It is commonly thought that children learning two languages simultaneously during infancy go through a stage when they cannot differentiate their two languages. Virtually all studies of infant bilingual development have found that bilingual children mix elements from their two languages. These results have been interpreted as evidence for a unitary, undifferentiated language system (the unitary language system hypothesis). The empirical basis for these claims is re-examined and it is argued that, contrary to most extant interpretations, bilingual children develop differentiated language systems from the beginning and are able to use their developing languages in contextually sensitive ways. A call for more serious attention to the possible role of parental input in the form of mixed utterances is made.