The idea Bem was arguing against
In the early 1970s most gender-role questionnaires worked like a single dial. Answer in the masculine direction and you scored low on femininity; answer in the feminine direction and you scored low on masculinity. One or the other, never both.
Sandra Bem thought that was wrong, and in 1974 she published a scale built to prove it. The Bem Sex-Role Inventory (BSRI) scores masculinity and femininity separately, which leaves room for people who are high on both. She called that combination androgynous, and argued it was an advantage: someone who can be assertive when a situation calls for it and gentle when it calls for that has a wider range of responses available than someone locked into one set.
How the 60 items are built
The inventory has 60 adjectives or short phrases, split into three groups of 20.
| Group | Items | What it holds |
|---|---|---|
| Masculinity | 1 to 20 | Traits judged more desirable for men in 1970s US culture, such as "independent", "assertive", "willing to take risks" |
| Femininity | 21 to 40 | Traits judged more desirable for women, such as "affectionate", "gentle", "sensitive to the needs of others" |
| Social desirability | 41 to 60 | Neutral traits desirable for anyone, such as "helpful", "sincere", "reliable". Used to check for response bias |
You rate how well each one describes you from 1 ("never or almost never true") to 7 ("always or almost always true"). Each dimension score is the average of its 20 items, so every score lands between 1 and 7. Nothing is reverse-scored.
Bem chose the masculine and feminine items by asking judges to rate how desirable each trait was for a man versus a woman. That method is the scale's weak point, and Bem knew it. The items describe a 1974 American stereotype, which is exactly what makes the long-term trend data interesting.
Turning two scores into a category
Bem's own procedure used a t-ratio: take the difference between your femininity and masculinity averages, scale it, and compare the result against a t value of 2.025 at 38 degrees of freedom, the 5% cutoff. Anything inside that band counts as androgynous, meaning the two scores are close enough that neither dominates.
The simpler version, and the one this site uses, compares the two averages directly:
| Pattern | Category |
|---|---|
| Masculinity clearly above femininity | Masculine |
| Femininity clearly above masculinity | Feminine |
| Both above the midpoint and close together | Androgynous |
| Both below the midpoint | Undifferentiated |
One caution before you read anything into a label. Bem's cutoffs were relative to whichever sample she was studying, and later researchers have pointed out that the exact boundary between categories is somewhat arbitrary. The two dimension scores carry more information than the category name does.
Reliability
In Bem's two original samples, alpha was .86 for masculinity in both, .80 and .82 for femininity, and .75 and .70 for the social desirability group. The androgyny difference score came in at .85 and .86.
Four-week retest correlations were .90 for masculinity, .90 for femininity, .93 for the androgyny score and .89 for social desirability.
The independence claim also held up in the data. Masculinity and femininity correlated close to zero: .11 for Stanford men and -.14 for Stanford women, then -.02 and -.07 in the second sample. If the two dimensions were opposite ends of one scale, those correlations would be strongly negative. They are not.
Social desirability correlated moderately with each dimension on its own, but the androgyny index correlated near zero with it (.04 to .08), which suggests the index is not simply picking up people who say yes to everything.
Norms from the original samples
| Group | Masculinity mean (SD) | Femininity mean (SD) |
|---|---|---|
| Stanford men (n = 444) | 4.97 (0.67) | 4.44 (0.55) |
| Stanford women (n = 279) | 4.57 (0.69) | 5.01 (0.52) |
| Foothill men (n = 117) | 4.96 (0.71) | 4.62 (0.64) |
| Foothill women (n = 77) | 4.55 (0.75) | 5.08 (0.58) |
Men averaged higher on masculinity and women higher on femininity, and the differences were large enough to be reliable. But the gaps are about half a point on a seven-point scale, not the mirror-image pattern the old single-dial model predicted. Among Foothill men, androgynous was the most common category at 44%; at Stanford it was a close second to masculine, 34% against 36%. For women the order flipped, with feminine first and androgynous second.
Shorter versions and what changed over time
Two shortened forms are in circulation. Bem herself published a 30-item version in 1981 with 10 items per dimension, and Campbell, Gillaspy and Thompson (1997) later confirmed its reliability in a confirmatory factor analysis, with alpha of .82 for masculinity and .89 for femininity. Mateo and Fernández (1991) built a 12-item version from a Spanish student sample, six items per dimension, which has been validated in several countries since.
Donnelly and Twenge (2017) pooled BSRI studies across four decades to see whether the scores moved. In the 1993 to 2012 window (34 samples, N = 8,027), women's femininity scores fell noticeably, with an effect size of d = -.26, while women's masculinity and both of men's scores stayed flat. Extending back to 1974 (94 samples, N = 24,801) adds a rise in women's masculinity of d = .23. Men's scores barely moved in either window.
Two readings are possible. Either American women became less willing to describe themselves with traditionally feminine traits, or the items themselves stopped matching how people think about gender. Bem's trait-selection method makes the second explanation hard to rule out.
Take the test
The CheckPsych version uses all 60 items and takes about 15 minutes. You get an average for each of the three groups, plus a category based on how your masculinity and femininity averages compare.
References
Bem, S. L. (1974). The measurement of psychological androgyny. Journal of Consulting and Clinical Psychology, 42(2), 155-162.
Bem, S. L. (1981). Bem Sex-Role Inventory: Professional manual. Consulting Psychologists Press.
Mateo, M., & Fernández, J. (1991). La dimensionalidad de los conceptos de masculinidad y feminidad. Investigaciones Psicológicas, 9, 95-116.
Campbell, T., Gillaspy, J. A. Jr., & Thompson, B. (1997). The factor structure of the Bem Sex Role Inventory (BSRI): Confirmatory analysis of long and short forms. Educational and Psychological Measurement, 57, 118-124.
Donnelly, K., & Twenge, J. M. (2017). Masculine and feminine traits on the Bem Sex-Role Inventory, 1993-2012: A cross-temporal meta-analysis. Sex Roles, 76, 556-565.
Carver, L. F., Vafaei, A., Guerra, R., Freire, A., & Phillips, S. P. (2013). Gender differences: Examination of the 12-item Bem Sex Role Inventory (BSRI-12) in an older Brazilian population. PLoS ONE, 8(10), e76356.
Ahmed, T., Vafaei, A., Belanger, E., Phillips, S. P., & Zunzunegui, M-V. (2016). Bem Sex Role Inventory validation in the International Mobility in Aging Study. Canadian Journal on Aging, 35(3), 348-360.
Disclaimer: this article and the BSRI are for general education and self-reflection. They are not a diagnosis. The inventory measures self-description against traits that were rated in the 1970s, so a category label says more about cultural stereotypes than about who you are. Scores are not a measure of mental health.