The bad strategy is controlling. As the label implies, it is a “You will do it because I tell you to do it” approach to child-rearing. The two chief ways in which parents can be controlling are harsh physical control and psychological pressure or manipulation (examples are shaming, inducing feelings of guilt, or withdrawing affection). An extensive literature documents a variety of negative consequences of controlling strategies, especially during adolescence. E.g., Karreman, van Tuijl, van Aken et al. (2006); Rothbaum and Weisz (1994); Kawabata, Alink, Tseng et al. (2011).
In 2016, the Dutch scholars cited earlier published a meta-analysis of the literature on gender-differentiating parenting regarding control strategies. Their conclusion was generically similar to those of the other meta-analyses: The evidence showed little difference in the way parents treat daughters and sons. Parents were slightly more controlling with sons, but the effect was negligible (d = –0.08). The difference was even smaller regarding autonomy-supportive behavior (d = –0.03). The analysis showed an effect over time, with boys receiving more autonomy-supportive parenting in studies from the 1970s and 1980s, while girls received more autonomy-supportive parenting in studies published from 1990 on. But the fitted effect sizes remained small throughout, with an absolute d of no more than about 0.10. Endendijk, Groeneveld, Bakermans-Kranenburg et al. (2016): Fig. 2.
The authors’ conclusion resembles those of others who have studied gender differentiation in parenting:
These findings question the importance of gender-differentiated parental control as a means of gender socialization and as a mechanism underlying gender differences in child behavior. However, the large differences between studies and the individual differences within studies suggest that some parents do treat their sons and daughters differently with regard to parental control. Parents’ gender stereotypes might explain why some parents do treat their sons and daughters differently and others do not, but this mechanism has yet to be confirmed empirically. (Endendijk, Groeneveld, Bakermans-Kranenburg et al. (2016): 23 of 33).
12. Eagly (1987). This abbreviated description is drawn from Eagly and Wood (2011). For a detailed description of social role theory, see Wood and Eagly (2012).
13. Eagly and Wood (2011): 468.
14. Eagly and Wood (2011): 459.
15. Bleier (1991) quoted in Halpern (2012): 178.
16. Summers’s remarks were subsequently transcribed and released by the Office of the President at Harvard. Summers (2005).
17. Stuart Taylor Jr., “Why Feminist Careerists Neutered Larry Summers,” Atlantic, February 2005.
18. Sam Dillon and Sara Rimer, “No Break in the Storm over Harvard President’s Words,” New York Times, January 19, 2005.
19. Taylor, “Why Feminist Careerists Neutered Larry Summers.”
20. Sam Dillon and Sara Rimer, “President of Harvard Tells Women’s Panel He’s Sorry,” New York Times, January 21, 2005; Dillon and Rimer, “No Break in the Storm over Harvard President’s Words.”
21. For those who want to look into sex differences more thoroughly, here’s a brief reading list.
The book that for me best integrates the findings from the technical literature with compelling narratives about real cases is Susan Pinker’s The Sexual Paradox: Men, Women, and the Real Gender Gap. Much has been learned since she published it in 2008, but none of her technical positions have been discredited and many have been reinforced by subsequent work.
Simon Baron-Cohen’s The Essential Difference: Male and Female Brains and the Truth About Autism (2003) is that rarity, a book that is both scientifically seminal and readable.
A new edition of David Geary’s magisterial Male, Female: The Evolution of Human Sex Differences is in press as I write. I expect it to be, like the earlier editions in 1998 and 2010, the definitive statement of the existing state of knowledge.
Diane Halpern’s Sex Differences in Cognitive Ability (2012) is focused on a narrower topic and is somewhat denser than the others, but she tells you all you need to know about her topic in one source.
If you want to work your way into a broader discussion of sex differences in smaller chunks, I recommend a series of columns that David Schmitt wrote for Psychology Today from 2013 to 2016, easily accessible online at www.psychologytoday.com. Engagingly written, they are also fully documented and accompanied by bibliographies.
22. As of early 2017, the Women’s Studies Online Resources webpage listed 673 American institutions of higher education that have women’s studies programs, departments, or research centers. I didn’t try to review the course offerings for all of them, but I did examine the course catalogs during the 2016–17 school year for the women’s studies programs at 13 of the most prestigious universities in the country—the eight Ivy League schools plus MIT, Stanford, Duke, the University of Chicago, and the University of California at Berkeley. I was searching for courses in women’s studies programs that provided systematic discussions of biological evidence for sex differences in cognitive repertoires. I found a single example: Cornell University’s course FGSS 3210, “Gender and the Brain,” cross-listed as biology course BIONB 3215, which tells prospective enrollees, “Reading the original scientific papers and related critical texts, we will ask whether we can find measurable physical differences in male and female brains, and what these differences might be.” I can’t guarantee that the online course listings were complete. And while I found no courses dealing with hormones or other genetically-grounded sources of male-female differences, I presume that these topics, along with differences in the brain, are sometimes raised in courses not specifically devoted to them. But compare what I found with what should be the norm. At a reputable university—and these 13 are at the top of the heap—to get a degree in women’s studies should include as an obvious requirement a solid foundation in evolutionary biology and in the differential biology of the two sexes, including biology above the neck. None of these 13 prestigious schools did.
23. Trivers (2011): 314–15.
1: A Framework for Thinking About Sex Differences
1. In the 1971 edition of the Oxford English Dictionary, the only definition of gender as a noun applied to the sexes was as a jocular transfer of the linguistic use of gender (masculine and feminine genders) to apply to humans. By the 1989 edition, it had added a new meaning: “In modern (especially feminist) use, a euphemism for the sex of a human being, often intended to emphasize the social and cultural, as opposed to the biological, distinctions between the sexes.” The OED dated the earliest known example of that meaning to 1963, but the rationale for using gender instead of sex was first introduced a decade earlier in Money (1952).
2. Thorndike (1911): 32.
3. Quoted in Baron-Cohen (2002): 251.
4. Baron-Cohen (2003): 61.
5. Baron-Cohen (2003): 26.
6. Baron-Cohen and his colleagues devised tests for measuring systemizing and empathizing, with scores labeled SQ and EQ respectively. In the largest sample of people (5,186 total) who were administered both tests, the effect sizes (a term explained later in this chapter) were +0.63 (females had a higher mean) for EQ and –0.47 (males had a higher mean) for SQ. Wright and Skagerberg (2012). Effect sizes in the samples used by Baron-Cohen and his colleagues have been +0.50 and +0.76 for EQ and –0.59 for SQ. Baron-Cohen, Richler, Bisarya et al. (2003): Table 1; Baron-Cohen and Wheelwright (2004): Table II.
7. Sawilowsky (2009).
8. Cohen used the descriptors to guide researchers in characterizing an expected d value when there was no prior research available. He repeatedly told his readers not to make too much of them. For example: “The terms ‘small,’ ‘medium,’ and ‘large’ are relative, not only to each other, but to the area of behavioral science or even more particularly to the specific content and research method being employed in any given investigation.” Cohen (1988): 25. And “A reader who finds that what is here defined as ‘large’ is too small (or too large) to meet what his area of behavioral science would consider appropr
iate standards is urged to make more suitable operational definitions.” Cohen (1988): 79.
9. Funder and Ozer (2019).
10. Funder and Ozer (2019).
11. See Rosnow and Rosenthal (2003) for a critique of Cohen’s guidelines antedating Hyde (2005). See Gignac and Szodorai (2016) for suggested guidelines corresponding to those proposed by Funder and Ozer (2019).
12. Hyde (2005): 581. I have reproduced the inequality symbols as they appear in the article, which apparently treats “<” as equivalent to “≤.”
13. Hyde (2005): 585–86.
14. Hyde (2005): 587.
15. Hyde (2005): 587.
16. Hyde (2005): 590.
17. In his technical discussion of this issue, Del Giudice put it this way: “When measuring a multidimensional construct, the overall difference between two groups is not the average of the effects measured on each dimension, but a combination of those effects in the multidimensional space: Many small differences, each of them on a different dimension, can create an impressive effect when all the dimensions are considered simultaneously. Crucially, such overall differences are likely to matter more than their individual components, both in shaping people’s perceptions and in affecting social interaction.” Del Giudice (2009): 268.
18. See Del Giudice (2009). He augmented the discussion in Del Giudice, Booth, and Irwing (2012). Janet Hyde responded in the comments to Del Giudice (2009) with “The Distance Between North Dakota and South Dakota.” Stewart-Williams and Thomas (2013) critiques the use of Mahalanobis D in its appendix.
2: Sex Differences in Personality
1. Adapted from McCarthy, Nugent, and Lenz (2017): Table 2. I have omitted neurological and neurodegenerative diseases that were included in the table: migraine, stroke, multiple sclerosis, Alzheimer’s disease, Parkinson’s disease, amyotrophic lateral sclerosis, and myasthenia gravis.
2. As with almost all sex differences, a minority literature disputes the magnitude of differences. In the case of depression, for example, see Martin, Neighbors, and Griffith (2013).
3. In statistics, factor has a technical meaning. In its most basic form, the statistical procedure (factor analysis) creates a first component—a factor—that explains as much of the variation among the observations as possible. The algorithms then create a second factor that explains as much of the remaining variation as it can, and so on through successive iterations, each of which produces a new factor, until all of the variation has been assigned to a factor.
4. Lewis Goldberg, who had been instrumental in resuscitating interest in personality studies after a lull in the 1960s and 1970s, coined the phrase “Big Five.” See his review of the development of the Big Five model in Goldberg (1993a).
5. Costa and McCrae (1985). The 1985 version had measures of neuroticism, extraversion, and openness. Agreeableness and conscientiousness were added later. The official name of the current version is NEO-PI-3.
6. For a full discussion of this issue, see Pettersson, Mendle, Turkheimer et al. (2014).
7. Other personality models include HEXACO, which adds an honesty-humility factor to the Big Five, and a model focused on three problematic aspects of the personality: narcissism, Machiavellianism, and sadism. It’s called variously the Dark Triad or the Dark Tetrad model. For a review of the literature, see Furnham, Richards, and Paulhus (2013).
8. Among the 30 FFM facets (the detailed characteristics that make up the five factors), here are the ones that showed absolute effect sizes of less than 0.20 in both the Costa and Kajonius studies: experiences anger or bitterness, assertive or forceful in expression, open to the inner world of imagination, open to new experiences in life, open to reexamining one’s own values, trust in others’ sincerity or intentions, values orderliness, believes in fulfilling moral obligations, uses self-discipline in fulfilling tasks, and thinks things through before acting. In the Del Giudice study using the 16PF, these were the factors showing an absolute difference of less than 0.20: Self-reliant, solitary, resourceful, lively, animated, spontaneous, abstracted, imaginative, absentminded, organized, perfectionistic, compulsive private, discreet, nondisclosing, socially bold, venturesome, and thick-skinned.
9. Effect sizes are reported for latent variables corrected for specific variance and measurement error. See the discussion in Del Giudice, Booth, and Irwing (2012).
10. Del Giudice, Booth, and Irwing (2012).
11. Del Giudice, Booth, and Irwing (2012).
12. Noftle and Shaver (2006). Averages of d are computed using the absolute value. In this case, the values were corrected for attenuation due to scale unreliability.
13. Del Giudice (2009): Table 1.
14. Connellan, Baron-Cohen, Wheelwright et al. (2000): Table 1.
15. Sagi and Hoffman (1976); Simner (1971); Hoffman (1973). This and subsequent citations in the list are drawn from Alexander and Wilcox (2012).
16. Hittelman and Dickes (1979); Leeb and Rejskind (2004); Lutchmaya, Baron-Cohen, and Raggatt (2001).
17. Cossette, Pomerleau, Malcuit et al. (1996).
18. Gunnar and Donahue (1980).
19. Mayes and Carter (1990).
20. Alexander, Wilcox, and Woods (2009); Benenson, Duggan, and Markovits (2004); Campbell, Shirley, and Heywood (2000). The doll-truck contrast shows up in nonhuman primates as well: Male vervet and rhesus monkeys prefer trucks while female ones prefer dolls. Alexander and Hines (2002); Hassett, Siebert, and Wallen (2008).
21. Mundy, Block, Delgado et al. (2007); Olafsen, Ronning, Kaaresen et al. (2006).
22. McClure (2000). Another 14 studies simply reported that the results were “nonsignificant” without including the information necessary to calculate effect sizes. Half of those studies had samples of 48 or fewer. Large effect sizes can be statistically insignificant with sample sizes that small. McClure calculated a lower bound effect size of +0.26 if all of the nonsignificant results had an effect size of zero—unrealistically low. The true value is somewhere between +0.26 and +0.92, probably well toward the +0.92 end of the range. Even the lower bound of 0.26 was statistically significant.
23. Costa, Terracciano, and McCrae (2001).
24. Costa, Terracciano, and McCrae (2001): Table 3. The only exception was agreeableness in Zimbabwe, which was a trivial –0.02.
25. McCrae and Terracciano (2005): Table 5. The inconsistent effect sizes, all of them only fractionally different from zero, were for Nigeria (N = 0, A = 0, E = –0.04, O = 0, C = 0), India (E = –0.05, O = +0.03), Botswana (E = –0.01), Ethiopia (A = –0.02), Russia (A = –0.02), and Uganda (O = 0).
26. The prediction is necessary, but it has also been explicitly acknowledged. See Eagly, Wood, and Johannessen-Schmidt (2004), quoted in Schmitt, Long, McPhearson et al. (2016).
27. Costa, Terracciano, and McCrae (2001): 327. For the sake of consistency in the interpretation of results, the countries in this and the subsequent discussion are limited to those for which the UN has calculated a score on the Gender Inequality Index. The median effect sizes for emotional stability, agreeableness, openness to emotion, and extraversion for adults across 21 countries were –0.51, +0.45, +0.26, and +0.23 respectively. The Costa study omitted the fifth factor, conscientiousness, because none of its facets showed consistent sex differences. The Costa study also addressed a problem with the facets for measuring extraversion and openness. The facets for measuring extraversion included warmth/affiliation, which is higher in females, and dominance/venturesomeness, which is higher in males. Behaviorally, these traits are completely different, but they more or less cancel each other out in the combined measure of extraversion. A similar problem occurs with openness to experience (women are higher on measures of openness to emotion and males are higher on openness to ideas). The Costa study dealt with this problem by creating measures of extraversion and openness that specifically focused on the warmth/affiliation aspect of extraversion and the emotional aspect of openness. The discussion in Del Giudice, Booth, and Irwing (2012) of these masking tendencies
when traits are aggregated into the Big Five includes citations of the relevant sources.
28. Indicators are given in Jahan et al. (2016): Statistical Annex, Tables 4 and 5. The UN also has a Gender Development Index based on women’s life expectancy, years of schooling, and women’s per capita gross national income. The correlation between the Inequality and Development indexes is –.66 (“high” means “bad” on the Inequality Index, “good” on the Development Index). Both indexes capture measures of health (maternal mortality rate versus life expectancy) and education (percent with at least some secondary education versus years of education). In deciding whether to combine the two indexes, the issue is how much is added by the measure unique to the Development Index, per capita gross national product. I judged that to be minor, outweighed by the potentially distorting effects of double-counting education and health. Data were downloaded from the Human Development Reports website, www.hdr.undp.org.
29. Absolute size, because it doesn’t make any difference whether females or males score higher on a personality trait—theoretically, sex differences on all personality traits should be diminishing.
30. Author’s analysis using GII scores and Costa, Terracciano, and McCrae (2001): Table 3.
31. For a more concrete sense of how these correlations translate into scores for nations at the extremes, compare the nations in the McCrae sample with the five lowest GII scores (Switzerland, Denmark, Iceland, Germany, Denmark) with the nations with the five highest GII scores (Burkina Faso, India, Uganda, Ethiopia, and Morocco). On all five factors, the mean effect sizes for the most gender-egalitarian countries ranged from two times to more than three times the effect size for the least gender-egalitarian countries: +0.46 compared to +0.22 for agreeableness, +0.18 compared to +0.09 for conscientiousness, +0.27 compared to +0.13 for extraversion, +0.36 compared to +0.12 for openness to emotion, and –0.58 compared to –0.26 for emotional stability. Author’s analysis using GII scores and McCrae and Terracciano (2005): Table 5.
Human Diversity Page 46