Gründe für Geschlechterunterschiede in STEM: Geschlechtsunterschiede bei kognitiven Fähigkeiten (Teil 2)

Eine interessante Studie bespricht eine Vielzahl von Gründen, warum sich Geschlechterunterschiede im STEM-Bereich ergeben.

It is a well-known and widely lamented fact that men outnumber women in a number of fields in STEM (science, technology, engineering and maths). The most commonly discussed explanations for the gender gaps are discrimination and socialization, and the most common policy prescriptions target those ostensible causes. However, a great deal of evidence in the behavioural sciences suggests that discrimination and socialization are only part of the story. The purpose of this paper is to highlight other aspects of the story: aspects that are commonly overlooked or downplayed. More precisely, the paper has two main aims. The first is to examine the evidence that factors other than workplace discrimination contribute to the gender gaps in STEM. These include relatively large average sex differences in career and lifestyle preferences, and relatively small average differences in cognitive aptitudes – some favouring males, others favouring females – which are associated with progressively larger differences the further above the average one looks. The second aim is to examine the evidence suggesting that these sex differences are not purely a product of social factors but also have a substantial biological (i.e. inherited) component. A more complete picture of the causes of the unequal sex ratios in STEM may productively inform policy discussions.

Quelle: Men, women and STEM: Why the differences and what should be done?

Die Einteilung in der Studie ist wie folgt:

1. Sex differences in preferences and priorities
2. Sex differences in cognitive aptitudes
3. Sex differences in variability
4. Bias and discrimination in the workplace
5. Policy implications
6. Levelling the playing field vs. equalizing sex ratios
7. Conclusion: Many factors at play

Ich dachte ich gehe diese Punkte mal einzeln durch, weil da viel interessantes dabei ist

Heute also:

Sex differences in cognitive aptitudes

Los geht es:

Sex differences in occupational preferences are not the only reason we might expect uneven sex ratios in certain STEM fields, even if discrimination were entirely removed from the picture. A second, more controversial suggestion is that STEM gender gaps are due in part to average sex differences in a small subset of STEM-relevant cognitive capacities, which result in somewhat more men than women having a suitable profile of aptitudes for working in some STEM fields.

Das ist in der Tat ein zumindest außerhalb der passenden medizinischen und biologischen Wissenschaften höchst politisch unkorrektes Thema: Solche Unterschiede dürften eigentlich nach Auffassung vieler nicht vorliegen und eine solche Ansicht zu äußern ist sexistisch. Für Leser dieses Blogs ist es hingegen wenig überraschend.

This statement could easily be misconstrued, so it’s important to be absolutely clear what is meant.

Ja, da wird gerne vieles falsch verstanden und unterstellt.

First, the claim is not that men perform better than women in every cognitive domain. On the contrary, men perform better in some domains, whereas women perform better in others. The best-known examples are that men score higher than women on most tests of spatial ability, whereas women score higher than men on most tests of language ability, including verbal comprehension, reading and writing (Halpern, 2012; Reynolds et al., 2015). Note that, although there are reliable sex differences in various specific cognitive abilities, most commentators agree that there are no average differences in general cognitive ability in representative samples (Halpern, 2012).

Das wird in der Tat gerne übersehen. Es gibt eine Vielzahl von Bereichen, in denen Frauen im Schnitt besser sind als Männer und viele spielen in klassische KIischees herein, etwa das der Mann seine Frau fragen muss, wo er die Schlüssel hingelegt hat.

Second, even in areas where men do perform better, the claim is not that all men – or even most – perform better than all or most women. As with occupational preferences, members of both sexes vary enormously in every cognitive aptitude, and the distribution for men overlaps almost entirely with that for women (Hyde, 2005). However, for some aptitudes, the distribution for one sex is shifted somewhat to the right of that for the other, such that the average score for the former is somewhat higher. In saying this, it is worth stressing that the average score does not describe all members of the group, or even the typical member, but merely represents the central tendency within a broad array of scores. Most people fall above or below the average.

Eben der Unterschied im Schnitt. Oder auch Normalverteilungen, die abweichende Mittelwerte haben, andere Ausformungen, aber üblicherweise überlappende Träger.

Third, the claim is not that these cognitive sex differences are especially large. On the contrary, at the centre of the distribution, they tend to be quite small (Hyde, 2005; Stewart-Williams & Thomas, 2013a, 2013b). The only reason they matter at all is that even small differences at the mean are associated with progressively larger differences the further from the mean one looks (see Figure 1). For jobs requiring normal-range abilities – including many lower-level STEM jobs – cognitive sex differences are likely to make little difference: the pool of potential female candidates is similar in size to the pool of potential males. However, for jobs requiring exceptional abilities, the sex ratio of possible candidates may be somewhat skewed in favour of one sex or the other – even when the relevant sex differences in the general population are small or even negligible (Halpern, 2012; Halpern et al., 2007; Steven Pinker, 2002).

Das ist ein Punkt, der in der Tat häufig nicht erkannt wird: Bei bereits einer leichten Verschiebung sind die Unterschiede in der „Spitzengruppe“ bereits sehr groß. Wenn also ein Bereich hohe Anforderungen an bestimmte Fähigkeiten stellt, dann wirkt sich der kleine Unterschied sehr stark aus.

Bei geringen Anforderungen hingegen kann der Unterschied evtl nicht deutlich werden.

Das kann man aus meiner Sicht zb beim Gewichtheben gut darstellen (auch wenn da ein großer Unterschied im Schnitt zwischen den Geschlechtern vorliegt, aber es ist dafür sehr anschaulich).

Wenn man die Unterschiede ermitteln will indem Männer und Frauen ein 2 Kilogewicht anheben sollen, dann wirkt sich der Kraftunterschied nicht aus und man könnte zu dem Ergebnis kommen, dass ein Kraftunterschied nicht vorliegt. Vielleicht heben Frauen das Gewicht bei der Messung dann sogar (zufälligerweise) etwas schneller auf und man könnte das als „Widerlegung“ anführen. Es ist aber tatsächlich schlicht kein sehr gut durchdachter Test.

Wiederholt man den Test mit 60 kg wird der Kraftunterschied hingegen schon sehr schnell wesentlich deutlicher. Widerholt man ihn mit Sportlern aus dem Bereich dann werden immer weniger Frauen bestimmte Gewichte heben können, bis schließlich nur noch Männer übrig sind.

Zur weiteren Verdeutlichung noch dieses Zitat:

Just in the single year 2017, Olympic, World, and U.S. Champion Tori Bowie’s 100 meters lifetime best of 10.78 was beaten 15,000 times by men and boys.  (Yes, that’s the right number of zeros.)

Der Record der besten Frau wurde also in einem Jahr 15.000 mal von Männern unterboten. Aber eben von keiner Frau.

Es folgt dann eine interessante Grafik, die das noch einmal verdeutlicht.

Figure 1. For two groups, A and B, for which the average score on a normally distributed variable is higher for the former than the latter, the ratio of A-to-B gets progressively larger the further above the mean one looks. For both groups, fewer and fewer people occupy each segment above the mean, with the percentage decline getting larger with every step. Thus, the number of people falling between 1 and 2 standard deviations above the mean is 40% of that falling between the mean and 1 standard deviation, whereas the number of people falling between 2 and 3 standard deviations above the mean is only 15% of that falling between 1 and 2 standard deviations. This is the case for both groups. However, because the lower-scoring group (B) starts this accelerating decline before the higher-scoring group (A), the percentage decline for the lower-scoring group is always larger than that for the higher-scoring one. The net result is that the ratio of A-to-B gets progressively larger for scores above the mean. Meanwhile, for the equal-but-opposite reason, the ratio of B-to-A gets progressively larger for scores below the mean. The difference depicted in Figure 1 is large – a full standard deviation between the mean for each group – and thus the skew quickly gets large as well. However, even small differences at the mean may be associated with high levels of skew at sufficiently extreme levels of any trait.

Hier sieht man, dass bei einer Abweichung von 1.0 (eine große Abweichung) das Verhältnis von Männer und Frauen schnell sehr sehr unterschiedlich wird. Bei den besten 2,1% schon 21:1. Und solche Abweichungen finden wir durchaus in relevanten Bereichen, etwa dem im ersten Teil besprochenen „Dinge-Personen“-Unterschied etc.

Fourth, the claim is not that women lack the cognitive talents to make it in STEM. Most people lack the cognitive talents, and of those who do possess them, some are men and some are women. The claim is simply that, because of small average differences in a small subset of abilities, somewhat more men than women may be suited to work in some areas of STEM – and for the same reason, somewhat more women than men may be suited to work in others.

Auch das geht gerne unter. Klar kann ich zB die Weltmeisterin der Frauen im Laufen nicht ansatzweise schlagen (siehe obiges Beispiel). Und die meisten Männer werden von ihr abgezogen werden. Aber innerhalb der Spitzengruppe der Männer und Frauen ist sie keinerlei Konkurrenz für die Männer.

The relevant differences

With these important qualifications in mind, let us now consider some of the sex differences in cognitive aptitudes that may help to explain the uneven sex ratios found in certain STEM disciplines.

Hier geht es also um die Unterschiede, die dann in den STEM-Fächern relevant sein könnten.

Spatial abilities

The first concerns sex difference in spatial abilities. As mentioned, the average score on most spatial tests is moderately higher for men than for women (visuospatial ability: d = 0.48; spatial visualization: d = 0.23; mental rotation: d = 0.66; Archer, 2019). This is a well-established finding (Voyer et al., 1995), especially with respect to mental rotation (Maeda & Yoon, 2013), and it seems reasonable to think that it might be part of the reason that somewhat more men than women gravitate to fields that require above-average spatial abilities – fields such as physics and engineering. Consistent with this assessment, more than 50 years of research in educational psychology indicates that spatial ability is indeed an important predictor of STEM success (Wai et al., 2009).

Zum räumlichen Denken hatte ich hier einige Artikel:

• Geschlechtsunterschiede beim räumliche Denken
• Räumliches Denken als Voraussetzung für die Naturwissenschaften
• Mathematik, negative Einstellungen zur Mathematik und räumliches Denken
• Räumliches Denken und mathematische Fähigkeiten
• Sexuelle Orientierung und Neurokognitive Eigenschaften
• Geschlechterunterschiede beim Zusammenbau von Ikeamöbeln
• Technisches Verständnis, Berufwahl und Testosteron

Einige Beziehen sich auch auf die nachfolgenden Eigenschaften. In der Tat sind viele der Eigenschaftsunterschiede im Schnitt  gut nachgewiesen und in einer Vielzahl von Studien immer wieder ermittelt worden.

Mechanical reasoning

A second, less widely known cognitive sex difference is that, on average, males as a group score higher than females as a group on tests of mechanical reasoning – that is, tests of the ability to solve problems involving mechanical principles and physical laws (Flores-Mendoza et al., 2013; Lemos et al., 2013). Unlike most cognitive sex differences, this one is rather large even at the mean (d = 0.8–1), and thus larger still at the right-hand tail of the distribution (Hedges & Friedman, 1993; Hedges & Nowell, 1995). As with the spatial sex difference, the sex difference in mechanical reasoning has clear implications for the gender composition of fields such as engineering and physics.

Es dürfte schwer sein bestimmte Fähigkeitsbereiche genau von einander abzugrenzen, einfach weil wir natürlich keinen Gehirnbereich haben, der für die Lösung moderner technischer Fragestellungen vorgesehen ist, evolviert sind solche Eigenschaften vielleicht eher für die Fähigkeit bestimmte Werkzeuge herzustellen oder Speere so zu werfen, dass sie das Ziel treffen oder etwas in der Art und sie werden eben heute für andere Bereiche ermittelt. Wenn „Mechanical Reasoning“ beispielsweise in dieser Hinsicht nur ein „schwierigerer Fall des räumlichen Denkens“ wäre dann wäre auch damit zu rechnen, dass der Geschlechterunterschied dort größer ist, genauso wie bei der 100 Kilohantel der Unterschied größer ist als bei der 10 kg Hantel, auch wenn vielleicht beide gleiche Muskelgruppen beanspruchen.

Mathematical ability

A third difference relates to mathematical ability; in this case, however, the findings are less straightforward. Sex differences in mathematical ability are highly variable: in some nations, boys do better on average; in others, girls do; and in others still, there are no overall sex differences (Else-Quest et al., 2010; Hyde et al., 2009; Stoet et al., 2016).

Was dann auch erklärt, warum Mathematik eine andere Geschlechterquote hat als etwas Maschinenbau.

At the same time, though, there are several maths-related differences which may be relevant to STEM outcomes.
First, whereas females tend to do better in tests of mathematical computation (d = 0.14), at least in childhood, males tend to do better – at least from adolescence – in tests of mathematical reasoning or problem solving (d = 0.3; Hyde et al., 1990; see also Benbow, 1988; Halpern, 2012).

Wie man sieht sind hier relativ geringe Unterschiede vorhanden, aber dennoch können sich diese entsprechen auswirken.

This is a potentially important finding because, as Hyde et al. (1990) note, mathematical reasoning ‘is critical for success in many mathematics-related fields, such as engineering and physics’ (p. 151). Second, despite small and cross-culturally variable differences in average scores, males tend to outnumber females at the highest levels of mathematical performance. Wai et al. (2010) report, for instance, that in the US since the 1990s, the male-to-female ratio among the top 1% on the SAT-Math has been around 1.1-to-1, among the top 0.5% has been around 1.6-to-1, and among the top 0.01% has been around 4-to-1. Similar ratios have been found in other countries and using other tests (Baye & Monseur, 2016; Makel et al., 2016; Reilly et al., 2015). The upshot is that somewhat more men than women are likely to have the mathematical acumen to work in maths-intensive STEM fields – even though, once again, many women do too, and most men do not.

Auch hier sind die Wechselwirkungen mit etwas den Interessen zu berücksichtigen, aber auch mit anderen Fähigkeitsausprägungen, etwa das es zwar mehr dumme Männer gibt, dafür aber auch mehr sehr intelligente etc.

Furthermore, several lines of evidence suggest that absolute levels of maths ability are not all that matters when it comes to occupational outcomes. The relative balance between people’s maths ability and their language ability matters too, and the sexes differ in this respect. First, more males than females exhibit ‘maths tilt’ (maths > verbal), whereas more females than males exhibit ‘verbal tilt’ (verbal > maths; Coyle et al., 2015; Lubinski et al., 2001; Stoet & Geary, 2018; Wai et al., 2018). Importantly, maths-tilt predicts interest in STEM, whereas verbal-tilt predicts interest in the humanities – even for people with high-level maths abilities.

Die Studie von Stoet und Geary hatte ich hier besprochen

Second, among the minority of people who possess exceptional mathematical abilities, the women are more likely to possess exceptional language abilities as well. This means that mathematically gifted women have more vocational options than their male counterparts, and consequently that fewer mathematically gifted women end up pursuing a STEM career (Wang et al., 2013; see also Breda & Napp, 2019). To the extent that this explains the gender gap in maths-intensive fields, the gap results not from mathematically gifted women having fewer options, but rather from them having more.

Aus der Studie von Breda:

Gender differences in math performance are now small in developed countries and they cannot explain on their own the strong underrepresentation of women in math-related fields. This latter result is however no longer true once gender differences in reading performance are also taken into account. Using individual-level data on 300,000 15-y-old students in 64 countries, we show that the difference between a student performance in reading and math is 80% of a standard deviation (SD) larger for girls than boys, a magnitude considered as very large. When this difference is controlled for, the gender gap in students’ intentions to pursue math-intensive studies and careers is reduced by around 75%, while gender gaps in self-concept in math, declared interest for math or attitudes toward math entirely disappear. These latter variables are also much less able to explain the gender gap in intentions to study math than is students’ difference in performance between math and reading. These results are in line with choice models in which educational decisions involve intraindividual comparisons of achievement and self-beliefs in different subjects as well as cultural norms regarding gender. To directly show that intraindividual comparisons of achievement impact students’ intended careers, we use differences across schools in teaching resources dedicated to math and reading as exogenous variations of students’ comparative advantage for math. Results confirm that the comparative advantage in math with respect to reading at the time of making educational choices plays a key role in the process leading to women’s underrepresentation in math-intensive fields.

Frauen haben also mehr Optionen und es dürfte auch dazu kommen, dass viele Berufe aus dem Bereich „Sprache“ auch eher mit Leuten zu tun haben und viele aus dem Bereich „Mathe“ eher mit Dingen

Systemizing and empathizing

A final cognitive difference bearing on the question of STEM sex ratios relates to Baron-Cohen’s (2003) distinction between systemizing and empathizing. Systemizing refers to the desire and ability to understand or build ‘systems,’ including mechanical systems like cars, physical systems like galaxies and abstract systems like logic and mathematics. Empathizing, in contrast, refers to the desire and ability to understand people: their thoughts, their desires, their feelings.

Männer schneiden also mit einem großen Unterschied in den Werten im Schnitt besser im Bereich der Fähigkeit ab, Systeme zu verstehen und aufzubauen (Woraus vielleicht auch der eher lösungsorientierte Ansatz vieler Männer kommt) während Frauen eher im Schnitt besser abschneiden beim verstehen von den Gedanken, deren Wünschen und deren Gefühle (weswegen es ihnen im Schnitt wichtiger sein kann, dass ihnen jemand zuhört als das er eine Lösung für ihre Probleme findet). Natürlich dürfte sich das dann auch auf den Wunsch niederschlagen in einem Bereich zu arbeiten, in dem man systembedingte Probleme lösen muss (zB „Optimierung von Lieferketten“ oder „Entwicklung neuer Maschinen in technischer Hinsicht“) bzw in dem man eher mit Leuten arbeitet (zB „Personalabteilungen und das Aussuchen der passenden Mitarbeiter“ oder Psychologie).

Virtually every human being possesses both abilities to a greater or lesser extent. On average, though, men score higher than women on tests of systemizing (d = 1.21), whereas women score higher than men on tests of empathizing (d = 0.87–0.91; Archer, 2019; Baron-Cohen et al., 2001). As usual, these differences are not overwhelmingly large among the majority in the normal range (see Stewart-Williams & Thomas, 2013a, 2013b, on the interpretation of effect sizes). However, among the minority of exceptional empathizers, women considerably outnumber men, and among the minority of exceptional systemizers, men considerably outnumber women (Baron-Cohen et al., 2014; Greenberg et al., 2018).

Die Werte sind nicht absolut riesig, aber sie sind schon groß. Und natürlich hat das Auswirkungen.

It seems reasonable to suppose that sex differences in systemizing and empathizing are part of the reason that more men than women gravitate to fields such as physics and engineering, whereas more women than men gravitate to fields such as psychology and education. Consistent with this suggestion, a survey of nearly half-a-million people by Ruzich et al. (2015) revealed that STEM workers score consistently above workers in non-STEM fields on measures of systemizing ability.

Das wäre dann der „Gegentest“: Wenn die Männer die in bestimmten Bereichen arbeiten würden, eher niedrige Werte in „Systemezing“ für Männer hätten, dann wäre die These, dass es auf diese Eigenschaft ankommt, schwer zu halten.

Most commentators agree that sex differences in cognitive aptitudes are too small to explain STEM gender gaps in their entirety, and that sex differences in occupational preferences are a much more important contributor (Ceci et al., 2009; Dekhtyar et al., 2018; Johnson et al., 2008; Wai et al., 2018). Still, the evidence for the cognitive differences is robust, and it is perfectly plausible that they help to shape men and women’s career choices and trajectories. Indeed, given the clear relevance of the aptitudes in question to STEM, it would be surprising if this were not the case.

Ich denke, dass gerade das Zusammenspiel beider den Effekt sogar noch verstärkt. Zumal beide Unterschiede relativ groß sind.

The nature and nurture of sex differences in cognitive aptitudes

Where do these cognitive sex differences come from? One common view is that they are primarily a product of widely held stereotypes of the sexes: stereotypes that females are bad at maths, for instance, or that males alone possess the kind of innate brilliance supposedly required for certain academic fields (Bian et al., 2017, 2018; Nosek et al., 2009). In a number of ways, such stereotypes may function as self-fulfilling prophesies. First, they may lead parents and teachers to steer girls away from activities related to male-dominated areas such as maths, physics and computer programming, or to overlook or undervalue female giftedness in these domains and thus fail to adequately nurture it (Cheryan et al., 2017; Eccles et al., 1990; Lavy & Sand, 2018). Second, the stereotypes may persuade girls themselves that these areas are unlikely to be their fortes, which may lead them to lose interest in them and devote more time to other activities. This time-allocation pattern will naturally help determine which skills they end up developing. Finally, stereotypes about females’ mathematical or intellectual abilities, when made salient, may cause girls and women to underperform in high-stakes, time-limited tests as a direct result of their anxiety about confirming the denigrating stereotypes (a putative example of a phenomenon known as stereotype threat; Nguyen & Ryan, 2008; Schmader et al., 2008; Spencer et al., 1999).

Auch etwas, von dem Studien etwa aus den Gender Studies lernen könnten: Einfach mal die anderen Theorien darlegen und dann ausführen, welche Argumente dafür und dagegen sprechen.

The sociocultural explanations are intuitively plausible, and they all have at least some research backing them up. At the same time, however, the explanations are vulnerable to a number of criticisms. To begin with, it is unclear to what extent current social influences actually point in the direction these explanations presuppose. According to one study, by four years of age, girls tend to assume that boys are academically inferior, and by seven, boys assume the same thing (Hartley & Sutton, 2013). Similarly, teachers tend to view their female students as superior at maths and reading, even when aptitude tests indicate that the boys are doing better (Robinson & Lubienski, 2011). Popular culture often mirrors these trends, with girls depicted as academically superior to boys (consider, for instance, Bart and Lisa from The Simpsons, and Ron and Hermione from the Harry Potter series; see Synnott, 2016, for discussion of modern cultural depictions of boys and men). Certainly, several studies suggest that people see extreme brilliance as a male trait more than a female one (Bian et al., 2017, 2018). Nonetheless, it is far from obvious that, on balance, stereotypes about academic ability favour boys more than girls. Moreover, although stereotypes can bias social perception and function as self-fulfilling prophesies, the effects tend to be weak and the main reason that stereotypes correspond to social realities appears to be that they reflect those realities, rather than that they create them (Jussim, 2015).

In der Tat sind viele Vorurteile über Männer und Frauen im Schnitt schlicht zutreffend. Weil sie eben eine biologische Grundlage haben. Interessant aber auch, dass Mädchen in vielen Bereichen in jungen Jahren schlicht als besser wahrgenommen werden. Vielleicht auch schlicht eine Folge der Entwicklungsverzögerung bei Jungs. 

As for stereotype threat, a recent slew of studies has failed to find evidence that situations likely to induce threat do in fact hamper females’ performance in maths or other cognitive domains. The studies in question include several meta-analyses (Flore & Wicherts, 2015; Stoet & Geary, 2012), a number of large, pre-registered replications (Finnigan & Corker, 2016; Flore et al., 2019) and an analysis of 5.5 million chess games played in international tournaments, which found that women’s performance was better, rather than worse, under conditions of stereotype threat (Stafford, 2018; although see Smerdon et al., 2020). Meanwhile, on the other side of the ledger, a recent study failed to replicate the finding that stereotype threat impairs men’s performance on tests of language ability (Chaffee et al., 2020). At the very least, the effects of stereotype threat are more modest and situationally contingent than was originally assumed.

Ja, der Stereotype Threat hat gerade eine schwere Zeit hinter sich, bei dem entsprechende Studien ihn gerade nicht replizieren konnten. Insofern eher eine schwache Sache.

Still, even if the current crop of sociocultural theories requires refinement, it seems undeniable that social factors play a role in shaping cognitive sex differences. Aside from anything else, the relevant abilities can be improved with practice and training (Uttal et al., 2013), and the magnitude of the sex differences varies from place to place and from time to time (Hoffman et al., 2011; Hyde et al., 2009). Social forces are clearly part of the story. The question is whether social forces are the entirety of the story – and the answer is that they’re almost certainly not.

Das man mit Training in bestimmten Bereichen besser werden kann ist ganz klar. Aber auch in den Bereichen ist es natürlich unwahrscheinlicher, dass jemand, der sich alles hart erarbeiten muss (während ihm vielleicht ein anderer Bereich leichter fällt) für einen Bereich entscheidet, wenn seine „Konkurrenten“ dort sich diesen Bereich anscheinend wesentlich leichter erschließen können und ebenso trainieren.

Various lines of evidence suggest, once again, that biological factors play a pivotal role as well.

First, many of the sex differences under discussion appear early in the developmental process. The sex difference in mental rotation, for instance, can be detected by three months of age (Moore & Johnson, 2011; Quinn & Liben, 2014; although see Miller & Halpern, 2014), and the sex difference in language ability can be detected by seven months (Bando et al., 2016).

Da müsste das Patriarchat schon ganz schön schnell zuschlagen und dafür sind keine Anzeichen vorhanden.

These findings do not rule out purely environmental explanations, but they do render them less plausible. At a minimum, they eliminate the possibility that social influences appearing after the age of one could provide a complete explanation for the differences. (Note, though, that later-appearing differences are not necessarily products of social causes alone; to some extent, they may be part of the natural maturational process. This is especially likely when the changes coincide with puberty; Berenbaum & Beltz, 2011.)

A second line of evidence for a biological contribution is that several of the traits under discussion have been linked, albeit somewhat tentatively, to sex hormones (Berenbaum & Beltz, 2011). Most of the relevant research has focused on spatial abilities. Various studies have found that girls and women with CAH do better than unaffected controls on a range of spatial tasks (Berenbaum et al., 2012; Hampson et al., 1998; Mueller et al., 2008; Resnick et al., 1986). Admittedly, not all studies have found this pattern (see, e.g. Hines et al., 2003; Malouf et al., 2006). One possible explanation for the mixed findings, however, is that most studies in the area have small sample sizes, and thus some fail to detect an effect, even when the effect is there (Berenbaum & Beltz, 2011). Consistent with this suggestion, a meta-analysis of CAH studies concluded that females with the condition have better spatial skills than unaffected controls (Puts et al., 2008; although see Hines, 2009). Whereas females exposed to high levels of testosterone in early life appear to have above-average spatial abilities, males exposed to low levels – including males with idiopathic hypogonadotropic hypogonadism (or IHH) – appear to have below-average abilities (Buchsbaum & Henkin, 1980; Hier & Crowley, 1982).

Die Eigenschaften zeigen sich in einer Vielzahl von Gruppen, die in Verbindung mit unterschiedliche, gerade pränatalen Testosteronständen stehen. Ich meine auch bei Schwulen und Lesben zeigen sich entsprechende Unterschiede und auch im Transbereich etc.

On top of that, the link between testosterone and spatial abilities has been demonstrated experimentally in various nonhuman mammals (summarized in Hampson et al., 1998), making it a reasonable default hypothesis for our own species.

Bei Hampson habe ich das hier gefunden:

In the past 30 years, studies in rodents and nonhuman primates have firmly established that the mammalian central nervous system is sensitive during specific periods in early development to androgens and their metabolites. The presence of high levels of androgens at these times induces permanent alterations in the ultrastructural organization of steroid-sensitive brain regions (including the hypo– thalamus, limbic system, and neocortex: Bachevalier & Hagger, 1991; Breedlove, 1992; Gorski & Jacobson, 1981; Juraska, 1991; Roof & Havens, 1992; Toran-Al– lerand, 1991), and irreversibly alters the behavioral propensities of the animal (Beatty, 1992; Goy & McEwen, 1980), promoting the emergence of a male-typical pattern of brain and behavior. These events occur in male animals under the influence of endogenous androgens, but can be induced experimentally in females by exogenous manipulations of hormones. These organizational effects of androgens were originally described for repro–ductive behaviors in rodents by Phoenix, Goy, Gerall, and Young (1959). Organ– izational effects on other sexually differentiated behaviors have now been identified, such as rough-and-tumble play (Goy & Resko, 1972; Meaney, 1988) and the tendency toward aggression (see Beatty, 1992, for a recent review). Whether early androgens or other sex steroids also influence central nervous system (CNS) differentiation in humans is not established, but evidence favors a neuroendocrine contribution to the development of sexual orientation, certain childhood play behaviors, and perhaps tendencies toward aggression in humans (for reviews, see Collaer & Hines, 1995; Hampson & Kimura, 1992) 

High levels of androgens during critical periods in early development may also promote a brain organization that facilitates the ability to process certain types of spatial information. Spatial ability has evolved as a sexually differentiated function in many mammalian species, with male animals typically showing greater profi– ciency than females on various tasks that involve processing or remembering spatial information, such as route learning in an unfamiliar environment (Beatty, 1984; Williams & Meek, 1991). This has been conceptualized as a product of sexual selection, resulting from ecological pressures for greater mobility in males (Gaulin & FitzGerald, 1986). Although it can be modified by learning and other sociocul– tural factors, vestiges of this sex difference are also seen on laboratory spatial tasks in humans (Linn & Petersen, 1985; Voyer, Voyer, & Bryden, 1995), particularly those involving dynamic stimuli (Hunt, Pellegrino, Frick, Farr, & Alderton, 1988), visualization in three dimensions (Sanders, Soares, & D’Aquila, 1982), or targeting accuracy (Watson & Kimura, 1991), tasks that most closely resemble the three-di– mensional natural environment in which spatial abilities likely evolved.

In rodents, hormonal manipulations in the neonatal period have been shown to systematically affect rate of learning in adulthood on spatial tasks such as the radial-arm maze or Morris water maze and to alter reliance on geometric versus landmark cues for spatial navigation (Roof & Havens, 1992; Williams, Barnett, & Meek, 1990; Williams & Meek, 1991). In untreated animals, faster maze learning

is typically found in males, but this sex difference is eliminated by castration of males, or by administration of masculinizing hormones to females, in the neonatal

period (Dawson, Cheung, & Lau, 1975; Roof & Havens, 1992; Stewart, Skvarenina,

& Pottier, 1975; Williams et al., 1990). Developmental alterations in neurons in the frontal cortex and hippocampus may be part of the neural substrate for these effects (Williams & Meek, 1991). In primates, spatial abilities per se have not been

examined. However, developmental effects of testosterone have been found in

rhesus monkeys on object discrimination learning tasks and on maturation rates in the regions of frontal and temporal cortex that subserve these tasks (Bachevalier & Hagger, 1991).

Nicht gerade die ausführlichste Auflistung und etwas allgemein, aber dennoch interessant.

In addition to the spatial-skills findings, testosterone has been linked to higher systemizing, and to lower social skills, empathizing and verbal ability (Auyeung et al., 2006; Chapman et al., 2006; Lutchmaya et al., 2001, 2002; although see Nadler et al., 2019, on the testosterone/low-empathy link). It is unclear at this stage whether hormones affect abilities directly, or do so indirectly through their effects on people’s activity preferences (see, e.g. Schmidt, 2011). Based on present evidence, however, a reasonable supposition is that they do both (Berenbaum et al., 2012).

Auch hier gibt es viele interessante Studien, die diesen Bereich betreffen.

Third, sex differences in cognitive abilities appear to transcend cultural boundaries. Across cultures, girls typically outperform boys on tests of linguistic ability (Stoet & Geary, 2013), whereas boys typically outperform girls on most spatial tasks (Cashdan et al., 2012; Lippa et al., 2010; although see Hoffman et al., 2011). Certainly, as mentioned, there is variation in the magnitude of these differences from nation to nation and from generation to generation, suggesting a sizeable role for malleable social factors. Still, the direction of the spatial/linguistic differences is essentially invariant. This is not what one would expect if these differences had a purely environmental origin.

Auch da ist es interessant, warum diese Unterschiede weltweit auftreten. Kulturell ist das relativ schwierig zu erklären. Gerade da Sprache in vielen Bereichen ja auch viel mit Macht zu tun haben kann und selbst bei einem umfassenden Patriarchat davon auszugehen wäre, dass Männer sich evtl beide Eigenschaften kulturell unter den Nagel reißen.

Fourth, efforts to eliminate the gaps quickly seem to reach the point of diminishing returns. Wai et al. (2010) looked at the ratio of boys-to-girls in the top 0.01% of US seventh graders taking the SAT-Math test between 1981 and 2010. In the early 1980s, the ratio was 13 boys for every girl. By the early 1990s, this had dropped to just four boys for every girl, perhaps as a result of increasing access to a good maths education for girls (see also Hyde et al., 1990). Since then, however, the ratio of boys-to-girls among the top maths performers has remained largely the same, despite intensified efforts to eliminate the remaining gap.⁵ Meanwhile, the ratio of girls-to-boys among the top performers on tests of verbal and reading ability has consistently favoured girls (Makel et al., 2016; Wai et al., 2010). Again, these findings are not what one would expect if the gaps were due entirely or even primarily to sociocultural causes.

Es passt hingegen sehr gut dazu, dass es biologische Grundlagen dieser Eigenschaften gibt. Natürlich zeigt dieser Anstieg der Frauen aber auch, dass Bildung etwas bringt und eine Angleichung herbeibringt. Das aber dann, als man schon solche Erfolge erreicht hat, der Anstieg nicht fortsetzt spricht für einen biologischen Grund.

Fifth and finally, a large meta-analysis by Xu et al. (2017; N = 254,231) concluded that gay men tend to have spatial and linguistic abilities comparable to those of straight women, whereas lesbians tend to have spatial abilities comparable to those of straight men (but female-typical linguistic abilities). These findings are not readily explicable on the assumption that social forces alone create the usual pattern of sex differences. Gay men were presumably subject to essentially the same gender-specific social forces as straight men, and lesbians the same gender-specific social forces as straight women. As such, the near-reversal of the usual spatial vs. language pattern is hard to reconcile with the claim that this pattern is due largely to social forces. Other variables, such as prenatal hormones, appear to play a larger role.

Ah, hier kommt es auch noch mal. Das wäre aus meiner Sicht auch eher mit den Hormonen zu erklären.

To be clear, none of this is to deny a role for social forces in general or for stereotypes in particular. Indeed, even if the stereotypes ultimately trace back to genuine, unlearned differences between the sexes, they could still further shape people’s interests and choices, and therefore help determine which skills they practice and hone (Ellemers, 2018). Nonetheless, it is difficult to explain all the data without assuming a non-trivial biological contribution.

Aus meiner Sicht spricht sehr viel dafür, dass die Bioloigie eine erhebliche Rolle spielt. Aber es bleibt ein sehr spannender Forschungsbereich.

 

Evolutionary rationale

It seems reasonable to conclude that the cognitive sex differences considered in this section are moulded to an important degree by unlearned biological factors. However, as with sex differences in occupational preferences, it is not at all obvious why this might be the case. Some argue that sex differences in spatial ability trace to the fact that Homo sapiens is an effectively polygynous species: that is, a species in which males have somewhat greater reproductive variability than females (Betzig, 2012; Labuda et al., 2010). In effectively polygynous species, the argument goes, males tend to have larger ranges than females, and therefore tend to evolve stronger spatial and wayfaring skills (Gaulin, 1992). The polygyny-related spatial sex difference may have been further amplified in our species by the fact that human males are specialized for hunting and tracking animals, and perhaps also for engaging in coalitional warfare with neighbouring groups (Silverman & Phillips, 1998; see Archer, 2019, for an overview of adaptationist explanations).

Many evolutionary psychologists are unpersuaded by these ideas, however. In their view, cognitive sex differences were not specifically favoured by natural selection, but instead are byproducts of other sex differences that were. Clint et al. (2012) argue, for instance, that the male advantage in spatial skills is merely a side effect of hormonal sex differences that were selected for other reasons. If this is right, then the spatial sex difference may have an innate basis but not be a direct product of natural selection.

At this stage, the ultimate origins of human cognitive sex differences are uncertain. What does seem certain, though, is that the differences are not solely a product of social forces. To some extent – perhaps to an important extent – they are a part of human nature.

Ergänzen könnte man sexuelle Selektion, und zwar intrasexuell und intersexuell,  etwa darauf, dass Männer weniger Schwäche zeigen, Probleme eher lösen etc.