More than half of the world’s population is bilingual, yet the neural architecture allowing these individuals to successfully communicate in two languages and switch back and forth between them is far from understood. In this dissertation, I present a series of studies that systematically target fundamental questions about bilingual language use across a range of conversational contexts, both in production and comprehension. The results unveil novel patterns of language control associated with naturalistic language use, which contrast with previous results from artificial settings. Following these findings, I propose a new theoretical framework of bilingual language organization, with an architecture that assumes a common selection principle at each linguistic level to account for attested features of bilingual speech in, but crucially also out of, experimental settings.