The “problem of number” supplement 1

Tate (2010) On-line supplement

Notes Concerning Statements in the Article

The term “genetics”—not “biology”—is used throughout the paper because evolved strategies have to be genetically transmitted to be subject to Darwinian natural selection. “Biology” can refer to non-genetic properties such as physiology, which is malleable within the lifetime of the organism based on environmental factors.

To further clarify the relative contribution logic for the self-reported sex variable, Buss (1998) argues that culture can impede the activation of evolved mechanisms, which would statistically attenuate effects based on self-reported sex. Yet, if making empirical predictions in a regression model, there should be a statistically detectable effect of the postulated evolved mechanisms using the proxy of self-reported sex, even if the effect itself might be weaker under some cultural conditions than under others.

Each discrete sex chromosome configuration and hormone insensitivity conditionlisted in the Introduction has an estimated incidence of between 1 in 2,000 and 1 in 500 (see Grovholt et al., 1998; Kruse et al., 1998; Thomas et al., 2001), and depending on the population considered can be as high as 1 in 90 (see Imperato-McGinley et al., 1979). Taking only 8 such instances as independent probabilities, one could derive a distribution of probabilities between 0.8% (if all average to 1 in 1,000) and 1.6% (if all average to 1 in 500). The rates become higher as additional genetic and hormone conditions are considered (see, e.g., Hines, 2003, for other conditions not listed here). It is well known in psychological research that accumulating systematic error variance (even at small rates) decreases construct validity and thus inferences about that construct’s effects (cf. Westen & Rosenthal, 2003).

For the results presented in Table 1, it is worth noting that Buss and Schmitt’s (1993) used the open-ended response procedure but analyzed pseudo-winsorized means (Pedersen et al., 2002), not ranks. The procedure is referred to as “pseudo-winsorized” because the Buss and Schmitt (1993) procedure does not involve determining the outliers from the distribution itself; instead, outliers in the Buss and Schmitt (1993) investigation were arbitrarily determined as values of 100 or above (see Pedersen et al., 2002). True winsorization requires the outliers to be determined from the distribution of obtained responses(Wilcox, 1996).

The effects in Studies 1 and 2a are described using the feminine identification anchor (rather than the masculine anchor) to foreshadow the results of later studies (specifically Study 2b).

The materials for Study 2b unfortunately did not contain a place to provide one’s age. Also, due to a printing error, some participants did not get the self-reported sex question on their materials, which accounts for the high number of “unspecified” responses in this sample.

References Unique to the Supplement

Buss, D. M., Larsen, R. J., Westen, D., & Semmelroth, J. (1992). Sex differences in

jealousy: Evolution, physiology, and psychology. Psychological Science, 3, 251-255.

Hines, M. (2003). Brain gender. New York: Oxford University Press.

Westen, D., & Rosenthal, R. (2003). Quantifying construct validity: Two simple

measures. Journal of Personality and Social Psychology, 84, 608-618.

Wilcox, R. R. (1996). Statistics for the social sciences. San Diego, CA: Academic Press.

Table 2 Supplement. As per editorial request, Table 2 has been decomposed to display correlations between predictor variables and the outcome separately for self-reported sexes.

Simple Correlations among Bipolar Gender Identity, Actual Number of Sexual Partners, and Desired Number of Partners for Study 2a Separated by Self Reported Sex

Measure / 1 / 2 / 3
1. Desired number of sexual partners / -- / .640** / .371*
2. Actual number of sexual partners / .425** / -- / .346*
3. Bipolar gender identity / .223 / .185 / --

Note: Self-reported sex is displayedas female in top diagonal, male in bottom diagonal. Gender identity was scored ascending from “very feminine” (1) to “very masculine” (7). Actual number of partners was coded in ascending order.Desired number of sexual partners was coded in ascending order based on ranked data. The correlation coefficients are therefore derived from Spearman’s rank correlation test (rs).*Correlations are significant at p < .05. **Correlations are significant at p < .01 (two-tailed).

Appendix: Bipolar Masculinity-Femininity Scale

Given the response options below, how would you describe yourself? (check only one)

○ Very feminine

○ Feminine

○ Somewhat feminine

○ Equally feminine and masculine

○ Somewhat masculine

○ Masculine

○ Very Masculine

○ Neither masculine nor feminine