categorization agreement of the personality attributes questionnaire and the bem sex role inventory

9
CATEGORlZATlON AGREEMENT OF THE PERSONALITY ATTRIBTJTES QUESTIONNAlRE AND THE BEM SEX ROLE INVENTORY JOHN P. GAA' AND DOV LIBERIIAN Unwersitg of Hauston Two studies investigated the extent to which the BSRI and the PA& assigned the same individuals to identical sex-role categories. In the first study both instruments were administered to 160 Ss and in the second to 132 Ss. In both instances less than 55% of the Ss were assigned to identical categories by both instruments. Because there are substantial differences in the ways in which the measures were conceptualized and constructed, imperfect agree- ment in sex-role categorization might be expected. However, findings indicate that care must be taken when com aring results of studies that employed different instruments or when labeling individuals as belonging to specific sex-role Categories. Within the past several years a shift in focus in sex-role identity research has occurred, which has placed increasing emphasis on the concept of androgyny. This emphasis has been seen with regard to conceptual issues, and with a concern for t,he measurement and identification of individual sex roles. Androgyny has been conceptualized as representing an integration of mas- culine and feminine sex-role characteristics. Traditional measures of masculinity and femininity tended to view sex-role identity as being bipolar, with masculinity and femininity representing opposite ends of a continuum. This approach, however, does not allow for the measurement of sex role in a manner consistent with the concept of androgyny. Partially because of this, the conceptualization of mas- culinity and femininity as representing independent and separate variables, rather than bipolar extremes, recently has gained wider acceptance. Both Bern (1974, 1977a) and Spence (Spence, Helmreich, & Stapp, 1974) have developed instruments for the measurement of sex role identity based on this assumption of the indepen- dence of masculine and feminine scales. The Bem Sex Role Inventory (BSRI) consists of 60 items, 20 from each of three scales: Masculinity (M), Femininity (F), and Social Desirability (SD). The personality characteristics that appear in the M and F scales were defined as masculine or feminine on the basis that each descriptor was judged to represent a desirable characteristic for a male or female. These descriptors were selected from among 200 personality descriptors deemed by Bern and her students to have positive value and to be either masculine or feminine in nature. Items included on the scales were selected from those terms (of the original 200) that were judged by Stanford undergraduates to be significantly (p <.05) more desirable for a man or significantly (p <.05) more desirable for a woman. The SD scale was intended to provide a scale that is neutral with respect to sex role and was drawn from 200 personality descriptors deemed to be characteristic of neither a feminine nor a masculine- role. ~ Individuals who are responding to the BSRI are directed to indicate 011 a 7-point scale the extent to which each item (personality characteristic) is des- criptive of themselves. Each of the seven points on the scale is labeled; the scale extends from 1 (never or almost never true) to 7 (always or almost always true). The short form of the Personality Attributes Questionnaire (PAQ) developed by Spence consists of 24 items, which include 8 items from each of three scales: Masculinity (M), Femininity (F), and Masculinity-Femininity (M-F). The short form of the PA& is drawn from the full PA&, which uses 55 items taken from ft T h e authors wish to indicate that each has contributed equally to the research. 593

Upload: john-p-gaa

Post on 06-Jun-2016

213 views

Category:

Documents


0 download

TRANSCRIPT

CATEGORlZATlON AGREEMENT OF THE PERSONALITY ATTRIBTJTES QUESTIONNAlRE

AND THE BEM SEX ROLE INVENTORY JOHN P. GAA' AND DOV LIBERIIAN

Unwersitg of Hauston

Two studies investigated the extent to which the BSRI and the PA& assigned the same individuals to identical sex-role categories. In the first study both instruments were administered to 160 Ss and in the second to 132 Ss. In both instances less than 55% of the Ss were assigned to identical categories by both instruments. Because there are substantial differences in the ways in which the measures were conceptualized and constructed, imperfect agree- ment in sex-role categorization might be expected. However, findings indicate that care must be taken when com aring results of studies that employed different instruments or when labeling individuals as belonging to specific sex-role Categories.

Within the past several years a shift in focus in sex-role identity research has occurred, which has placed increasing emphasis on the concept of androgyny. This emphasis has been seen with regard to conceptual issues, and with a concern for t,he measurement and identification of individual sex roles.

Androgyny has been conceptualized as representing an integration of mas- culine and feminine sex-role characteristics. Traditional measures of masculinity and femininity tended to view sex-role identity as being bipolar, with masculinity and femininity representing opposite ends of a continuum. This approach, however, does not allow for the measurement of sex role in a manner consistent with the concept of androgyny. Partially because of this, the conceptualization of mas- culinity and femininity as representing independent and separate variables, rather than bipolar extremes, recently has gained wider acceptance. Both Bern (1974, 1977a) and Spence (Spence, Helmreich, & Stapp, 1974) have developed instruments for the measurement of sex role identity based on this assumption of the indepen- dence of masculine and feminine scales.

The Bem Sex Role Inventory (BSRI) consists of 60 items, 20 from each of three scales: Masculinity (M), Femininity (F), and Social Desirability (SD). The personality characteristics that appear in the M and F scales were defined as masculine or feminine on the basis that each descriptor was judged to represent a desirable characteristic for a male or female. These descriptors were selected from among 200 personality descriptors deemed by Bern and her students to have positive value and to be either masculine or feminine in nature. Items included on the scales were selected from those terms (of the original 200) that were judged by Stanford undergraduates to be significantly ( p <.05) more desirable for a man or significantly ( p <.05) more desirable for a woman. The SD scale was intended to provide a scale that is neutral with respect to sex role and was drawn from 200 personality descriptors deemed to be characteristic of neither a feminine nor a masculine- role. ~

Individuals who are responding to the BSRI are directed to indicate 011 a 7-point scale the extent to which each item (personality characteristic) is des- criptive of themselves. Each of the seven points on the scale is labeled; the scale extends from 1 (never or almost never true) to 7 (always or almost always true).

The short form of the Personality Attributes Questionnaire (PAQ) developed by Spence consists of 24 items, which include 8 items from each of three scales: Masculinity (M), Femininity (F), and Masculinity-Femininity (M-F). The short form of the PA& is drawn from the full PA&, which uses 55 items taken from ft

T h e authors wish to indicate that each has contributed equally to the research.

593

594 Journal of Clinical Psychology, Ju ly , 1981, Vol. 37, No. 3.

pool of over 130 items (initially identified by Rosenkrantz, Vogel, Bee, Broverman. and Rroverman, 1968) that were identified by college students as personality char- acteristics that differentiate between men and women. The items employed on the instrument were selected from those for which a consistent stereotype (male or fcmale) was identified by both male and female Xs. Items then were assigned to one of the three scales on the following basis: Masculine items are those identified 8s characteristically socially desirable for both sexes, but as occurring more fre- quently in males ; Feminine items are those identified as characteristically desirable for both sexes, but as occurring more frequently in females; and l\lasculinity- Femininity itcnis are those identified as socially desirable for one sex, but not the other. For the short form of the PA&, eight items were chosen for inclusion on each scale to which they belonged. Individuals who are responding to the in- strument are instructed to indicate on a 5-point scale the point on a bipolar de- scriptive continuum that best describes them. Unlike the BSRI, only the end points of each continuum are labeled.

BSRI and PA& total scale scores are determined in an identical manner. Responses values for the individual items (1-7 on the BSRI and 0-4 on the PA&) on a given scale simply are totaled in order to determine the scale score, which results in a possible 32 points on each scale of the PA& and a possible 140 points on each scale of the BSRI.

Additionally, the way in which Ss are assigned to a sex-role category is identical for both instruments. An individual is categorized as androgynous when scores on both the p t l and F scales are above the median score, as masculine when the M score is above the median and the F score is below the median, as feminine when the F score is above the median and the M score is below the median, and as undifferentiated when both the M and F scores fall below the median. Although the sex-role categorization procedure was not initially the same for both the in- struments, Bem (1977a) modified the categorization format for the BSRI to match that used by Spence in the PA&.

A number of conceptual and instrumentation similarities appear to be ap- parent in the BSRI and PA&. Both instruments are seen as measuring sex role identities existing in relation to discrete roles, as opposed to a bipolar conceptuali- zation. Additionally, both use the same labels for their two primary scales (Mas- culinity and Femininity) and for the categories of identity that may be derived froni use of the scales (Androgyny, Masculinity, Femininity, and Undifferentiated). Certainly, Bem’s adjustment (Beni, 1977a) of the categorization format, labels, and procedures to match those of the PA& has increased the perceived similarity of the instruments. Because of these common aspects, the assumption that the BSRI and PA& measure the same construct and, therefore, should categorize individuals identically in terms of sex role is one that might easily be made. How- ever, there are a number of aspects related to the two instruments that bring into question the appropriateness of such an assumption.

Although many similarities of conceptualization exist, i t is not clear that Bem’s and Spence’s perceptions of the constructs of masculinity and femininity are congruent. For example, Bcin has conducted a number of studies designed to establish behavioral correlates of various sex role categories and feels that identity categories relate to a wide variety of “sex typed” behavior. However, Spencc and Helmreich (1978) have indicated that “Knowledge of the degree to which an individual possesses or manifests masculine or feminine personality character- istics does not permit inferences about how sex-typed the individual is in interest patterns, for example, or in any other category of role related behaviors [p. 141.’’ Additionally, the diff crences in the original formulations of the measures, which resulted in the BSRI initially employing relative balance of M and F scores to determine the identity categories of Masculine, Feminine, and Androgynous, may well have had an impact on scale development, which might influence the degree to which the BSRT and the PA& measure identical constructs,

C’ategorization Agreement 595

Diff ereiices in criteria aiid procedures discussed cwlicr for inclusion of itcins on scores that bear identical labels (Rlasculinity and Femininity) also mny affect, the probability that thc measures will categorize individuals idcntically. Indeed, RS Spe‘nce and Helmreich (1978) have pointed out, “The possibility exists, thcn, that the presence of items on the BSRI R I and F scales that would not clearly meet our criterion for inclusion on these scalcs may sometiincs causc somewhat different results t o be obtained with the PA& and the BSRI . . . [p. 251.”

The present paper presents two sstudics dcsigiied to investigatr the extent to which the BSRI and the PAQ do in fact measure the samc construct by cxamiii- ing the degree to which individuals are assigned idcntica! sex rolc categories on the basis of responses to the instruments. Failure of thc instruments to arrive at such agreement would have potentially major implications for attempts to intcgrate results of studies employing the different instruments.

STUDY O N E

A ~ F T H O I ) Subjects

The BSltl and the PA& \\ere administered in random order to 160 under- graduate student volunteers. All Ss were juniors or seniors in teacher education programs at) n major southwestern university.

Analysis Scores for each S on the BSKI and the PA& werc computed, and Pearson

product-moment correlations between the respective A l and the respective F scales of the two instruments were detcrmincd. In accordance with the suggestion of Bem (197713)) sample medians were established for both the RSItI and the PA&, and Ss were classified as Androgynous, Masculine, Feminine, or Undif- ferentiated using these medians. This procedure calls for the computation of sample medians and employing them in making categorical assignments following normal procedures for the instruments. Computed medians for the sample were 101 on the Masculine scale of the BSRI and 108 on the Feminine scale of the BSRI, and 21 on the Masculine scale of the PA& and 26 on the Feminine scale of the PA&. Using the computed medians, Ss then mere assigned the appropriate sex role cate- gory on each instrument using the procedures described above. Finally, cross tabulations were computed, and the percentage of categorical agreement across instruments was established.

RESULTS Correlations between raw scores on the BSRI and the PAQ were .56 betweerr

the Masculine scales and .59 between the feminine scales. The cross tabulation of categorical assignments of individuals based on scores

of the BSRI and PA& indicate that substantial variation in categorical assign- ment exists. On the BSRI, 20.6% of the Ss were classified as ,4ndrogynous, 20%) as Masculine, 20.6% as Feminine, and 38.87; as Undifferentiated. Scores on the PA& resulted in 24.4y0 of the Ss being categorized as Androgynous, 31.3% as Masculine, 21.3% as Feminine, and 33.1Y0 as Undiff crentiated.

A separate consideration of each of the four identity categories yields some interesting comparisons. Of 39 Ss identified by the PA& and the 33 identified by the BSRI as Androgynous, only 19 were identified as Androgynous by both instruments. Indeed, of the 39 Ss labeled as Androgynous by the PA&, 9 were classified as Undifferentiated on the basis of BSRI scores, while 8 were labeled Feminine and 3 as Masculine. Of the 33 Ss categorized as Androgynous by the BSRI, 4 were categorized as Masculine, 8 as Feminine, and 2 as Undifferentiated on the basis of PA& scores.

596 Journal of Clinical Psychology, Ju ly , 1981, Vol. 37, No. 3.

Masc.

TABLE 1 COMPARISON OF CATEGORICAL ASSIQNMENTS

__--__-- -

PAQ Count Row pet. Row Col. pct. Total pct. Andr. Masc. Fern. Undif. total

Andr. 19.0 4 .0 8 . 0 2 . 0 33.0 57.6 12.1 24.2 6.1 20. t i 4'8.7 11.8 25.5 3.8 11.9 2 .5 5 .0 1.3 3.0 18.0 2 . 0 9 .0 32.0 9 . 4 56.3 6.3 28.1 20 * 0 7.7 52.9 5 .9 17.0 1 .o 11.3 1.3 5.6

HSKI Fern. 8.0 4 .0 13.0 8 .0 33.0

24.2 12.1 39.4 24.2 20.0 20.5 11.8 38.2 15.1 5 .0 2 . 5 8 . 1 5 . 0

Undif. 9 .0 8 . 0 11.0 34.0 62.0 14.5 12.9 17.5 54.8 38.8 23.1 23.5 32.4 64.2 5 .6 5 . 0 6.9 21.3

Column 39.0 34.0 34.0 53.0 160.0

totd 24.4 31.3 21.3 33.1 loo. 0

Thirty-four Ss were labeled as Masculine by the PA& and 32 by the BSRI; of these, only 18 were identically identified by both. Of the 34 identified as Mas- culine by the PA&, 4 were categorized as Androgynous, 4 as Feminine, and 8 as Undifferentiated on the basis of BSRI scores. Three of the 32 Ss categorized as Masculine by the BSRI were labeled as Androgynous by tjhe PA&, 2 as Feminine, and 9 as Undifferentiated.

The PA& also categorized 34 Ss as Feminine, while the BSRI placed 33 Ss in this category; yet only 13 Ss were identified as Feminine by both instruments. Eight of the 34 identified as Feminine by the PAC) were categorized as Androgynous by the BSRI, 2 as Masculine, and 11 as Undifferentiated. Of the 33 Ss labeled as Feminine on the basis of the BSRI, 8 were categorized as Androgynous, 4 as Masculine, and 8 as Undifferentiated by the PA&.

More individuals were assigned to the Undifferentiated category than to any other. The PA& classified 53 Ss as Undifferentiated, while the BSRI assigned 62 Ss to the category. However, only 34 individuals were placed in this category by both instruments. Of the 53 identified as Undifferentiated by the PAC), 2 were labeled Androgynous, 9 as Masculine, and 8 as Feminine by the BSRI. Nine of the 62 Ss categorized as Undifferentiated on the basis of their BSRI scores were classified as Androgynous by the PA&, 8 were classified as Androgynous by the PA&, 8 were classified as Masculine, and 11 as Feminine.

Overall, 52.5% of the Ss (84 of 160) were classified identically by both of the instruments, with the most overall agreement in thc Undifferentiated category, and the least with respect to the Feminine category.

Categorization A greemen t 597

DISCUSSION Whilc raw scorc correlations between respective Masculinc and Feminine

scales on the two instruments were fairly high, thc failure of the BSRI and PA& to identify large percentagcs of individuals as bclonging to the samc scx-role catego- ries raises important questions about thc constructs that are being incasurcd by the instruments and the extent to which it can be assumed that studics that employ different instrunicnts can bc equated.

However, several aspccts of thc study secm to bring into question thc extent to which the obtained results might be rclated to the spccific saniplc corisidercd in the study. Thc incdians for the sample xere well above thosc reportcd by Bcm for the BSltI and thc median indicated by Spence for the Feminiric scale of thc PA&. Although thc correlation of thc two Feminine scales is close to that reported by Spence and Helmrcich (l97S), there existed a wide difference between the correlation of the Masculine scales in the prcsent study and thosc reported by Spence and Helmreich (1978). Because of the possibility that thesc factors inter- acted in such a way as to rcducc the degree to which thcre is categorical agreement across instruments, it was decided to repeat the study with some modifications.

STUDY Two METHOD

Subjects The BSRI and the PA& were administered in random order to 132 volunteers.

All Ss mere graduate studcnts in a college of education at a major university in the southwest. Analysis

Scores for each S on the BSRI and the PA& were computed using the scoring formats proposed by Bein (1977a) and Spence (Spcncc et al., 1974), and Pearson product-moment correlations were computed between both the respective hI and F scales of the two instruments.

Once again, sample medians were computed for the BSRI and the PA&. Using these sample medians, Ss mere classified on each instrument as Androgynous, Masculine, Feminine, or Undiff ercntiated. Coniputed sample mediaiis were 103 on the Masculine scale of the BSRI and 104.36 on the Feminine scale of the BSRI; and 21.33 on the IIasculine scale of the PA& and 24.50 on thc Feminine scale of the PA&. Cross tabulations then were computed, and the percentage of cate- gorical agreement across instruments was determined.

Additionally, scparate saniplc mcdians for males and fcinales were established for both the BSIZI and PA&. Sample medians for males were 111.0 on the Masculine scale of the BSIII and 99.5 on the Feminine scale, and 22.8 on the Jlasculine scale of the PAQ and 24.2 on the Feminine scale. Female sample medians were computed to be 98.8 on the nlasculine scale of the BSRZ and 108.5 on the Feminine scale, and 20.5 on the Masculine scale of the PA& and 25.0 on the Feminine scale. Rlales and Females then were assigned separately to sex role categories on the basis of responses to each instrument. Cross tabulations were computed, and the percentage of categorical agreement across the two instrumcnts was determined.

RESULTS Correlations between raw scores on the BSRI and the PA& were .79 betwcen

Masculine scales and .7 1 between Feminine scales. The cross tabulation of the categorical assignments based on total sample

medians for the BSRI and PAQ again indicate a wide variation in those assign- ments. On the basis of BSRI scores, 21.2% of the Ss were classified as Androgynous 30.3% as Masculine, 28.0% as Feminine, and 20.5y0 as Undifferentiated. Scores on the PAQ resulted in 23.5y0 of the Ss being categorized as Androgynous, 25% as Masculine, 26.t5?&o as Feminine, and %yo as Undifferentiated.

598 Journal of Clinical Psychology, J u l y , 1981, Vol. 37, No. 3.

TABLE 2

COMPARISON OF CATEGORICAL ASSIGNMENTS

Row Count Row pct. Col. pct. Total pct. Andr. Masc. Fern. Undif. total

PAQ

Andr.

Max.

BSRI Fern.

Undif.

Column

total

14.0 50.0 45.2 10.6 10.0 25.0 32.3 7.G

5 . 0 13.5 16.1 3.8 2.0 7 .4 6.5 1.5

31.0 23.5

9 . 0 32.1 27.3

6 .8 21.0 52.5 63.6 15.9

.o

.o

.o

.o 3 . 0

11.1 9 . 1 2 . 3

33.0 25.0

4 . 0 14.3 11.4 3 .0 5 . 0

12.5 14.3 3.8

21.0 56.8 60.0 15.9 5 .0

18.5 14.3 3 .8

35.0 26.5

1 . 0 3 .6 3 . 0

.8 4.0

10.0 12.1 3.0

11.0 29.7 33.3

8 . 3 17.0 63.0 51.5 12.9 33.0 25.0

28.0 21.2

40.0 30.0

37.0 28.0

27.0 20.5

132.0 100.0

A separate consideration of Ss within each of the identity categories reveals interesting assignments patterns across instruments. Of the 31 Ss labeled as Andro- gynous on the basis of PA& scores and the 28 Ss identified by the BSRI, onlv 14 were classified identically by both instruments. Indeed, of the 31 Ss labeled as Androgynous by the PA&, 2 were classified as Undifferentiated on the basis of BSRI scores, while 5 were labeled as Feminine and 10 as Masculine. Of the 28 Ss categorized as Androgynous by the BSRI, 1 was categorized as Undifferenti- ated, 4 as Feminine, and 9 as Masculine on the basis of PA& scores.

Thirty-three Ss were labeled as Masculine by the PA& and 40 by the BSRI; of these, 21 were classified identically by both. Of the 33 identified as Masculine by the PA&, 9 were categorized as Androgynous, none as Feminine, and 3 as Un- differentiated on the basis of BSRI scores. Ten of the 40 Ss categorized as Mas- culine by the BSRl were labeled as Androgynous by the PAQ, 5 as Feminine, and 4 as Undifferentiated.

The PA& also categorized 35 Ss as Feminine, while the BSRI placed 37 Ss in this category. Twenty-one Ss were classified identically by both instrumeqts. Four of the 35 Ss identified as feminine by the 1’AQ were categorized as Andro- gynous by the BSRI, 5 as Masculine, and 4 as Undifferentiated. Of the 37 Ss labeled as Feminine on the basis of the BSRI, 5 were categorized as Androgynous, none as hlasculine, and 11 as Undifferentiated by the PAQ.

The PA& classified 33 Ss as Undifferentiated, while the BSRI assigned 27 Ss to that category. However, only 17 individuals were identified as such by the PA&, 1 was labeled Androgynous, 4 as Masculine and 11 as Feminine by the BSRI. Two of the 27 Ss categorized as Undifferentiated on the basis of thc RSRI were classified as Androgynous by the PAQ, 3 were classified as Masculine, and 5 as Feminine.

Categorization Agreement 599

TABLE 3

COMPARISON OF CATEGORICAL ASSIGNMENTS: MALE

Count Row pct. Col. pct. Total pct.

Row Andr. Masc. Fern. Undif. total

PA&

Andr.

Masc.

BSRI Fem.

Undif.

4 . 0 36.4 36.4 7 .4 6.0

25.0 54.5 11.1

~~ -

5 .0 45.5 22.7 9.3

15.0 62.5 68.2 27.8

~~

2 . 0 18.2 14.3 3.7 2 . 0 8.3

14.3 3 .7

~ ~~~

.o 11.0

.o 20.4

.o

.o 1.0 24.0 4 . 2 44.4

14.3 1.9

.o .o 6.0 .0 6.0

.o .o 100.0 .o 11.1

.o .o 42.9 .o

.o .o 11.1 .o 1.0 2 .0 4 . 0 6.0 13.0 7.7 15.4 30.8 46.2 24.1 9 .1 9 . 1 28.6 85.7 1.9 3.7 7 .4 11.1

Column 11.0 22.0 14.0 7 .0 54.0

total 20.4 40.7 25.9 13.0 100.0

Examination of the diagonal of Table 2 indicated that approximately 55% (72 of 132) of the Ss were categorized identically by the PA& and the BSRI. Examination of cross tabulations for males and females scparately shows similar percentages of males and females identically categorized by the two measures (57.4% of males and 53.8% of females). However, patterns of categorical agree- ment seem to differ markedly for males and females. For example, the BSRI and the PA& each identified 11 male Ss as Androgynous, but only 4 Ss were identified mutually as such by both instrumcnts; yet of the 17 female Ss identified by the BSRI and of the 20 Ss identified by the PA& as Androgynous only 10 were cate- gorized identically.

DISCUSSION Correlations between scales of the BSRI and the PA& were quite high; .79

between the masculine scales and .71 between the feminine scales. These compare with correlations obtained by Stapp and Karmer (unpublished manuscript) and cited by Spence and Helmreich (1978): .75 and .73 for males and females, respec- tively, for the masculine scaIes and .57 and .59 for males and females, respectively, for the feminine scales of the instruments. Nonetheless, only about 55% of the individuals in the study were categorized identically by both the PA& and thc BSRI, which again raises questions about both the degree to which the instruments are measuring identical constructs and the efficacy of the categorization scheme.

SUMMARY Although results varied in specifics across the two studies, there are several

aspects that seem to be consistent. In both studies the fairly to very high cor-

GOO Journal of Clinical Psychology, Ju ly , 1981, Vol. 37, No. 3.

TABLE 4

COMPARISON OF CATEQORICAL ASSIGNMENTS: FEMALE

Count Row pct. Col. pet. Total pct.

R O W Andr. Masc. Fem. Undif. total

PA&

Andr. 10.0 4 . 0 2 . 0 1.0 17.0 58.8 23.5 11.8 5.9 21.8 50.0 36.4 9 . 5 3.8 12.8 5 . 1 2.6 1.3 4.0 6 . 0 3.0 3.0 16.0

25.0 37.5 18.8 18.8 20.5 20.0 54.5 14.3 11.5

5.1 7 .'7 3 .2 3 .8

Masc.

BSRI Fern.

Undif.

5 .0 .o 15.0 11.0 31.0 16.1 .o 48.4 35.5 39.7 25.0 .o 71.4 42.3 6.4 .o 19.2 14.1 1.0 1.0 1 . 0 11.0 14.0 7.1 7.1 7.1 78.6 17.9 5.0 9 . 1 4 . 8 42.3 1.3 1.3 1 . 3 14.1

Column 20.0 11.0 21.0 26.0 78.0

total 25.6 14.1 26.9 33.3 100.0

relations betwcen identically labeled scales seem to be indicative of a relatively high degree of overlap in the constructs measured by the BSItI and the PA&. However, the failure of the two instruments to identify large percentages of in- dividuals as belonging to the same sex role categories raises important questions about the measurement of the construct of androgyny, and the extent to which i t can be assumed that studies that employ differing instruments are measuring the same construct.

The results of both studies support the findings of Wang-McCarthy, Jose, and Crosby (Note 1) with regard to the degree to which the PA& and the BSRI categorize individuals identically. Their results, which indicated that 54% (52 of 96) of the Ss were categorized identically, fall between the percentages (52.5% and 55y0) of identical categorization established by the present studies.

It must be remembered, however, that there are major differences in the way in which the measures were conceptualized and constructed and that these dif- ferences are such that i t might be expected that an imperfect agreement would be obtained.

Although both instruments now employ identical category labels and cate- gorization procedures, Bem's (1974) conceptualization of the constructs a t the time of the BSRI was developed was very different. Initially, Bem employed only three sex role categories (Androgynous, Masculine, and Feminine) and saw androgyny as representing the balance of masculine and feminine characteristics. Hcr more recent conceptual revisions (Bem, 1975; Bem & Lenney, 1976) have revised categories and procedures for assignment to categories, but have not ex- amined any aspects of the instrurncnt that might have been affected by the changes in the conceptual framework.

Cateqorization Agreement GO 1

Differcnccs in response format also may contribute to disagrecincnt in cate- gorical assignmcnt. The BSRl prescnts a list of trait dcscriptors and asks individuals to ratc how characteristic the descriptor is of thcm, while the PA& presents a series of bipolar scales and asks individuals to indicate the point on thc scale that is most descriptive of themselves. This difference is more complex than may he initially evident in that both the judgment task and the assignment values diffcr. Additionally, differences in types of itcms included on parallel scales of the two instruments may further confound response diff erences and indicate that undcr- lying constructs are not identical,

Results of a recent study (Liberman & Gaa, 1980) indicatcd that a response pattern tendency exists for the BSRI, which may result in categorization assign- ments being influenced by factors other than identity. Although no data are avail- able that relatc to this aspect of the PA&, there is the possibility that these re- sponse pattern tendencies also may contribute to non-identical categorizations.

The levels of agreement demonstrated in the present studies are such that interpretation is difficult. It is evident that there is a great deal of overlap in the constructs as they are being measured. However, it is not clear that the under- lying constructs as conceived by Bern and Spence are identical in spite of agree- ment on terminology, scoring, and classification formats.

The findings further seem to indicate that care must be taken when assumptions are made about personality characteristics that pertain to an individual based upon his/her sex role assignment. One must be careful to indicate which instrument was used in making that assignment and also must realize that the use of different instruments for measuring sex role might well result in any given individual being assigned differing role categories.

In any case, the results indicate that great care must be taken in instances in which results of studies that employed differing instruments are compared or integrated. This concern is a major one when research is undertaken which examines identity differences across groups, but becomes even more crucial as the instruments may he used to label individuals as belonging to specific sex role categories. Ad- ditionally, the present study raises the issue of the degree of agreement between the BSRI and PA& with regard to the underlying conceptualization of sex role identification.

REFERENCE NOTE 1. WANQ-MCCARTHY, W. J., JOSE, P. E., & CROSBY, F. Effects of sex role self-conce t and gender

on verbal assertiveness. Paper presented at the annual meeting of the American k.ychologica1 krsociation, New York, September 1979.

REFERENCES BEM, S. L. The measurement of psychological androgyny. Journal of ConacUing and Clinical Psy-

chology, 3974, @, 155-162. BEM, S. L. Sex-Role adaptability: One consequence of psychological androgyny. Journal of Per-

sonality and SociCrl Psychology, 1975, 31, 834-643. BEM, S. L. On the utility of alternative procedures for assessin psychological androgyny. Journal

of ConauUing and Clznkl Psychology, 1977, @, 196-205. (a$ BEM, S. L. The Bem Sex Role Inventory (BSRI). In J. E. Jones & 3. W. Pfeiffer (Eds.), The 1975

Annvol Handbook for Group Fclcilitoora. La Jolla, Calif.: University Associates, 1977, 83-85. (b) BEM, S. L., & LENNEY, E. Sex-typing and the avoidance of cross-sex behavior. Journal of Per-

sonality and Social Peychology, 1976, 55, 48-54. LIBEWAN, D., & GAA, J. P. Response tendency on the Bern ,Sex Role Inventory. Journal of P8y-

SPENCE, J., & HEWREICH, R. Masculinaty and femininity. Austin: University of Texas Press, 1978. SPENCE, J., HELMREICH, R., & STAPP J. The Personal Attributes Questionnaire: A measure of

s x role stereot pes and masculiiity-femininity. JSAS Catalog of Seleeted Documents in Psy-

ROSENKRANTZ, P. S., VOQEL, S. R., BEE, H., BROVERMAN, I. K., & B R O V E ~ A N , D. M. Sex-role stereotypes and self-concepts in college students. Journal of ConauUing and Clink1 Psychology,

&logy, 1980, 106, 259-263.

&logy, 1974, K 43.

1968, S8, 287-295.