The last two decades have spawned many research studies examining issues related to gender differences in the learning of mathematics. While there is still some debate about the nature of these differences and where and how they manifest themselves, some general conclusions are possible. These are well documented in the research literature and tend to be specific to either performance variables or affective variables( 情感变量 ). On affect, there is considerable evidence that males are more positive about personal aptitudes in mathematics when compared to females, despite the fact that similar performance differences are not substantiated by research evidence. Brandon, Newton and Hammond (1987) concluded that boys' achievement levels surpassed that of girls' levels at some point but that this point varied considerably from study to study. Generally, some of this emergent male superiority is attributed to greater exposure to formal mathematics training (e.g., Ethington and Wolfe, 1984, 1986). The greatest differences between boys and girls can be found in their attitudes to and self-confidence in mathematics rather than in actual achievement (Boli, Allen and Payne, 1985; Eccles, 1982). Some studies (Leder, 1988; Mura, 1987; Thomas and Costello, 1988) provided evidence of boys’ perceived superior competence and girls undervaluing their achievements in mathematics. Nevertheless, it is not the purpose of this paper to review the considerable literature on self-concept or to validate self-concept as a construct. There is however, systematic research over the past decade, which has provided the necessary scientific credibility (see for example Marsh, 1988, 1990; Marsh and O'Neill, 1984; Marsh et al., 1983) across age groups. To date however, there are no studies which specifically target the self-concept of teachers, particularly self-concept of teachers of mathematics. Nor has the link between the affective domain and performance been adequately established but there is increasing evidence of strong correlations between academic self-concepts and academic achievements which are content specific (Marsh, 1992). The direct link between self-concept and achievement may be tenuous but the evidence that teacher attitude affects student performance is stronger. Midgley, Feldlaufer and Eccles (1989) clearly demonstrated that teacher efficacy has a strong relationship with students' self-perceptions about abilities in mathematics. This association was found to be strongest for low achieving students who are taught by a teacher with low mathematics efficacy. The research reported here is based on a larger study (Relich, Conroy and Webber, 1991) in which it was hypothesized that gender differences in self-concept would be found between male and female teachers of mathematics and that therefore female teachers would present inappropriate role models for their students, especially their female students. Gender differences were found at the primary but not the secondary level. Furthermore, this difference was found to be attributable to the teachers' level of study of mathematics, that is, males were more likely to have studied mathematics at a higher level than females and therefore registered higher mathematics self-concept scores. What was not clear is why such differences emerged at the primary level. The major purpose of this research was to probe for the formulation of teacher self-concept in mathematics and the impact that such self-concept may have on how the teacher perceives the nature of mathematics and mathematics learning as well as his or her role as a teacher of mathematics. Such insights may provide important and relevant information for the development of pre-service and in-service courses for teachers of mathematics, especially at the primary level.