An urban independent middle school grouped its 63 sixth and seventh graders into single-sex mathematics classes (SSMC) to improve girls' achievement in mathematics (AIM) and attitudes toward mathematics (ATM) with no negative impact on boys. Researchers analyzed AIM, ATM, and interactions/instruction. AIM measures included Metropolitan Achievement Test-7, textbook unit tests, and teacher-constructed tests. T tests, alpha=.05, showed no significant differences in mean scores for males and females. Quantitative and qualitative analyses of ATM as measured on the Modified Fennema-Sherrnan Mathematics Attitude Scales (1993) and through standard open-ended interviews (Isaac & Michael, 1995) indicated positive ATM and SSMC. Observations, field notes, and videotapes provided data for analyzing expectations, interactions, and pedagogy. Using Cazden's (1986) definitions of teacher talk, significant differences were noted--an objectivist approach in boys' classes and a constructivist approach in girls' classes. Within a single-sex class, a range of learning styles calls for a variety of instructional approaches. (Contains 43 references.) (Author)