An important question in instructed second language (L2) acquisition regards the relative effectiveness of comprehension-based instruction and production-based instruction for learning L2 grammatical forms. While recent meta-analyses (Shintani, 2015; Shintani, et al., 2013) show an immediate advantage of comprehension-based instruction for receptive knowledge and a long-term advantage of production-based instruction for productive knowledge, questions remain regarding the underlying cognitive mechanisms associated with language production, and how they may be particularly beneficial to the acquisition of L2 grammatical forms. In this talk I will discuss two recent studies from our lab: 1) A training study in which beginning L1 English-L2 learners of German learned German grammatical gender marking (e.g., ein blauer Becher “a.MASC blue.MASC cup.MASC” vs. eine blaue Schüssel “a.FEM red.FEM bowl.FEM”) via comprehension-based vs. production-based training units and (2) a structural priming study that measured the production of double object vs. prepositional object dative constructions (e.g., The boy gives the girl the book/The boy gives the book to the girl) among intermediate L1 Korean-L2 learners of English. Based on results from these two studies I will argue that learning is enhanced when learners must overtly produce targeted grammatical forms rather than simply comprehending these forms—and especially production-based activities that encourage learners to evaluate whether their own self-generated productions match predicted target forms. I attribute this advantage for production-based training to the cognitive mechanisms that underlie language production, including utterance planning and lexical retrieval, and the ways in which production-based training encourages learners to “notice the gap” between their own productions and target forms (e.g., Potts et al., 2019; Schmidt, 2001; Swain, 2005). Over time, these processes support the creation of stronger linguistic representations in memory than activities that only require learners to recognize target forms and map those forms to their intended meaning.