1. Janet Shibley Hyde University of Wisconsin
Men Are from Earth, Women Are from Earth: Science vs. the Media on Psychological Gender Differences
Janet Shibley Hyde
University of Wisconsin

2. Collaborators
Mathematics Performance: Elizabeth Fennema, Sara Lindberg, Marcia Linn
Sexuality: Jenni Petersen
Self-Esteem: Kristen Kling
Moral Reasoning: Sara Jaffee
Temperament: Nicole Else-Quest
Special thanks to NSF for funding, REC

3. The Differences Model

5. The Oppositional Model

6. Innate, Biological Causes

8. Biological Causes
“What we’ve found is that the female brain is so deeply affected by hormones that their influence can be said to create a woman’s reality.” (p. 3)
“The female brain has tremendous unique aptitudes… a nearly psychic ability to read faces and tone of voice for emotions… All of this is hardwired into the brains of women.” (p. 8)

9. Innate, Biological Causes

10. Innate, Biological Differences

12. The Deficit Model
Harvard President Lawrence Summers, who claimed, in a controversial speech, that women do not have the math ability to succeed in science and engineering (2005)

13. Meta-Analysis: A Method for Assessing Psychological Gender Differences
A quantitative literature review
A method for quantitatively combining the results of numerous studies on a given question
Tells us not only whether there is a difference, but also how big it is

14. Steps in a Meta-Analysis
Locate all prior studies on the question
Extract statistics from each study and compute an effect size, d, for each study
Compute a weighted average d, averaged over all studies
Test the set of effect sizes for homogeneity. If nonhomogeneous, conduct moderator analyses to account for variations.

15. Effect Size: The Size of the Gender Difference d = MM – MF sw (Cohen)

16. Cohen’s Guidelines for Interpreting Effect Sizes
d = .20 small
d = .50 medium
d = .80 large
(Hyde: d ≤ trivial)

17. Stereotypes of Gender Differences in Abilities
Verbal
Mathematical
Spatial

18. Meta-Analysis of Gender Differences in Mathematics Performance (1990)
100 studies
Testing of more than 3 million people
Hyde, Fennema, & Lamon, Psychological Bulletin, 1990

19. Gender Differences in Math Performance
All Studies
d = +.15
Samples of the General Population
d = -.05
(Hyde, Fennema, & Lamon, 1990)

20. Almost complete overlap between distributions

21. Age x Cognitive Level Age Computa- tion Concepts Problem Solving
5 – 10
-.20
-.02
0.00
11 – 14
-.22
-.06
15 – 18
0.0
+.07
+.29 NA
+.32
Note that girls are actually better than boys at computation through middle school and there’s no difference in high school. No gender difference in understanding mathematical concepts at any age. No gender difference in problem solving in elementary school or middle school. Small gender difference favoring males emerges in high school and we need to pay attention to it. Explanations on next slide.

22. Possible Reasons for the Gender Gap in High School, 1990
Course choice (Eccles)
Mathematics
Science
Stereotype threat (Steele)

23. Ethnicity d No. of studies African Americans -.02 21 Hispanics 0.00 20
Asian Americans
-.09 4
White Americans
+.13 13
Australians
+.11 7
Canadians
+.09 5
Mixed or unreported
+.15
184
The much-discussed gender difference is not present in ethnic groups other than whites – see African Americans and Hispanics.

24. Cross-national Trends in Math Performance
Note that gender differences are tiny compared with cross-national differences.
5th graders, word problems
Lummis & Stevenson, 1990

25. Strongest predictors of mathematics achievement at age 10 (Melhuish et al., Science, 2008)
0.5
Effect Size
0.1
SES
Birth weight
Gender
Mother’s Educ
Father’s Educ
Family income
Home learn env
Elem sch qual
Preschool qual

26. Gender Differences in Height
d = 2.0 (Niewenweg et al., 2003)

27. Quiz Question
In the U. S. today, what percentage of bachelor’s degrees in mathematics go to women?

28. Answer:
48%

29. New Meta-Analysis State Assessments, 2008
Annual assessments by states of all children’s mathematics performance (and other areas) mandated by No Child Left Behind (NCLB)
Contacted departments of education in all 50 states asking for data needed to compute d
Responses from 10 states
Testing of more than 7 million children
Hyde, Lindberg, Linn, Ellis, & Williams, Science, 2008

30. Grade d
Grade 2
0.06
Grade 3
0.04
Grade 4
-0.01
Grade 5
Grade 6
Grade 7
-0.02
Grade 8
Grade 9
Grade 10
Grade 11
Overall
d = .0065

31. Conclusion
Girls have reached parity with boys in math performance at all grade levels: Gender similarities
Why the change over time?
Gender gap in taking high school math has disappeared

32. From The Onion
Researchers for the National Science Foundation have found that boys and girls
now perform equally in standardized math tests. What do you think?
Max Thomas
Cashier
“Great, that’s all I need.
My wife knowing the exact
moment I arrive in Boston
if my train left New York
traveling at 60 miles per
hour.
Luke Casey
Industrial Loom Operator
“But linear algebra was
the only thing that ever
made me feel like a
man.”
Michelle Banks
Horticulturist
“All I know is, every
time I try to solve a
complex math
problem, my breasts
get in the way.”
July 29, 2008

33. Rants Results were generally enthusiastically received, but…
“In regard to your study… the cynic would point out that you are a female psychologist, someone with a political motive to provide a self-fulfilling prophecy for the self-esteem challenged.” Mark, 7/25/2008
“….. If women want to major in mathematics, physics, and chemistry in college in order to get a teaching certificate, then fine. But I would urge you to discourage, strongly, women from pursuing graduate degrees in these subjects and, instead, leave them to men.” Norman, 8/8/2008

34. The SAT-Math Well-publicized gender difference In 2010, SAT Math
Males M = 534 (SD = 103)
Females M = 500 (SD = 101)
Sampling is completely uncontrolled
In 2010, SAT taken by
827,000 females
721,000 males
Male sample is more selective
Gender difference on SAT Math is in part a sampling artifact
d = .34

35. The Role of Culture Percentage of U.S. PhD’s in Mathematics
Awarded to Women
Green & LaDuke, 2009

36. Why should we care about math?
For the individual, it’s crucial for access to prestigious, high-paying jobs in science, technology, and engineering. Much demand for workers.
For the nation, we can’t afford to waste 50% of our talent (women) in the global economic competition.

37. Another Gender Stereotype: Verbal Ability
Another meta-analysis (Hyde & Linn, 1988)

38. Gender Differences in Verbal Ability
Another stereotype is called into question.
(Hyde & Linn, 1988)

39. Self-Esteem
Does girls’ self-esteem take a nosedive at the beginning of adolescence?
Best-selling authors claim that girls’ self-esteem takes a “nosedive” at the beginning of adolescence.

40. Meta-Analysis of Gender Differences in Self-Esteem (Kling, Hyde et al
216 effect sizes
Testing of 97,000 respondents
Additional data from National Center for Education Statistics (NCES) – well-sampled testing of 48,000 American adolescents

41. Gender Differences in Self-Esteem
(from meta-analysis)
d = +.04 to +.24
(from NCES data, adolescents)
(Kling, Hyde, et al., Psychological Bulletin, 1999)
The gender difference in self-esteem is small (averaged over all ages, although NCES is all adolescents)

42. Gender Differences in Self-Esteem
Age group d
No. of studies
7-10
+.16 22
11-14
+.23 53
15-18
+.33 44
19-22
+.18 72
23-59
+.10 16
60 or older
-.03 6
We don’t see a huge gender difference emerging at the beginning of adolescence. The gender difference in high school is not large but worth paying attention to. Note that the gender difference seems to go away in adulthood.

43. Gender Differences in Self-Esteem
Ethnic Group d
No. of Studies
Whites
+.20 52
Blacks
-.04 11
Again, the small phenomenon that’s present for Whites is not present in other ethnic groups.

44. Gender Differences in Sexuality, 1993 to 2007 Petersen & Hyde, Psychological Bulletin, 2010

45. Study 1 840 articles with usable data 1,441,333 participants
82 different countries

46. Which behaviors show the largest gender differences
Which behaviors show the largest gender differences? Which attitudes show the largest gender differences?

47. Attitudes (16) Behaviors (14) Petting Sex Frequency Sex Incidence
Oral Sex
Anal Sex
Casual Sex
Number of Partners
Masturbation
Pornography Use
Extra-relational Sex
Age at First Intercourse
Condom Use
Same-Gender Behavior
Cybersex
Attitudes (16)
Premarital Sex
Extramarital Sex
Masturbation
Casual Sex
Condoms
Double Standard
Fear/Anxiety/Guilt
Sexual Satisfaction
Sexual Permissiveness
Sex when Couple Engaged
Sex when Couple Committed
Gay Rights
Gay Marriage
Homosexuals
Gay Men
Lesbian Women

48. Large
Medium
Small 35

49. Medium
Small
Small

50. Moderators Gender Empowerment (GEM)
United Nations Development Programme (UNDP)
For each nation, ratio of women to men for:
Percent of parliamentary seats
Estimated income
Percent of senior officials, legislators, prof/tech positions
US GEM = .76, Turkey GEM = .29
Prediction: larger gender differences in nations with more gender inequality (social structural theory, Eagly & Wood)

51. Results for Gender Empowerment
GEM predicts the size of the gender difference in casual sex across nations
β = -.37**
GEM predicts the size of the gender difference in masturbation across nations
β = -.57**

52. The Gender Similarities Hypothesis (Hyde, American Psychologist, 2005)
Men and women are very similar on most (not all) psychological variables.
Evidence
Over 46 meta-analyses and 124 effect sizes for gender differences,
30% of d values near 0: 0 – 0.10
48% of d values near .20: – 0.35
= 78% of gender differences are small or close to 0
Massive support for gender similarities

53. Costs to Overinflated Claims of Gender Differences
Education: Single-sex classrooms and schools, in the absence of empirical support

54. Costs to Overinflated Claims of Gender Differences
Education: Emphasis on girls’ self-esteem problems leads us to ignore boys’ self-esteem problems
And boys’ self-esteem problems can be very dangerous to others – e.g., school shootings
Single-sex sex school and classrooms are currently a popular fad – but they rest on the assumption of enormous psychological gender differences – differences so large that boys and girls can’t be educated in the same classroom.

55. Costs to Overinflated Claims of Gender Differences
Workplace
How can we claim that women are vastly different from men, but can do the same jobs as men, and deserve equal pay?

56. Costs to Overinflated Claims of Gender Differences
Clinical practice
Couple therapy: costs of assuming that men and women have vastly different communication styles
Faulty attributions to gender
“We can’t communicate because you’re a woman and I’m a man” vs.
We need to improve our communication skills.

57. In Conclusion “Men Are from Mars, Women Are from Venus” is false.
The truth: Men are from earth and women are from earth.

58. Thank you!
Janet Hyde

59. Gender Differences in Temperament (Else-Quest, Hyde et al., 2006)
What are the early, fundamental differences?

60. Gender Differences in Temperament: Effortful Control
Attention
-0.23
Effortful control
-1.01
Inhibitory control
-0.41
Low intensity pleasure
-0.29
Perceptual sensitivity
-0.38
These gender differences are substantial. Girls have somewhat better attention skills and are better at controlling their behavior. Consistent with the finding of the preponderances of boys among those with ADHD.

61. Gender Differences in Temperament: Negative Affectivity
Emotionality
0.01
Anger
0.04
Pleasure
-0.09
Sadness
-0.10
Fear
-0.12
These gender differences are small – gender similarities

63. Gender Differences in Moral Orientation
Care orientation: d = -.28
Justice orientation: d = +.19
(Jaffee & Hyde, Psychological Bulletin, 2000)

64. Gender Differences in Spatial Ability (Linn & Petersen, 1985)
Spatial Perception
+.44
Spatial Visualization
+.13
3-dimensional Mental Rotation
+.73

65. 3-Dimensional Mental Rotation

66. Gender Differences in Temperament: Surgency
Activity
0.33
Shyness
-0.10
High intensity pleasure
0.30
Impulsivity
0.18
Smiling
0.01

67. Selectivity of Sample The Greater Male Variability Hypothesis Sample d
Number of studies
General
-.05
184
Moderately selective
+.33 24
Highly selective
+.54 18
Precocious
+.41 15
Gender differences are larger at the upper tails of the distribution. Greater male variability hypothesis – male distribution has greater variance than female distribution. Even if this is true, it doesn’t explain why males are more variable.
The Greater Male Variability Hypothesis

68. National Assessment of Educational Progress (NAEP)
NAEP categorizes items as easy, medium, or hard.
Took hard items and coded for item complexity
Analyzed hard items at Levels 3 and 4 for gender differences
Result, grade 12
d = 0.07
Hyde et al. (2008)

69. Can Women Be Found among the Mathematically Elite?
Previous analyses have examined high scorers: the top 5% or 1% of the entire distribution
Doesn’t get at those who are profoundly gifted in mathematics
Another data set: the Putnam Mathematical Competition
Taken by 3,500 undergrad math students in U.S. and Canada
Majority can’t solve any of the 12 problems; the top 25 scorers solve 5 or more problems

70. Women Among Top 25 in Putnam, 1992-2007
Name
Year
Birth Country
IMO Medals
Olena Bormashenko
2004
Russia
1 gold, 1 silver
Ana Caralana
2003, 2004
Romania
1 gold, 2 silver
Ioana Dumitriu
1995, 1996
Julie Kerr
1992
USA
Suehyun Kwon
2003
South Korea
1 gold
Alison Miller
Greta Panova
2001
Bulgaria
Dana Pascovici
Melanie Wood
2001, 2002
2 silver
Wai-Ling Yee
1999
Canada
Inna Zakharevich
IMO = International
Math Olympiad
Andreescu, Gallian, Kane, & Mertz, Notices of the AMS (2008)

71. How to Interpret the Putnam Data?
Women exist among those who are profoundly gifted in mathematics
Is the glass half full or half empty? Focus on women who made it, or preponderance of men?
Clear role of culture in discovering and nurturing mathematical talent among girls and women

72. The Role of Culture
r = .44
r = .44
Hyde & Mertz, PNAS, 2009

73. But, says Larry Summers…
Two separate issues
Gender differences/similarities in the general population – average differences
Gender differences in the upper tail of the distribution, the highly talented
How can there be differences in the tail with no gender difference in average scores?
Gender differences in variance

74. The Greater Male Variability Hypothesis
Assuming d = 0, VR = 1.2
Green = female
Orange = male
Brown = overlap
Hyde & Mertz, PNAS, 2009

75. The Greater Male Variability Hypothesis
Originally proposed more than 100 years ago
Variance ratio
VR = VarM / VarF
VR > 1.0 means greater male variability

76. Grade VR Grade 2 1.11 Grade 3 Grade 4 Grade 5 1.14 Grade 6 Grade 7
1.16
Grade 8
1.21
Grade 9
Grade 10
1.18
Grade 11
1.17
Hyde et al. (2008)

77. Theoretical Distributions
If d = 0.05 and VR = 1.12, persons above 95%ile
Males: Females = 1.34
99.9%ile, exceptional talent
Males: Females = 2.15
Hedges & Friedman (1993)

78. Actual Distributions: Percent of Youth above Percentile, Grade 11
Above 95%ile
Above 99%ile
Ethnic
Group F M
M/F
Asian
American
5.7%
6.3%
1.09
1.4%
1.3%
0.91
White
5.4%
7.8%
1.45
0.9%
1.9%
2.06
Above 95%ile: n = 219 Asian Americans, n = 3473 Whites, State of Minnesota
Hyde et al. (2008)