Divergent hypotheses exist concerning the types of knowledge underlying early bilingualism, with some portraying a troubled course marred by language delays and confusion, and others portraying one that is largely unremarkable. We studied the extraordinary case of bilingual acquisition across two modalities to examine these hypotheses. Three children acquiring Langues des Signes Québécoise and French, and three children acquiring French and English (ages at onset approximately 1;0, 2;6 and 3;6 per group) were videotaped regularly over one year while we empirically manipulated novel and familiar speakers of each child's two languages. The results revealed that both groups achieved their early linguistic milestones in each of their languages at the same time (and similarly to monolinguals), produced a substantial number of semantically corresponding words in each of their two languages from their very first words or signs (translation equivalents), and demonstrated sensitivity to the interlocutor's language by altering their language choices. Children did mix their languages to varying degrees, and some persisted in using a language that was not the primary language of the addressee, but the propensity to do both was directly related to their parents' mixing rates, in combination with their own developing language preference. The signing-speaking bilinguals did exploit the modality possibilities, and they did simultaneously mix their signs and speech, but in semantically principled and highly constrained ways. It is concluded that the capacity to differentiate between two languages is well in place prior to first words, and it is hypothesized that this capacity may result from biological mechanisms that permit the discovery of early phonological representations. Reasons why paradoxical views of bilingual acquisition have persisted are also offered.