An aspect of language that shows large variability in learning success is that of collocations, a specific type of multi-word unit. Collocations are predictable and expected for the native speakers of a language (e.g., “catch a cold”, “run a business”). But for L2 speakers, learning incongruent collocations (those that have partially overlapping lexical make-up across the two languages) is highly problematic. Prior learning studies have suggested that input conditions that cause interference should be avoided. An alternative approach inspired in the literature on desirable difficulties is that, to learn and process incongruent L2 collocations efficiently, L2 speakers must learn to inhibit the equivalent L1 collocations. I will report on two studies: The first study tested the counter-intuitive prediction that learning would be improved when practice conditions induce L1-related interference. In a second study I propose a new approach to quantify within-language competition in collocations.