Theories of language production are monolingual but the world is multilingual. In the domain of word-form encoding, it is clear that languages rely differentially on different phonological units, challenging the generality of the monolingual theories. To address this, we propose the proximate units principle, which holds that the initial selection of sub-lexical phonological units (syllables, morae, phonemic segments, etc) is crucial both to understanding language specific processing, and to identifying what is language general in word production. We define proximate units and the role they play in speech planning and execution. The proximate units principle is consistent with much of what is already known about word form encoding across languages but also makes new predictions and can bring greater clarity to interpretations of experimental and speech error data.