Native prosodic structures constrain L2 word recognition: Evidence from Bengali-English bilinguals

Bilingual word recognition is assumed to be modulated by a word’s segmental and meaning similarity across languages, labelled cognate in psycholinguistics, usually conflating borrowed and inherited words. We conducted an ERP fragment priming study with Bengali-English bilinguals. English words borrowed into Bengali (doctor, Bengali: [ˈɖaktar]) were compared with those which were not (river). The stimuli varied in fine metrical details, one-foot (dóctor) or two-feet (éxpèrt) whilst stress placement was kept constant. Crucially, two-feet English words are always one-foot in Bengali [ˈeksparʈ]. Behavioural results (RTs) showed that although loan status did not affect priming, mismatch in feet significantly reduced the effect. In the ERP data, only one-foot words elicited significant priming effects. Furthermore, different ERP components were modulated depending on loan type. Thus, loan status alone is not sufficient to understand L2 word processing; the influence of the native metrical structure (preference for a single foot) constrains processing of all words.