Wie man Geschlechterunterschiede mit Statistik verschwinden lässt

Eine interessante Besprechung zur „Gehirnmosaik Studie“ (eine Studie, die einem aus Genderkreisen sicherlich noch häufiger entgegengehalten wird) findet sich hier:

In der Einleitung stellt er erst einmal ein paar Zahlen dar und erläutert was aus diesen folgt:

Sexual diversity researchers sometimes focus on the ways men and women differ in their affects, behaviors, and cognitions. There are many reasons for this. One is that knowing about psychological sex differences is actually quite important for improving men’s and women’s sexual health (de Vries & Forger, 2015; Lawrence & Rieder, 2007; Morrow, 2015). But determining how “big” a psychological sex difference is, and even whether it is a “real” and “important” sex difference, is not all that simple. How do we know when a sex difference is so? What metric should we use?

Probably the most common metric for evaluating a psychological sex difference involves looking at the “effect size” of the difference. Effect sizes are often expressed in terms of a “d” statistic. A positive d value of +0.50 typically indicates men are moderately higher on a psychological measure, a negative value of -0.50 indicates women are moderately higher. Below are some d values of varying sizes reliably observed in empirical studies of men and women:

d = -0.20 has been observed for sex differences in trust (Feingold, 1994). This size of sex difference is considered “small” and indicates 58% of women are higher than average man in trust (based on Cohen’s U₃).

d = +0.50 has been observed for sex differences in spatial rotation skills (Silverman et al., 2007). This size of sex difference is considered “moderate” and indicates 69% of men are higher than average woman in spatial rotation skills.

d = +0.80 has been observed for sex differences in physical aggression (Archer, 2004). This size of sex difference is considered “large” and indicates 79% of men are higher than average woman in physical aggression.

d = -1.00 has been observed for sex differences in tender-mindedness (Feingold, 1994). This size of sex difference indicates 84% of women are higher than average man in tender-mindedness.

d = +2.00 has been observed for sex differences in throwing distance among children (Thomas & French, 1985). This size sex difference indicates 98% of boys throw farther than the average girl.

Was sagen nun diese Unterschiede aus:

Which of these sex differences should qualify as a “real” sex difference? All of them, none of them, only those beyond ±0.20? How do we know when sex differences are so?

As I have noted in earlier posts (see here and here), sex differences with larger d values are not “more real” than smaller-sized sex differences. All men do not have to be taller than all women for a sex difference in average height to be “real” and have important social consequences (Prentice & Miller, 1992). Sex differences with larger dvalues also are not necessarily more attributable to evolution or biology, and smaller sex differences are not necessarily more cultural or due to learning than larger sex differences. Nature does not work that way.

In psychological science, most researchers who use the d statistic have concluded (so far, new evidence can always change our views) that many psychological sex differences are small, some are moderate, and a few are large. Those are just the facts. And these facts make sense from an evolutionary perspective, as sex differences are expected only in those domains where men and women have faced different Darwinian pressures (Buss, 1995; Okami & Shackelford, 2001).

Bereits kleine Unterscheide können sich also erheblich auswirken. Ich hatte dazu auch schon einmal etwas anhand des „Cola-Pepsi“ Beispiels geschrieben.

Die weitere Forschung wird dann wie folgt dargestellt:

Empirically, Hyde (2014) reviewed 100s of studies on psychological sex differences and concluded there are relatively moderate to large sex differences in spatial rotation abilities, agreeableness, sensation seeking, interests in things versus people, physical aggression, certain sexual behaviors (e.g., masturbation and pornography use), and attitudes about casual sex. Smaller sex differences exist in measures of gregariousness, reward sensitivity, conscientiousness, negative affectivity, relational aggression, and self-esteem.

Some of these sex differences persisted or varied in size across cultures and time periods, others did not (see Lippa, 2009). Ellis (2011) examined psychological sex differences across cultures and found evidence 65 apparently universal sex differences, prevalent across all cultures. These sex differences were universal across all cultures, with not a single replication failure across 10 studies (probably too tough a criterion leading to an under-reporting of actual psychological sex differences).

Schmitt (2014) found most psychological sex differences are larger in more genderegalitarian cultures, and smaller in more patriarchal cultures, directly refuting the view psychological sex differences are learned because they emerge more strongly in patriarchal cultures that have more extreme sex-role socialization (see also Udry, 2000). Indeed, psychological sex differences seem to emerge most strongly in the most gender-egalitarian of cultures (e.g., Northern Europe), cultures that allow people to pursue the lived experiences they most want to pursue (see also, Schwartz & Rubel-Lifschitz, 2009).

 

Da finden sich einige Studien, die hier auch schon besprochen worden sind und die ein entsprechendes Bild bestätigen, nachdem biologische Unterschiede vorhanden sind, die sich auf das Verhalten auswirken.

Die eigentliche Kritik lautet dann wie folgt:

The Joel et al. (2015) study was certainly impressive, involving a remarkable collection of datasets across a wide variety of variables. Unfortunately, the specific techniques Joel et al. (2015) used for determining whether a sex difference is “real” were highly problematic. For instance, Joel et al. defined a sex difference as occurring only when it had “a high degree of internal consistency in the form of the different elements of a single brain (e.g., all elements have the “male” form).” (Joel et al., 2015, p. 1). That is, for a sex difference in the brain to “really” exist, men must have relatively masculine brains in each and everyrespect, and women must have relatively feminine brains in each and every respect. Otherwise, no sex difference. Really?

Same thing for psychological differences. In one analysis, a sex difference was only deemed to occur when men were “stereotypical” in all gender-stereotypical activities (i.e., men were in the top third of the 5 masculine activities of boxing, construction, playing golf, playing video games, and watching porn; and men were also in the bottom third of the 5 feminine activities of scrapbooking, taking a bath, talking on the phone, watching talk shows, and using cosmetics). Conversely, women had to be the bottom third of all masculine and top third of all feminine activities to count toward the existence of a sex difference. Not one deviation from gender-stereotypicality was allowed. Not one.Otherwise, no sex difference. Using this incredibly extreme criterion, only 1.2% of men and women counted as demonstrating a “sex difference” in gendered behavior (i.e., “In other words, even when considering highly stereotypical gender behaviors, there are very few individuals who are consistently at the ‘female-end’ or at the ‘male-end’” Joel et al., 2015, p. 4).

This criterion for deciding whether a sex difference exists is, quite frankly, rather strange. It is as though these authors purposefully designed their statistical analyses to allow them to declare, „these aren’t the sex differences you are looking for“ (articulated in the voice of Sir Alec Guinness).

Es freut mich, dass sich das etwas mit meiner Kritik deckt, denn auch ich fand diese Bündelung „Ein Unterschied liegt nur dann vor, wenn die Gehirne in jeder Hinsicht dem Extrem entsprechen“ arg konstruiert und auf ein bestimmtes Ergebnis abgestellt.

Weiter im Text:

Such a technique may be informative for evaluating the strawman view that all sex differences must entirely and completely co-vary together if it is a “real” sex difference (which is very unlikely given what sexual diversity researchers know about developmentalneuroscience, epigenetics, strong modularity, and so forth; see here). Logically, the Joel et al. (2015) view on real sex differences would even argue men’s and women’s faces are not really different from each other because not all men’s faces are entirely masculine and not all women’s faces are entirely feminine (and not all facial attributes perfectly co-vary together). This last finding would be interesting, implying that not all facial features are masculinized or feminized at the same time (sensitive periods vary; see Sisk & Zehr, 2005) or via the same mechanisms (e.g., organizational, activational, or direct genetic effects; McCarthy & Arnold, 2011; Ngun et al., 2011). But this interesting point should not be mistakenly viewed as suggesting men’s and women’s faces are, on average, identical. They are not.

Das wird dort noch an einer Grafik näher erläutert. In der Tat ist es ein ziemlicher Strohmann, auf den da eingedrescht wird, denn das was sie dort wiederlegen wird eben so gar nicht behauptet.

Every human being’s degree of masculinity and femininity is, indeed, a mosaic. No one is completely masculine or completely feminine in every possible way, and this is true of our brains, our bodies, and our behaviors. As Kinsey noted long ago, humans are not discrete creatures, “The world is not to be divided into sheep and goats.” These insights are a far cry, though, from the conclusion that psychological or neurological sex differences are “extremely rare” (Joel et al., 2015, p. 5) and that we should entirely abandon sex as a variable of interest in our sexual science research programs. That conclusion is, well…the word abracadabra comes to mind.

A much more useful tool for evaluating the degree (and the existence) of sex differences across a large number of variables is to use multivariate statistics designed for this very purpose (Del Giudice, 2013). Doing so has led sexual diversity researchers to conclude there are very large psychological sex differences. Del Giudice, Booth, and Irwing (2012), for instance, utilized a multivariate method to examine several personality dimensions simultaneously (controlling for collinear overlap among dimensions). From a multivariate perspective, lots of small ds can be “additive” and create “planetary-sized” sex differences when examined together. Del Giudice et al. (2012) found exactly that, with less than 10% overlap in men’s and women’s personality traits when looking across 15 personality dimensions simultaneously.

Del Giudice (2012) hatte ich hier besprochen. Viele kleine Unterschiede bewirken eben insgesamt große Unterschiede, genauso wie jemand der Hamburger etwas lieber als Gemüse, Bier etwas lieber als Wein und legere Kleidung lieber mag als sich fein herauszuputzen eben eher in ein anderes Restaurant gehen wird als jemand, bei dem dies andersherum ist.

Similarly, Conroy-Beam et al. (2015) examined sex differences across 18 mate preferences simultaneously and found only 23% overlap in men’s and women’s overall mate preference distributions. If one can think in terms of multiple regression and multidimensional space, this approach is a fairer evaluation of whether men and women differ in a particular domain (such as the domain of “personality” or “mate preference” psychology).

Diese Studie hatte ich hier besprochen.

Once again, it is probably not true that psychological sex differences should be described as extremely rare or nonexistent. Men and women are members of the same species, but psychologically there are important differences that should not be overlooked if we are to maximize everyone’s medical, mental, and sexual health (Lawrence & Rieder, 2007). As esteemed researcher Margaret McCarthy(link is external) has rightly noted about this new study, “By studying male versus female brains, we have a great tool for exploring the biological basis of those differences…[Joel’s] call for us to abandon the monikers of male and female or men and women I think is too far.”

In der Tat. Es ergibt sich schlicht nicht aus ihrer Studie, auch wenn sie in den Medien leider so gelesen wird.

Too far, statistically contrived, and scientifically dangerous. The views of Joel et al. (2015) represent the polar opposite of where sexual science is going (Cahill, 2014; Ruigrok et al., 2014), as evidenced by the National Institutes of Health’s recent declaration that biological sex differences are so large and consequential that both males and females must be included in studies involving cells and non-human animals and in preclinical research (seehere(link is external)). Let us hope sexual scientists keep their eyes on where science is really going, rather than the abracadabra analyses that misleadingly convey there are no important sex differences. Don’t be fooled, these are the droids we’re looking for.

Sehr deutliche Worte. Und das zu recht.