Do gender gaps exist in classroom performance? The conventional notion holds that girls tend to do better in liberal arts subjects, whereas boys typically perform better in math and science. But the data do not necessarily bear out this widely accepted notion. To understand more about the test performance of males and females and the challenges that new assessment methods pose for fair testing, researchers from Educational Testing Service (ETS) recently completed a four-year study that analyzed results from more than 400 different assessments and 1,500 data sets involving millions of students. The authors, ETS Distinguished Research Scientist Warren W. Willingham and ETS President Nancy S. Cole (1997), found that gender gaps in the United States are not nearly as large or as pervasive as one might think. Their study exposes several myths about differential performance by gender and provides interesting avenues for further exploration.
The Willingham and Cole study looked at assessment performance by gender at the 4th, 8th, and 12th grade levels. For each grade, they looked at all available assessment data, being careful to control for the nature of the construct being measured, the comparability of the male and female samples, and the particular cohort of samples being assessed.
In addition, they needed to standardize results to assure comparability across assessments. For example, they wanted to compare SAT scores that are reported on 200-800 scales with ACT scores that are reported on scales ranging from 1-32. They chose to use the standard mean difference, or statistic D, to allow such comparisons. The statistic D is simply the mean score of females minus the mean score of males, divided by the average standard deviation. Thus, a positive D indicates that females have a higher average score, a negative D indicates males have a higher score, and a D of zero indicates no difference by gender. In general, an absolute value of D less than .2 indicates an insignificantly small difference (if any), .2 to .5 is considered small, .5 to .8 is considered medium-sized, and anything more than .8 is large.
Not Their Parents' Gender Gap
By comparing statistic Ds across assessments, Willingham and Cole found that gender differences among subject areas are smaller than one might expect. For example, the study combined data from 74 assessments at the 12th grade level to compare performance in 15 subject areas. For nine of those 15 areas, the D was between -.2 and .2, indicating essentially no difference between the sexes. Only two of the 15 subject areas were outside the "small" range of -.5 to .5. ("Mechanical/electronic" favored boys, with <DATE TYPE="2">D</DATE> less than -.9, and "verbal-writing" favored girls, with D greater than .5.)
And what of the belief that girls excel in liberal arts, boys in science and math? Although there was some evidence for these gaps in 1960, they appear to have narrowed considerably in math and science, with data from 1990 showing no male advantage in either subject (D between -.2 and .2). However, the 1990 data did show a slight writing advantage for girls (D equal to approximately .4) that remained largely unchanged from 1960.
Focusing on Subskills
The authors, however, caution against making simplistic assertions across broad subject areas like mathematics, science, and writing. For example, when they broke down math into its component subject areas, they found that females did slightly better on some subskills (such as math computation), and that males did better on others (such as math concepts). The same held true for language subskills. This finding suggests that when curriculum developers and teachers consider ways to address possible gender gaps in major subject areas, they should focus attention on the subskill level.
Gaps That Appear in Adolescent Years
Another major finding was that although the gaps across subject areas are minimal at the 4th grade level, gender differences appear to grow by the time students reach 12th grade. Exactly when the majority of this growth occurs appears to depend on the subject area. For example, girls' advantage in writing grows substantially between 4th and 8th grade, whereas boys' advantage in natural science does not occur until sometime between 8th and 12th grade.
Does the fact that boys and girls appear on relatively equal footing in 8th grade, only to have some gaps appear later, show that the gaps are the result of unequal educational opportunities? Not necessarily. Wilder (1997) notes that the growth of such gaps is likely the result of a complex web of factors that are virtually impossible to unravel. These factors might include biological differences, such as the hormonal changes that strike boys and girls with the onset of puberty. They might also be related to culturally defined sex role expectations. Wilder cites evidence of children being both more motivated to succeed and more successful in subject areas "when they regard achievement in those subjects as consistent with their own sex role." Another factor might be that girls' and boys' interests diverge as they get older, causing them to cultivate some skills more than others. Willingham and Cole note that girls were far more likely than boys to report "talking/doing things with parents" as a leisure activity (D = 1.19), whereas boys were more likely to report "using personal computers" (D = -.7).
Exactly how these complex factors interact is still unclear. It does appear safe to say, however, that what gender gaps do exist are probably the result of far more than just differential course-taking patterns or culturally defined gender expectations. Moreover, Willingham and Cole emphasize that individual uniqueness overshadows the impact of any possible gender differences. Students' natural interests, skills, and aspirations are likely to exert a far greater influence on their academic achievement than will their gender.
