South African pupils performed well below the TIMSS international average in 1995 and 1999 and significantly below all other countries (including the other African countries) in the 1999 study. Path analysis, namely Partial Least Square (PLS) analysis, was applied to the South African TIMSS-R data to explore the effect of contextual factors at school level and classroom level within South African schools on the aggregated pupils' performance in mathematics. The results from the combined school- and classroom-level model revealed a relationship between the location of the schools, teachers' attitudes and beliefs, teaching load, lesson planning, and class size; all of which had direct effects on the South African pupils' aggregated performance in mathematics and in total explained 27% of the variance in the mathematics scores.