Correction of errata in Study 1:

Lippa, R. A. (2005). Subdomains of gender-related occupational interests:

Do they form a cohesive bipolar M-F dimension? Journal of Personality, 73, 693-730.

 

Richard A. Lippa

California State University, Fullerton

 

 

 

 

 

 

 

 

 

Richard Lippa

Psychology Department

California State University, Fullerton

Fullerton, CA 92834

e-mail: rlippa@fullerton.edu

---

 

Study 1

Data

           The data came from Loehlin and Nichols' (1976) classic behavior genetic study of 839 same-sex twin pairs (351 male pairs and 488 female pairs).²   In the early 1960s these twins completed during their senior year of high school a number of self-report scales and questionnaires, including Holland's Vocational Preference Inventory (VPI; Holland, 1958).  The VPI contains a section that asks respondents to rate their dichotomous interest or noninterest in 160 occupations.  Because data from members of twin pairs were not statistically independent, analyses were conducted separately on first members and on second members of twin pairs.   Results were very similar for the two groups, and therefore results will be presented here only for the first members of twin pairs.

Results

           Occupational preference items that correlated with sex at a level of .15 or greater were identified as sex-linked, and by this criterion, 43 of 160 occupations were classified as masculine and 32 of 160 occupations were classified as feminine.  Two principal components analyses with varimax rotations were conducted, one on masculine items and the other on feminine items.   The analysis of masculine items yielded five interpretable factors, and the analysis of feminine items yielded four interpretable factors.³   Table 1 shows occupations that loaded .4 or higher on each of the five masculine factors in a five-factor solution and on each of the four feminine factors in a four-factor solution.  A small number of items that loaded highly on more than one factor were eliminated.

--------------------------------------------------

Insert Table 1 about here.

--------------------------------------------------

           Not all masculine and feminine subdomains were equally represented in the VPI list of 160 occupations.  For example, there were only three sports-related occupations, but there were thirteen blue-collar realistic occupations.  Because the VPI contains a large, diverse set of occupations, the number of occupations representing each factor probably roughly represents social reality -- e.g., there are many more blue-collar realistic occupations than sports-related occupations in the United States economy.  The items listed under each factor in Table 1 were used to create scales of gender-related occupational subdomains.  Scale scores were the mean of all items loading on a given factor.  The reliabilities (α) of each scale are presented in Table 1.

           An overall masculine occupations scale was computed by averaging individuals' preference for the 43 male-preferred occupations, and similarly an overall feminine occupations scale was computed by averaging individuals' preference for the 32 female-preferred occupations.  The reliability of the masculine occupations scale was .84 for all participants, .73 for males, and .75 for females, and the reliability of the feminine occupations scale was .86 for all participants, .79 for males, and .69 for females.

           Table 2 shows the intercorrelations of masculine and feminine occupational subdomains scales for men (above the diagonal) and for women (below the diagonal). Masculine subdomain scales tended to show both positive and negative correlations with one another, feminine subdomains tended to show positive or zero correlations with one another for men and positive and negative correlations with one another for women, and masculine subdomain scales tended to correlate negatively with feminine subdomain scales.  

--------------------------------------------------

Insert Table 2 about here.

--------------------------------------------------

           To reveal the structure implicit in the intercorrelations in Table 2, principal components analyses were conducted on masculine subdomain scales, feminine subdomain scales, and overall masculine and feminine occupations scales.⁴  These analyses were conducted for males and females combined and for males and females separately. Two principal components were extracted, solutions were left unrotated, and factor scores for participants were computed in all three analyses.  Because masculine and feminine subdomain scale items were selected because of their correlation with sex, it seemed likely that the first principal component in the combined-sex analysis would reflect a sex difference factor (i.e., a factor that tapped the variance shared in common by all items).  However, it was not guaranteed that this same component would emerge in analyses for males only and for females only.

            Figure 1 plots masculine and feminine subdomain scales in the two-dimensional spaces generated by the three principal components analyses.  The first principal component accounted for 43% of the variance in the combined-sex analysis, 35% of the variance in the males-only analysis, and 33% of the variance in the females-only analysis.  The revealed structure was very similar in all three analyses: masculine subscales tended to lie on one side of the first principal component, and feminine subscales tended to lie on the other side.  The overall masculine and feminine occupation scales, which served as marker variables, showed that the first principal component was indeed a "sex difference" factor in the combined-sex analysis.  However, this same factor was present in the analyses for males only and females only, and in these analyses the first principal component is best described as an M-F dimension.   It is worth noting that subdomain scales formed a circular arrangement in factor space that was consistent with Holland's hexagon model; however, unlike most investigations of Holland's model, all the occupational scales studied here were constructed to be gender-related.

--------------------------------------------------

Insert Figure 1 about here.

--------------------------------------------------

           To probe the consistency of the structure generated by the three principal components analyses, participants' factor scores from the combined-sex analysis were correlated with factor scores from the males-only analysis and from the females-only analysis.   In both cases, the correlation of first principal component factors scores was close to unity (r = .996 for males and r = .994 for females, p's < .001).   These extremely high correlations show that the first principal component identified an individual difference dimension that was virtually identical in all three principal components analyses, and they underscore the fact that the "sex differences" dimension in the mixed-sex analysis was the same as the M-F dimension in the males-only and females-only analyses.

           The second principal component in all three analyses can be interpreted as a version of the "Ideas-Data" dimension, with "educated realistic" and artistic occupations (which entail more creativity and theoretical thought) at one side of the second component, and "blue-collar realistic," "masculine business," "helping," and "children-related" occupations (which involve more practically oriented work) at the other side.   Participants' second principal component factors scores from the combined-sex analysis correlated highly with the second principal component factor scores from the males-only analysis (r = .93, p < .001) and from the females-only analysis (r = .98, p < .001).  The presence of this stable second factor in all three analyses suggests that there were systematic non-gender-related influences on the intercorrelations shown in Table 2.

           For males and females combined, masculine occupations were strongly negatively correlated with feminine occupations (r = -.87, p < .001).   Masculine and feminine occupations were also strongly negatively correlated for males only (r = -.77,  p < .001) and for females only (r = -.78,  p < .001) . 

Discussion of Study 1

           Analysis of the Loehlin and Nichols' data showed that coherent subdomains of masculine and feminine occupational interests could be identified from a large set of occupational preference items.  At the same time, structural analyses of masculine and feminine occupational subdomain scales showed that masculine subdomains scales tended to lie on one side of the first principal component, feminine subdomains tended to lie on the other side, and therefore that masculine and feminine subdomains tended to be bipolar opposites.

           The structure of occupational subdomain scales was highly consistent in all analyses -- for both sexes combined, for males only, and for females only.  In all cases, the first principal component was gender-related, reflecting sex differences in the combined sex analyses and individual differences in M-F in males-only and females-only analyses.   The second principal component appeared to tap a version of the "Ideas-Data" dimension, and the presence of this second dimension suggests that intercorrelations of masculine and feminine subdomains scales were systematically influenced by factors other than the masculinity or femininity of occupations – that is, when masculine scales (or feminine scales) sometimes correlated negatively with one another, it was because of differences on the second but not on the first principal component.  Participants' overall preference for masculine occupations showed a strong negative correlation with their overall preference for feminine occupations, and this was true whether correlations were computed for both sexes combined or within each sex.

---

Table 1

 

Masculine and Feminine Subdomain Items and Scale Reliabilities, Study 1

 

Masculine Subdomains

 

Blue-Collar Realistic	Educated Realistic	Masculine Business	Risk-Taking	Sports
Airplane        mechanic Auto mechanic Carpenter Construction       inspector Crane operator Factory foreman Locomotive       engineer Machinist Power shovel       operator Power station       Operator Tool designer Truck driver Wrecker        (building)	Aeronautical     design     engineer Chemist Electronic      technician Experimental     lab engineer Interplanetary     scientist Physicist	Bank examiner Banker Business     Executive Financial     Analyst Sales manager Stock and bond     salesman Tax Expert	Army general Army officer Aviator Counter-    intelligence    man Racing car     driver Test pilot	Professional     athlete Referee     (sporting     events) Sports promoter
α = .86, .83, .86	α = .82, .81, .78	α = .74, .77, .66	α = .71, .66, .62	α = .70, .73, .59

 

Feminine Subdomains

 

Note: Reliabilities (α's) are for all participants, males, and females.

---

Table 2

 

Intercorrelations of Masculine and Feminine Subdomain Scales, Study 1

 

 

 

	Blue-collar Realist	Educat-ed Realist	Masc Busi-ness	Risk-Taking	Sports	Helping	Creat Arts	Child-Related	Comm Arts
Blue-collar Realist	----	-.02	-.14*	.19***	.08	-.38***	-.43***	.05	-.14**
Educat-ed Realist	.02	----	-.29***	.05	-.15**	-.19**	-.20***	-.11*	-.11*
Masc Busi-ness	.21***	-.12**	----	-.12*	-.03	-.07	-.13*	-.10	-.06
Risk-Taking	.02	.17***	-.27***	----	-.01	-.24***	-.30***	-.21***	-.23***
Sports	.10*	-.09	-.03	.15**	----	-.06	-.21***	.04	-.23***
Helping	-.40***	-.18***	-.15**	-.19***	-.13**	----	.21***	.12*	.00
Creat-ive Arts	-.35***	-.17***	-.31***	-.10*	-.17***	-.10*	----	.00	.32***
Child-Related	.07	-.34***	.09*	-.33***	.07	.24***	-.16***	----	.00
Com-mercial Arts	-.24***	-.15**	-.17***	-.09*	-.08	-.06	.34***	-.12*	----

 

 

Note: Correlations for males are above the diagonal, and correlations for females are below the diagonal.

 

*  ----- 2-tailed p < .05

** -----2-tailed p < .01

*** --- 2-tailed p < .001

---

 

Figure 1 (Revised)

 

Plots of masculine and feminine subdomain scales in two-dimensional component space, for all participants (top), males (middle), and females (bottom) in Study 1