This study examines the effect of teaching mathematics in French on mathematics achievement evaluated in English. In this context it analyzes the effect of increased intensity of bilingual education on mathematics achievement. It also analyzes the effects of language of testing in the context of French immersion at the intermediate level. The participants in the study are two cohorts of French immersion pupils followed from Grades 4-7. The treatment group received 80% of the core academic curriculum, including mathematics, in French and 20% in English. The comparison group received 50% of the core academic curriculum in French and 50%, including mathematics, in English. Achievement in mathematics was measured for both groups at the end of Grade 6. Analyses of covariance showed an advantage in mathematics for the 80% French group compared to the 50% French group. These results provide further evidence for Cummins' threshold hypothesis and interdependence hypothesis.