The role of ethnic resources in the educational success of immigrants is highly disputed. Combining arguments of segmented assimilation and Coleman's concept of family social capital, this study investigates whether speaking one's language of origin at home relates to achievement by facilitating the mobilization of resources or their transmission from parents to children. Mediating and moderating mechanisms are disentangled and empirically questioned in regression models that predict the mathematical competences of immigrant students from Turkey, Poland and the former Soviet Union. Based on data from the German National Educational Panel Study, the results contradict the assumption that foreign language use will contribute to education through mechanisms of social capital. Regardless of parental human capital or the group of origin, the language used at home does not affect the students’ learning when German language proficiency is accounted for.