halpern article

following article by Diane Halpern was originally published in Skeptic, Vol. 2, No. 3, pp. 96–103, which was based on a lecture delivered by Dr. Halpern at the Skeptics Distinguished Lecture Series at Caltech, Sunday, April 18, 1993. We present it here in response to the brouhaha surrounding the comments by Harvard University President Lawrence H. Summers, who told Harvard faculty members that the likeliest explanation for the gender differences in math and engineering is inherent differences in cognitive abilities, not upbringing, education, or career choices. As always at Skeptic, we prefer to set politics aside and follow the science.

Dr. Diane Halpern is an international authority on the scientific study of gender differences, cognitive abilities, critical thinking, and family dynamics. She is the author of Sex Differences in Cognitive Abilities, Enhancing Thinking Skills in the Sciences and Mathematics, Thought and Knowledge: An Introduction to Critical Thinking, and Changing College Classrooms. She is Professor of Psychology at Claremont McKenna College and was the 2004 President of the American Psychological Association.

---

Sex, Brains & Hands —
Gender Differences in Cognitive Abilities

by Diane F. Halpern

When it comes to gender differences in cognitive abilities anyone who maintains a reasonable amount of skepticism may already be viewing this subject with the same open mindedness that you would apply to recent Elvis sightings. So much has been said and written on sex differences in cognitive abilities that it is difficult to separate the various claims and come up with empirically supported conclusions. My plan is to present some of the theories and research that have explored individual differences in cognition, and discuss what we know and what we do not know.

I am a cognitive psychologist and it is my interest in how we think that is the thread tying these seemingly diverse topics of “sex, brains, and hands” together. Like any detective, I have followed some intriguing clues about individual differences in human cognition and have reached some controversial conclusions. For the last several years I have been involved in what I have called “trial by media” or “science by press release.” One of the problems in discussing sex differences in thinking is that the public has received so much misinformation from the press, who are more interested in grabbing the reader’s attention and meeting a deadline than in understanding complex issues. Reporters tend to prefer misleading headlines that have more to do with selling newspapers than with the actual content of the articles. This is not good for science. It is not an unbiased process, and I have begun speaking out against it, especially when I found myself being misquoted and quoted out of context.

When I went into cognitive psychology I did not plan to conduct controversial research. It started when I was teaching courses in cognitive psychology and the psychology of women, and the same question about the relationship between sex (or if you prefer, gender) and cognitive abilities came up in both classes. It seems that almost everyone is interested in this topic, which is probably why it has received so much press coverage in the last several years.

In order to answer the question of how women and men differ in their thinking, I began to review and synthesize the research literature on sex differences in cognitive abilities. If you ever try a computerized search in this body of literature you will be overwhelmed with the number of citations on these topics. About 14 years ago when I was conducting my research for the first edition my book Sex Differences in Cognitive Abilities, I had planned to show the weakness of the evidence in support of biological bases for any claimed cognitive differences between the sexes. If there were any differences, it seemed to me that they must be small and insignificant. Instead I found that the differences are sometimes large and that some of the biological data used to explain the differences were too strong and too consistent to ignore. I also found that the effects are not simple, and that other variables influence the findings. We usually talk about laterality as left or right handedness, but it is really a continuous variable that extends over many different indices of right or left sidedness. Laterality interacts with sex so that the kind of answers we get to questions about sex differences in cognitive abilities depends upon what I call sex bilaterality interactions. That is, some of the results seem to depend on both one’s sex and one’s preferred hand.

In discussing gender differences in cognitive abilities, I sometimes feel like a dentist who unexpectedly hits a nerve while drilling in the mouth of a sleeping giant. The reason for the intense nature of this controversy is easy to understand. There are serious social and political ramifications to concluding empirically that there are systematic sex and laterality differences in cognitive abilities. Such conclusions have a tremendous potential for misuse and abuse. They could be and have been used, for example, to justify discrimination, and or affirmative action based on one’s sex and preferred hand. Since sex and handedness are biologically determined variables that are not subject to individual control, then one is “stuck” with what one has. If these variables are also linked to thinking skills, then it implies that some aspects of intelligence and cognition are biologically determined. From there it is a small step to concluding that if some groups (e.g., gender or preferred hand use) think differently, there is nothing society (the environment) can do about it. As we know, this sort of information has been misused to discriminate against groups of people in the past. So I certainly do understand the controversies involved. But that does not mean we should not study the issues.

The study of sex differences has also been criticized as being inherently sexist, because it creates an emphasis on the way women and men differ, while ignoring the multitude of similarities. This is undoubtedly true. But I find the reasons for conducting such research to be much more persuasive than those against doing so. First, arguments against studying individual differences are frequently based on the assumption that if the truth were known, women’s deficiencies would be revealed. In my text, Sex Differences in Cognitive Abilities, I call this the “women have less fallacy.” This is simply not true because researchers have shown that there are areas in which females, on the average, excel, and areas in which males, on the average, excel. But differences are not deficiencies. The study of sex differences, like any of the other individual or group differences that psychologists study, is not a zero-sum game where one group gains only at the expense of another. The problem lies not in the fact that people are different. It is in the value that we attach to these differences. Second, it is only through such studies that similarities can be revealed. We cannot understand the ways in which people are similar without also examining the ways in which they differ. You simply cannot study similarities without studying differences. Perhaps most importantly, sex differences research is the only way that we can empirically determine if common myths and stereotypes about men and women have any basis in fact. The sole alternative to knowledge is ignorance, and ignorance does not counter stereotypes or dispel myths. High quality research is the only way we can determine whether, when, and how much women and men differ. It is the only way that we can reject false stereotypes and understand legitimate differences.

The question of gender differences is really a set of questions and not a single question. I have organized the issue around five questions:

1. Are there sex differences in cognitive abilities? In other words, is there sufficient solid empirical evidence that females and males, on the average, perform differently on valid tests of cognitive abilities? The other questions are only meaningful if we conclude that, in fact, some differences exist.
2. If there are differences, when in the life span do these differences appear? Are they present at very young ages or only later in life?
3. How large are the differences? In other words, even if there are statistically significant differences, are they large enough to be of any practical importance?
4. Are the differences due to factors inherent in the biology of maleness and femaleness, or are they due to cultural and environmental experiences and expectations? Theoretically, this is the most important question and one that is familiar to all students of psychology. There are mountains of literature on this subject. When I put the books and journal articles together that I have used for the background to my research, I can measure them in yards. Despite all this research, we are still wrestling with that age old question of whether nature or nurture plays a greater part in sex-related cognitive differences. These are exceptionally controversial and extremely political questions. And like all loaded questions the answers we get sometimes backfire. Results can and have been used in ways that support discrimination, and programs used to redress discrimination. Science in this area is not impartial.
5. The final question is an applied one, and for many reasons may be the most important one. As concerned citizens, parents, researchers, educators, and especially scientists, math educators, and skeptics, what should we be doing with this knowledge?

I am going to start my discussion with the first question. Are there valid sex differences in cognitive abilities? I have told my students that all difficult questions in life have exactly the same answer: “It Depends!” The answer to this first question depends on whether, where, and when we find sex differences in cognition, and what variables cause these differences. First, it depends upon the specific cognitive ability that we are examining. The majority of the literature in this area has investigated differences in three different cognitive domains: verbal, visual-spatial, and quantitative abilities. But terms like verbal, visual-spatial, and quantitative are category headings used to organize and study cognition. They are not unitary constructs. Verbal ability for example, applies to all the components of language usage, including skills like word fluency, grammar, spelling, reading, vocabulary, verbal analogies, and language comprehension. Examples of items used to tap verbal ability include selecting words that most nearly are the same in meaning; a variety of vocabulary type questions; verbal analogies (e.g., an igloo is to Indian, as tepee is to: ice, canvas, eskimo, or home); reading comprehension, where you would have a complex passage and are asked questions about the material that has been read; simple grammar questions (e.g., which is correct: “give the money to Bob and me,” or “give the money to Bob and I”?). You can see that these are certainly not tapping the same sort of skills even though they all involve language.

Spatial abilities are also not unitary, and there are a least four separatable components to spatial abilities. One of these components involves spatial perception. If you were a subject in an experiment looking at some of these differences, you might be given what is called the “rod and frame test.” (It is an old test that has been around many decades and has a terrible history of being misused in psychology). If you were a subject in this experiment, you would be sitting in a darkened laboratory with a tilted frame that is glowing in the dark. You would have a knob in front of you, and by turning the knob you would adjust the position of the rod within the frame to the vertical. What we find is that some people are good at aligning it to the vertical, while others are influenced by the tilt of the frame. That is a test of spatial perception.

Another common spatial test is “mental rotation.” For example, if you were to rotate one figure of the two figures over letter A, would they be exact or different, similarly for figure B. This is a mental rotation test.

The next component is spatial-visualization. These are imbedded figures tests that might involve a booklet where the subject must trace the figure on the left that is imbedded in the one next to it. The fourth measure of spatial ability is called spatial-temporal, involving movement over space and time. It tests judgments about the speed and direction of movement.

Quantitative ability is also a heterogenous field. Consider the differences among tasks like simple rote multiplication, word problems, and other more advanced topics in mathematics, some of which are spatial in nature, like calculus, topology, and geometry. Whether and when you find sex differences depends upon the ability being assessed. Differences among the type of test given, the nature of the subject pool, and numerous other factors have generated numerous contradictory findings and unreplicated claims (and some name calling for good measure). The short answer to what is really a long question is yes, there are some tests of verbal, visual-spatial, and quantitative abilities that show consistent sex differences. But the short answer does not do justice to the literature because there are so many tests that do not show such differences.

Sex differences are most reliably found in the tail ends of mental ability distribution (by this I mean a “bell-shaped curve” which is normally distributed with the upper and lower ends as the tails). Consider the highly publicized studies by Benbow and her colleagues concerning mathematically precocious youth who score extremely high on the mathematics portion of the SAT test. As most of you probably know, and I am going to assume you know this because it has been carried in every major newspaper, news magazine, radio, and television show, males substantially outnumber females among this elite group of young people. The statistics are startling: sex differences in the ratios of males to females are two to one for those scoring over 500, four to one for those scoring over 600, and 13 to one for those scoring over 700! This does not mean, of course, that there are no girls in this group or that girls cannot attain the highest levels of math achievement. Obviously the girls are there, but they are there in reduced numbers relative to the number of boys.

Interestingly, Benbow also found that her sample of young people who are extremely gifted in mathematics is disproportionately left-handed. (This work has been replicated by other researchers.) Benbow and her colleagues have found that these differences have remained stable for the last twenty-five years. (It is interesting to note that we have considerably less data about verbally precocious youth. This is significant because verbal ability is necessary for comprehension and communication in every field of study including mathematics.) There are also disproportionately more males at the low end of cognitive abilities distribution, with males overrepresented in some categories of learning disabilities and retardation. The low end of verbal abilities provides a very clear example of this. Stuttering, a disability of the production of fluent speech, is overwhelmingly (but not exclusively) a male problem. Approximately four to five percent of the population are considered stutterers. Of this large number there are three to four times more male stutterers than there are female stutterers, and correspondingly stuttering is much more common among left handers.

Similarly, dyslexia, a severe reading disability found in individuals whose other cognitive abilities are within normal ranges, is also predominantly, though not exclusively, a male problem. Approximately two percent of the school population is dyslexic. (It might interest you to know how they get this figure. There is funding for two percent of the school age population, so two percent are dyslexic. If a kid changes school districts he or she could become dyslexic or nondyslexic depending where the cut is made. Moderate dyslexia is five times more likely to occur in males than in females, and severe dyslexia is ten times more likely to appear in males. Dyslexia is also more likely for left-handers than for right-handers. These data demonstrate strong sex-related differences, but we need to keep in mind that differences are much smaller for the vast majority of the population that does not fall into the tails of the distribution. So for most of the population, the differences are much smaller in size.

When in the life span do these differences appear? You can probably guess the answer to this question: It depends! It depends upon the type of test that is given and who is being tested. Some of the differences show up very early in life. There are reports that girls, on the average, talk sooner than boys. They have what we call “longer mean utterance length’s – a psychological term describing how children string words together in the language acquisition process in order to communicate. Girls develop the use of passive voice and other complex grammatical constructions and advanced comprehension at an earlier age than boys. We are talking about the quality, not the quantity of what is being learned and produced. It is interesting to note that there is one traditional verbal area in which males excel, at least at adolescence, and that is in solving verbal analogies.

Consider, for example, the figure below which displays the time trend in sex differences on the verbal portion of the SAT test.

First notice that, in general, SAT-verbal scores have gone down. In 1967, girls were scoring higher than males, on the average. Around 1971–1972 the lines cross and males are now significantly out-scoring females on the verbal portion of the SAT. This is due, at least in part, to the fact that more girls are now taking the SAT test, which we would expect would lower the average score for girls. So that is thought at least to be part of what is happening in explaining these data. Peterson and her coauthors have investigated the nature of gender differences in visual-spatial abilities. In an extensive review they concluded that reliable gender differences are found at around age seven or eight. These differences increase at around age 18, and they continue throughout the life span. More recent studies have shown reliable sex differences in visual spatial tasks by age four and one half, prior to kindergarten, which is probably as early as it can reliably be measured. Other researchers are using more physiological measures (e.g., brain activation) to test children at even younger ages, but it is still to early to determine if sex differences in spatial ability can be found among toddlers.

Developmental trends in quantitative abilities are harder to pin down. There seems to be a clear advantage in arithmetic for girls in the early elementary school years, with several tests showing girls out-scoring boys in computational arithmetic. All of those tests that children take nationally and internationally, in the early grade school years show that girls, on the average, score higher than boys. But, the trend reverses in early to mid-adolescence, when the advanced math courses are introduced. This finding is complicated by the fact that there are a number of other changes that are occurring at early to mid-adolescence, any one of which could help explain the data. As to the trend of the mathematical portion of the SAT, it is dramatically flat, if one can have a dramatic non-trend. What we see here from 1967 on for a 25-year period, males are scoring 47 to 50 points higher than females on the average across all those years.

The literature on aging shows that verbal abilities in general tend to stay high into old age. Spatial skills seem to decline at a more rapid rate for everyone as we age. The aging literature, of course, has to be interpreted with extreme caution because of the obvious problems in cross generational comparisons. Sex differences for cognitive abilities in the elderly are extremely difficult to study. For starters, it is very difficult to get large samples of old people, particularly older men. As you probably know, men on the average die six to seven years younger than women.

At least some of the cognitive differences are quite large, although most of the differences are not. The effect size for tasks involving rapid mental rotation, for example, is among the largest effect sizes in the psychological literature. It is almost one standard deviation. This means the two sex distributions are almost one standard deviation apart. That is a lot. Conceptually, this effect size is as large as the difference in IQ between college freshmen and their professors. (When I tell this to college freshmen they are, amusingly, unimpressed!) Or alternatively, this is as large as the difference in height between 13-year old and 18-year old girls.

Although there is less research on handedness with these tasks, it seems that left handers also excel at mental rotation tasks. There are also reports of very large effects on spatial-temporal tasks favoring males, and tests of associational fluency favoring females. The effect size for something like associational fluency (finding words with similar meaning) is almost one and a quarter standard deviation units apart. Cohen, who is the guru of effect-size statistics, has interpreted effect size of this magnitude as so large that tests of statistical significance are not needed. The data are self-evident. They need nothing more sophisticated than a binocular test of significance. (If you are not familiar with this high level statistical jargon, a binocular test of statistical significance consist of looking at the data with both eyes open and concluding that they are different.) There is little overlap in the distributions. These effects are much larger than those found in psychological research in other areas and fields of study.

The largest differences are found on timed tests, particularly reaction time tests, which are frequently the dependent measures in mental rotation studies. Although the effect sizes are large compared to the other topics psychologists study, when the dependent measure is reaction time, we are dealing with differences that are measured in fractions of a second. We measure reaction time typically in milliseconds, so the unit is a thousandths of a second. Practically speaking I do not know what it means to say there is a 40 millisecond, or a 200 millisecond, even a 500 millisecond difference, or whether these differences in fractions of a second add up to some practical significance at the end of the day. That is not the kind of question for which I think we have answers. Even when you have these larger effect sizes their practical significance is unknown.

The division of abilities into verbal, visual-spatial, and quantitative has been useful, but there are alternative ways of investigating the thinking process. One way is to think about what it is an individual does when she or he is engaged in a particular task. I may be more useful in understanding the data to look at the underlying cognitive processes. I have summarized some of the tasks in which males and females tend to differ. In general, those tasks in which females tend to excel and exhibit large differences involve generating synonyms, producing language fluently, and computing and solving anagrams. The underlying cognitive process for these tasks seems to involve rapid access to and retrieval of information in memory. By contrast, the tasks in which the literature shows that males excel are verbal analogies – the mapping of meaning in relationships, mathematical problem solving, mental rotation and spatial perception, and using dynamic visual displays. Here the underlying cognitive processes involve manipulating and maintaining a mental representation. There is a large body of literature in cognitive psychology related to the issue of sex differences that looks at those processes involved when individuals use their short term visual memory to access information from long-term memory, so this conceptualization fits with the mainstream cognitive literature. Theoretically the most interesting question to ask is why do these differences exist? I am certain that the differences are due in large part to socio-cultural factors. If you look at cross-country data, and within countries analyzing for socio-economic status, virtually every investigation shows large main effects for culture. Undoubtedly much of the difference is due to variables like culturally determined sex roles, expectations, and learning histories, which include the kinds of toys we are given as children and the adult roles to which we aspire. Thus, data must always be interpreted in the context of the society in which they are collected.

Although I believe that we cannot underestimate the importance of environmental variables, what I would like to do for the rest of this presentation is summarize some of the evidence of the biological explanations for at least some portion of the sex differences found with cognitive tests. The socio-cultural influences are relatively noncontroversial in that virtually every researcher acknowledges that they are important. In addition, numerous biological explanations have been proposed. In considering these biological hypotheses, however, I keep hearing a little voice – something I read when I was an undergraduate by a psychologist named Weinsteen who offered this sternly worded caveat – biology has always been used as a curse against women. I try to keep her warning in mind whenever I review biological theories of cognitive sex differences.

Some have suggested that psychology should not study the biological basis of sex differences because biologically-based theories legitimize negative stereotypes of women. I respond to these critics by noting that silence does not counter stereotypes, ignorance does not promote equality, and differences are not deficiencies. We have had stereotypes a lot longer than we have had research. I think it is time to look at what research has to say. I understand, however, the concerns of those who fear biologically-based theories. Some of the theories have been ludicrous, for example, the hypothesis that women have smaller and therefore inferior brains, an idea very popular around the turn of the century or the mistaken notion that women should eschew serious academic pursuits because studying these topics would use blood that was needed for menstruation.

One hypothesis that has garnered recent support concerns sex differences in lateralization and/or structures mediated by prenatal hormones, prenatal stress, and or sex differences in adolescence maturation rate. Let me provide a very brief introduction to this very complex area. A large body of research has revealed that for most right-handed people, the right hemisphere tends to be more dominant for nonlinguistic spatial tasks, and the left hemisphere more specialized for verbal tasks. About half of all left-handers show this pattern of dominance, with the remainder showing either reverse dominance or equal representation of these tasks in both hemispheres. Thus, hand preference became a rough and imperfect indicator of brain organization. (You may be interested to know that critics of the original studies criticized this work because the researchers used Caltech students for subjects. The critics said Caltech students were atypical and results obtained with such unusual subjects could not be generalizable back to the general population.) Hand preference research parallels sex difference research in many ways, some of which I have already mentioned. Left-handers are over-represented in certain categories of mental retardation, are more likely to be dyslexic, and more likely to have stuttering problems. They are also more likely to be among talented adolescents identified as mathematically precocious, and they are over-representated, relative to their proportion in the general population, in architecture and mathematics. So what we have again is differences, not who or which group is better. Given that the type of abilities that differ by hemisphere of representation are the same ones that differ by sex, it seemed to be only a short leap to then argue that the sexes differ in the way their hemispheres specialize these abilities. There is a large body of experimental research using such paradigms as dichotic listening, direction of eye movement during cognitive tests, post mortems, EEGs, split brains, WADA Test, patients following localized brain surgery, and speeded tapping and hand movement. The results that came from these diverse paradigms are not entirely consistent, and I think it would be surprising if they were because they are so different in terms of what they are looking at. However, when differences are found, they usually support the notion that females maintain a more bilateral organization of cerebral function, at least for verbal tasks, and males more often demonstrate greater cerebral lateralization. Other researchers have demonstrated differences in the way cognitive structures are distributed by function within each hemisphere. Another sex-related brain difference that may be important in cognition (and has been replicated several times) is the finding that there is a portion of the corpus collosum – a thick band of neural fibers that connect the two halves of the brain – that is larger in females than in males and larger in left-handers than in right-handers. The most recent research in this area is showing that prenatal ovarian hormones are important determinants in the size of the corpus collosum. Although it is a long leap to extrapolate from brain structure to ability and behavior, numerous researchers have suggested exactly this sort of link.

The idea that the brain is a sex-typed organ has generated a great deal of interest. There is a large and growing body of literature that suggests that cognitive abilities vary both as a function of one’s sex and preferred hand, that is, whether you are more or less left or right sided. Some of the most recent research is showing that prenatal hormones, the ones that direct and reflect the sexual differentiation of the fetus, are the same ones that determine handedness. Consider, for example, a large study in which the researchers reported sex by handedness interaction on cognitive tests. They used three large samples in different geographical areas of the country, so they had built-in two replication samples. They used multiple measures of spatial and verbal ability, and they found that while, overall, males performed better than females on 14 out of 15 of the different spatial tasks, across three geographically distinct samples, they found that left-handed males performed poorer than right-handed males on all 15 of these tests across all three samples. On the other hand, left-handed females performed better than right-handed females on 12 of these tests. Reverse results were found with verbal abilities with right-handed females out performing left-handed females, and left-handed males out performing right-handed males. It is not important that you keep the specific sex by handedness interactions straight; what is important is with replications and large numbers of tests, many psychologists are finding differences that depend on one’s sex and one’s laterality. Sex by handedness interactions have been noted by numerous other investigators, although they are not all easy to interpret.

These results are particularly important because we have no reason to believe that sex role pressures, learning environments, or any other psychosocial variable differs as a function of laterality. That is, there is no environmental hypothesis that we have that can explain these results. We do not socialize left-handed girls differently from right-handed girls, or left-handed boys differently from right-handed boys.

There are several theories that have been designed to explain some of these sex by laterality differences. The most popular one is by Geschwind and Galaburda, who proposed a biological theory of cognitive sex differences. They believe that prenatal hormones are important determinants of brain development. By itself, that is not a very controversial position. Geschwind and Galaburda also found strong positive relationships among left-handedness, high levels of prenatal testosterone, both chemically induced (people taking drugs) and secondary to maternal stress, and allergies such as asthma, hay fever, and other immune disorders, particularly those involving the thyroid. We also know that there is a greater proportion of males than females who are left-handed, which would be predicted by this theory, because males are exposed to greater levels of prenatal testosterone. The next plausible question is whether there are any data that might support this relationship among sex, handedness, prenatal hormones, and cognition. The answer is yes.

As I already mentioned, Benbow and her colleagues found large and consistent sex differences favoring males among those who are most gifted in mathematics. This difference is found in early adolescence prior to differential course taking, and prior to, in most cases, the onset of adolescence. Using the same subject pool, Benbow has recently documented physiological correlates of extreme mathematical giftedness that includes significant increase in left-handedness, allergies, myopia, and relatively late puberty, on the average. The underlying theoretical position is that the same prenatal hormone that determines the sex of the developing fetus also influences other organs that are being formed at the same time, notably the left hemisphere. According to Geschwind and Galaburda, the left hemisphere matures at a slower rate than the right, therefore is more vulnerable to a whole variety of influences. The theory proposed by Geschwind and Galaburda, and others, is that high levels of prenatal testosterone cause slow neuronal growth in the left hemisphere and impair development of important immune system structures (the thymus). With this model, they predicted and found positive associations among being male, left-handed, immune disorders, and anomalous right-hemisphere abilities.

It is very difficult to explain associations among a set of variables as diverse as these without some sort of unifying theory that would help to tie them to some common origin or common influence. Unfortunately this gets even more complicated. Other evidence in support of Geschwind and Galaburda’s hypothesis were provided by Sanders and Ross-Field. I believe they were among the first researchers who reasoned that male homosexuality might also be determined by the same prenatal variables that are involved in cognitive sex differences. This possibility lead to the prediction that male homosexuals, as a group, would resemble females in their cognitive abilities more than they would resemble heterosexual males. Using several different tests of spatial ability, they found that their samples of male homosexuals demonstrated spatial abilities similar to that of the female samples. Both male homosexuals and females were significantly lower in their visual spatial abilities than the heterosexual males. They replicated this finding in three different experiments, and it has now been replicated by several other investigators. It seems that many people are supporting the same finding. There are also several reports in the literature showing that male homosexuals and male and female transexuals, are more likely to be left-handed than any other groups. This suggests again that these variables are related in ways that are not easy to unravel.

We certainly do not have a tight package of explanations, but we have, perhaps, some hint at what is happening. Additional support for the notion that sex hormones affect cognitive processes come from the highly publicized studies (front page news, above-the-fold kind of studies), that say there are slight variations in cognitive performance for menstruating women, as a function of the portion of the menstrual cycle they are experiencing. Recent studies of girls with Congenital Adrenal Hyperplasia (CAH), a condition in which girls are exposed to high levels of adrenal androgens prenatally also support this relationship. CAH is usually detected soon after birth and corrected. So we have a group of girls who differ from normal girls only in that they were exposed to high levels of androgens prior to birth. Relative to control groups, the CAH girls score extremely high on tests of visual-spatial ability. They also show strong preferences for what experimenters identify as boy-typical toys (transportation and construction toys). Other researchers found that males with extremely low levels of testosterone in adolescence have extremely poor visual-spatial abilities. These results, along with other experimental findings, support the crucial influence of sex hormones on cognitive abilities. These data are consistent with a huge nonhuman animal research literature that shows the gonadal hormones play a major role in the development of sex differences in behavior in the brain in a variety of other species. They are not readily amenable to psychosocial explanations. They are also not popular data. I am more comfortable making fun of studies, like the one that show prenatal hormones affect the toy preferences of toddlers, than I am with seriously trying to explain them. I grew up in the decade of the 1960s and at that time, I had very clear notions about the determinants of sex differences. I was especially interested in what caused human behavior – which is why I became a psychologist. I argued my beliefs with fervor and I have never been surer about anything as I was in the 1960s. I recited what I call “the old party line.” First, there are no sex differences, other than those involved in reproduction. I remember arguing with one of biologist, telling her that I wanted to know about the important differences. She looked at me and said, “reproduction is important.” So I realized then that I would have to clarify my interests. I believed that any evidence showing cognitive sex differences could be explained by experimenter bias, by flaws in the data, sloppy researchers, etc. In the 1960s I believed that I could explain away any study that found differences. But as the data was accumulated and the evidence that there are sex differences in cognition became impossible to ignore, I changed my explanation of choice and I espoused the point of view that the differences were too small of be of any practical significance. I really perfected my “small-effects size” argument. It seemed that just as I perfected that argument, large-effect sizes began creeping into the literature. When this response would no longer work, I once again knew with certainty that such large differences that were found could be attributed completely to differential socialization practices. In this way I was able to maintain for many years a tidy explanation of how and why females and males differ with regard to variables that are unrelated to reproduction.

What do I know now? Well, I know a lot more, but am certain about a lot less. I discovered that explanations of cognitive sex differences are much more complex than some single point along a continuum with biological at one end and psychosocial at the other. We will never be able to say, for example, it is 40% of one and 60% of the other. I now know that psycho-bio-social interactions are needed; ones that, in fact, recognize the reciprocal effects that psychology, biology, and sociology have on each other. I know that we need a theory that recognizes that experience alters the biological underpinnings of behavior. Our experience changes our biology, which in turn, influences the types of experiences to with we are exposed. Our knowledge of the way brain structures and organization direct cognition is still sketchy and incomplete. It is probable, if not absolutely certain, that current theories will be replaced with more sophisticated ones as our knowledge about the relationship between brain structure and organization and cognitive abilities increase. None of this is meant to imply a new breed of biological determinism. Biological theories do not imply inevitable or immutable outcomes. We also have to learn from our past – theories that seemed promising at first have not held up under repeated investigation. The same fate may befall today’s theories of differences in brain structure and organization.

For our colleagues who are in the business of education, I want to stress that all of the evidence supports the notion that most cognitive skills are readily educable or trainable. Despite the intriguing nature of the recent biological hypothesis, it is important to keep in mind that the single most important determinant of whether girls take higher level mathematics or not is parental attitude, not ability. The most important determinant of achievement in any field is educational level. As a society we should be concerned with developing the intellectual potential of every individual to its fullest.

The reason for this concern is not merely altruistic. In an article in Science, Stein reported that mathematical achievement among the top 5% of 12th graders is lower in the United States than any other industrialized nation. Eighth graders in the United States are below international norms in solving problems that require higher order thinking skills. In the United States more than two-thirds of all the bachelors degrees and 80% of the doctorates in mathematics are earned by Asian and white males. If biological variables are involved in determining quantitative or mathematical ability, or any other ability for that matter, their effects are much too small to account for this order of magnitude. Biological theories also cannot explain the scarcity of males from other minority ethnic groups in higher mathematics and in the sciences. The biology of femaleness and maleness does not change as a function of ethnicity. Even the most able girls begin dropping out of advanced mathematics as soon as these courses become optional. By the 7th or 8th grade we are losing our very best. It is important to encourage all able students, but particularly girls to stick with math. As citizens and as educators we have an obligation to help girls and boys realize that math counts. It is also important to provide visual-spatial training as early as possible in the elementary school years. The importance of visual-spatial skills has been recently been shown by Casey, for scores on test of visual-spatial ability can predict whether girls will select math or science majors in college. We have remedial reading classes filled mostly with boys, but almost no curriculum to encourage the development of visual-spatial skills that the data suggest would be needed more often by girls.

Finally, we must not fall prey to the dangers of self-fulfilling prophecies. The data presented here represents average differences, based on large samples of males and females. No single individual is average. Group average data have little to do with individual performance. There is considerable between-sex overlap in all of the cognitive abilities, with large numbers of males demonstrating high verbal abilities, and large number of females demonstrating high visual-spatial quantitative abilities. The literature concerning cognitive sex differences has been proliferating in recent years because the questions are of profound human interest. But the most important issue is not how women and men differ on the average. We should keep in mind the words of the 18th-century British writer who was once asked, “Which is smarter, men or women?” He replied: “Which man, which woman?”