The input factors that may cause variation in bilingual proficiency were investigated in 38 French-English bilinguals aged six to eight, of middle-to-high socio-economic status, attending an international state school in France. Data on children's current and cumulative language exposure and family background were collected through questionnaires given to parents and children. Language proficiency was measured using the standardised French and English versions of the Peabody Picture Vocabulary Test and the Student Oral Language Observation Matrix, a rating scale of oral competence, completed by children's French and English teachers. The results indicated significant correlations between the language proficiency measures and current amount of overall exposure to each language, as well as current input and output quantity. Cumulative length of exposure was also a significant predictor for all the language proficiency measures. Furthermore, the child's dominant language was a reliable indicator of variables related to language use, including the child's preferred language for speaking and reading and the language used with peers in the school playground. The implications of these findings are discussed, highlighting particularly the need to find ways of promoting the child's weaker language.