The Shape of the Bilingual Mental Lexicon: Testing the Cognate Continuum

Items in the mental lexicon have three storage and processing strata, the concept, lemma, and lexeme, which equate to semantic, syntactic and phonological information. Lexical items relate to each other at each stratum. Bilingual lexicons, which contain items from all languages, may contain cognates, items sharing concepts and with overlapping lexemes. Because cognates likely relate at the lemma level also, this research proposed the Cognate Continuum, a categorization of cognates and noncognates in the bilingual mental lexicon. The Cognate Continuum includes three sets of cognates: (i) true cognates have the closest phonology and syntax. (ii) lemma cognates have close syntax with differing phonology. (iii) lexemic cognates have close phonology with differing syntax. The relationship of cognate pairs has been found to lead to differential speed and accuracy in processing, known as the cognate effect. Degree of phonological overlap has been shown to modulate the cognate effect, but degree of syntactic overlap has not previously been tested. Cognates’ perceived phonological similarity was tested in a Norming Study. In the language identification (LID) and self-paced listening (SPL) tasks, processing differences between these types of cognates was tested. In the LID, participants had to identify the language of real-word cognates and noncognates. Because all form representations are theoretically stored in a single lexicon, participants would have to suppress activation of one of the cognate pair’s lexemes. Thus, participants were expected to inhibit cognates, with greater inhibition for greater similarity, resulting in the following pattern: (a) noncognates \u3e\u3e lemma cognates \u3e\u3e lexemic cognates \u3e\u3e true cognates However, results indicated that only degree of phonological overlap affected cognate processing; further, facilitation was found instead of inhibition contra prior studies (Dijkstra et al., 2010). In the SPL task, participants listened to sentences presented in segments, pressing a button to advance. Facilitation was expected, and degree of overlap was again expected to mediate the effect, resulting in the following pattern: (b) true cognates \u3e\u3e lexemic cognates \u3e\u3e lemma cognates \u3e\u3e noncognates As expected, results indicated that both degree of syntactic and phonological overlap affected cognate processing. Overall, results suggest that syntactic overlap modulates the cognate effect in an intrasentential context, but not in isolation. Moreover, this suggests that words’ lemmas are not accessed without a syntactic context, highlighting the need for specification of lemmas in bilingual lexical comprehension models, with consideration given to task schema