Data-driven learning (DDL) provides second-language (L2) learners with numerous samples of natural language from which they can detect patterns and understand usage. While largely restricted to the textual domain, the increasingly available number of multimedia corpora and concordance tools provide opportunities for learners to search and listen to natural speech. This entry reviews the theoretical underpinnings for applying DDL to pronunciation as well as the empirical work that provides guidance for its implementation. It also outlines a research agenda for future investigations of DDL for L2 pronunciation that is likely to enhance our understanding of how to best implement multimedia corpora considering design factors in a wide range of learning contexts.