An important question about early bilingualism that concerns both parents and researchers is the degree to which one language may interfere with another. This question rests on an implicit assumption that learning more than one language must always produce confusion and/or interference between (or among) the languages. Although many naturalistic studies have addressed this issue, no firm answers are yet available from the conflicting results obtained. Several factors appear to be responsible for the contradictory evidence, including the small numbers of subjects in each study, the large number of different language combinations of varying similarity that have been examined, and the variety of linguistic input situations (e.g., sequential bilingualism, language separation between home and school, parent/language separation), that have been observed. There is a consensus that children's language mixing can be eliminated if parents adhere strictly to the principle of one parent/one language enunciated by Grammont (Ronjat, 1913). However, this claim has not been directly examined. The primary goal of the present study is to characterize the linguistic input available to a child growing up with two languages. Analyses of the data show that a large proportion of parents, even those firmly committed to maintaining a strict separation of language by parent, model linguistically mixed utterances for their children. This finding suggests that children's early language mixing does not reflect interlinguistic confusion. Rather, it suggests that the child is formulating hypotheses about language based on the data available, i.e., that using the language of both father and mother in a single utterance is acceptable.