Light Verb Constructions (LVCs) are combinations of a partially delexicalized verb and a noun indicating an action or an event (e. g., give a description). Studies modelled on the Meaning-Text theory and qualia roles of the Generative Lexicon model propose that LVCs combine nouns and verbs according to shared underlying, underspecified semantic features. This paper explores these features’ role in translating English LVCs into Spanish. Data come from the parallel corpus P-ACTRES 2.0, including fiction and nonfiction materials. Results indicate that translation choices for LVCs with have, take, make/do, and give are governed by the same notions of transference, inception, and volition that determine their combinatorial compatibility. Register also seems to influence the choice. This information may prove significant to machine translation, bi/multilingual writing support, and post-editing.