For example, in a mathematics class for mature women in Australia, Vicky Webber focused a series of classes around an article attacking single mothers' welfare payments as a burden to the Australian taxpayer. Fuelled by anger, the women developed a sophisticated mathematical analysis of the misuse of numerical data in the article.

It is critical to consider how mathematics achievement is assessed. That girls do better in classroom assessment and boys on standardized tests raises important questions (Burton, Kimball). In the United States the exclusive reliance on standardized test results such as the SAT in determining college admissions and scholarships clearly discriminates against females. As a result some colleges are beginning to use both SAT results and high school grades, a procedure which reduces the over prediction of male and the under prediction of female performance in university. The issue of balancing assessment between classroom performance and standardized exams has been raised as well in Europe (Burton) and Australia (Willis) where the inclusion of course work as a component of the national or provincial assessment for high school students has had the result of raising all students' scores, especially those of girls.

Interestingly, there has been significant political opposition to the inclusion of course work based on charges of lowered standards and cheating. Assumptions underlying this position include a symbolization of standardized tests as tough, detached, hard, objective, i.e., masculine and elite; and classroom grades as soft, easy, subjective, deceptive, i.e., feminine and mediocre. Contrast these arguments with the existence of the previously . mentioned tutoring programs that coach privileged, primarily white U.S. students on how to raise their SAT scores.

This is not seen as cheating, although it gives an unfair advantage to a small group of students. In establishing more equitable methods of assessment based on different patterns of achievement by females and males, a benefit may also accrue to other marginalized students who are disadvantaged by traditional assessment procedures. What is important is that assessment measures fairly reflect the mathematical knowledge of all students. If the goal is a truly equitable mathematics education, it must be equitable not only for white middle-class females, but for all students. The establishment of gender-inclusive curricula and gender-fair assessment is one place to begin and such changes may benefit many male students as well. Conversely, teaching styles, content and assessment patterns that work for more than white, privileged male students will of necessity benefit female learners. Both of these attempts are worthy of feminist attention.

Meredith Kimball is a professor of psychology , and women's studies at Simon Fraser University in B.C. She has conducted research for several years on the math accievement of women. A longer version of this paper will appear as a chapter in a forthcoming book by the author, Feminist : Visions of Gender Similarities and Differences, to be published by Haworth Press. Comments and questions can be sent to Dr. Kimball at the Department of Psychology, Simon Fraser University, Burnaby, B. C. V5A 1 S6, Canada.

1. Statistical significance is determined by a formula that includes the number of women and men in the study, the average score for each group, and the range of scores within each group. The conventional level for declaring a result significant is p =.05. What this means is that there are five chances out of 100 that the researcher would find a difference this large or larger if women and men really are the same, and ninety-five chances out of 100 that s/he would not. Basically, the researcher takes a gamble based on probability and declares a difference as real if the probability that she is wrong is small (5% or ; less).
   
   
2. Technically, an effect size is determined by dividing the difference between the two means, by the combined standard deviation for the two groups. This means that an effect size is dependent both on how large the difference between the means is and how much variability there is within each group. For example, an effect size of.5 reflects that the means differ by half of the combined standard deviation.