Can learning explain cognate effects in bilingual comprehension and production?

Most theories of bilingual word recognition and production assume parallel, on-line activation of both languages, even in one-language contexts. A considerable amount of the evidence for parallel activation comes from the study of translation equivalents with similar form and meaning across two languages (cognates), which bilinguals process differently to translation equivalents with no form similarity across languages (non-cognates). The on-line account has been queried by Costa et al. (2017), who suggest that the cognate effect can be explained by learning: on-line cross-talk during second language acquisition would lead to different representations for cognates compared to non-cognates in the bilingual mental lexicon. In this chapter, we focus on these two hypothesised origins of the cognate effect and consider the extent to which cognate effects can be explained by learning and on-line activation.