This paper suggests that large-corpus linguistics might help resolve the impasse, caused by mainstream linguistics' favoring of formalism over functionalism, regarding three conditions a valid language science has to fulfil: coverage of language data, data convergence, and linguists' consensus on how to proceed.