Description:
<jats:title>Abstract</jats:title><jats:p>This paper adopts a construction grammar approach to test the efficiency of data‐driven learning (DDL) when applied to constructions. High‐intermediate learners of English as a foreign language took part in an experiment consisting, for each construction, in (i) a pre‐test, in which the learners had to produce as many sentences as possible illustrating the constructions, (ii) a DDL intervention, which involved various concordance‐based activities, and (iii) several post‐tests, similar to the pre‐test, performed at regular intervals. The analysis looks at the number of (valid and invalid) constructions produced as well as the native‐like nature of these constructions, as established through a comparison with native corpus data. The analysis also examines the evolution of these results over the different post‐tests. Finally, it considers learners’ attitudes towards the DDL interventions, as evaluated through a questionnaire.</jats:p>