Masculinity, Femininity, and Androgyny Viewed and Assessed ... · + BjMS + B6FS + A. (1) It...

12
Journal of Personality and Social Psychology 1983, Vol. 44, No. 2, 428-439 Copyright 1983 by the American Psychological Association, Inc. 0022-3514/83/4402-0428$00.75 Masculinity, Femininity, and Androgyny Viewed and Assessed as Distinct Concepts David Lubinski, Auke Tellegen, and James N. Butcher University of Minnesota The objective of this study (N = 172) was to evaluate (a) the equivalence of the scales of the short Bern Sex-Role Inventory (BSRI) and the Extended Personal Attributes Questionnaire (EPAQ); (b) the construct validity of the short BSRI and EPAQ as measures of well-being; (c) the validity of the concept of androgyny as an intrinsically interactive (rather than simply additive) concept; and (d) the utility and meaning of two special EPAQ measures: unmitigated agency (M~) and unmitigated communion (F c ~). The short BSRI and EPAQ were found empir- ically interchangeable when placed in a multitrait-multimethod matrix and two extrinsic convergent validation rectangles. Results of a hierarchical multiple regression analysis with interaction terms obtained with the Differential Person- ality Questionnaire (DPQ) provided only partial support for masculinity and femininity as measures of psychological well-being and no support for the sig- nificance of androgyny treated as an interaction of masculinity and femininity. Unmitigated agency and communion did not show the expected negative cor- relations with the mutual mitigation (interaction) of masculinity and femininity. Implications of this study for future investigations of androgyny and other in- tegrative concepts denned by interactions are discussed. As the concept of androgyny has evolved (e.g., Bern, 1974, 1977, 1979b; Lubinski, Tellegen, & Butcher, 1981; Spence & Helm- reich, 1979a; Spence, Helmreich, & Strapp, 1975; Strahan, 1975), so has its assessment. For example, the Bern Sex-Role Inventory (BSRI; Bern, 1974) has been refined to the short BSRI (Bern, 1979b), and the Personal Attributes Questionnaire (PAQ; Spence et al., 1975) has been expanded to the Extended PAQ (EPAQ; Spence, Helmreich, & Holahan, 1979). The short BSRI and the EPAQ cur- rently represent the two most refined inven- tories that purport to measure androgyny and are certainly the most widely used. Both in- ventories classify individuals into four groups—masculine, feminine, androgynous, and undifferentiated—using a median-split method. In the early to middle 1970s, androgyny was claimed by some to be an index of ad- justment and psychological health (e.g., Bern, We wish to thank Barbara Lubinski and Harold Brull for their help in the data collection process. Requests for reprints should be sent to Auke Tellegen, Department of Psychology, University of Minnesota, Elliott Hall, 75 East River Road, Minneapolis, Minne- sota 55455. 1974; Kaplan & Bean, 1976; Rawlings & Carter, 1977). Bern (1974), one of the ad- vocates of this position, proposed that the BSRI be used to classify individuals with re- spect to psychological health and that an- drogynous individuals are exemplars of psy- chological health. Kaplan and Bean (1976) suggested that psychotherapeutic techniques be developed to change both masculine and feminine sex types to a more androgynous orientation, and a rash of literature sprang from this new psychological "typology," re- ferring to androgyny as an "ideal psycholog- ical state" (e.g., Gilbert, 1981; Kaplan, 1976; Osofsky & Osofsky, 1972; Rebecca, Hefner, & Oleshansky, 1976; and Sturdevant, 1980, to mention a few). Given this enthusiasm, it is curious that (with respect to the short BSRI and the EPAQ) no empirical evidence has been provided to justify adding androgyny as a variable to the masculinity-femininity do- main, in other words, evidence substantiating the construct validity of androgyny with re- spect to psychological well-being. In the most recent theoretical discussion of androgyny (Bern, 198la, 1981b; Spence & Helmreich, 1981), no mention was made regarding how the concept of androgyny re- 428

Transcript of Masculinity, Femininity, and Androgyny Viewed and Assessed ... · + BjMS + B6FS + A. (1) It...

Page 1: Masculinity, Femininity, and Androgyny Viewed and Assessed ... · + BjMS + B6FS + A. (1) It contains the terms that will enable us to test concurrently (a) the construct validity

Journal of Personality and Social Psychology1983, Vol. 44, No. 2, 428-439

Copyright 1983 by the American Psychological Association, Inc.0022-3514/83/4402-0428$00.75

Masculinity, Femininity, and AndrogynyViewed and Assessed as Distinct Concepts

David Lubinski, Auke Tellegen, and James N. ButcherUniversity of Minnesota

The objective of this study (N = 172) was to evaluate (a) the equivalence of thescales of the short Bern Sex-Role Inventory (BSRI) and the Extended PersonalAttributes Questionnaire (EPAQ); (b) the construct validity of the short BSRIand EPAQ as measures of well-being; (c) the validity of the concept of androgynyas an intrinsically interactive (rather than simply additive) concept; and (d) theutility and meaning of two special EPAQ measures: unmitigated agency (M~) andunmitigated communion (Fc~). The short BSRI and EPAQ were found empir-ically interchangeable when placed in a multitrait-multimethod matrix and twoextrinsic convergent validation rectangles. Results of a hierarchical multipleregression analysis with interaction terms obtained with the Differential Person-ality Questionnaire (DPQ) provided only partial support for masculinity andfemininity as measures of psychological well-being and no support for the sig-nificance of androgyny treated as an interaction of masculinity and femininity.Unmitigated agency and communion did not show the expected negative cor-relations with the mutual mitigation (interaction) of masculinity and femininity.Implications of this study for future investigations of androgyny and other in-tegrative concepts denned by interactions are discussed.

As the concept of androgyny has evolved(e.g., Bern, 1974, 1977, 1979b; Lubinski,Tellegen, & Butcher, 1981; Spence & Helm-reich, 1979a; Spence, Helmreich, & Strapp,1975; Strahan, 1975), so has its assessment.For example, the Bern Sex-Role Inventory(BSRI; Bern, 1974) has been refined to theshort BSRI (Bern, 1979b), and the PersonalAttributes Questionnaire (PAQ; Spence et al.,1975) has been expanded to the ExtendedPAQ (EPAQ; Spence, Helmreich, & Holahan,1979). The short BSRI and the EPAQ cur-rently represent the two most refined inven-tories that purport to measure androgyny andare certainly the most widely used. Both in-ventories classify individuals into fourgroups—masculine, feminine, androgynous,and undifferentiated—using a median-splitmethod.

In the early to middle 1970s, androgynywas claimed by some to be an index of ad-justment and psychological health (e.g., Bern,

We wish to thank Barbara Lubinski and Harold Brullfor their help in the data collection process.

Requests for reprints should be sent to Auke Tellegen,Department of Psychology, University of Minnesota,Elliott Hall, 75 East River Road, Minneapolis, Minne-sota 55455.

1974; Kaplan & Bean, 1976; Rawlings &Carter, 1977). Bern (1974), one of the ad-vocates of this position, proposed that theBSRI be used to classify individuals with re-spect to psychological health and that an-drogynous individuals are exemplars of psy-chological health. Kaplan and Bean (1976)suggested that psychotherapeutic techniquesbe developed to change both masculine andfeminine sex types to a more androgynousorientation, and a rash of literature sprangfrom this new psychological "typology," re-ferring to androgyny as an "ideal psycholog-ical state" (e.g., Gilbert, 1981; Kaplan, 1976;Osofsky & Osofsky, 1972; Rebecca, Hefner,& Oleshansky, 1976; and Sturdevant, 1980,to mention a few). Given this enthusiasm, itis curious that (with respect to the short BSRIand the EPAQ) no empirical evidence hasbeen provided to justify adding androgyny asa variable to the masculinity-femininity do-main, in other words, evidence substantiatingthe construct validity of androgyny with re-spect to psychological well-being.

In the most recent theoretical discussionof androgyny (Bern, 198la, 1981b; Spence& Helmreich, 1981), no mention was maderegarding how the concept of androgyny re-

428

Page 2: Masculinity, Femininity, and Androgyny Viewed and Assessed ... · + BjMS + B6FS + A. (1) It contains the terms that will enable us to test concurrently (a) the construct validity

MASCULINITY, FEMININITY, AND ANDROGYNY 429

lates to psychological health. Yet, Bern(1981b) writes, "Androgyny was a conceptwhose time had come, a concept that ap-peared to provide a liberated and more hu-mane alternative to the traditional, sex-biasedstandards of mental health" (p. 362). In viewof these considerations, the main focus of thisarticle is an attempt to tie down the conceptof androgyny empirically and to determinethe relation between this popular notion (asmeasured by the short BSRI and the EPAQ)and several indexes of psychological well-being.

It should be noted that Spence and Helm-reich have never claimed that their positivelyvalued EPAQ measures (M+ and F*") shouldbe used to construct a new model of mentalhealth; in fact, they have warned investigatorsagainst such extrapolations (Spence & Helm-reich, 1979b). Spence and Helmreich (1979b)view their M+ and F+ scales as facets of themore global domain of masculinity and fem-ininity that Bern (1979b) speaks of, namely,instrumental (M+) and expressive (F+) per-sonality traits, respectively. With this inmind, Spence and Helmreich (1979b) havegone on to say that "the relation betweenandrogyny in this personality sense and othertypes of androgyny . . . [e.g., Bern's, 1974,1979a] must be determined empirically" (p.1035). Therefore, most of our discussion ofandrogyny as a "new model of mental health"is in response to Bern's (1974, 1979a) theo-retical considerations.

Spence and Helmreich have, however, in-vestigated the relation between their M+ andF1" measures and indexes of psychologicalhealth in normal and clinical populations(e.g., Holahan & Spence, 1980; Spence et al.,1979). Typically, these studies (in concor-dance with the findings of Antill & Cun-ningham, 1979) reflect a substantial degreeof construct validity for masculinity as anindicator of psychological well-being but littleor no support for femininity as such an in-dicator; moreover, in none of these studieswere attempts made to assess the possibleemergence of androgyny as an important dis-tinctive dimension.

Therefore, because the overall tenor of thesex role literature suggests that androgyny isan ideal psychological state, we have chosento inquire into the construct validity of both

the short BSRI and the EPAQ with respectto subjective indicators of psychological well-being, even though the theoretical consider-ations of the authors of the EPAQ are moreconservative than Bern's. Furthermore, as weshall see, inclusion of the EPAQ will allow usto examine some of Spenee and Helmreich'sown distinctive ideas concerning androgyny.1

The specific purpose of the present studywas to gain a better understanding of theshort BSRI and the EPAQ by comparing andcontrasting these two instruments to one an-other and to a more comprehensive inven-tory of self-view measures—the DifferentialPersonality Questionnaire (DPQ; Tellegen,Note 1). Particular attention is devoted toevaluating the construct validity of masculin-ity (M), femininity (F), and androgyny (aswell as traditional notions of sex role iden-tification) in relation to measures of subjec-tive psychological well-being. In order to as-sess the utility of androgyny as a conceptdistinct from F and M, it will be operation-alized and evaluated as an intrinsically in-teractive variable (rather than as a simplyadditive composite of F and M).

Androgyny was first cast as an interactiveconcept by Lubinski et al. (1981) using theshort BSRI. Our rationale was twofold. First,Bern has suggested that even though M andF are both related to adaptiveness and psy-chological health, extreme degrees of M andF in isolation become negative and even de-structive, whereas in combination, M and Ftemper each other, one neutralizing the morenegative manifestations of the other. We ar-gued that this tempering implies an interac-tion and can be assessed as such in an analysisof variance (ANOVA) or, as we prefer, a regres-sion model. Second, and more important, ifandrogyny is to achieve the status of an em-pirically useful concept (regardless of the do-main of behaviors to which it is hypothesizedto be related), it must have unique predictiveproperties relative to masculinity and femi-ninity. Lacking such "surplus" properties,

1 Although the EPAQ contains six masculinity-femi-ninity measures, when mentioning the EPAQ masculin-ity and femininity measures, we will always be referring,unless we specify otherwise, to their positively valued(M+ and F+) measures because these two scales are em-ployed to construct the fourfold typology, which is ofprimary interest to us.

Page 3: Masculinity, Femininity, and Androgyny Viewed and Assessed ... · + BjMS + B6FS + A. (1) It contains the terms that will enable us to test concurrently (a) the construct validity

430 D. LUBINSKI, A. TELLEGEN, AND J. BUTCHER

androgyny would be predictively and con-ceptually redundant, and it would be enoughto interpret findings using the concepts of Mand F.2

Our earlier evaluation of androgyny as aninteractive concept produced disappointingresults (Lubinski et al., 1981). None of thetested M X F interactions contributed signif-icantly to the prediction of relevant psycho-logical criteria. In the present sl^udy, however,we investigate further the utility of two well-known contemporary formulations of an-drogyny.

By employing the entire DPQ with its par-ticular emphasis on markers of positive andnegative affectivity, we hoped to determinethe extent to which androgyny (as measuredby the short BSRI and the EPAQ) may beviewed as an indicator of psychological health.This would require that the appropriate M XF interaction share variance with these in-dicators. A second extension of our earlierstudy is to evaluate Spence et al.'s (1979) newmeasures: "unmitigated agency" (M~) and"unmitigated communion" (Fc~).

Our basic approach to data analysis wasto conduct hierarchical multiple regressionanalyses with interaction terms, using theshort BSRI and the EPAQ separately as pre-dictors of seven measures of subjective well-being. Our regression equation is as follows:

Y = B,M + B2F + B3S + B4MF

+ BjMS + B6FS + A. (1)

It contains the terms that will enable us totest concurrently (a) the construct validity ofmasculinity (M) and femininity (F) as linearpredictors of psychological well-being (Y),namely, BiM and B2F; (b) the validity of theconcept of androgyny conceived as an in-trinsically interactive (rather than simply ad-ditive) concept, namely, B4MF; and (c) thevalidity of the traditional assumption that itis the masculine man and feminine womanwho typify psychological well-being, namely,B5MS and B6FS, respectively, where S refersto gender.3 (For further clarification regard-ing the specific theoretical significance ofeach of these interactions, see Lubinski et al.,1981.)

Although our rationale for operationaliz-ing Bern's androgyny as an interactive con-

cept seems clear (cf. Lubinski et al., 1981,pp. 723-724), readers may wonder why weare placing this requirement on the EPAQ,because Spence and Helmreich (e.g., 1978,p. 109) appear to reject the surplus meaningthat has accrued to androgyny. In fact, Spenceand Helmreich (1979b) state explicitly,

We introduced the term androgyny simply as a conve-nient label to identify individuals who score relativelyhigh on both the M and the F scales of our particularinstrument.. . . Androgyny, as we have used the term,has no particular theoretical import, being intended toindicate nothing more than a relatively high degree ofboth instrumental and expressive personality traits, asdenned by the PAQ. (p. 1035)

It would seem, therefore, that Spence andHelmreich view androgyny as simply an ad-ditive composite of M and F. However, insome of their discussions at least (Spence etal., 1979, pp. 1674-1675), they appear toadvance an androgyny concept that is notsimply additive but interactive, that is, dis-tinct from M and F.

2 Interestingly, our operational definition of androg-yny is similar to Kaplan and Bean's (1976) conceptualdefinition. They contend that "androgyny . . . is not asimple or even complex union of the totalities of tra-ditionally defined masculinity and femininity. It's a thirddimension" (p. 383). By operationalizing androgyny asa partial product, and entering the term representing it(MF) as an independent variable, androgyny becomes athird dimension distinct from masculinity and feminin-ity. This, among other things, enables us to assess theconstruct validity of androgyny empirically, both con-vergently and discriminantly (e.g., Campbell & Fiske,1959; Cronbach & Meehl, 1955). This approach alsoresolves the problem noted by Spence and Helmreich(1981), namely, that of anchoring a variable empiricallyto a pair of relatively independent measures.

3 Our interpretation of these interactions differs dras-tically from that of Flaherty and Dusek (1980, p. 991).These authors interpret their significant F X S interactionin conjunction with a significant M main effect as in-dicating that androgyny exists in females but not males.We disagree. Flaherty and Dusek's F X S interaction il-lustrates what we have called the traditional notion thatfemininity is adaptive in females but not in males. Thistype of interaction, however, should be distinguishedfrom those that corroborate the idea of androgyny. Toargue that androgyny operates in one gender but not theother amounts to hypothesizing that androgyny itself ismoderated by gender in relation to relevant criteria. Forits empirical evaluation, this hypothesis would call foradding the product term B7MFS as an independent vari-able in the appropriate hierarchical analysis (Cohen,1968). This three-way interacton was computed for alldependent measures employed in this study for both theEPAQ and the short BSRI and in each case was foundnot to be significant.

Page 4: Masculinity, Femininity, and Androgyny Viewed and Assessed ... · + BjMS + B6FS + A. (1) It contains the terms that will enable us to test concurrently (a) the construct validity

MASCULINITY, FEMININITY, AND ANDROGYNY 431

Furthermore, Spence and Helrnreich haverecently constructed measures of unmiti-gated agency and communion that appear toreflect an interactive concept (M~ and Fc~,respectively;,see Spence et al., 1979). Themotivation for the construction of thesescales was provided by Bakan (1966). Ac-cording to Spence et al. (1979, pp. 1674-1675),

He [Bakan, 1966] proposes that a strong sense of agency,unmitigated by a sense of communion, is destructive tothe individual and to society. Similarly, communionmust be mitigated by agency if the individual is to func-tion effectively. The developmental task of males is thusto learn to balance masculine agency with some degreeof communion, and of females, to balance femininecommunion with some degree of agency. . . . In ex-panding our instrument to include the negatively tonedmasculine and feminine attributes, it was our intent todevelop scales that conceptually parallel the [positivelyvalued] M and F scales, containing negatively valuedmasculine characteristics that reflect Bakan's "unmiti-gated agency" and negatively valued feminine charac-teristics that reflect "unmitigated communion."4

We believe that the idea of an M X F in-teraction is implied by these considerations.Were we not to regard mitigation as an in-teractive concept, then the extent to whichM contributes to predicting relevant psycho-logical criteria would not change as a func-tion of F, and vice versa. But in order to saythat the tendencies associated with highstandings on M and F mitigate one another—supposedly allowing full expression of posi-tive masculine and feminine attributes butinhibiting the more negative expressions thatare present when one has high standing ononly M or F—we would have to posit an in-teraction. Without it, the very ideas of un-mitigated agency and communion, alongwith their measures, Fc~ and M~, would bemeaningless. In fact, to be valid, Fc~ and M~should behave as measures of mitigation inreverse; that is, they should show a clear andnegative correlation with the interactive in-tegration of F and M, assessed directly withthe multiple regression procedures describedin Lubinski et al. (1981).

Specifically, the validity of Fc~ and M~may be assessed by computing the partialcorrelations between M~ and Fc~ with theappropriate M X F interaction, after con-trolling for M, F, and sex of subject.5 If M~(reversed) and Fc~ (reversed) "tap" mitiga-

tion, then we would expect to find ah appre-ciable negative partial correlation betweenthese measures and the M X F interactionvariable.

MethodSubjects

A total of 172 college students (87 men and 85 women)were recruited from a number of introductory psychol-ogy classes at the University of Minnesota: Only studentsunder 30 years of age were included in the sample.

ProcedureIn groups of about 85, students were administered a

test battery consisting of three personality inventories:the short Bern Sex-Role Inventory, the Extended PersonalAttributes Questionnaire, and the Differential Person-ality Questionnaire (Tellegen, Note 1). A brief descrip-tion of the three instruments follows:

The short BSRI. The shprt Bern Sex-Role Inventory(Bern, 1978, 1979b) consists of two 10-item scales-Masculinity and Femininity—and 10 filler items. TheBSRI classifies individuals, Using a median-split method,as androgynous (high M, high F), masculine (high M,low F), feminine (low M, high F), and undifferentiated(low M, low F).

TheEPAQ. The Extended Personal Attributes Ques-tionnaire consists of six (three positive and three nega-tive) masculinity-femininity scales. The four scales an-alyzed in this study may be described as follows: Mas-culinity (M+) is designed to measure a facet of positivelyvalued masculinity, namely, instrumentality; femininity(F+) is designed to measure a facet of positively valuedfemininity, namely, expressiveness (the EPAQ F+ and M+

scales are the old PAQ F and M scales); negative mas-culinity (M~) is intended to reflect unmitigated agency;and negative femininity (Fc~) is intended to reflect un-mitigated communion. The latter two are measures ofnegatively valued characteristics specifically designed tobe counterparts to the former two (see Spence et al.,1979, particularly pp. 1674-1675). Like Bern, Spenceet al. (1979) use the median-split method, on their pos-itively valued measures (M+ and F1"), to classify individ-uals as masculine, feminine, androgynous, or undiffer-entiated.

As described in some detail in Lubinski et al. (1981),subjective well-being involves two broad and relatively

4 The concept of unmitigated communion is an ex-tension of Bakan's (1966) ideas introduced by Spence,Helmreich, and Holahan (1979). Bakan himself speaksonly of unmitigated agency not unmitigated commu-nion,

5 For both of these analyses, we use a partial corre-lation because the critical M X F interaction can onlybe captured after partialing out its constituents. Also,gender was added as a third main effect to circumventthe possibility of the M X F interaction's absorbing someof the variance due to sex differences, which would com-promise its interpretability.

Page 5: Masculinity, Femininity, and Androgyny Viewed and Assessed ... · + BjMS + B6FS + A. (1) It contains the terms that will enable us to test concurrently (a) the construct validity

432 D. LUBINSKI, A. TELLEGEN, AND J. BUTCHER

independent dimensions, namely, positive and negativeaffectivity (see Costa & McCrae, 1980, for a similarview). In this study, factor scores representing these twohigher order dimensions and derived from the Differ-ential Personality Questionnaire (DPQ) are our primarycriteria.

The DPQ. The 11 primary scales of the DPQ weredeveloped factor analytically in a series of studies aimedat converging, through a process of successive approxi-mations, on major self-view dimensions (Lubinski et al.,1981; Tfellegen, 1981, Note 1). Jointly, the 11 scales markthree higher order factors that can be interpreted asbroad self-view dimensions, namely, Positive Affectivity,Negative Affectivity, and Constraint. Of the primaryscales, Well-Being, Social Potency, and Achievement areprimarily associated with Positive Affectivity; Stress Re-action, Alienation, and Aggression, with Negative Af-fectivity; and Control, Authoritarianism, Harm Avoid-ance, and Aggression (negatively), with Constraint.These higher order dimensions can also be discerned toother broad-range personality inventories, but from aninterpretive point of view, they emerge especially clearlyfrom the primary DPQ scales. Factor scores representingall three higher order dimensions were computed.

Students were given as much time as they needed tocomplete the questionnaires and spent, on the average,approximately 55 minutes on the task.

ResultsThe results are reported in three parts.

First, in order to determine the extent towhich corresponding scales of the short BSRIand the EPAQ may be considered concep-tually equivalent, a multitrait-multimethodmatrix is presented and discussed. Second,the correlational profiles between these scalesare also inspected to assess, in the context ofall 11 DPQ measures, the amount of extrinsicconvergent validity shared by these two in-ventories. These profiles further define mas-culinity and femininity through construct ex-plication. Third, multiple regression analysesare reported, using first the short BSRI andthen the EPAQ to predict seven measures ofsubjective psychological health. The depen-dent measures chosen for this analysis aretwo factor scores—Positive Affectivity andNegative Affectivity—and the major first-or-der markers of these higher order dimensions,namely, Well-Being, Social Potency, andAchievement for Positive Affectivity andStress Reaction and Alienation for NegativeAffectivity. Finally, the view of the EPAQscales (M~ and Fc~) as measures of unmiti-gated agency and communion, respectively,is tested using partial correlations.

The top of Table 1 presents a multitrait-multimethod matrix (Campbell & Fiske,

1959), organized around traits rather thanmethods. Four coefficient alpha reliabilityestimates (Cronbach, 1951) were computed,two in each of the heteromethod-monotraittriangles. Although all of the reliability esti-mates were encouragingly high, the shortBSRI scales appeared more reliable than theEPAQ measures (BSRI-M = .85, BSRI-F =.89, EPAQ-M+ = .77, EPAQ-F+; = .77). Thetwo validity coefficient estimates (feminin-ity = .75, masculinity = .72) were also im-pressive (in fact, when corrected for atten-uation, these two coefficients rose to .89 and.90 for masculinity and femininity, respec-tively). In addition, discriminant validity isalso displayed in that the heterotrait-mono-method entries (.29 and .16; M = .23) weresignificantly smaller than the convergent va-lidity coefficients.

The lower part of Table 1 contains twoextrinsic convergent validation rectangles, anextension of the multitrait-multimethodmatrix proposed by Fiske (1971). Accordingto Fiske (1971, p. 245), "as a rule, measuresof the same construct are not interchangeableempirically and must not be considered con-ceptually equivalent until a high degree ofconvergence in correlational patterns withother variables, or extrinsic convergent vali-dation, has been demonstrated empirically."The two extrinsic convergent validation rect-angles compare the correlational profiles ofthe four masculinity-femininity measureswith all 11 scales of the DPQ and its threefactor scores. Inspection of these immedi-ately reveals the substantial degree of overlapof these two inventories. One may, in fact,conclude that their corresponding measuresare conceptually near equivalent.

The two rectangles (in Table 1) also con-tribute to the construct explication of mas-culinity and femininity by allowing us to de-fine these M-F measures in the context of afairly comprehensive set of major self-viewdimensions. (Because the amount of extrinsicconvergent validation for these two inven-tories was high, the corresponding M-Fscales can be given the same interpretations.)

Measures of masculinity seem to be pri-marily associated with markers of positiveaffectivity and may be defined, specifically,as indicators of subjective well-being, socialpotency, achievement, and stress reaction

Page 6: Masculinity, Femininity, and Androgyny Viewed and Assessed ... · + BjMS + B6FS + A. (1) It contains the terms that will enable us to test concurrently (a) the construct validity

MASCULINITY, FEMININITY, AND ANDROGYNY 433

(reversed). Thus, high masculinity scores re-flect a view of oneself as interpersonally ef-fective and dominant, whereas absence ofsuch self-attributions is associated with lowscores on masculinity. Indeed, an inspectionof the content of the masculinity items in-dicates that the domain sampled could quiteadequately be labeled dominance-poise (e.g.,short BSRI-M: "dominant," "forceful,""strong personality"; EPAQ-M+: "feels su-perior," "self-confident," "stands up well un-der pressure," "competitive"). Therefore,some association between these masculinitymeasures, and markers of positive affectivity,is expected. The femininity measures, on theother hand, tap a different region of the self-view domain and may be defined as markersof well-being, social closeness, and aggression(reversed). The association of femininity withthese three measures indicates that, unlike

masculinity, femininity is primarily relatedto "nurturance and accommodating warmth"(Lubinski et al., 1981). The item content ofthe two femininity scales also seems to war-rant this conclusion (e.g., short BSRI-F:"warm," "understanding," "eager to soothhurt feelings," "sympathetic"; EPAQ-F+:"understanding of others," "warm in relationwith others," "helpful to others," "able todevote self completely to others").

It appears, then, that the attempt at con-struct explication is illuminating. In the con-text of 11 self-view dimensions, the measuresin question—masculinity and femininity—seem to share most of their variance with twoquite distinct subsets.

Turning now to the multiple regressionanalyses, recall that these were performed inorder to evaluate interactive relations be-tween the F and M scales in relation to spe-

Table 1A Multitrait-Multimethod Matrix Between Corresponding M-F Measures of the Short BSRIand the EPAQ Followed by Two Extrinsic Convergent Validation Rectangles

Measure

Masculinity

BSRI-M EPAQ-M

Femininity

BSRI-F EPAQ-F

BSRI-MEPAQ-M

BSRI-FEPAQ-F

(.85).72

.29

.14

(.TIT.25.16

(.89).75 (.77T

Ul \^

Well-BeingSocial PotencyAchievementSocial ClosenessStress ReactionAlienationAggressionControlHarm AvoidanceTraditionalismAbsorption

Positive AffectivityNegative AffectivityConstraint

.33

.63

.36

.21-.42-.14

.03

.03-.14-.06-.02

.53-.21-.10

.46

.55

.40

.20-.53-.11

.02

.04-.14

.11-.03

.59-.24-.01

.33

.10

.15

.42-.17-.26-.39

.22

.19

.29

.26

.36-.24

.30

.30

.13

.12

.45-.07-.28-.40

.15

.20

.27

.31

.35-.20

.26

Note. Each heteromethod-monotrait triangle is enclosed by a solid line. Both triangles contain an italicized validitycoefficient and two alpha coefficients (in parentheses). Below the multitrait-multimethod matrix are two extrinsicconvergent validity rectangles. For all correlations above, rm = .13, r.01 = .18, r.m = .23. BSRI = the short BernSex-Role Inventory. EPAQ = Extended Personality Attributes Questionnaire. DPQ = Differential Personality Ques-tionnaire.

Page 7: Masculinity, Femininity, and Androgyny Viewed and Assessed ... · + BjMS + B6FS + A. (1) It contains the terms that will enable us to test concurrently (a) the construct validity

434 D. LUBINSKI, A. TELLEGEN, AND J. BUTCHER

cific dependent variables. The results, re-ported in Table 2, were obtained by first en-tering the three main effects (masculinity,femininity, and gender) in an incrementalstepwise fashion and then entering the threeinteraction terms (M X F, M X S, F X S) us-ing the same procedure. Results with theshort BSRI are reported first.6

Positive Affectivity

Masculinity accounted for 28% of the vari-ance, and when femininity was added, theamount of variance accounted for increasedto 33% (R = .57). None of the remainingvariables increased the R2 significantly.

Negative Affectivity'Femininity accounted for 6% of the vari-

ance, and when masculinity was added, theamount of variance accounted for increasedto 8%. The only other significant entry is the(M X F) interaction, increasing the varianceaccounted for to 10% (R = .32), but the di-rection of the partial correlation was oppositeto what was expected (rY4.i23 = -16).

The same multiple regression analysis wasconducted using the EPAQ masculinity andfemininity measures.

Positive AffectivityMasculinity was entered first, accounting

for 35% of the variance. When femininity wasadded, the amount of variance accounted forincreased to 41% (R = .64).

Negative AffectivityMasculinity was entered first, accounting

for 6% of the variance. When femininity wasadded, the amount of variance accounted forincreased to 9%. One other variable, the(M X F) interaction term, accounted for asignificant portion of the variance. However,the interaction again was in the wrong direc-tion (rY4.i23 = 18).

To scrutinize these data further, the sametype of regression analysis was performed forboth the short BSRI and the EPAQ, using themajor markers of Positive and Negative Af-fectivity separately as criterion variables (withWell-Being, Social Potency, and Achievementprimarily defining Positive Affectivity andStress Reaction and Alienation defining Neg-

ative Affectivity). As expected, the results ofthese analyses for the main effects correspondclosely to those reported in Table 1, and nodetailed discussion is presented. To be sure,in some of these analyses the (M X F) inter-action term shared a significant portion ofvariance with the criterion. However, as inthe case of Negative Affectivity, all of thesesignificant interactions were in the wrongdirection. (The only significant M X F inter-action effects involved the following scales:For the short BSRI, Well-Being, R2 change =.03, p < .05; Alienation, R2 change = .07,/K.001. For the EPAQ, Alienation, R2

change = .09, p< .001.)7

6 Before reporting the results of our regression anal-yses, the three hypotheses about to be tested, along withtheir empirical requirements for corroboration, shouldbe discussed further. For both inventories, measures ofM and F must, in addition to accounting for a significantamount of the criterion variance, be in the predicteddirection. This stipulation also pertains to the partialcorrelations as they are entered in the stepwise proce-dure. (This is an important requirement because evenif the simple correlations of M and F with a measure ofpsychological health are positive, one of the predictorsmay become negatively ̂ correlated with the criterionwhen the other predictor is controlled.) The same logicapplies to the M X F interaction term; it too must rep-resent a relation that is in the predicted direction. Thus,M, F, and M X F should have positive partial correlationswith Positive Affectivity and negative partial correlationswith Negative Affectivity. The final hypothesis to betested is the traditional notion that the masculine manand feminine woman are the ones who typify psycho-logical health. This hypothesis will receive empirical sup-port (gender is coded: females = 2, males = 1) if the F XS partial is positively correlated with Positive Affectivityand the M X S partial is negatively correlated with Pos-itive Affectivity, whereas the inverse of these directionswould be expected for Negative Affectivity. Here, too,the direction of these interactions may be evaluated bymeans of partial correlation. On the other hand, underthe traditional view, the M X F interaction should notbe significant. (The partial correlations discussed aboveare given in Table 2 under the column labeled Partial r.)

7 We recognize that the linear-by-linear interactiontested by us is only one of several possible patterningsof masculinity and femininity. However, this form of in-teraction appears to capture the current interaction con-ception of androgyny (i.e., M and F "temper" [Bern,1979a] or "mitigate" [Spence et al., 1979] each other).An alternative idea, one discussed in Bern's (1974) earlierwork, namely, the "balance model," would require adifferent analysis in which the discrepancy between Mand F^e.g., |M — F|) would be evaluated as an indepen-dent vanablerAtttoough apparently not currently advo-cated, this alternative model was also tested for each ofthe eight dependent measures discussed above using boththe short BSRI and the EPAQ with uniformly negativeresults.

Page 8: Masculinity, Femininity, and Androgyny Viewed and Assessed ... · + BjMS + B6FS + A. (1) It contains the terms that will enable us to test concurrently (a) the construct validity

MASCULINITY, FEMININITY, AND ANDROGYNY 435

In our. final analysis, we assessed the Va-lidity of M- and Fc~ as measures of unmit-igated agency and communion, respectively.Partial correlations showed that M- (or un-mitigated agency) is essentially independentof "mitigation," or the M X F interaction(i.e., rAM-;M,F,s = -05, where A stands for the"androgynous" M X F variable). Fc~ dis-played a low correlation with mitigation;however, this correlation indicated a positiverelation between maladaptive femininity andandrogyny (i.e., TAFC-.M.F.S = -26, p < .001).

Discussion

The purpose of this study was to collectand evaluate evidence that would (a) deter-mine the extent to which correspondingscales of the short BSRI and the EPAQ maybe considered empirically near equivalent;(b) clarify the nature of masculinity and fem-ininity and androgyny in the context of sev-eral general self-view dimensions, particu-larly in reference to subjective indicators ofpsychological well-being, possibly in inter-action with the individual's gender; and (c)

assess the utility and meaning of Spence andHelmreich's new measures, unmitigatedagency (M~) and unmitigated commu-nion (Fc~).

In our first analysis, a substantial degreeof overlap was found between correspondingscales of the short BSRI and the EPAQ. Theyshowed an impressive convergent and dis-criminant pattern when placed in a multi-trait-multimethod matrix. An even morestriking display of similarity was obtained byplacing pairs of corresponding scales in twoextrinsic convergent validation rectangles.Our findings suggest that researchers shouldbe able to u^e these inventories inter-changeably.

Construct explication shows that mascu-linity and femininity, as measured by theshort BSRI and the EPAQ, have an affinityto different groups of self-view dimensions.The correlational profile for femininity de-fines it as being primarily related to a clusterof variables that can be labeled as "nurtur-ance-warmth." Masculinity, on the otherhand, was defined by a cluster of variablesreflecting "dominance-poise."

Regressions of Masculinity, Femininity, Gender, and All Possible Pairwise Interactionson Positive Affectivity and Negative Affectivity

Positive affectivity Negative affectivity

Step

123456

Variable

MFS

M X FM X SF X S

Partial r

.53

.25-.02-.09-.04

.01

R

.53

.57

.57

.58

.58

.58

R2 change

BSRI

.28***

.05***

.00

.01

.00

.00

Variable

FMS

M X FM X SF X S

Partial r

-.24-.15-.02

.16-.06

.01

R

.24

.28

.29

.32

.32

.32

K2 change

.06**

.02*

.00

.02*

.00

.00

EPAQ

123456

MFS

M X SF X S

M X F

.59

.32-.05-.07

.03

.01

.59

.64

.64

.65

.65

.65

.35***

.07***

.00

.00

.00

.00

MFS

M X FF X S

M X S

-.24-.17-.03

.18-.07

.01

.24

.29

.30

.34

.35,35

.06***

.03**

.00

.03*

.00

.00

Note. These data were obtained by entering the three main effects in an incremental stepwise fashion and thenentering the three interaction terms using the same procedure. The column labeled Partial r represents the partialcorrelation for that variable and the criterion variable after controlling for all of the preceding variables in theequation. Thus, the first entry in every column is merely the simple correlation. N = 172.' M = masculinity; F =femininity; S = gender. BSRI = the short Bern Sex-Role Inventory. EPAQ = Extended Personality AttributesQuestionnaire.* p < .05. ** p< .01. *** p < .001.

Page 9: Masculinity, Femininity, and Androgyny Viewed and Assessed ... · + BjMS + B6FS + A. (1) It contains the terms that will enable us to test concurrently (a) the construct validity

436 D. LUBINSKI, A. TELLEGEN, AND J. BUTCHER

This result is consistent with our earlierfindings using the short BSRI and other stud-ies using the BSRI "old version" (e.g., Ber-nard, 1980; DeGregorio & Carver, 1980;Wiggins, & Holzmuller, 1978). Moreover,Antill and Cunningham (1979) and Spenceet al. (1979) showed that M+ shares muchvariance with markers of Positive Affectivity,whereas the relation between F+ and thesemarkers was quite weak. (For a brief sum-mary of the above studies, see Lubinski et al.,1981.)

The masculinity scales of both inventoriesin fact received a good deal of construct val-idation in view of their pattern of positiverelations to markers of Positive Affectivity(i.e., Well-Being, Social Potency, andAchievement) and negative relation to StressReaction. Femininity did not fare as well,showing a corroborative but less impressivepattern of relations with indicators of Weil-Being, Social Closeness, and Alienation. Inthe broad context of these personality di-mensions, a fairly consistent definition ofmasculinity and femininity emerges. Mas-culinity appears to be primarily character-ized by a dominant and "assimilative" style,and femininity by qualities of nurturance andwarmth, an "accomodative" style. (Consis-tent with this interpretation are the recentfindings of Harrington and Andersen, 1981,who report a substantial relation betweenmasculinity and several measures of creativ-ity but negative or negligible correlations be-tween femininity and these measures.) How-ever, in view of our regression analyses, itdoes not appear that in relation to indicatorsof subjective well-being the combination ofthese two distinct and relatively independentdimensions interact synergistically to pro-duce an important third dimension of an-drogyny.

Our failure to "triangulate" androgyny asa significant distinct concept was disappoint-ing. Each time androgyny appeared to emergewith predictive properties distinct from mas-culinity and femininity, its association withthe criterion was in the wrong direction.These findings are, of course, mostly relevantto Bern's ideas because Spence and Helm-reich do not share Bern's view that androgynyepitomizes psychological health. However,similarity of the results obtained with the

EPAQ increases the generalizability of ourfindings and indicates that claims for an-drogyny, however measured, as a constella-tion specifically indicative of psychologicalhealth so far have not received empirical sup-port.

Not only were the relations found in thisstudy in the wrong direction but the evidencewas also negative in our first study (Lubinskiet al., 1981). Recognizing that one disconfir-mation is generally more decisive than onecorroboration in evaluating the verisimili-tude of a theory (Meehl, 1979), androgynyseems to be in some difficulty, at least withrespect to subjective psychological well-being.

On the whole, and in the light of the find-ings reported above, the fourfold typologydescribed and advocated by Bern (but notSpence and Helmreich) for classifying indi-viduals with respect to psychological healthseems to have little utility. Furthermore, theresults clearly indicate that the suggestion todevelop psychotherapeutic techniques withinthis framework is unwarranted (e.g., Kaplan& Bean, 1976). For the same reasons, we areskeptical about the heuristic merits of a clas-sification of counselor personalities based onthe fourfold typology (e.g., Highlen & Rus-sell, 1980). Such a classification, at the veryleast, is premature because there is simply noevidence to warrant the expansion of mas-culinity and femininity into a functional ty-pology at this point. Moreover, in view of thenonsignificance of the M X S and F X S in-teraction terms, we conclude that with re-spect to subjective well-being, neither thecontemporary formulations of androgynynor the traditional assumptions regarding sexrole identification are accurate.

If traditional formulations of sex role iden-tification are inaccurate and androgyny as aninteractive variable does not share variancewith indicators of psychological well-being,can anything of a prescriptive nature be con-cluded from the contemporary sex role lit-erature? The most relevant and provocativefindings are the strong relation or overlap ofinstrumentality (or positively valued mas-culinity as assessed by the short BSRI-M andEPAQ-M+ scales) to indicators of psycholog-ical well-being and the fact that the strengthof this relation is the same in men andwomen.

Page 10: Masculinity, Femininity, and Androgyny Viewed and Assessed ... · + BjMS + B6FS + A. (1) It contains the terms that will enable us to test concurrently (a) the construct validity

MASCULINITY, FEMININITY, AND ANDROGYNY 437

This may suggest the straightforward con-clusion that "masculinity" in the sense ofdominance-poise should be encouraged andreinforced in women and men alike. More-over, the class of "feminine" behaviors, suchas nurturance and warmth (as assessed by theshort BSRI-F and the EPAQ-F+), may becultivated independently of dominance-poise.But the available evidence indicates that nur-turant qualities are not as strongly associatedwith a sense of well-being as dominant qual-ities are.

On the other hand, there is no evidencethat with respect to subjectve well-being,masculinity and femininity become dysfunc-tional when an individual has a high standingon one variable but not on the other. As as-sessed by the short BSRI and the EPAQ (i.e.,M+ and F+) scales, masculinity and feminin-ity do not temper (Bern, 1979a) or mitigate(Spence et al., 1979) one another to enhancean individual's overall subjective well-being.

The possibility remains, of course, that Mand F interact in relation to a more restrictedrange of variables than would be required bythe more encompassing concept of androg-yny advanced by Bern. For example, Spenceand Helmreich (1979a) report what appearsto be an M X F interaction in a study of at-titudes toward women. Harrington and An-dersen also report some significant M X Finteractions in the prediction of certain in-dexes of creativity. (However, this finding isoffset by the negative correlations of femi-ninity with these measures of creativity,which are substantial compared to the vari-ance accounted for by the M X F interac-tion.)8

Because no evidence was found indicatingthat the single M~ (reversed) and F<f (re-versed) scales measure the mitigation of pos-itively valued masculinity and femininity, wesuggest that researchers who wish to deter-mine the extent to which masculinity andfemininity mitigate one another, or assessandrogyny, continue the study of F X M in-teractions, using the multiple regression pro-cedure discussed in this study.

One attractive feature of this technique,not illustrated in the present article, is thatit enables investigators to determine whethermasculinity or femininity function as mod-erator variables even when not correlated

with the criterion under consideration. Forexample, although most studies indicate thatfemininity shares little variance with mea-sures of psychological health, femininity mayhave moderated relationships with masculin-ity to produce a significant M X F interaction(Saunders, 1956; Zedeck, 1971). Thus, al-though neither masculinity nor femininityseems to function as a moderator variablewith respect to subjective indicators of psy-chological well-being, the possibility remainsan empirical question for other relevant psy-chological criteria. Also, by adding the prod-uct term M X F X S as an independent vari-able, researchers may determine whether an-drogyny (i.e., M X F) operates in one genderbut not in the other.

Finally, we suggest that the multiple regres-sion techniques discussed in this article beviewed as a general data-analytic device forinvestigating the construct validity of thosehigher order psychological concepts that aredefined by nonadditive combinations of cer-tain first-order personal characteristics. (SeeHeilbrun, 1978, p. 106, and Diamond, Royce,& Voorhees, 1981, p. 188, for two contem-porary theoretical formulations that mightbenefit from this type of analysis.) The sur-plus meaning of these higher order conceptsmay be captured empirically in a regressionanalysis by entering their components andtheir relevant combinations in the proper se-quence. For example, we have suggested thatandrogyny is only one example of a broaderclass of "fulfillment" or "self-actualization"concepts (Lubinski et al., 1981). All such con-cepts appear to imply that a "fully function-ing" person has integrated seemingly contra-dictory or unrelated qualities as synergisticconstituents of an organized whole that con-sequently "is more than the sum of its parts."We again suggest that this model can also beused to assess other higher order concepts(fulfillment or otherwise) that imply nonad-ditive combinations of first-order compo-nents.

8 Stokes, Childs, and Fuehrer (1981) also recently re-ported a significant M X F interaction in the predictionof "disclosure to intimates." However, instead of enteringthe product term, M X F, in the appropriate hierarchicalfashion, they entered it first in their stepwise regression.Consequently, the results of their analysis are uninter-pretable (cf. Lubinski, 1983).

Page 11: Masculinity, Femininity, and Androgyny Viewed and Assessed ... · + BjMS + B6FS + A. (1) It contains the terms that will enable us to test concurrently (a) the construct validity

438 D. LUBINSKI, A. TELLEGEN, AND J. BUTCHER

Reference Notei

1. lellegen, A. Brief manual for the Differential Person-ality Questionnaire. Unpublished manuscript, Uni-versity of Minnesota, 1982.

References

Antill, J. K., & Cunningham, J. D. Self-esteem as a func-tion of masculinity in both sexes. Journal of Con-sulting and Clinical Psychology, 1979, 47, 783-785.

Bakan, D. The duality of human existence. Chicago:Rand McNally, 1966.

Bern, S. L. The measurement of psychological androg-yny. Journal of Consulting and Clinical Psychology,1974, 42, 155-162.

Bern, S. L. On the utility of alternative procedures forassessing psychological androgyny. Journal of Con-sulting and Clinical Psychology, 1971,45, 196-205.

Bern, S. L. The short Bern Sex-Role Inventory. Palo Alto,Calif.: Consulting Psychologists Press, 1978.

Bern, S. L. Beyond androgyny: Some presumptuous pre-scriptions for a liberated sexual identity. In J. Sherman& F. Denmark (Eds.), Psychology of women: Issues inpsychology. New York: Psychological Dimensions,1979. (a)

Bern, S. L. Theory and measurement of androgyny: Areply to the Pedhazur-Tetenbaum and Locksley-Col-ten critiques. Journal of Personality and Social Psy-chology, 1979, 37, 1047-1054. (b)

BemrS. L. The BSRI and gender schema theory: A replyto Spence and Helmreich. Psychological Review, 1981,88, 369-371. (a)

Bern, S. L. Gender schema theory: A cognitive accountof sex typing. Psychological Review, 1981, 88, 354-364. (b)

Bernard, L. C. Multivariate analysis of new sex role for-mulations and personality. Journal of Personality andSocial Psychology, 1980, 38, 323-336.

Campbell, D. X, & Fiske, D. W. Convergent and dis-criminant validation by the multitrait-multitnethodmatrix. Psychological Bulletin, 1959, 56, 81-105.

Cohen, J. Multiple regression as a general data-analyticsystem. Psychological Bulletin, 1968, 70, 426-443.

Costa, P. T., & McCrae, R. R. Influence of extraversionand neuroticism on subjective well-being: Happy andunhappy people. Journal of Personality and SocialPsychology, 1980, 38, 668-678.

Cronbach, L. F. Coefficient alpha and the internal struc-ture of tests. Psychometrika, 1951,16, 297-334.

Cronbach, L. F., & Meehl, P. E. Construct validity inpsychological tests. Psychological Bulletin, 1955, 52,281-302.

DeGregorio, E., & Carver, C. S. Type A behavior pattern,sex role orientation, and psychological adjustment.Journal of Personality and Social Psychology, 1980,39, 286-293.

Diamond, S. R., Royce, J. R., & Voorhees, B. The factor"fulfillment model" and the concept of multivariatepersonality type. Personality and Individual Differ-ences, 1981,2, 181-189.

Fiske, D. W. Measuring the concepts of personality. Chi-cago: Aldine-Atherton, 1971.

Flaherty, J. F., & Dusek, J. B. An investigation of the

relationships between psychological androgyny andcomponents of self-concept. Journal of Personalityand Social Psychology, 1980, 38, 984-992.

Gilbert, L. A. Toward mental health: The benefits ofpsychological androgyny. Professional Psychology,1981,12, 29-38.

Harrington, D. M., & Andersen, S. M. Creativity, mas-culinity, femininity, and three models of psychologicalandrogyny. Journal of Personality and Social Psy-chology, 1981,47, 744-757.

Heilbrun, A. B. An exploration of antecedents and at-tributes of androgynous and undifierentiated sex roles.Journal of Genetic Psychology, 1978,132, 97-107.

Highlen, P. S., & Russell, B. Effects of counselor genderand counselor and client sex role on females' coun-selor preference. Journal of Counseling Psychology,1980, 27, 157-165.

Holahan, C. K., & Spence, J. T. Desirability and unde-sirable masculine and feminine traits in counselingclients and unselected students. Journal of Consultingand Clinical Psychology, 1980, 48, 300-302.

Kaplan, A. G., Androgyny as a model of mental healthfor women:. From theory to therapy. In A. G. Kaplan& J. P. Bean (Eds.), Beyond sex-role stereotypes: Read-ings toward a psychology of androgyny. Boston: Little,Brown, 1976.

Kaplan, A. G., & Bean, J. P. (Eds.). Beyond sex-rolestereotypes: Readings toward a psychology of androg-yny. Boston: Little, Brown, 1976.

Lubinski, D. The androgyny dimension: A comment onStokes, Childs, and Fuehrer. Journal of CounselingPsychology, 1983, 30, 130-133.

Lubinski, D., Tellegen, A., & Butcher, J. N. The rela-tionship between androgyny and subjective indicatorsof emotional well-being. Journal of Personality andSocial Psychology, 1981, 40, 722-730.

Meehl, P. E. Theoretical risks and tabular asterisks: SirKarl, Sir Ronald, and the slow progress of soft psy-chology. Journal of Consulting and Clinical Psychol-ogy, 1979, 46, 806-834.

Osofsky, J. D., & Osofsky, H. J. Androgyny as a life style.Family Coordinator, 1972, 21, 411-419.

Rawlings, E. I., & Carter, D. K. Psychotherapy for women:Toward equality. Chicago: Charles C Thomas, 1977.

Rebecca, M., Hefner, R., & Oleshansky, B. A model ofsex-role transcendence. Journal of Social Issues, 1976,32, 197-206.

Saunders, D. R. Moderator variables in prediction. Ed-ucational and Psychological Measurement, 1956, 16,209-222.

Spence, J. T., & Helmreich, R. L. Masculinity and fem-ininity: Their psychological dimensions, correlates andantecedents. Austin: University of Texas Press, 1978.

Spence, J. X, & Helmreich, R. L. On assessing "an-drogyny." Sex Roles, 1979, 5, 721-738. (a)

Spence, J. X, & Helmreich, R. L. The many faces ofandrogyny: A reply to Locksley and Colten. Journalof Personality and Social Psychology, 1979,37,1032-1046. (b)

Spence, J. X, & Helmreich, R. L. Androgyny versusgender schema: A comment on Bern's gender schematheory. Psychological Review, 1981, 88, 365-368.

Spence, J. X, Helmreich, R. L., & Holahan, C. K. Neg-ative and positive components of psychological mas-

Page 12: Masculinity, Femininity, and Androgyny Viewed and Assessed ... · + BjMS + B6FS + A. (1) It contains the terms that will enable us to test concurrently (a) the construct validity

MASCULINITY, FEMININITY, AND ANDROGYNY 439

culinity and femininity and their relationships totieu-rotic and acting out behaviors. Journal of Personalityand Social Psychology, 1979, 37, 1673-1682.

Spence, J. X, Helmreich, R., & Stapp, J. Ratings of selfand peers on sex role attributes and their relation toself-esteem and conceptions of masculinity and fem-ininity. Journal of Personality and Social Psychology,1975, 32, 29-39.

Stokes, J., Childs, L., & Fuehrer, A. Gender and sex rolesas predictors of self-disclosure. Journal of CounselingPsychology, 1981, 28, 510-514.

Strahani R. F. Remarks on Bern's measurement of psy-chological androgyny: Alternative methods and a sup-plementary analysis. Journal of Consulting and Clin-ical Psychology, 1975, 43, 568-571.

Sturdevant, S. Therapy with women: A feminist philos-

ophy of treatment. New York: Springer Publishing Co.,1980.

Tellegen, A. Practicing the two disciplines for relaxationand enlightenment: Comment on "Role of the feed-back signal in electromyograph biofeedback: The rel-evance of attention" by Quails and Sheehan. Journalof Experimental Psychology: General, 1981,110,217-226.

Wiggins, J. S., & Holzmuller, A. Psychological androgynyand interpersonal behavior. Journal of Consulting andClinical Psychology, 1978, 46, 40-52.

Zedeck, S. Problems with the use of "moderator" vari-ables. Psychological Bulletin, 1971, 76, 295-310.

Received June 30, 1981Revision received January 18, 1982 •