How cross-language similarity and task demands affect cognate recognition

This study examines how the cross-linguistic similarity of translation equivalents affects bilingual word recognition. Performing one of three tasks, Dutch–English bilinguals processed cognates with varying degrees of form overlap between their English and Dutch counterparts (e.g., lamp–lamp vs. flood-vloed vs. song-lied). In lexical decision, reaction times decreased going from translation equivalents without any cross-linguistic orthographic overlap to very similar but non-identical cognates. Identical cognates showed a large discontinuous processing advantage and were subject to facilitation from phonological similarity. In language decision, the effect of orthographic similarity reversed: a cognate inhibition effect arose, the size of which increased with orthographic similarity. Here identical cognates were markedly slower than other cognates. In progressive demasking, no orthographic similarity effect was found for non-identical cognates, but a semantic similarity effect arose. In addition, there was a facilitation effect for identical cognates of low English frequency. The task-dependent result patterns are interpreted in terms of four accounts of cognate representation and provide evidence in favor of a localist connectionist account.