This paper reports findings from the first two years of a four-year longitudinal study into the ways that students' attitudes towards, and achievement in, mathematics are influenced by ability-grouping practices in six schools. Through the use of questionnaires administered to the whole cohort of 943 students, interviews with 72 students and approximately 120 hours of classroom observation, the relative achievement in, and the changes in attitudes towards, mathematics are traced as the students move from year 8 to year 9, with students in four of the six schools moving from mixed-ability grouping to homogenous ability groups or ‘sets’. Ability-grouping was associated with curriculum polarisation. This was enacted through restriction of opportunity to learn for students in lower sets, and students in top sets being required to learn at a pace which was, for many students, incompatible with understanding. The same teachers employed a more restricted range of teaching approaches with ‘homogeneous’ groups than with mixed-ability groups which impacted upon the students' experiences in profound and largely negative ways. Almost all of the students interviewed from ‘setted groups’ were unhappy with their placement.