The pace at which new L2 words or expressions are acquired is influenced by the degree of engagement with them on the part of the learner. Several researchers with a Cognitive-Linguistics (CL) background have, since the 1990s, proposed ways of exploiting non-arbitrary aspects of vocabulary as stimuli for such engagement. Their proposals have been backed up by the results of several quasi-experimental studies. It must be acknowledged, however, that many of these are small-scale, some show only small effect sizes, and some are hard to interpret due to confounding variables. Taken collectively, the reported experiments are nevertheless beginning to constitute a body of evidence in favour of CL-informed instruction that is hard to dismiss, so there is reason to believe that this kind of instruction deserves a niche in second language programmes. However, a judicious implementation of CL ideas stands to gain considerably from a closer alignment with ‘mainstream’ second language vocabulary research. Insights to be taken on board from the mainstream concern issues of selection, the desirability of distributed learning, and the need to cater for complementary types of knowledge.