the re sults of the pre se nt study substantially unde re stimate the e ffects of social desirability on this attitude set. It is difficult to de te rmine , base d on one se t of results, which of these se emingly contradictory e xplanations is most like ly.
110
Th eriau lt an d Holm berg
The third main lesson, of course, is that the effects of social desirability on wome n were ve ry differe nt from what we had anticipate d. We had e xpected that libe ral, feminist response s would be see n as more socially acce ptable on a colle ge campus, certainly at least for uppe r-year stude nts or for stude nts who had take n a course on ge nde r issue s. Appare ntly, we should have paid close r attention to social philosophe rs such as the Kinks, who open their song “ Young Conse rvative s” with the line “ Have you he ard the word? The re volution ’s ove r” (Davie s, 1983) and proceed to be moan the death of libe ralism in the young. Judging from our sample , young wome n today do inde ed seem to belie ve that a hint of social and moral conservatism is more socially desirable than hard-core libe ral views. This finding could be attributable to a widespread conse rvative upswe lling in our socie ty; howeve r, male responde nts did not show the same patte rn of socially-de sirable responding. Given that these same young women te nd to see a traditional patte rn of inte ractions betwee n the gende rs as more socially de sirable , we suspe ct that we are see ing empirical e ffects of the much-vaunte d “ backlash against feminism ” . These young women probably conside r the major battle s of the feminist movement to be won. It is worth noting that many of the items on the SRIS de aling with women ’s role in the workplace and division of labor within the home were discarde d afte r the pilot study because the y simply showe d no variation. Responde nts in that study ove rwhe lmingly agre ed that women have as much right to a career as men, and that men and women should share house work and pare nting dutie s. The se matters are conside red non-controve rsial. In an e nvironme nt where such practical issues are assume d to have be en resolve d in an egalitarian manne r (the re ality may be quite diffe re nt), wome n may se e no proble m pre se nting the mselves on less crucial or le ss personally rele vant issues as “ old-fashione d girls ” , comfortable with more traditional interactions betwe en the ge nde rs, and not insisting on such “ radical feminist ” ideas as abortion on de mand or salarie s for home make rs. The sample ’s fe male re sponde nts see m to prese nt a mixture of conservative morality and romanticism that may be quite out of step with social reality in present-day North America. The lack of progre ss in the are a of job equity, the feminization of pove rty, and the steady rates of viole nce against women and their childre n by men all sugge st that complace ncy on the part of young women is ill founde d. The se inte rpre tations of the wome n ’s re sponse s must only be presente d with caution, howeve r. Furthe r studie s are ne ede d to se e if the issues raise d here replicate . First, our post-hoc correlational analyse s re pre se nt multiple comparisons; proble ms with Type I error are a concern.6 Se cond, our re sponde nts atte nd a small, rural unive rsity in Atlantic Canada, whose 6
Note that we present the e xact p-value for significant corre lations so that re aders can apply the alpha value of their choice.
Old-Fash ion ed Girls
111
stude nt body comes pre dominantly from small towns. Results at large urban unive rsities, or in othe r re gions of North America, could be quite differe nt. Finally, we do not wish to make our responde nts look more conservative than the y are . O verall, response s for both men and women do indicate acceptance of feminist vie wpoints and reje ction of rape myths; the correlations with social de sirability represe nt a shading of re sponse s in a traditional dire ction, or a traditiona l re sponse by some re sponde nts, not a whole sale reje ction of feminist beliefs. Eve n in instance s where more “ fe minist” views were definite ly re jected, according to the re sponde nts ’ nume rical response , their comme nts sugge ste d that their views were not comple tely anti-fe minist. For example , many re sponde nts re jected the notion that more daycare should be made available to “ free women from the constant caring of the ir childre n ” , disliking the implication that childre n are a burde n, but did not ne cessarily reje ct widespread availability of daycare . In short, the exact nuance s of re sponde nts ’ answers are comple x, and more work must be done exploring how responde nts unde rstand and respond to spe cific items. The main lesson of this pape r, howe ver, is that social desirability may have an effect on those re sponse s, and this e ffect is not always a simple one . If re searche rs inte reste d in gende r-related issues want the be st possible insight into the ir results, the y would do well to conside r the inte ractive effects of gende r and social desirability on their particular study.
REFERENCES Albright, D. G., & Chang, A. F. (1976) . An examination of how one ’s attitudes toward women are re flected in one ’s de fensiveness and self-estee m. Sex Roles, 2, 195-198. Anderse n, M. L. (1993) . Thinking about women: Sociological perspectives on sex and gender. New York: Macmillan. Bee re , C. A., King, D. W., Beere, D. B., & King, L. A. (1984) . The Sex-Role E galitarianism Scale: A measure of attitudes towards equality between the se xes. Sex Roles, 10, 563-576. Brownmiller, S. (1975) . Against our will: Men, wom en and rape. Ne w York: Simon & Shuster. Buhrke , R. A. (1988) . Factor dimensions accross different me asure s of se x role ideology. Sex Roles, 18, 309-321. Burt, M. R. (1980). Cultural myths and support for rape . Journal of Person ality and Social Psychology, 38, 217-230. Cota, A. A., & Xinaris, S. (1993) . Factor structure of the Sex-Role Ideology Scale . Sex Roles, 29, 345-358. Crowne, D. P., & Marlowe, D. (1960) . A new scale of social desirability independent of psychopathology. Journal of Consulting Psychology, 24, 349-354. Edwards, A. L. (1957). The social desirability variable in personality assessm ent. Westport, CT: Greenwood Pre ss. Davies, R. (1983) . Young conservatives. London: Davray Music, Ltd. Dobash, R. E., & Dobash, R. (1979). Violence against wives: A case against the patriarchy. Ne w York: The Free Press. Evans, R. G. (1979) . The relationship of the Marlowe-Crowne Scale and its compone nts to defensive prefe re nces. Journal of Personality Assessm ent, 43, 406-410.
112
Th eriau lt an d Holm berg
Fischer, D., & Fick, C. (1993) . Measuring social desirability: Short forms of the MarloweCrowne Social Desirability Scale. Educationa l and Psychological Measurem ent, 53, 417-424. Greenwald, H. J., & O ’Connell, S. M. (1970) . Comparison of dichotomous and Likert formats. Psychological Reports, 27, 481-482. Greenwald, H. J., & Satow, Y. (1970). A short social desirability scale . Psychological Reports, 27, 131-135. Holcomb, D. R., Holcomb, L. C., Sondag, K. A., & Williams, N. (1991) . Attitudes about date rape: Ge nder differences among college students. College Student Journal, 25, 434-439. Holcomb, D. R., Sarvela, P. D., Sondag, K. A., & Hatton-Holcomb, L. C. (1993) . An evaluation of a mixed-gende r date rape intervention workshop. Journal of Am erican College Health, 41, 159-164. Jean, P. J., & Re ynolds, C. R. (1984) . Sex and attitude distortion: Ability of fe males and male s to fake liberal and conservative positions regarding changing se x roles. Sex Roles, 10, 805-815. Jones, E. E ., & Sigall, H. (1971) . The bogus pipeline: A new paradigm for measuring affect and attitude. Psychological Bulletin, 76, 349-364. Kalin, R., & Tilby, P. J. (1978) . Deve lopment and validation of a sex-role ideology scale. Psychological Reports, 42, 731-738. Kashy, D. A., & DePaulo, B. M., (1996) . Who lies? Journal of Personality and Social Psychology, 70, 1037-1051. Miller, S. L., & Simpson, S. S. (1991) . Courship violence and social control: Does gende r matter? Law and Society Review, 25, 335-365. Paulhus, D. L. (1984). Two-component models of socially desirable responding. Journal of Personality and Social Psychology, 46, 598-609. Spence, J. T., He lmreich, R., & Stapp, J. (1973). A short ve rsion of the attitudes toward women scale (AWS). Bulletin of the Psychonom ic Society, 2, 219-220. Spohn, R. B. (1993) . Social desirability correlates for acce ptance of rape myth. Psychological Reports, 73, 1218. Ward, S. K., Dziuba-Leathe rman, J., Gerard Stapleton, J., & Yodanis, C. L. (1994) . Acquaintance and date rape: An annotated bibliography. Westport, CT: Greenwood Pre ss.