Factor analysis in assessing validity

MISSING IMAGE

Material Information

Title:
Factor analysis in assessing validity applications to the Bem Sex Role Inventory
Uncontrolled:
Bem Sex Role Inventory
Physical Description:
vii, 150 leaves : ;
Language:
English
Creator:
Belcher, Marcia Jean Macklin, 1949-
Publication Date:

Subjects

Subjects / Keywords:
Sex role   ( lcsh )
Psychological tests   ( lcsh )
Factor analysis   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1981.
Bibliography:
Includes bibliographical references (leaves 145-149).
Statement of Responsibility:
by Marcia Jean Macklin Belcher.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 000295264
notis - ABS1607
oclc - 07863928
System ID:
AA00002188:00001

Full Text














FACTOR ANALYSIS IN
APPLICATIONS TO THE


ASSESSING VALIDITY:
BEM SEX ROLE INVENTORY


MARCIA


JEAN


MACKLIN


BELCHER


A DISSERTATION PRESENTED T
OF THE UNIVERSITY
PARTIAL FULFILLMENT OF
FOR THE DEGREE OF DOC


O THE GRADUATE CO
OF FLORIDA IN
THE REQUIREMENTS
TOR OF PHILOSOPHY


UNIVERSITY


OF FLORIDA


1981


UNCIL















ACKNOWLEDGEMENTS


received


impossible


much


guidance


adequately


support


thank


from


everyone.


so many,


These


acknow-


ledgements,


therefore,


will


only


touch


highlights.


Completion


this


task


would


not


have


been


possible


without


my parents


Because


them,


have


grown


believe


can


accompli


sh whatever


want


to accomplish.


They


applauded


when


things


went


well


and


commiserated


with


me when


they


did


not.


They


believed


me.


committee


provided


invaluable


guidance


support.


Though


they


were


not


always


sure


of what


was


doing,


they


trusted


me.


Drafts


were


returned


promptly


with


valuable


comments.


In particular,


want


thank


chairman,


Linda


Crocker,


who


stuck


with


through


thick


thin,


high


low


Jamie


Algina


provided


direction


guidance


data


analysis.


must


also


thank


Michael


Levy,


boss


during


most


writing


stage


this


dis


sertation.


allowing


to work


flexible


hours,


he contributed


enormous


progr


ess.


This


dissertation


was


typed


on a word


processor.


For









Most


all,


my housemate,


I want


. Sue


thank


Kinzer.


my daughter,


Though


Erika


Erika,


decided


and


that


would


never


write


a di


ssertation


(too


much


work),


sup-


ported


through


mine.


was


always


there,


even


when


could


talk


about


was


sser


station.


washed


clothes,


prepared


meals


cared


daughter


More


than


anyone


knew


how


much


effort


went


into


document.


knew


because


was


there


every


step


way















TABLE OF CONTENTS

CHAPTER PAGE

I STATEMENT OF THE PROBLEM.......................... 1

Purpose of the Study............................ 2
Theoretical Rationale........................... 3
Significance of the Study....................... 6
Summary......................................... 7

II REVIEW OF THE LITERATURE.......................... 9

Development of the BSRI......................... 9
Reliability................................... 11
Methodological Criticisms..................... 12
Typical vs. Ideal Ratings of Self
and Others.................................... 15
Validity and the BSRI........................... 21
Using Behavioral Measures..................... 21
Personality Measures.......................... 23
Factor Analyses of the BSRI..................... 26
Using Self-report Data....................... 26
Factor Analyses Using Desirability
Ratings..................................... 34
Summary......................................... 37

III METHODOLOGY....................................... 40

Design................... ................. 40
Subjects........................................ 41
Instrumentation ............. ..... 41
Data Analysis................................... 42
Limitations.............................. 45
Summary................. .. .................. 46

IV RESULTS........................................... 48

Results of Male vs. Female Self-report
Ratings....................................... 49
Results of Desirability vs. Sslf-rsnnrf










DISCUSSION


AND


CONCLUSIONS.


scuss
scuss


Male/Female


Self


-report/Des


Comparisons.


irability


Compari


sons


Implications


BSRI


Users


1mphI


cation


Studies


Future
BSRI..


Validation


Future
Summary


Studi


Sex


ceptions


APPENDICES


BEM


SEX


AND


ROLE


SELF


INVENTORY


-REPORT


DESIRABILITY


INSTRUCTIONS


.................... 100


Desirability


Self


-report


INTERITEM


Instruc


tions


tructions.


CORRELATIONS


BASED


..................... 101
...................... 103


ON THE


STANDARDIZED


VARIANCE-COVARIANCE


MATRICES.


...... 105


PRIMARY


FACTOR


MATRICES


BY GROUP


.................. 126


BIBLIOGRAPHY


.................... 145


BIOGRAPHICAL


SKETCH.


S............................. 150
















Abstract


of Dissertation Presented


he University
Requirements


of Florida


the Graduate Council


in Partial Fulfillment of


Degree of


Doctor


of Philosophy


FACTOR ANALYSIS


IN ASSESSING VALIDITY:


APPLICATIONS TO THE BEM SEX ROLE


INVENTORY


By

Marcia Jean Macklin Belcher


June,


Chairman:


1981


Linda Crocker


Major


Department


Foundations of Education


The measurement


sex


role


perceptions plays


important


role


evaluations of


sex


equity


in education.


purpose


this


study was


assess


factorial


invariance of


Bem Sex Role


Inventory


( BSRI~


conditions


males


females


using


a self-report


format;


females


using


a self-report


format and


females


rating


woman.


desirability


Responses


same


were obtained


traits


from


an American


females


males


first


phase of


this


study.


second


phase,


responses


were collected


from


females


Subjects


were


undergraduate


students


University


of Florida.


Rssnnnstnq


Ii III*~ --N *~ ~ Eu I


r~rra


C nLa..


An flit 6


urr


r I I ilr


I


* rir
















CHAPTER I

STATEMENT OF THE PROBLEM


In recent


years


an emphasis has been


placed


providing


equal


educational


opportunities


for male


female


students.


Backed


federal


(P.L.


93-380


among


others


efforts


have


been made


to develop programs which


overcome


vocational


real


perceived barriers


development


student


to personal


because of gender


While


equity


been


variously


defined,


one component has


dealt

made


with

that


affective domain.


fostering


egalitarian


The

sex


assumption has


roles


been


the cause of


sex


equity will be


advanced.


Numerous


scales


have


been


developed


to measure


these


perceptions.


Currently,


one


instrument


seems


to be


the most widely used.


That


instrument


the Bemr


Role


Inventory


(BSRI).


The BSRI


also


has


been


used


studies


in Australia


(Feather,


1978;


Rowland,


1977


England


(Whetton &


Swindells


, 1977),


New


Zealand


(Hughes,


1979) ,


Germany


(Hogan,


It has


1979),


been


and Sweden


adanted


(Carlsson & Magnusson,


use wi th


1980)


elucabhl mental iv


-U I


I I -


L









Facilitators


(Jones


Pfeiffer,


1977)


along with


suggestions


Though


incorporating


primarily


developed


into


training materials.


as a research


instrument


test hypotheses


about


androgyny


(Bem,


1974) ,


the BSRI


also been


used


as an affective


evaluation


instrument,


particularly


in vocational


education


(e.g.,


Romero &


Romero,


1978;


Campbell,


1978;


Blimline,


1976).


listed


Sourcebook


of Measures of Women


s Educational


Equity


(Parks


et al.,


1979)


as a potential


evaluation


tool.


Throughout


raised


this


about


flurry


validity


activity,


questions have been


scores obtained


on sex-role


instruments,


after


including


a review of


the BSRI.


findings on


Kelly


several


and Worell


instruments,


(1977),

called


a moratorium on


elopment of new sex-role


instruments


until


validity


those


already


use


could be


better


assessed


. Thomas


(1978),


in a


review of


sex-


role


research


studies


instruments


in psychology,


anthropology,


the operational


sociology,


definition of


education,


sex


roles


also concluded


that


reliability


and validity


of sex-role


instruments


needed more careful


attention.


Purpose


the Study


Duroose of


this


study was


LAi n-. '- 'a t. S


L^ J


1-WU









Should


separate


norms


developed


for men


women on


the BSRI?


i.e.,


are


factor


structures of


responses


the BSRI


invariate


men and women?


same


constructs


(factors)


being


used when


rating


a general


referent


(e.g


, American man,


American


woman)


same


as when


factors


rating


present


oneself?


when


More


using


specifically,


directions


are


scale


development


BSRI


as when


responding


self-report


Theoretical


Rationale


Traditionally,


sex-role was considered


a bipolar,


unidimensional


concept,


and measurement


instruments


were


developed


this


assumption.


Items were given oppo-


site weightings


depending


on whether


they were


labelled


masculine or


feminine.


label


"feminine"


was


assigned


depending


on whether more males


females


endorsed


item


(e.g.,


Gough,


1952).


Bemr


(1974)


rejected


these


assumptions


and procedures


when developing


BSRI.


Instead of


conceiving


sex


role


as bipolar


unidimensional,


she developed


separate


orthogonal


sca


masculinity


femininity.


Instead


differential


endorsement


by mal


females,


. .a a S a a -


"masculine"


I 1 _


i. .I









desirability


in American


society


a man or


a woman.


Items which


received


significantly


different desirability


ratings


sexes


were


assigned


appropriate


scale.


BSRI


was


then


used


as a self-report measure


where


respondent


were


asked


indicate on


7-point


scale


how true each


traits was


themselves.


While there

to scales on the


reason why


basis of physical


items

sex,


should be


there


assigned

a logical


flaw


in Bemr


s procedure which has


received


little


atten-


tion.


task


rating


desirability


a series of


traits


abstract


referent may


be conceptually very


different


from


task


rating


oneself.


the concep-


tual


basis


similar,


self-ratings


then


confusion


stereotypic


to what


ratings


being measured


likely


result.


In only


one


study


(Pedhazur


& Tetenbaum,


1979)


this problem been


investigated.


Because of methodo-


logical


problems


with


their


study,


Pedhazur


and Tetenbaum


were


unable


results.


to draw


After


firm conclusions


factor


analyses of


based


desirability


their


ratings


self-report


ratings,


however,


they


tentatively


concluded


that


resulting


factor


structures were different


enough


to warrant


further


attention.









invariance of


factor


structures


for males and


females.


Since Bemr


(1974)


recommended


same


scoring


procedure


norms


both


sexes on


the BSRI,


question directly


addressed


validity


score


interpretation.


factor


After


loading matrices,


visual


inspection of


Pedhazur


the orthogonal


Tetenbaum


(1979)


concluded


that


factor


structures


were different


males


females,


especially


the way


that


items Bem


labelled masculine were conceptualized.


Another


study


(Sassenrath


& Yonge,


1979) ,


however,


concluded


that


self-


report


factor


structures were


invariate


two sexes.


There


are other


indications


that


the BSRI


may measure


different


things


males


females.


With


sex


role


classification


(based


on BSRI


SCO


res)


used


inde-


pendent


variable,


different


results have been reported


men


women on both behavioral


Lenney,


1976)


observations


and self-report measures which


(e.g.,


should


Bem &


relate


sex


role


(e.g.,


Bem,


1977).


Consistent differences


have


also


been noted


between


self-ratings


ratings of others


when


being

al.,


that


rated

1968;


was


(e.g.,

Deutsch


known was


McKee


& Gilbert,


sex of


lerriffs,

1976).


the other


1959;

Not s


persons


Rosenkrantz


surprisingly,


self-ratings


are


less


stereotyped,


raising


further


ques-


. *


..


..


..









stereotyped


(e.g.,


Elman


al.,


1970).


Evidently,


however,


ratings


both men


and women


are


pulled


direction


the most


socially desirable


pole


(e.g.,


Rosenkrantz


al.,


1968).


Given


that


gender


sex


role have


been so closely


tied


past,


seems


evident


that


the constructs


used


to measure


sex


role


should


be closely


scrutinized


gender


differences.


there


are differences


for men


women on


sex


role


itself,


also possible


that


differences


exist on perceptions


traits which


underlie


sex


role.


Furthermore,


task


rating


desirability


oneself on


a general


same


referent


traits may


task


based on different


rating


sets


of underlying


constructs.


Significance of


the Study


Federal


law mandates


sex


equity


education.


aspect of many


educational


programs


has been


the promotion


egalitarian


sex


roles.


Evaluation


efforts


assessing


value of


these


programs


have


relied heavily on


measurement


sex-role


perceptions


through


use of


instruments


such


BSRI.


Assuming


that


this


use will


continue,


important


that


its validity


be carefully


asses


sed.


users of


BSRI,


it would


valuable to









females might


help clarify


the discrepant


results


males


females which


often have


been


reported.


test


developers,


it would


be valuable


to know


whether


tests constructed


using


one


rating


system


(e.g.,


generalized


desirability


rating


could


then be


used with


results


based on


another


rating


system


(e.g.,


self-rat-


ings).


not,


this often-used method


establishing


scales


would


require


serious


reconsideration.


Identification of


different


factor


structures


males


females


ratings would


call


self-ratings


into question


and desirability


implicit assumption


that

all


traits

tasks.


have

Such


similar meaning


to all


findings would have


respondents


ramifications


research


to compare


evaluation


the BSRI


effects of


effort


subscores of


an educational


in which


males and


treatment.


examiner


females

These fi


wishes


to judge


ndings


would


also


have


implications


research


studies


in which


respondents


are


asked


to rate


themselves


then


rate


"ideal"


"typical"


structures occur


under


individual.

r different


different


instructions,


factor

then direct


comparisons


scores obtained


under


two conditions


would


be of questionable value.


Summ,,rv









assessing masculinity


femininity.


Items were


assigned


to scales


based on


desirability


ratings


a man or woman


in American


society.


The BSRI


was


then


used


as a


self-


report measure.


purpose of


this


study was


investigate whether


same


underlying


constructs


were measured when different


sexes


responded


inventory


when


rating


task


was changed


from desirability


ratings


to self-report


ratings.


Aside


from providing


new


information on


validity


BSRI


scores,


this


study has general


implications


development,


norming,


inter-


pretation of


other


high-inference,


self-report measures.















CHAPTER


REVIEW OF THE LITERATURE


The


supporting


literature


present


study


been organized


First,

(BSRI),


reviewed


the development of the

assumptions underlying


three major

Bemrn Sex Role


sections.

Inventory


its development,


methodological


criticisms


are


reviewed.


Because Bemr


(1974)


assigned


items


to scales on


basis


desirability


ratings,


second


section,


the meaning


"desirability"


ratings of


third


differences


a generalized


section


found


referent


focuses on studies


between self-ratings


are discussed.


related


construct


validity


BSRI.


The construct


validity


BSRI


been


assessed


using


behavioral measures,


personality measures,


factor


analysis.


Relevant


studies


from each


area


are


summarized.


Particular


attention


is paid


to previous


factor


analytic


studies


which


have


have


presented


included


separate


factor


results


analyses of


males and


desirability


females or


ratings.


Development of


the BSRI


.m


---









"that


they might


assertive


yielding,


be both masculine


both


feminine,


instrumental


both


expressive"


(Bern,


1974,


. 155)


depending


tuation.


Conversely,


believed


strongly


sex-typed


individuals


might


exhibit a more


limited


repertoire of behaviors


across


situations.


The

differed


development


from previous


Bem Sex Role


instruments


design


Inventory

ed to ass


ei


(BSRI)

ss sex


roles.


Previous


sca


les were


based on a


bipolar


concep-


tualization of


1952)


masculinity


Measured on a


femininity


single continuum,


(e.g.,


such


Gough,


scales were


composed


items based on


differential


endorsement


males


females


they


rated


thems


elves


(e.g.,


Strong,


1943;


Gough,


1952)


Bemr


(1974,


155),


however,


conceived


"sex-typed person


someone who


internalized


Soc


iety


sex-typed


standards of


desirable behavior


for men


and women.


selected


items


scale


using


criteria

ability


significant


ratings when a man w


differences

as rated an


between desir-


d when a woman was


rated,


agreement


between male


female


judges on


how desirable


a characteristic was


a man or woman;


female


judges


gave


a significantly


different


desirability


rating


a characteristic


than male


judges


--








final


masculinity


version


BSRI


scale contained


consisted of


20 positive


traits


items.


judged


to be more desirable


men


than women.


femininity


scale consisted of


20 positive


traits


judged


to be more


desirable


women


than men.


The


social


desirability


scale contained


10 positive


traits and


10 negative


traits


judged


neutral


with


respect


to sex.


While p

masculine or


classify


depending


femininity


androgyny.


previouss


feminine,


individuals


scales


BSRI


classified persons

scores could be u


as masculine,


the difference


scores.


The closer


between


A t-ratio was


a person


feminine,


their


used


score was


as either


sed


androgynous


masculinity


index of


to zero,


more


androgynous.


Reliability


Bemr


(1974)


reported


internal


consistency


estimates


using Cronbach


alpha


samples of


masculinity,


.80-


femininity,


.70-


social


desirability.


Test-retest


reliability


coefficients


over


four-week


period


ranged


from a


low of


social


desirability


(1977)

(1977)


to a


reported

obtained


high


similar

test-ret


androgyny


reliability

est reliabil


scores.


coefficients. R

ity coefficients


Hogan


:owland









females,


estimates


were


for masculinity,


femininity,


androgyny.


Several


from a


researchers


developmental


have


standpoint.


examined


Hyde


scores


Phillis


the BSRI


(1979)


tested


subjects


ranging


aaes


from 13


to 85 years.


They


found


age differences


in androgyny with


greater percentages


of androgynous males


the older


age groups.


Results


females


were


the opposite direction.


Another


study


(Abrahams,


Feldman,


& Nash,


1979)


focused on different


life


situations


as an


indicator


change


in sex


role


scores.


They


found


differences


predicted direction


between groups of men


and women who were


living


together,


who


just married,


who


small children.


Thus,


appears


that


test-retest


reliability


estimates


are


adequate when


undergraduate college


students


are


being


assessed.


When


scores


are


viewed across


life


span,


however,


fluctuations occur


based


on age


life


situation.


Methodological


Criticisms


Bemr


s methodology


in constructing


scoring


the BSRI


been


criticized on


several


fronts.


terms of


scoring,


Strahan


(1975)


pointed


that


the observations









subject


were


response


further


noted


feeling


for e

that


particularly


ach masculine


"masculine"


item would be


t-ratio confounded mean


that day,

elevated.

level


differences


variability


ratings


between


subjects.


Since


Bemr


(1974)


stated


that


one of her


reasons


selecting


between


t-ratio was


subjects,


use


these criticisms


in drawing


served


inferences


to weaken her


stance.


Spence,


scoring


Helmreich,


question on


Stapp


conceptual


(1975)


rather


attacked


than on


methodological


grounds.


They


showed


that


individuals


scoring


low on both


scales differed on


self-esteem from


individuals


scoring


high on


both


scales.


They


recommended


use of


a median


split method


classifying


respondents


into


four


types.


Bemr


(1977)


agreed


that


distinction between


high-high


scorers and


low-low


scorers


was warranted


supported Spence et


(1975)


use


the median


split method.


Kelly,


Furman,


Young


(1978)


in comparing


four


sex-


role


inventories concluded


that


little data existed


justify


combining


scores


into


broad


typologies


(e.g.,


masculine,


feminine,


androgynous).


They


suggested


multiple


regression


preferred method


combining


w w









1977;


Strahan,


1975;


Kelly


Worrell,


1977;


Pedhazur


Tetenbaum,


1979)


have made


similar


recommendations.


Pedhazur


and Tetenbaum


(1979)


criticized Bemr


using


item-by-item


t-tests


a basis


item


selection


since


tests might


capitalize on chance


findings.


Using


discriminant


function


analysis,


they


found


that


items,


masculine


feminine,


provided correct classification


97.7% of


the cases


the de


sirability


ratings


men and


women


Use of


60 of


items provided


correct


classification.


Bemr


(1979)


replied


that


criteria


requiring


significance


from each


four


groups of


judges


in order


item


to be


included greatly


reduced


possibility


item being


selected


chance.


noted


that


use of


item-by- item significance


tests


established


test construction


practice.


Other


included


studies


the BSRI


designed


have


to cross-validate


produced mixed


items


results,


especially


"neutral"


social


desirability


scale.


Edwards a

been able


nd Ashworth


replicate


(1977)


Bemr


first r

(1974)


reportedd

results


that

for


they

only


had

two


items:


masculine


feminine.


Several


items were


rated


a direction opposite


that


found


Bem.


However,


their


instructions


-iuedaes


rati no


- I TI Z


scale differed


v


C









elicited.


Their


scale


also contained


"undesirable"


markers


where


Bemr


scale did


not.


When Walkup and Abbott


(1978)


used Bemr


results


instructions


18 of


scale,


20 masculine


they


replicated Bem


items and


19 of


feminine


items.


However,


only


half


items


included


social


desirability


scale were


neutral


with


respect


sex.


Pedhazur


Tetenbaum


(1979)


obtained


similar


results.


They noted


that masculinity


femininity


direction,


item ratings were


but 16 of


scaled


neutral


the predicted


items could be


classified


sex-


linked.


Strahan


(1975)


Pedhazur


and Tetenbaum


(1979)


found


Bemr


(1974)


instructions


judges


rating


desirability


to be


vague.


Strahan noted


that


judgments


could


made


either


normatively


or prescriptively,


things


are or


they


should


The question of


desirable


whom,


individual or


society,


was


also


raised.


Typical


vs.


Ideal


Ratings of


Self


and Others


Questions


regarding


Bemr


s use of


desirability ratings


establishing


scales


really


concern


related


issues:


What


does


"desirable"


mean


instructions:


"Indicate


how


desirable


in American


society


-


i


S k


i CI









When Bern


had a


(1974


series


desirability


ratings


completed


items with


development of


significantly


a man


a woman.


BSRI,


different


When used


as a self-report measure,


however,


found only


small


differences


between males


females on


the way


they


rated


themselves.


Pedhazur


and Tetenbaum


(1979)


reported


same


findings.


The question of


was


what


addressed by Gilbert


"desirable"


et al


meant


. (1978).


in BSRI


College


ratings


students


were


asked


to describe


either


what


a man/woman


in our


society


like,


what


is desirable


a man/woman


in our


society,

These in


or what a man/woman


istructions


in our


corresponded


society

typical,


should be

desirable,


like.

and


ideal


ratings.


Results


indicated


that


items


selected


Bem were


based on prescriptive


(ideal)


rather


than


normative


(typical)


judgments.


Some differences


interpretation of


"des


irable"


were


found


for male


female


judges.


In particular,


females did


differentiate


between


typical,


desirable,


ideal


rating masculine


items


a woman.


Downing


role on


(1978)


stereotyped


was


interested


ratings


the effect


a male


sex


female.


used


the BSRI f


self-ratings


and classified


subjects


___









al.,


1968).


found


that


the discrepancy between


ratings of


a male


and a


female


varied


according


sex


role


rater.


Though


androgynous persons were


least d

female,


liscrepant


absolute


their

terms


ratings of


a healthy male


groups described


a healthy


adult male


as different


from a


healthy


adult


female.


Downing


had hypothesized


that


this difference would


occur


androgynous


persons.


concluded


that


"when


information


level


low,


androgynous


individuals describe


unknown others


feminine


stereotypically


individuals do"


just


(Downing,


as masculine and


1978,


Researchers who


have


used


other


checklist


instruments


similar


the BSRI,


have


noted differences


between


ideal


ratings


typical


ratings


oneself


other


Using


Sarbin


s Adjective Checklist,


Sherriffs


and McKee


(1957)


found


that when


asked


to check


adjectives charac-


teri


stic


themselves,


there was


a marked


reduction


from


number


adjectives


checked


a previous


study


where


respondents


were


asked


to check


adjectives generally


describing men


and women.


finding was


particularly


true


men.


The


authors


noted


that


"instructions


describe men or


women


general


[italics


theirs]


almost


invite


subjects


to dismiss


individual


differences while









McKee


and Sherriffs


(1959)


found


that


in describing


their


ideal


selves,


both men


and women selected masculine


feminine


items.


Women,


however,


were more


likely


to select


masculine


items


ideal


self-descriptions


than men were


to select


feminine


items.


In 1976,


Deutsch


and Gilbert


repeated McKee


and Sherriff


(1959)


study using


BSRI


found


basically


same


results.


this


study,


sequence of


ratings was


varied


tested


order


differences;


none were


found.


role


stereotypes


self-ratings


have


also


been


studied


using


a bipolar


format.


In one


study,


a Stereo-


type Questionnaire was


administered with


three


sets of


instructions.


Subjects were


asked


to "imagine


that


are


going


to meet


a person


first


time


the only


thing


know


in advance


that


person


is an adult


male"


marked


(Rosenkrantz


each


et al.,


item on


1968,


extent


288) .


they


Subjects


expected


then


item


characterize


an adult


a male.


female.


Subjects


same

were


instructions were given


also asked


to describe


themselves.


While


the order


of presentation


of male


female


instructions


was


varied,


subjects


rated


themselves

were given


last.


last


authors explained


order


to establish


that self-ratings


a masculine-feminine









themselves


falling


between


their


responses


for males


females.


While


both


sexes


agreed on


stereotype of


an adult male


female,


self-ratings were different.


Male


subjects


masculine


than


rated t

female


themselves


subjects


significantly more


High correlations,


ranging


from


.84,


were


found


between social


desirability


ratings


and masculine,


feminine,


self-


ratings.


Elman


et al.


(1970)


using


a portion of


the Rosenkrantz


et al.


(1968)


Stereotype


Questionnaire


also compared self-


ratings


to ratings


others.


Again,


real


self-ratings


were


seen


falling


between


typical


ratings of


sexes.


Ideal


self-ratings were more


positive


than


their


ideal


ratings


same


sex.


From


studies which


have


used multiple


ratings


(McKee


& Sherriffs,


1959;


Deutsch


& Gilbert,


1976;


Elman et


al.,


1970;


Rosenkrantz


et al.,


1968)


it can be concluded


that masculine


feminine


stereotypes emerge


both


self-descriptions


descriptions


typical male


female


The self-description,


however,


is not as


extreme, f

and woman.


Galling


between


Frieze et


the description of


(1978)


in commenting o:


typical man

n several


e above


studies


noted


.


(I









situations,
experience.


people draw most heavily


Thus,


when


asked


on prior


to "describe"


person
highly


about


whom nothing was


stereotypic


On the other
themselves.


hand,


known but


responses would be


people


Therefore,


sex,


probable.


know most about


the motivation


stereotype


Also,


self-descriptions


stereotypic


when other
have more


(Frieze


cues
leeway
al.,


would be


thinking may


are


reduced.


less prevalent


available or when subjects


to respond


1978,


to ambiguous stimuli


285).


Indeed,


less


stereotyped


responding


did occur


when


people


rated


their


ideal


selves or


ideal others,


ratings


which


relied more on


desirability.


Several


personal

1 studies


preference

(e.g., Ro


social


senkrantz


et al.,


1968;


Downing,


1978)


noted


that


the most socially


desirable


pole was


selected


Though


affect


seems


self-ratings,


these


likely


ser


ratings.


that multiple


no evidence


is availa


ratings could

ble to confirm


this.


fact,


(Elman


et al.,


1970;


four


studies


Rosenkrantz


previously


et al


cited,


, 1968)


only


asked


self-ratings


last.


Yet


four


found basically


same


results.


checked


Furthermore,


order


effects


Deutsch


found


and Gilbert


none


(1976)


using


probability


level


.10.


The


have direct


research


reviewed


relevance


supports


this


study:


three conclusions which


Differentiation








possible


American


BSRI)


that what


society


has changed


"desirable"


(Bem'


a woman


instructions


since


in modern


developing


items were originally


assigned


to subscales.


self-ratings


Furthermore,


ratings of


different concepts may


desirability for


underlie


a woman or


man.

when


Thus, c

subjects


comparison of


respond


under


factor


these


structures obtained


different conditions


seems


warranted.


Although


it has


not been confirmed


that


asking


same


subjects


to provide desirability


self-ratings may


affect


the outcome,


seems preferable


obtain


these


ratings


from separate groups of


subjects


(randomly


assigned


to condition)


to avoid possible


carry-over


effects


from


preceding


instructions.


Validity


the BSRI


Using


Behavioral Measures


It will


recalled


that


Bemr


(1974)


hypothesized


that


androgynous


individuals would


display


situationally


appropriate


behaviors


across


a variety


settings while


sex-typed


individuals would


be unable


to do so.


test


hypothesis,


Bemr


(1975)


used her


instrument


to divide


college


groups.


students


Subjects


into masculine,


were


feminine,


then observed


androgynous


a conformity


situation


(rated


masculine)


and while


playing


with a


kitten









were


less clear-cut


expressive,


feminine domain.


While


feminine


androgynous males


play with


a kitten


more


than masculine males,


the pattern of


findings


females


was


very


ambiguous.


In order


domain,


to clarify findings


further


related


experiments were carried


the expressive


(Bem,


Martyna,


Watson,


1976).


first


study,


masculine,


feminine,

interact


and

with


androgynous males


infant.


A second


females were


study had


asked


subjects


listen


to a


lonely


student.


Initial


results


interactions


with


infant


showed


no differences by


sex or


sex


role.


A supplementary


analysis


using


a median


split


method


rather


than


t-ratio


classifying


subjects did


indicate


that


feminine


androgynous


subjects were more


nurturant

feminine


than masculine

subjects were a


subjects when


subjects.


Iso more


listening


Androgynous


nur turant


lonely


than masculine


student.


Ickes and


Barnes


(1978)


using mixed


sex dyads


found


that


less


interaction


interpersonal


attraction occurred when


both


members


were


sex-role


stereotyped.


To expand on


previous


findings,


Bem and Lenney


(1976)


hypothesized

appropriate


that


sex-typed


activities


individuals would prefer


and resist


sex-inappropriate


sex-









differing


amounts


money.


Each


subject


then


engaged


three


masculine,


three


feminine,


three


neutral


activity


ults


indicated


that


sex


-typed


individual


were


ess


willing


engage


sex-


inappropriate


activities


were


more


uncomfortable


when


they


so.


A main


effect


sex


was


also


reported,


with


males


generally


being


ess


willing


engage


sex-


inappropriate


activity


than


femal


es.


In general,


then,


use


BSRI


ass


ifying


subjects


into


sex-role


groups


upheld


these


studi


es.


Findings


are


clearer,


however,


males


than


females,


indi


eating


that


instrument


behavioral


measures


used


are


more


valid


college


students


than


their


female


counterparts.


Personality


Measures


Based


on Bemr


(1974)


conceptualization


androgyny,


follows


that


people


typed


androgynous


(high


masculinity


femininity)


should


score


higher


than


sex-


typed


individuals


factors


contributing


to positive


mental


health


(e.g.,


self


-es


teem).


Results,


however,


have


always


been


predictable


supportive


constructs


upon


which


BSRI


was


based.









sex


types


were


accounted


differences


masculinity.


concluded


Jones,


that


Chernovetz,


flexibility


and Hansson


adjustment


(1978)


were generally


associated with masculinity


rather


than androgyny


both


sexes


Using


an Australian


sample,


Antill


and Cunningham


(1979)


found


significant


positive correlations


between


masculinity

femininity


scores

scores


and

and


self-esteem scores,


self-esteem scores


none


both


between


sexes.


When


assessed


type,


masculine


androgynous groups


scored


higher


on self-esteem than


feminine group.


Bem


(1977)


also


noted


that


relationship between


sex


role


self-esteem may


different


males


females.


Using multiple


regr


ess


ion analysis,


found


that men who


were


high on masculinity were


also


high on self-esteem


regardless of


their


femininity


scores.


High


self-esteem


women,


however,


was


related


to high


scores on both


masculinity


femininity.


Eman and Morse


(1977)


found


that


androgynous


persons


expressed


higher


degrees of


self-


esteem,


self-acceptance,


acceptance of


others


than did


masculine,


feminine,


or undifferentiated persons.


Kellv


et al.


(1977)


took


a slightly


different


approach


relating


androgyny


positive-valued


traits


adjustment.


as done with


Rather


the BSRI,


than


they


use only


asked









both masculinity


femininity)


endorsed


the most


undesirable masculine


feminine characteristics while


androgynous


men


different


endorsed


females,


with


least.

the only


The


pattern was


significant


finding


being


that


feminine-typed women


were


less


likely to use


undesirable


masculine


self-descriptors.


They


concluded


that


androgyny may be qualitatively


different


men


women.


Several


researcher


have


tried


to relate


BSRI


scores


social


desirability


acquiescence.


Moreland


Montague


(1977)


correlated


scores on


several BSRI


factors


scores on


the Marlow-Crowne Social Des


inability Scale


find


significance.


No significant


results were


found


by Millimet


and Votta


(1979)


sex-type


using


responses


the Couch


Keniston Agreement Response


Scale


the dependent


desirability


a factor,


variable.


Therefore,


it must affect


social


types


equally.


Though


concluded


should


findings


that Bem


highest


are


completely


s contention


self-esteem


that

not


clear-cut,


androgynous


upheld.


it must


persons


Rather,


masculinity


primarily


related


to self-esteem.


Furthermore,


different


patterns of


results


are


found


males a


female


none


. 0- .I r.


findings nffer


additional


.e









Factor


Analyses


the BSRI


Using Self-report Data


It will be


recalled


that Bemr


(1974)


hypothesized


stence of


separate orthogonal


dimensions


for masculinity


femininity.


Therefore,


factor


analytic


studies


should


confirm


this.


To date,


however,


studies


have


shown


BSRI


to be more


factorially complex


than


this


More


than


two dimensions


form or


40-item


have


form


been


(social


found whether


desirability


60-item


scale


deleted)


have


been


employed.


three


studies,


separate


factor


analyses


males


females


have


been performed.


Pedhazur


and Tetenbaum


(1979)


found


different


factor


structures


males


females.


Sassenrath


Yonge


(1979)


reported


similar


results


males


females.


third


study by Whetton


and Swindells


(1977)


provide enough data


adequately


assess


similariti


factor


structures


between


sexes


studies


and did


are described


address


in more


the question.


detail


These


below.


Pedhazur


and Tetenbaum


(1979)


used


a sample of


female


171 male graduate


students


enrolled


two New


.. a -


97,-~.-I r'; i-., "-4I~n ryr.4- q..--- .j n h a Af.


a *


,,


-- --


IlL r A L1A r" A I I A A *I A rl









orthogonal


and


oblique


(delta


= 0)


rotations.


Since


solutions were very


similar,


only


orthogonal


solutions


were


reported.


Pedhazur


Tetenbaum


(1979)


identified


four


interpretable


factors


from each


analysis,


accounting


the common


factor


variance


for males


females.


females,


one


18 of


factor which was


0 masculine


labelled


items


"assertiveness.


loaded


The


feminine


item


"shy"


also


loaded


negatively on


this


factor.


Males,


however,


differentiated


assertiveness


between


factors with


six


items


loading


above


.35 on both


factors.


"Shy"


also obtained


a negative


loading


for males,


they


include


"analytic"


added


"loyal


For

items.


male


females,

It was la


factor


second


belled


factor


consisted of


"interpersonal


interpersonal


feminine


sensitivity.


sensitivity was


very


similar.


Except


the exclusion of


"loyal"


"feminine,


" all


items


loading


above


sexes


agreed.


The


third


factor


females,


which


was


labelled


"self-


sufficiency,


reliant,


" consisted


independent,


three masculine


self-sufficient)


items


which


(self-


also









items


were


included


as part of


one of


larger


assertiveness


included with


factors.


that


feminine


factor


items were


males.


Consistent


with other


studies,


a bipolar


masculinity-


femininity factor


also emerged.


females,


this


factor


consisted


Males,

with


items--"masculine"


however,


included


masculine"


five


grouping


items,


"feminine.


grouping


"gullible"


"athletic"

"childlike"


(included with


female


self-sufficiency factor)


with


"feminine.


Pedhazur

females differ


and Tetenbaum


the way


(1979)


they


concluded


conceptualize


that males


items


which Bern


(1974)


labelled


"masculine.


They


also concluded


that males


females differ


the way


they view


bipolar


conceptualization of masculinity


femininity.


Pedhazur


and Tetenbaum did


state


a particular


criterion


number


factors


to rotate;


they


note, however,

resulted in a


that


inclusion


less clear


facto


of a fifth f

r structure.


:actor


Given


rotation

the


number


this


of high double


study,


loadings,


possible


especially


that


males


factors were









obtained


from separate correlation matrices.


Meredith


(1964)


concluded


that


factors


from separate groups could


be matched


unless


each measure was expressed


same


unit


across


groups.


also offered


a proof


that more


differences would


arise when comparisons were


based on


orthogonal


structures


than when based on oblique


factor


pattern structures.


contrast,


Sassenrath


Yonge


(1979)


reported


that


after


separate


analyses


of males


females,


they


combined


because


and Tetenbaum

undergraduates


school in

they used


(1979)

(533


California


results were


they


females,

SLike


40-item version


so similar


elicited


. Unlike


responses


361 males)

Pedhazur an


Pedhazur


from


who were


id Tetenbau


the BSRI.


It wa


attending

m (1979),

s unclear


how their


procedures differed.


They


state whether


principal


axes or


principal


components


analysis was


used.


They


rotated


those


factors


with


eigenvalues greater


than


1.00


using


a Varimax procedure.


factors were


identified


which


accounted


the variance


(Sassenrath


Yonge,


1979).


Use of


term


"variance"


rather


than


"common variance"


may


indicate


that


principal


components


analysis


was


used.


the discrepancy


findings


between


studies


may


be due


the use


sexes









result


blurring


structure


(Guertin


Bailey,


1970,


148-153).


Speculation


aside,


Sassenrath


and Yonge


(1979)


first


factor


corresponded


closely


to Pedhazur


and Tetenbaum'


interpersonal


sensitivity


factor.


It consisted of


feminine


items


(compassionate,


warm,


sympathetic,


sensitive


to needs of others,


understanding,


tender,


gentle,


affectionate,


eager


to soothe


hurt


feelings,


loyal,


cheerful,


loves


children).


Sassenrath


consisted mainly of


Yonge


s second


masculine


items.


third


Factor


factors


labelled


"dominance,


" consisted of positive


loadings


above


.35 on


the masculine


items


"defends


beliefs,


" "assertive,


" "has


strong personality,


" "force


ful,


" "has


leadership


abilities


aggressivev


" "takes

e," "acts


risks,


" "dominant


leader,


" and


," "takes

negative


stand,"

loadings on


feminine


labelled


items


"autonomy,


"shy"


"soft-spoken.


" consisted of


" Factor


the masculine


items


III,


"self-


reliant,


" "independent,


" "makes


deci


sons


eas


ily,


" "self-


sufficient,


" "individualistic,


" and


"ambitious.


With


exception of


item


"soft-spoken,


" these


items


also


loaded


study.


above


Though


.35 on


in both


Pedhazur


studies


and Tetenbaum


the items


(1979)


loaded between











Sassenrath


Yonge


(1979)


found


two other


factors


consisting of

"competition,


"aggressive,


factor


; masculine

" consisted


items.


" "competitive,


consisted


One,

items


" and


items:


labelled

"athletic,


"ambitious.


"has


Another


leadership abilities"


and

thes


"acts

e item


as a leader.

s appeared a


With


is part


the exception of


larger


"athletic,


assertiveness


factor


both males


females


in Pedhazur


Tetenbaum


study.


A bipolar


masculinity-femininity


factor


also appeared


in Sassenrath


Yonge


results.


Like


female


bipolar


(1979),


factor


their


reported


factor


Pedhazur


consisted


and Tetenbaum


items with opposite


loadings--masculine


feminine.


Sassenrath


Yonge


(1979)


results


supported


Pedhazur


and Tetenbaum


findings


regarding


femininity


items.


They


also


added confirming


evidence of


a bipolar masculinity-


femininity


dimension.


However,


results


regarding


factor


structure of


the masculine


items clouded


rather


than


clarified


previous


findings.


In both


studies,


numerous


double


loadings occurred,








factors.


Other methodological


criticisms


leveled


against


the Pedhazur


and Tetenbaum


(1979)


study


can also be


applied


the work


of Sassenrath


Yonge


(1979).


Calling


previous


work


of Whetton


Swindells


(1977)


offers


little


help


clarifying matters.


They used


item version,


obtaining


different


factors.


Their


sample was


small,


consisting


of 114


females and 118 males.


Most


frustrating,


however,


was


selectivity with


which


they presented


their


data.


Though


they


reported


finding


factors


variance,


which


which


they


accounted


chose


accounted


over


70% of


report only


nearly


half


the common


largest


the common


factors


variance


"moreover


could


be paired meaningfully with


factors


from


the other


sex


factor


analysis"


(Whetton


Swindells,


1977,


152).


They


state a


cut-off


criterion


an item


loading


justify


inclusion


a factor


either


sex.


With


address


so much un

factorial


certainty,

results of


appeared


this


best not


study.


No other


analyses


researchers


males


have


performed


females


separate


BSRI.


Even


factor


though


sample


ize,


sample composition,


inclusion of


variables


other


than


items


themselves,


the method of


factor


analysis


have


varied


considerably


from study to study









same


items


almost


always


fail


to load


on any


factor


(e.g.,


yielding,


flatterable,


does


use


harsh


language,


analytic).


The


BSRI


two-dimensional;


typically


four


five


factors


are


found.


items


U, sclie


"feminine,


" sometimes


conjunction with several


others,


constitute


a bipolar


masculinity-femininity


factor.


Several


feminine


items


(e.g


, shy,


soft-spoken,


gullible,


childlike)


load


as negative


aspects of


a factor


containing mainly masculine


items.


items which


comprise


the masculinity


scale


are


more complex


factorially


than


items which


compri


femininity


scale.


No one


has directly


asses


impac t


including


or deleting


20-item soci


desirability


scale on


resulting


studies,


factor


structure.


is clear


that


From


when


surveys of


social


several


desirability


items


are


included


a number


items designated as


feminine


a number


items


included


Soc


desirability


scale


load on


same


factors.









their


results was


sparse.


would


seem


important


assess


differences


between males


females on


factor


structure of


social


inability


items


as well


as on


the masculinity


femininity


items.


Therefore,


item form was


used


this


study.


also


important


note


previous


factor


analytic


studies


have


suffered


certain


limitations


used which


Factor


Analyses


present

Using


factor


study was designed


Desirability


analytic procedures


to overcome.


Ratings


noted


previously,


rather


than


assigning


items


scales on


basi


differential


endorsement by males


females,


desirability


female


Bemr


of a


in American


(1974)


list


asked


judges


traits


society.


to rate


either


Items were


a male or


assigned


to scales


basis


differential


mean


responses depending


whether


a man


a woman


was


being


rated.


The


BSRI


subsequently


been


used


as a self-report measure.


Only


Pedhazur


and Tetenbaum


(1979)


have


assessed


factor


structure of


the BSRI


when


desirability


ratings were


used


rather


than


self-report


ratings.


60-item


form of


the BSRI


was


administered


1,464


graduate


students


using


Bemr


s original


instructions.


Ratings


a woman were


completed


426 males


females.


Ratings


a man were


_


--










Pedhazur


and Tetenbaum


(1979)


then performed


three


separate


factor


analyses:


a man,


a woman,


and an


adult.


Squared multiple correlations were


used as


initial


communality


estimates.


Both


Varimax orthogonal and


oblique


(delta


= 0)


methods


were


used


to rotate the


factor


structures.


Since


results were


so similar,


only the


orthogonal


structures


were


reported.


Pedhazur


and Tetenbaum concluded


factor


structures


were


similar


whether


ratings


a man,


woman,


adult were


being


elicited.


In each


case,


they


found


three


factors.


One


factor


they named


"assertiveness"


"instrumentality.


When


referent was


a man,


20 masculine


items


loaded


this


factor


above


.40,


no other


items


loaded


When


referent


was a


woman,


all masculine


items


except


"masculine"


loaded


along


with


social


desirability


items


"conscientious"


"conventional"


(negative


loadings)


an adult


referent,


all masculine


items


loaded


except


"masculine"


"athletic.


A second


factor


was


named


"interpersonal


sensitivity.


When a man was


referent,


included


15 of


feminine


items


plus


loadings on


9 of


items


included


SOC


desirability


scale


(helpful,


happy,


reliable,


* t -


I


a


- % ** fl fl S a I Ins, .qtt.1 a C.. a~fl I a -


i


..


,L'I


-i


|









adult,


feminine


Soc


desirability


items


loaded;


"conscientious" was


included


and


"adaptable"


excluded.


item


"feminine"


loaded


when


referent was


a woman.


third


factor was


labelled


"immaturity"


Pedhazur


and Tetenbaum


(1979).


When


a man


was


referent,


consisted


items


loadings


"moody,


above


" "theatrical,


social


" "unpredictable,


desirability


" "jealous,


"secretive,


" "conceited,


" "solemn,


" "inefficient,


" and


"unsystematic.


feminine


items


"flatterable,


"gullible,


" and


same pattern was


found


were

when a


also


included.


woman was


Basically,

referent,


except


"solemn"


"flatterable"


were excluded and


"shy"


was


added.


an adult,


same


pattern


for men


was


found


except


"flatterable" was


excluded.


Pedhazur


and Tetenbaum


(1979)


concluded


the gender


referent


significantly


alter


factor


structure.


While


items Bemr


labelled masculine grouped


as one


factor,


items


labelled


feminine or


social


desirability


factor


did


was


group


positive


factorially


tone


as Bem named


the other was


them.


not.


Pedhazur


Tetenbaum


(1979)


perform


separate


analyses


males


females


see


they


conceptually


/-v r a..ftq nir/- ata.


childlikek"


; L, ,rr,


Arre


IC









Pedhazur


and Tetenbaum


(1979)


were


unable


to compare


their


results


using


self-ratings


their


results using


desirability


ratings


because different


forms of


instrument


were


used


different


subjects.


They noted,


however:


it seems
dimension


plausible
s that un


derl


speculate
ie trait


that


ratings


differ


when


task


to rate oneself


and when


to rate
abstract
findings


would


have


subscales


desirability
p. 1012).


the desirability


referent.


are obt


Assuming


ained


to question


self


ratings


traits


that


future


similar


research,


validity


-ratings on
(Pedhazur &


one


f arriving
basis of


Tetenbaum,


present


study was


designed


to permit


the comparison


between


factor


structures


underlying


self-ratings and


factor


structures


underlying


desirability


ratings


that Pedhazur


and Tetenbaum were


unable


to make.


Summary


development


the BSRI


differed


from


that


previous


sex


role measures.


Instead of


conceptualiz


sex


role


as bipolar,


separate masculinity


femininity


scales


were developed.


Items were


assigned


to scales depending on


whether


they were


judged


more desirable


men or


women


American


society.


Several


researchers


., Deutsch


Gilbert,


1976;


Downing,


1978)


have


found


that


differ-


1979,









Studies


the construct


validity


the BSRI


have


included


use of behavioral measures,


personality measures,


factor


BSRI


upheld,


analysis.


classifying


particularly


behavioral measures,


subjects


males.


use of


into sex-role groups was


terms of personality


measures,


however,


results


were


as clear-cut.


Though


Bemr


(1974)


contended


that


androgynous


persons


should


highest


self-esteem,


masculinity was


a better


a majority


predictor


studies


than


reported


androgyny.


that


Several


studies


(e.g.,


Kelly


et al.,


1977;


Bem,


1977)


reported


different


findings

factor s


patterns


tend


results


to support


structures


male


males


need


female


females.


investigating

responses on


These


the

the


BSRI.


Factor


analyses


have


confirmed


underlying


dimensions of


the constructs measured by


BSRI


as Bem


(1974)


originally


structured


subscales.


studies

factor


which


addressed


structure,


one


the question of


(Pedhazur


sex


Tetenbaum,


differences

1979)


concluded


that


(Sassenrath


differences


Yonge,


1979)


existed while a


reported


second


no differences.


Methodological


shortcomings


these


studies were


identified.


Comparisons


between


self-report


structures









number


studies which


have


addressed


the question.


Such


equivocal


findings


support


need


comparisons of


factor


structures obtained


from males


females and


comparisons of


instructions


factor


used


structures


to obtain


obtained


desirability


under


ratings


the various


self-


report


ratings on


BSRI.















CHAPTER III

METHODOLOGY


The


purpose of


this


study was


answer


following


questions


related


scores obtained


appropriate


from


self-report


use


BSRI:


ratings on


interpretation of


factor


the BSRI


invariate


structure


for males


females?


invariate


factor


self-ratings


structure of


desirability


the BSRI


ratings


women?


Design


Each


the above


questions was


addressed


as a


separate


study.


In each


case,


subjects completed


only one


form of


the BSRI.


The question


male-female


differences


factor


pattern


structures was


addressed


first.


Students were


asked


gender.


females


to rate


Since


themselves


respondent


than males,


when


BSRI


pool had


females


indicate


larger


had


their


proportion


responded,


procedures


second question were


implemented while


I- -. -. -. S


rl


II









Since group administration


the BSRI


was


used,


males


were


asked


take


complete


a one-page white


form.


Females


were


asked


take


complete


two-page


blue


form.


The


blue


form denoted


that


subject was


participating


second


study


comparing


self-report and


desirability


ratings.


The


women


who participated


second study


received


instructions


to either


rate


themselves or


desirability


items


a woman


in American


society.


forms were


ordered


such


that


every


other


form contained


desirability


instructions.


this way,


subjects


were


randomly


assigned


treatment


(instruction)


conditions


second


study.


Subjects


Undergraduates


enrolled


Introduction


to Psychology,


Human Growth and Development,


and Adolescent Psychology


were


subjects


first


study


(221 males,


females).


A total


of 445


females


from


same


population


were


used


second


study.


All


subjects were


students


the University


of Florida.


Data


were collected during


spring,


summer,


fall


1980,


during


the winter


quarter


1981.


1, '.. 4. te ru A. a 1. C a








desirability


examples


ratings,


were


Bemr


utilized


(1974)


original


eliciting


instructions


responses.


Subjects


were


also


asked


to complete


information on


their


sex,


age,


major,


and post-baccalaureate


plans.


No personal


identifiers were


requested.


A copy


instrument


can


found


in Appendix A.


Data Analysis


suggestions made


by Mulaik


(1972)


assessing


invariance of


factor


structures


separate groups


were


followed


this


study.


In particular,


Mulaik


(1972,


356-357%


cited


several


common mistakes which


can


lead


inappropriate comparisons


factors across groups.


They


included


deriving


factors


from correlation


instead


variance-covariance matrices,


comparing


factor


structure


orthogonall)


instead of


factor


pattern


(oblique)


coeffi-


clients,


using


different


criteria


rotation,


using


different methods of


extracting


factors prior


to rotation,


failing


to extract


enough


factor


to get an


accurate


fix on


common


factors


variable


failing


initially


transform the compared


factors


to be


as much


alike


as possible.


He based his conclusions


recom-


mendations on


previous


work


of Meredith


(1964)


Joreskog


(1969).










were


formed


males,


females


total


group.


A diagonal matrix was created by taking


square


root of


the diagonal


from


total


group matrix


(i.e.,


standard


deviation


each


item based on


total


group).


The


subgroup variance-covariance matrices were


postmultiplied by


inverse of


pre-


that diagonal matrix.


Using matrix


notation,


equation


was:


= T-T*A*T-1


where


= the
for


standardized


a given


variance-covariance matrix


subgroup


(e.g.,


males);


inverse


standard
diagonal


T where T contains


deviations


zeros


each


item


the off-diagonals;


variance-covariance matrix


subgroup.


a given


result was


standard


zed matrix which


differed


from a correlation matrix with


"l"s


the diagonal


only to


the extent


that


item variances


a subgroup


differed


from


item variance


total


group


(T).


Specifically,


variance


item i


2 2
was S2 /Si after
im IT


was


standardized.


Thus,


variance


item


was


same


males


total


group of males


females,


/s2
IT


= 1.00.


variance


total


males only,


variance on

2/ST was


item


less


exceeded


than


1.00


Dividing


total


variance would make


a difference


A--~~~~ A-- -


S


-1


r


E,


r


L1- L _


r









group


variance would be


larger


than


variance


each


subgroup.


The effect would


to decrease


"standard


variance"


a similar


item


fashion,


a subgroup


the covariance


less


than 1.0.


items


was


standardized


Sjm/SiTST'


Again,


standardized


covariance


term was


equal


when


item


standard


deviations were


group,


same


differed when


that


subgroup and


subgroup and


total


total groups


had


different


variances


items


standardized variance-covariance matrices were


then


submitted


to a


factor


analytic procedure


using


BMD.


Iterations were


performed


to estimate communalities.


Principal


axes were


then obtained


followed


direct


oblimin


(Jennrich


Sampson,


1966)


rotation


using


gamma


equal


zero.


number


factors


to rotate was chosen


inspecting


results


decrements


the latent


roots,


size of


latent


roots


simple


structure


resulting


factor


patterns.


each


factor


structures


to be compared,


one


structure was obliquely


rotated


while


same


number


unrotated


principal


axes was obtained


the other


group.


Fixing


the oblique


structure,


principal


axes


nI-har~~ rrtui'r emrtn-AIriko-# nar


nCkav


~hC~~a~


Cn


1 n r ~ C


nnir









(B'*B)


'*A and


BT=B*T where


= the
- the


BT= the


unrotate
rotated
rotated


principal


oblique
factors


axes


factors


for
for


subgroup
subgroup


subgroup


similarity


resulting


factor


structures was


assessed


calculating


coefficient of


congruence


(Harmon,


(1972,


1960)


355)


for

note


each matched pair


S


that


factors.


is common practice


Mulaik

to accept


factors


as similar


index


.90 or


better.


Given a


factor matrix,


and


a second


factor matrix,


coefficient of


congruence was calculated


using


following


formula:


CCab


= Sum


(a.b.)
JI3


2
((Sum a.)


(Sum b.))1I2


where


are


elements of


the column


vectors a and


corresponding


observed


variable.


According


Mulaik


(1972,


355),


index


equivalent


to finding


the cosine of


angle


between


the column


vectors of


factor


matrices.


Limitations


This


study was


limited


to undergraduate


students


enrolled


the University


of Florida.


In particular,


different


results might


have


been obtained


a non-college











example,


researcher might


number


choose


factors


a slightly


to rotate.


different


Another


number


factors


to rotate


therefore would obtain


different


results.


Results


also depend


upon


number


type of


variables


included


analysis.


Nesselroade


Baltes


(1970),


using


random data,


found


that


coefficients


congruence


increased when more


factors were


rotated and


decreased


with more


variables


included


analysis.


procedure


used


s study


assess


invariance


between


factor


structures


depends


upon


the same


number


factors


that


being


the optimal


included


number


each


factors


group.


differs


the extent


each group,


procedure


described here


will


differ


from a method


which


accounts


differing


numbers of


factors.


Finally,


factor


analysis


assumes


a linear


relationship


between


variables


(Guertin


& Bailey,


1970).


Any


other


relationship will


inaccurately


represented by


a factor


analytic


structure.


Summary


This


study


addressed


two questions:


factor


structure of


self-report


ratings


the BSRI


invariate









The


60-item


form of


the BSRI


was


completed by


undergraduate males


answering


limited


and


203 undergraduate


first question.


to females;


The


undergraduate


females


second question was


females completed


60-item BSRI


using


self-report


instructions


and 228


completed


the BSRI


using


desirability


rating


instructions


a woman


in American


society.


Separate


factor


analyses were


performed


each


question.


Following


standardization of


variance-


covariance matrix


each


group,


principal axes analysis


with


iterations


to obtain


communalities


were obtained.


Each


primary factor


matrix was


rotated


using


the direct


oblimin


procedure.


each question,


one


primary


factor


matrix was


fixed


principal


axes matrix for


second


sense.


group was


Two


rotated


factors were


to best


judged


least


similar


squares


the coefficient


congruence was


or better.















CHAPTER IV

RESULTS


The


purpose of


this


study was


inspect


and draw


conclusions on


similarity


underlying


constructs


using


responses


Bem Sex Role


Inventory


(BSRI).


Specifically,


first


study


invariance of


primary


factor


matrices


was


assessed


males


females.


second


study,


invariance of


primary


factor


matrices


was


asses


females


who either


completed self-report or


desirability


ratings


a woman


in American


society.


Responses of


221 males


females comprised


the data


base


first


study;


second


study,


females


females completed


completed


desirability


self-report

ratings. Fo


ratings


allowing


a principal


matrix


to similarity.


axes


solution,


one group was


oblique


fixed


Similarity


primary factor


a second matrix


resulting


rotated


structures


was


assessed


judged


using th

similar


ie coefficient of


the coefficie


congruence.

nt reached


Factors were

the criterion


.90.









that different


factor


structures


existed


male and


female graduate education majors


using


40-item


form of


the BSRI.


Sassenrath


Yonge


(1979) ,


however,


reported


only


one


factor


matrix,


stating


that


similar


results


had


been obtained


male


female


undergraduates


Neither


study


considered


problem of metric which


inherent


separate


factor


analyses


of males and


females


from


correlation matrices.


In addition,


both


studies


used


orthogonal


rotation.


Meredith


(1964)


pointed out


that


general


good matching


between


factor


structures cannot


obtained


orthogonal


rotation


used.


Results of Male


vs.


Female Self-report Ratings


The


self-report


item means


and variances


ratings can be


found


males


in Table


females'


The


groups were


highly


similar


their


mean


responses with


exception of


following


items:


athletic,


feminine,


masculine,


competitive.


As Pedhazur


and Tetenbaum


(1979)


noted


with


their


sample,


both


groups may be


responding


in a socially


desirable manner.


Lowest mean


ratings


were


found


those


items


considered


least


positive.


A total


factors


were


rotated


each


group.


These ten


factnrs


sr~rtnin nia


r- -**


Ad


1


E


| r


r


nF
































4,
U

00

Os."
E41U
>~


Hrr- Va ,rf I) 0 N00Nr-moC~1 1 oVaflr40i


S a


tnoQ1r-


* *


H~~~~r( HJ nrllrrI HU C'HU Hr-J


Hr-


0omc d~m~FdO4MN c fl'o u o o tO ND O~D


Ln t-NO


H H~r~ICJHNCIHH


HH r r4


r-I i


tO
r-I C

(E(


I~~~~~~f NrOM0 WDfO r LAOO


LAOOOHCnWN4'LONLODHOD0O~'OOLAONm
Ofl'NOLOfl4DrfltDHNC4WOWO~O


a a


9


* a S S
U~ LA U~ tO







r-a~on
~Ln ID tO
* S S a


H~r-I HC'H


C4 r-- H-I 0 -I


Hi


M N G %I U oinV n 'N C w m0DOI~Ir(U Oar N Hn


OC'4t~OH


4il'V


a. a


0) 0


sC


.l _
J-*c


Pt


* f


Jml










secretive 3
makes decisions easily 4
compassionate 5
sincere 6
self-sufficient 5
eager to soothe hurt 6
conceited 2
dominant 3
softspoken 3
likeable 5
masculine 1
warm 5
solemn 3
willing to take a stand 5
tender 5


friendly
aggressive
gullible
inefficient
acts as a leader
childlike
adaptable
individualistic
doesn't use harsh
language
unsystematic
competitive
loves children


tactful
ambitious


gentle
conventional


6
3


.85
.39
.92
.22
.45
.06
.47
.76
.72
.70
.97
.75
.57
.19
.60
.14
.80


2.37
1.91
.79
.60
1.32
.92
1.47
1.93
2.24
.57
1.20
.70
1.67
1.46
.89
.58
1.91


4.26
4.75
5.36
5.76
5.61
5.45
2.97
4.38
4.20
5.57
5.89
5.44
4.22
5.52
5.03
5.83
4.42


2.17
1.38
1.04
.83


2.31
1.67
.99
.77
1.20
1.26
1.84
1.99
2.33
.82
4.84
.80
1.67
1.37
1.18
.72
1.98
2.19
1.19
1.85
1.53
1.26
1.37


1.87


3.84
2.37
4.41
3.11
5.58
5.52


4.74
3.00
4.46
6.16
5.34
5.81
5.91
4.70


3.57
2.36
2.38
1.14
1.41
1.03
.75
1.75


3.90
3.05
5.69
5.59
5.07
5.66
5.45
4.44


2.86
2.19
1.69
1.95
1.33
1.29
1.00
2.17


3.37
2.27
2.39
1.64
1.39
1.17
.93
1.98






































tn m err FrH t rn wVm fl Ot N


43
C
v-a-,
4.'0
00


O rl fl'NmN fl


WHNriMcO


* 0 S 9 9


If CS ~rI fiC3cl N~ LA N


QJ4J


NO LA Cv Lfl LA N,


tor in









noted


after


that


eight


size


factors


latent


for males


roots drop


seven


below


females.


decrease


size of


latent


roots


fairly


steady


after


nine or


factor


this


reason


factors


were


rotated


because of


finding


by Guertin et


(1981)


that overrotation


less


harmful


to the


factor


structure


than


underrotation


because of Mulaik


(1972)


admonition against


rotating


factors.


The decision


to which


factor


matrix


to retain


fixed


was made


after


inspecting


results of


rotation


both males


females.


female


factor


matrix was


chosen


target matrix


because


fewer


double


loadings


above


.30 were observed


intercorrelations


between


factors were


lower


generally


(see Table


factor


intercorrelations).


Data


presented


in Tables


-16,


listed


appendix,


impact


indicate


that


resulting


rotation

factor s


to similarity


structures.


Table


an

14


displays


primary factor


loadings


female


self-


reports.


Discussion


will


based on


this


table


conjunction with


results


of Table


which


lists


primary


factor


loadings


male


self-reports when


matrix


been


rotated


- --.L- I 1


fAmal 1


fsar-nr mni- r


___ _


Cc


. 1 r









Table


Factor


correlations


primary


factors


males


females


actor


241
034
141
-027


140
104
-264
082
-121


-017
-017
053
-010


-23
07
-15
02
-10
00
05


-093
-045
105


NOTE:
Male
table


Female


factor


factor


intercorr


intercorrelations a
relations are below.


correlations carried


ire


three decimal


listed
Decimals


above the diagonal.
points removed from


places.









1979;


Sassenrath


& Yonge,


1979)


have compared male


female


factor


structures without


rotating


to similarity.


It should


noted


that


the present


study,


this


procedure would


have


resulted


in more


pronounced


differences


(i.e.,


comparison of Tables


than


when


rotation


to similarity was


used


(i.e.,


comparison of


Tables


and 15).


coefficient


congruence


was computed


each


pair


of matched


factors.


A criterion of


.90 was


estab-


lished


similar


acceptance


Against


level


this criterion,


judging


one of


factors as


factors


was


similar


males


females


Coefficients


ranged


from a


high


to a low of


.53.


The coefficients of


congruence


loadings of


each


factor


pattern matrix


are


presented


in Table 4.


Discussion of


results


will be


based on


this


table.


Factor


named


interpersonal


sensitivity,


consisted


items


with


loadings


above


females;


items


loaded


above


males


when


factor pattern matrix


was


rotated


to similarity.


All


items


belonged


to either


feminine or


social


desirability


scale.


both


groups,


this


factor


consisted


of such


items


as:


svmnathet ic.


Cnmna n r i n a i -


In v Q-a 4 tn a r na r na 4 a


Ch


a ~nar






















































OC r-' N
NM.JOON
in in in
* *


NeVe.Irn


N LA tO O Nl CO


Me. IV Ro o3


rflc M O Mt~lo tfl FlOHOM


0OLA M0I p e


vI


Cr1
V
'C
was

WV,
r ta


WV,
tic

OH
we;
wrz4

4-i


CflWCI2


C)
.1-I

C
I0
H-H
zJ-J
(j~j Fl


0)

0
aJZ
I
0

C

Un ---.3 I'fl























































(J O cNO1O0


NY) l r-4r4C'2'Dt


CNO in


\DO CQf l


tfltOHtOCflflO
~'nmc~oui~
NNNLflr1C~4H
0 0 0 9 0

z


IAm N H


NF~(fltfl ~


I
>W2
.4-i


COO
0CC)
L40'r4
44 Ci) Li
-a t


I
C
a)

0
a.
r14w


I
4-;

a)
'ar-I
tJIo
Cr1
04-i


Ow
,% *


I


'4-I
4-SW

EU v-i
'riIU

ore


I

'ow

r40
4-Jet-I
N- 4


I
C)
rz4.r4
4-)
10

HG)
.04-i
. --


I
U

C
r140U2
0~~


00 If
















cm
-o

(OH-
WkQGA
I

~O0
RI
rt~j
mm
(DI


Dirt


a

OH -
Onji.'.
Cl. e-.0

~II

0
I


(OW

IC'.
I
imw
Ut
0
m
I


o U'o
01 n


(A), Ul


.PU1U1O%


N ~bJ


WtNa tot
































































tAn

cnn
9




I
>1
SJ

tlrlr4
cog
I-lOrd
'-'woo


>9
.4.)

'-4 0,-4
Or1w4
00.0
woo
>1 cr4
4J>iW
cr4 .lJ a)
C.r4V
r4.rIr4

0'r4'r4
00)0

ClOG)
4J4JJ~S

000
4bjJjJ

vvv
0)00
ccc
r4Hr4
mw(s)
www
'COO

002W
000


wow
.I~i 4.) As)
or4 r1 r4

ccc
000

4~jJ.iJ
000
PC"
434-IA.)

wino
WOW
4-34)4)
morn
Duo
cr4 r4r4
~OVV
ccc.
*r4',-4~r4 0)
r4
- 0
- n -









feminine


items


this


factor.


Recall


that


neither


study


included


social


desirability


items


that were


part


items


this


study.


This


factor was


judged


similar


two groups


(Cc=.925).


Factor


named


assertiveness,


consisted of


items


female matrix


items


males.


The majority


items


belonged


masculine


scale.


High


loadings


were


found


items


such


as:


acts


as a leader,


leadership


assertive,


abilities,


willing


take


forceful,

a stand,


dominant,

and has s


aggressive,


trong


personality.


notable


discrepancy between


matrices


was


that males


included


feminine


items


"soft-


spoken"


and


"shy"


with


negative


loadings


above


.30.


This


factor


findings

a marked


conformed


fairly well


masculine


divergence


items.


between


to Sassenrath


Pedhazur


their


female


Yonge


and Tetenbaum noted

and male groups on


a factor


which


contained


some of


these


items.


Using


criterion of


.90,


this


factor


was


judged


as similar


males


females


this


study


(Cc=.856).


Factor


3 was


named


self-sufficiency


this


study;


five


items


loaded


above


females


seven


males.


Some common


items


to both


groups wer


self-


-~~~~~~ .L S


I I


t 1


- 1 I II I









females.


similar


to a factor


reported by Sassenrath


Yonge.


this


study,


factor


not meet


similarity criterion


(Cc=. 825)


Factor


entitled


competitiveness,


consisted of


a mix


four


items


from all


three


scales.


Two masculine


items


highest


loadings:


competitive


athletic.


males,


only


"athletic"


loaded


above


.30.


This


factor


was


judged


similar


two groups


(Cc=.684).


Factor


named positive


outlook,


consisted of


five


items


from


feminine


social


desirability scales.


females,


this


factor


was


bipolar


with


positive


loadings


above


items


"happy"


"cheerful"


negative


loading


"moody.


males,


this


factor was


strongly


bipolar.


A similar


factor


previously


been


reported.


It was


judged


dissimilar


two groups


(Cc=.768).


Factor


6 was


introversion


factor;


consisted of


four


items


females:


solemn,


secretive,


soft-spoken,


shy.


males,


only


"shy"


loaded


above


.30.


This


factor


been


reported


by previous


researchers.


factor


did


not meet


similarity


criterion


(Cc=.712).


Factor


was


rather


difficult


to name but


tentatively


*









greater


than


.30.


factor


was


judged as


similar


two groups


(Cc=. 661).


Factor


tentatively


titled


unpredictableness,


was


small


factor


with


high


loadings


two social desirability


items:


unpredictable


unsystematic.


males,


only


"unpredictable"


had


a loading


greater


than


.30.


The


factor


was


similar


males


females


(Cc=.771).


Factor


named


leadership,


high


loading s


four


masculine


females,


items:


leadership


items;


most had


it consisted


already


loaded


of positive


makes decisions easily,


abilities,


acts


and willing


on another


loadings


as a leader,


take


factor.

these


has


risks.


loadings


above


.30 were


found


for males on


this


factor.


similar


factor


was


reported by


Sassenrath


and Yonge


(1979).


factor


not meet


similarity


criterion


(Cc=.532).


Factor


entitled


steadfastness,


consisted of


loadings


three


items


females:


conscientious,


reliable,


strong


personality.


No loadings above


.30 were


found


males on


this


factor.


The


factor


was


found


to be


similar


two groups


(Cc=.648).


a-~~~~~~~ *1 3 rK s4- riitn4. .. ra A 4 en. ..9%MW.n S


shnou a1


nrat a


A~1~ ...t


fuar.4-nr


ff 1* *


I


L ^"









alone.


The optimal structure


males


the one


reported


in Table 16 of


the Appendix.


The greatest


divergency


noted


find


last


one definite


four


pattern


factors.


While


emerging,


is difficult


groupings of


the social


desirability


items


appear


quite different


sexes.


Results of


Desirability vs.


Self-report Ratings


item means


variances


self-report


desirability


ratings


can


found


in Table


In general,


few mean


differences


were


noted between


self-report and


desirability


ratings


items


Bemr


(1974)


classified


"feminine.


Items which


belonged


Soc


desira-


ability


scale


and were negative


tone


had


lower mean


responses


desirability


than


self-report,


indicating


that


these


items were


rated


less desirable


but more


frequently true.


both groups,


however,


mean


responses


were


fairly


low,


indicating


that


both


cases


desirability


might


be depressing


responses.


For masculine


items,


lower


mean


tions


responses were


than


noted


self-report


under


the desirability


instructions.


This


instruc-


finding


indicates


that


college women may view themselves as more


masculine


than


they


believe


desirable


in American


society.









Table


Item means


ratings


Item


Desir
Mean


. (N=228)
Variance


Self
Mean


(N=217)
Variance


Total
Variance


self-reliant


yielding
helpful
defends


own


1.61


6.15


beliefs


eerful.


6.06


moody
independent
shy
conscientious


4.34


2.27


2.04


2.28


1.23


athletic


affectionate
theatrical


asser


tive


flatterable


happy


1.43
2.49


1.82


4.05
4.83


1.33
1.89


1.76
1.86


has strong
loyal


personality


unpredictable
forceful


1.95


feminine
reliable


6.43
4.07
3.87


2.01


2.32
1.04


analytical
sympathetic
jealous


1.58
1.00


4.92
5.99


1.71


leadership
sensitive
others
truthful
willing to


ability


1.88


to needs of


6.09


take


understanding


5.20


risks


4.81


6.29


1.94


variances


female


fi Fl 1 f~ r Pnn r t









secretive
makes decisions easily
compassionate
sincere


self-sufficient


eager to soothe hurt
conceited


3.07
4.68
6.15
6.21
4.45


1.66


dominant


softspoken
likeable
masculine


4.31
6.13
1.50
6.00
3.27


warm


solemn


willing to take a stand 4.23


tender


5.93
6.15


friendly


aggressive
gullible
inefficient


acts


a leader


1.62
3.74


childlike


adaptable
individualistic
doesn't use harsh
language
unsystematic
competitive
loves children


tactful


ambitious


gentle
conventional


5.29
4.74


2.27
1.39
.66
.80
1.82
1.18
1.11
1.56
2.27


1.81
.91


1.98
2.16
1.21
2.08
1.61
1.33


5.29


3.70
6.00
5.40
4.64
6.09
4.62


1.83
1.89
1.11
1.57
1.96
.86
1.66


4.11
4.47
5.93
6.20
5.75
5.99
2.63
3.79
4.04
5.76
1.91
5.78
3.74
5.38
5.60
6.07
4.01
3.59


4.57
3.04
5.60
5.67


4.57
3.01
4.58
6.08


5.84
5.90
4.64


1.97


2.39
1.71


1.04
1.06
1.82


1.84
1.13
1.69


2.68
1.67
1.16


2.48


1.08


1.35


1.91


1.18
1.94
1.89
1.07
1.18


1.31


1.89
1.22
1.87


3.37


3.16
1.97
2.19
1.29
1.45
1.89


1.49
1.33
1.08


1.81


1.73









cumulative


variance


accounted


shown


in Table


latent


roots dropped


desirability


ratings


below


after


after


seven


five


factors


factors


self-


report.


The drop


size of


latent


roots was


stable


after


nine or


accounted


factors.

total va


These


iriance


ten

for


factors

the


desirability


ratings


total


variance


self-report


ratings


Higher


intercorrelations


among


factors


were


found


desirability


structure


than


self-report


structure


(see


Table


possible


that


fewer


factors


needed


to be


rotated


the desirability


matrix


than


self-report matrix.


results of


the oblique


rotation of


desirability


factors


were


fixed,


self-report matrix


was


rotated


to similarity.


Discuss


ion will


based


factorial


results of


desirability


structure with


comparisons made


primary factor


loadings


female


self-report


ratings when


rotated


to similarity.


factors,


only one met


similarity


criterion


.90.


Coefficients


congruence


ranged


from


a high


to a


low of


(see


Table


. All


loadings


above


each matrix


are


also displayed


in Table










Table 6.


Latent roots of the unrotated orthogonal factors for female


desirability and self-report ratings


Latent


Factor


root


Desirability
Cumulative % of


total variance


Self-report


Latent


root


Cumulative % of
total variance


9.65
8.96
3.45
1.06
1.03


7.17
5.52
2.35
1.99
1.50
1.36
1.15
.90
.83






Table


7. Factor correlations for rotated factors for desirability and
self-report ratings


Factor 1
1 -
2 -033
3 406
4 114
5 -276
6 -160
7 -045
8 -070
9 243
10 -081


3
-089
189


-051
152
344
377
071
199
264


-055
-093
017
140
-050
270
035


4
126
128
217

132
194
-147
254
047
-106


5
236
093
-096
-003

256
037
203
-051
-004


6
167
097
-080
-022
042

095
161
039
094


7
191
211
232
230
119
016

-011
096
170


8
-116
-000
-011
-039
-043
-206
110

064
048


9
-117
-014
233
212
-150
-050
114
019


10
201
059
-041
168
051
001
025
-113
-039


NOTE: Female self-report factor intercorrelations are listed above
the diagonal. Desirability factor intercorrelations are below. Decimal
points removed from table; correlations carried to three decimal places.














































4' inU ~I


9I 00 9o 9 0


('4 ID tO ON


04'O\


uw ir-t o


C4W O r-4


HfOO\


On AH


i-BOO


rH 01 r1
40 0 0
i^r-rr-


(flr4


I
4)
C
C)
Z*r4
0
Ir$
0)4-4
>4-I

t~AI


WI
'4-4 $4
Ow
-'-'V

0)0)
4~rH


0 I

C

-- #0
. cFI


WWIrr


I
0
*rt

(A)

H
''U
f-4~

4-4.-
A'.-


4) 0)
IC
oflno
.~ Rfl In














































r-u~
Mn'

*



















ON
r-to
'nan
*


fl'CflOr4(ThNHO
r4CNOCINNNIOW
n'n'ulUlr.2Lne4mrI
* 9 *



















NNfl'MO(NItfl(flO


* 9 9 9 9 9


r4N~.D
1000~
(N CN H
* *





















U1 U~I gO
* 6


or'


.-4N
nfl
9


eur Hr'-Ln' .oO


Mrbn rfl(flfl'O


Fz4(J2


I
0)
(nj-i
0
IC
0)0
Hr4
.04-i
mci


0201
4-By
I C
HO~0

tI-4*rI tfl


I
0
(OH
.0
(alto
Oji
1>0
(O.r'.r4
001W


I
di
4-i
0
C

.5-I
IC)
0W

















































r4CD r4 OC~l


Hr-4r(~DUI) H


oI r-Ui


r-r-norm


'Otto


04l HO In 04


Hl~ O~


I
.cn

CI
~z404J
r-4C


I
F140
-.4
141
0)0


a-~4J
WOO)
~ i'-.-


I % -r


020
*r4
I4~1
ala)
r-4.C

00
* a


* a -


S. I









Not


named


Items:


None


above


.572


Note:


"S" des


ignates
ignates
ignates


sca


that


that an


that


item
item
item


assigned
assigned
assigned


the masculine


feminine
social d


scale.
scale.


esirability









own


beliefs,


acts


as a leader,


leadership abilities,


ambitious,


individualistic,


strong personality,


analytical,


dominant,


willing


take


a stand,


aggressive,


forceful,


competitive.


list


included


18 of


masculine


which


items


from Bemr


had previously


been


scaling.


ass


In general,


signed


items


self-sufficiency


factor


lower


loadings


self-report


than on


desirability matrix.


Usinc


a criterion of


.90,


factor


not meet


similarity


criterion


desirability


self-report


ratings


(Cc=.863).


Factor


entitled


interpersonal


sensitivity,


was


also


similar


factor


same


name


reported


males


females.


A total


items


loaded


above


desirability


ratings on


this


factor


including


these


feminine


items:


warm,


tender,


gentle,


compassionate,


eager


to soothe


hurt


feelings,


sensitive


needs


of others,


understanding,


sympathetic.


A number


of social


desirability


similar


items


results were


also


loaded


found.


The


above


.30.


factor


self-report,


was


judged as


similar


two conditions


(Cc=.913) .


Factor


tentatively named


resourcefulness,


included


loadings


above


five


items.


desirability


ratina


. the


factor


included


these


items:


adaptable.


u









ratings.


The


factor


not meet


similarity


criterion


two groups


(Cc=.715).


Factors


4 and


both


consisted mainly


items with


negative connotations.


Since one


factor


seemed composed


mainly


items


with


an outward direction while


items


loading


second


factor


were more


passive,


factors


were


named


active


negative


affect


passive


negative


affect


respectively.


Factor


active


negative


affect,


was composed mainly


items


from


SOC


desirability


sea


desirability


ratings,


consisted of


items


including:


jealous,


secretive,


conceited,


moody,


and unpredictable.


self-report


ratings,


only


three


items


loaded above


.30.


Similarity was


Factor


items.


found


passive


soc


this


negative


desirability


factor


affect,


ratings,


Cc=


780)


included


five


included


a mix


items Bemr


desirability


(1974)


ass


scales.


signed


items


feminine


loading


above


social


.40 were


"gullible,


"does


use


harsh


language,


"inefficient,


"unsystematic.


Despite


loadings


above


most of


same


items on


self-report


ratings,


factor


did


meet


similarity


criterion


two groups


(Cc=


743k








above


.30.


factors did


not meet


similarity


criterion


two groups


(Cc=. 694).


Factor


named


steadfastness,


consisted


five


items.


desirability


rating


loadings


above


.30 were


found


social


esirability


items


"conscientious,


"helpful,


" and


"reliable,


" and


feminine


items


"loyal"


"sympathetic.


Only


"conscientious"


loading


above


self-report


factor


The


factor


reach


similarity


criterion


(Cc=.739).


Factor


8 also did


not meet


similarity


criterion


(Cc=.617).


desirability


ratings,


consisted of


loading s


above


three


items:


solemn,


conventional,


yielding


No loadings


above


.30 were


found


self-report


items on


factor.


Factor


9 consi


stedC


loadings


above


two


items,


"happy"


and


"cheerful,


" for


desirability


ratings


none


self-report.


factors


two groups


meet


similarity


criterion


(Cc=.691)


Factor


10 was


a "junk"


factor


with


no loadings above


either


ratings.


It did


not meet


similarity


criterion


(Cc=.57).









cluster


items.


together


Perhaps most


this case,


notable


nor


that


femininity


a bipolar masculinity-


femininity factor


now


emerged.


Comparisons of


Female Self-report Samples


Given


that


female


self-report


structures


different


patterns of


rotated with


loadings when


the males


with


they were


desirability


standardized


ratings,


one


subsequent question


became


how different


structures


actually were.


Therefore,


female


self-


report samples were


standardized on


each other


rotated


to similarity.


The


results


can


found


in Table


including


coefficients of


congruence.


Using


criterion of


factors were


applied


judged


to previous


as similar.


samples,


These


two of


factors were


interpersonal


sensitivity


factor


the positive


outlook


factor.


however,


coefficients of


congruence were


judged by where


a definite


drop occurred


factors,


first


five


factors


(i.e.,


those with coefficients


greater


than


.85)


would


judged


similar.


Reviewing


previous


comparisons


with


criterion of


.85,


only


first


factors


(i.e.,


interpersonal


sensitivity


asse


rtiveness)


would


judged


similar


both


"'


w. --


w
















































cc


C~OD0C4


HJ C0 H~


0( 0 y I0 0
oooo^>KO0i mm
^P^fMrN c nr
* *


C4U~
C
>1*~
vv
il-Jo'
CO v-i


OI U Hi


9 9


S* *


NrD H V 1CO lH%


0 in I


I


(12(12


I'
wrz4

WI
.Cr-I


0I
U) 4-' V
C

CCJJ
S.r4W








tender-


warm-


gentle- F
affectionate-


flatterabi


compassionate-
friendly- S
feminine- F
likeable- S
has strong
personality-


Self-sufficiency
Items:
self-sufficient-
adaptable- S
independent- M
self-reliant- M


individuals


tic-


.652
.645
.611
.598
.399
.353
.350
.314
.165


.491
.428
.446
.440
.159
.311
.240
.203
.340


.398


.697
.601
.594
.579
.541


.851


.455
.572
.532
.461


willing t
risks-
decisions


take


easy-


analytical-


.382
.328
.316


.088
.218
.212


.864


Positive Outlook


Items:


cheerful-
happy- S
moody- S
friendvly-
likeable-


.747
.688


-482


.404
.361


.503
.575
.544
.281
.251


.905


Not


named


Items:


gullible-
childlike-


.613
.528


.288
.229








Table


- Continued


Study 2
loadings


Study 1
loadings


Coefficient of
congruence


ineffic


ient-


theatrical- S
unsystematic-
jealous- S


.508
.336
.324
.316


.368
.230
.414
.172


.781


Not named


Items:
defends


beliefs-


.427


.439


yielding- F
unpredictable-
forceful- M
has strong
personality-


willing to
a stand-


-,408
.393


-.284


.168


.383


.367


.208


take


loves children-


dominant-


.338
.338
.307


.403
.148
.319


.817


Not


named


Items:


athletic- M
masculine- MN
competitive-
feminine- F


.687
.479
.405
.361


.398
.338
.392


-.045


.767


Not named


Items:


feminine-
secretive-


conventional-


.230


.598
.448
.448



















































HC'Joom


0CN


COtnr-a\ w


4-)
eM
H
U, *.rI
'-4 0) .0
Or-itO
12 C)
WW
a) C)
CU,',
""4 C
r4rir4
~C'0

0220
000
2tI-~u2
000
Sc"
4-) 4~) .3-)
000
4-) 4-) 4~3

0202W

CCC
000
HHIH
00)0)
.0.0.0

222
WOO)
43434-)


CCC
000
4-) 4-' 4-)
000

434343


II
020


.Url


I
102

00

043
Jaw
Cr4
Eli 0


I
C)
*r40)
4~2
tofu'
Eu-I
Iwo


020202
WOW
43434-)
000
CCC
.9-4 .r$ -.4
mww
00)0.
vv V U)
H
z 2 2 dii









replicable,


female


five


self-report


factors would


ratings,


be considered


two would be


stable


replicable


under


each


condition.


criterion


similarity was


further


lowered


a most


liberal


one of


.75,


nine


factors would


judged


similar


female


self-report


structures


when


standardized


on each


other.


Using


same criterion,


five of


factors would


then be


judged


as similar


the male/female comparison.


self-report/desira-


ability


case,


comparison


however,


three of


loadings on


ten would be


last


similar.


factors were


In each


not as


high


or consistent,


indicating


that


these


factors may not


stable or


replicable


first


factors.


Summary


first


study,


invariance of


factor


structures


males


females on


self-report


ratings was


ass


essed


procedures


suggested


by Mulaik


(1972)


were


followed


in comparing


factor


structures


these


groups.


The criterion


similarity


two matched


factors


was


a coefficient


of congruence of


.90.


Although


primary


factors were


rotated


to similarity for


groups,


only


one of


factors was


judged


similar


n'rk4


C n an* nn*a --


F~nChr


A ~mh~


FI A YI I* : L :


~AL CI: r LI









second


study,


using


females


only,


invariance of


report


society.


factor


desirability


Again,


structures was


ratings


primary


assessed


a woman


factors were


self-


in American


rotated


similarity


two groups,


again


only


one of


factors was


judged


similar.


This


factor was


interpersonal


sensitivity.


When


samples were


similarity,


self-report


standardized


two of


ratings


on one


from the


another


factors were


female


rotated


judged


as similar


using


criterion of


.90.


This


finding


indicated


that


the criterion of


may


have


been


stringent.


When


the criterion was


lowered


.85,


five


factors


were


judged


as similar


female


self-report


samples.


The


five


factors were


labelled


interpersonal


sensitivity,


outlook.


self-report


assertiveness,


both


self-sufficiency,


the male/female


comparisons,


only


positive


the desirability/


factors--interpersonal


sensitivity


assertiveness--were


judged


similar


using


the more


liberal


criterion


.85.


Thus,


factor


substantial


matrices occurred


discrepancies

d. While only


between


first


primary

five









male


self-report


female


self-report


to desirability


ratings.


was concluded,


therefore,


that different


constructs


underlie each


series of


ratings.


Only by


including


made


limited


between males


set of


items can valid


females or


between


comparisons


self-report and


desirability


ratings on


BSRI.
















CHAPTER V

DISCUSSION AND CONCLUSIONS


The


purpose of


this


study was


to assess


factorial


invariance


the BSRI


two separate comparisons:


undergraduate males


manner;


undergraduate


females

e female


responding i

s responding


n a self-report

in a self-


report manner


rating


the desirability of


items


woman


in American


society.


first


study


implica-


tions


interpretation of BSRI


scores


they


are


now


used


counseling,


research,


program evaluation


efforts.


second


part


study


implications


instrument development


self-report


inventory


which


items were originally


assigned


to scales using


desirability


ratings.


Results


these comparisons


will


summarized


discussed


in order.


Discussion of Male/Female Self-report


Comparisons


factor


factors


assertiveness,


analyses of


tentatively named


self-sufficiency,


60-item BSRI


interpersonal


resulted


sensitivity,


competitiveness,


positive


outlook.


Ia. J


itroversion.


frnlsfitl nas. -


nnnrscii ctah1 psne .








best


fit,


factor


was


judged


to be


similar


samples


the coefficient


congruence between matched


primary factors


equalled or


exceeded


.90.


With


this


criterion as


a guide,


one


factor was


identified as


similar


the male/female comparison.


This


factor,


entitled


interpersonal


sensitivity,


been


replicated previously


a number


researchers


(e.g.,


Pedhazur


& Tetenbaum,


1979;


many


Sassenrath


items Bemr


& Yonge,


(1974)


1979).


designated


While


factor


"feminine"


included


(e.g.,


warm,


sympathetic,


compassionate),


factor


cannot be called


"feminine"


since


item


"feminine"


consistently failed


load


this


factor.


possibility was


entertained


that


this criterion


(Cc=. 90)


may


have


been


too stringent.


assess


this


possibility,


female


self-report


samples were compared


standardizing


female


self-report


samples using


their


pooled


variance


rotating


them to


best


fit.


this case,


when


results


should


have


been highly


similar


two groups


(from


same


population),


only


two of


factors


were


judged


similar


using


a congruence


criterion of


.90.


coefficients of congruence


were


therefore


reinterpreted


light of


fact


that


some


factors might








noted


that a


occurred


distinct


after


drop


first


size of


five


factors.


the coefficients


This drop corre-


sponded


to a


cut-off


criterion of


the coefficient


congruence.


Applying


new criterion of


assess


factor


similarity,


two


factors were


judged


similar


male/female


comparison


(as opposed


five


judged


similar


female/female


previously mentioned


self-report comparison).


interpersonal


Besides


sensitivity factor,


assertiveness


factor,


comprised mainly


items Bem had


labelled

dominant,


"masculine"

aggressive,


(e.g.,

acts


willing


a leader)


take a

was n


stand,


ow judged as


similar


for males


females.


Again,


assertiveness


factor


was


labelled


"masculine"


because of


loading


that


item on


factor.


Previous


researchers have disagreed on


assertiveness


factor


the composition


similarity


factor


for males


females.


This


factor


was


similar


one


found by Sassenrath


Yonge


(1979)


using


a 40-item


form of


the BSRI.


They


concluded


factor


structure was


similar


for males


females when


factors were


rotated.


However,


Pedhazur


Tetenbaum


(1979)


concluded


that


factor was dissimilar


using


40-item form.









emerged


if Pedhazur


and Tetenbaum had


rotated more


than


four


factors.


present


study


a clear


structure


emerged


assertiveness


self-sufficiency for


both


sexes,


possibly because more


factors


were


rotated.


Using


female/female


self-report comparison as


guide,

failed


several

to be


other


judged


factor


similar


appeared


to be


for males


stable but

females.


was concluded


that


self-sufficiency was


a stable


factor


females


(Cc=.86)


that


factor


differed


mal


es.


Following


same


logic,


positive outlook


factor


also differed


for males


females.


This


study was


what was.


notable


for what was


In particular,


this


study


found,

did not


as well

reveal


a bipolar


masculinity-femininity


factor which had


emerged


previous


studies


(e.g.,


Gaudreau,


1977;


Waters et


al.,


1977;


Pedhazu r


& Tetenbaum,


1979;


Sassenrath


Yonge,


1979).


A form of


this


factor


emerge when


female


self-report


groups were


compared;


factor


was


not,


however,


considered


stable.


Since


this


factor was


typically


reported


a later


factor


analysis,


seems


likely


that


factor


emerged


either


as a


result of


rotating


factors or


from


the context


in which


flflPT


S- -


was


presented.


Direct


comparison


between


findinas of


F









these


findings


indicate


that


separate


scoring


procedures


scale,


be developed


males


comprised mostly of Bemr


females?


feminine


items,


A single


could be


used


to measure


interpersonal


sensitivity


both males


females.


Given a


certain amount of measurement


error,


"assertiveness"


could


also comprise


a common


scale


for male


females.


desirability


A number


items,


of Bemr


however,


s masculine


social


had quite different


factorial


representations


for males


females.


Bemr


(1979)


announced


that


the BSRI


was


being


revised


shortened


based


on s


several


factor


analyses.


Care


should


exercised,


however,


that


selected


items do not differ


factorially


BSRI


for males


females.


now constructed


should


Scores on


not be


60-item


interpreted


to have


similar meanings


for males


females.


Discussion of


Self-report/Desirability Comparisons


Factor


analyses


60-item BSRI


resulted


factors


self-report/des


inability comparisons.


five major


named:


factors based on


assertiveness,


desirability


interpersonal


ratings were


sensitivity,


resourcefulness,


and active


and passive


negative


affect.


primary


factor


structures met


similarity


criterion


only


one


factor:


interpersonal











When


the criterion


similarity


this


study was


lowered


.85,


asse


rtiveness


factor was


also


judged


similar


the desirability


self-report conditions.


should


factor


noted


differed


that


from


some of


those


items


assigned


loading on


this


this


factor


male/female comparison.


notable


that


no self-


sufficiency factor


emerged


under


desirability


instructions


even


though


this


factor


appeared


was


judged


similar


female/


female


self-report comparison.


Thus,


self-


sufficiency


appears


to be


factor


self-ratings


females


disappears


under


the des


irability


rating


condition.


Only


Pedhazur


and Tetenbaum


(1979)


have


performed


factor


analyses


desirability


ratings on


the BSRI.


Their


interpersonal


sensitivity


.40)


same


factor


items


high


reported


loadings


this


(above


study.


Their


feminine


factor,


however,


items


was much more


social


desirability


inclusive,


items


with more


loading


factor.


Pedhazur


and Tetenbaum


(1979)


assertiveness


factor


Again,


might


was


inclusive of


it could


have


argued


resulted


the one


that


a different


reported


rotation of more


picture


study.


factor


However,









Implications


BSRI


Users


illustrate


some


practical


implications


that can be


drawn


from this


study for


users of


the BSRI,


four


possible


situations


are described


below


in which BSRI


scores


are


commonly used.


Comments


are offered


appropriateness


these


common


practices


view of


findings


this


study.


one


level,


suppose


a counselor


were


faced


with


task


discussing


individual'


s scores on


the BSRI


with


that


person.


Would


it be


appropriate


to discuss


results


terms


by computing


of masculinity,


subscale


scores


femininity,


as originally


androgyny


hypothesized?


Although


these


subscales may


still


used,


the counselor


should


aware


that


traits


"masculinity"


"femininity"


are


unitary constructs.


example,


masculinity may be


factorially


decomposed


into


traits


such


"assertiveness,


" self-sufficiency,


" etc.,


femininity


might


be decomposed


into


traits


such


"interpersonal


sensitivity"


"positive outlook.


Since,


in general,


women


rated


themselves


low on


item


"masculine"


and men


low on


item


"feminine"


desirability


ratings were


"masculine,


" these


terms


are


highly


influenced by


sander


and social


'I


Psirah il itv.









"masculine"


"feminine,


" especially


term did not


correspond


their


gender.


One


respondent


even


described


study


"un-American.


When


the BSRI


was explained


terms of


"interpersonal


sensitivity"


"assertiveness"


dimensions,


however,


no problems


ensued.


Thus,


counselor


might


find


discussion of


individual


results more


acceptable


informative


to counselees


terms other


than


"masculinity"


"femininity"


were employed.


Teachers


might


also


find


such


terms more


acceptable


to parents and


students.


another


level,


suppose


a researcher wants


compare


views


a single


group composed


solely


either males or


females on what


is desirable


a man and


what


desirable


a woman


in American


society.


Assuming


that


gender


is constant


rating


instructions


are


constant,


indication


results of


that


such


present


comparisons are


study offer


inappropriate.


variety


compared


ways


with


which


the BSRI,


sex


this


role


stereotypes


probably


remains


can be


use most


compatible


with Bem'


original


development


procedure


es.


this


study,


under


desirability


instructions,


most


items


from Bem'


masculine


subscale


loaded


on a


single


factor


(assertiveness)


and most


items


from Bemr


feminine


subscale









"socially


desirable"


also


loaded


the major


feminine


factor,


this might


indicate that


-item version


(which


excludes


social


desirability


scale)


might be


preferred


such


studies.


Suppose,


compare


however,


scores


that


women


researcher


wanted


self-perceptions


to their


stereotypes


ideal


woman.


researcher


obtains


different


results


between


self-perceptions


stereotypes,


these


differences can


only


meaningfully


interpreted


similar


constructs


underlie


both


sets


ratings.


Results


present


study


indicate


that


similar


constructs do


underlie BSRI


ratings obtained


under


different


instruc-


tions.


Only


two of


factors compared were


judged


similar


self-report


desirability


ratings.


Thus,


comparisons of


scores


obtained


under


these


sets of


directions


Finally,


compare


should


be discouraged.


consider


researcher who desires


the mean self-perceptions of


females against


mean


self-perceptions


males.


Would


such a comparison be


valid?


Again,


findings of


this


study


indicate


that


different


constructs


underlie


self-perceptions of males


females.


factors compared


from responses of


mAlne


fomn1 aC


Ca ,' an, 1I*n


a n ^A


nntl~


n: mr' I~r


erlh









imply


similar


positions on


underlying


attitude


continuum.


Thus,


interpretation


findings


from


suich


study


could


problematic.


Implications


Future


Validation Studies of


the BSRI


This


study


conceptualized

of the BSRI to


probably


date


controlled

. Unlike


been


the most


factor

previous


carefully


analytic

studies,


examination

pains were


taken


to obtain


chance


similar


an adequate


sample


findings


to provide every


two groups


to arise.


Substantial


discrepancies


desirability/self-repo


between male/female

rt instructions were


perceptions

still


observed.


therefore


recommended


that


future


explora-


tions of


factor


structure of


the BSRI


based on


separate


male


female


samples.


This


study was


restricted


to a


comparison


factor


structures


separate


subgroups.


One


strength of


study


care which


was


taken


to make


comparison groups


similar


as possible,


i.e.,


standardization on


pooled


total


group variance,


use of


oblique

controls

structure


factors,

weaken

e for a


rotation

attempts


particular


to similarity.

o describe the

group when st


However,


optimal


:udied


these


factor

itself.


i* nr


e4- annArA ; -4rC(


" Ar


*n~ fAILfL -: .. -nl. n ., n f









different


factor


structures


could


result


from a


single


group


factor


analysis.


From viewing


primary factor


matrices given


to similarity


in Appendix C,


impact on


obvious


factor


that


rotation


structure.


Also,


procedure


used


this


study


required


an equal


number


factors


each


group.


desirability


ratings


been


considered


alone,


probably fewer


factors could


have


been


rotated


than were


required


this comparison.


these


reasons,


readers


are cautioned


take


factor


structures


reported


this


study


as an optimal


description of


underlying


dimensions of


the BSRI.


proper


procedure


optimal


factor


location would


separately factor


self-report


analyze


desirability


BSRI


ratings


males


females or


beginning with a


raw


score


matrix or


correlation matrix of


items


that


group only.


Using


standard


factor


analytic procedure


would


allow


the optimal


structure


to emerge


without


restrictions


placed


upon


samples


employed


this


study.


While


a valuable


greatest


factor


tool


use


analysis


ass


lies


as demonstrated


essi


this


validity


in clarifying


study


instrument,


the constructs


being


measured an


assessing the


impact


of other


variables


---