We tested whether learning associated to lexical selection is error-based, and whether lexical selection is competitive by assessing the after-effects of producing words on subsequent production of semantic competitors differing in degree of error (translation equivalents). Speakers named pictures or words in one language (part A), and then named the same set of pictures (old set) and a new set in another language (part B). RTs for the old set (i.e., translation equivalents) were larger than for the new set (i.e., items which not have been named previously in another language). Supporting that learning is error-based, this cost was mostly larger after naming in a language with a higher degree of error (L2 vs. L1). Supporting that lexical selection is competitive, after naming in a language with a high degree of error (L3), the cost was larger for naming in another language with a high degree of error (L2 vs. L1).

Elin Runnqvist was supported by a predoctoral grant (FI) from the Catalan government (AGAUR) during the initial phases of this study, and later by a postdoctoral grant from the People Programme (Marie Curie Actions) of the European Union’s Seventh Framework Programme (FP7/2007-2013) under REA grant agreement n˚ PIEFGA-2012-329339. Kristof Strijkers was supported by an ANR grant from the French Ministry of Research (grant number: ANR-16-CE28-0007-01). Finally, this research also received support by grants of the Agence National de la Recherche (ANR) awarded to the Institute of Language, Communication and Brain (ILCB; ANR-16-CONV-0002) and the Brain Language Research Institute (BLRI; ANR-11-LABX-0036). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.