Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Orthography outweighs phonology in bilingual cognate similarity judgments
Poster Session F, Friday, October 2, 2:45 - 4:45 pm, Wangari Maathai
Kalil Warren1, Annie Xie1, Danielle Fahey2, Mila Tasseva-Kurktchieva1; 1University of South Carolina, 2University of Alabama
Introduction. Cognates, words sharing form and meaning across a bilingual's two languages, are processed faster and more accurately than non-cognates, an effect attributed in models of bilingual lexical access (e.g., BIA+) to co-activation of shared sublexical representations. Yet the relative contribution of orthographic versus phonological overlap to perceived cross-linguistic similarity, and how this depends on a listener's language experience, remains unclear. We asked whether subjective phonological similarity judgments of auditorily presented cognate pairs are driven by phonological distance, orthographic distance, or both, and whether French proficiency moderates the use of phonological information. Methods. Thirty-five English–French bilinguals rated the sound similarity of 255 French–English cognate verb pairs in 3Sg form (e.g., approves / approuve) on a 7-point scale during a speeded auditory task; each trial presented both members of a pair. Item-level predictors included normalized Levenshtein distances computed over orthographic (NLD-O) and phonological/IPA (NLD-P) forms, and log-transformed lexical frequency (COCA, Lexique3). Subject-level predictors included French and English proficiency (TruthNorth), and self-rated native language, and length of exposure. Ratings were analyzed with cumulative link mixed models (ordinal regression) with crossed random intercepts for subject and item; reaction times were analyzed with linear mixed models. Results. Orthographic distance significantly predicted similarity ratings (β = 0.13, p = .002): more orthographically similar pairs were rated as more phonologically similar. Phonological distance, by contrast, showed no reliable main effect on ratings (β = −0.02, p = .69), despite the task being explicitly auditory, suggesting that orthographic representations intrude on phonological similarity judgments. English lexical frequency independently predicted higher similarity ratings (β = 0.16, p = .002); French frequency did not. In reaction times, greater phonological similarity was associated with faster responses (p < .001), confirming a processing advantage for more similar cognates. Critically, French proficiency moderated the use of phonological distance: higher-proficiency listeners showed a significant positive phonological-distance effect on ratings (interaction β = 0.067, p = .003) that was absent in lower-proficiency listeners. Native-English speakers and those with longer English exposure gave systematically lower similarity ratings overall (p < .05), consistent with stronger first-language phonological representations producing finer cross-linguistic discrimination. Conclusion. Perceived phonological similarity in cognate processing is not a fixed property of the stimuli but depends jointly on orthographic co-activation and the listener's phonological experience. The dominance of orthographic over phonological distance, even in an auditory task, is consistent with models in which orthographic and phonological codes are obligatorily co-activated during bilingual lexical access. The proficiency-dependent emergence of a phonological-distance effect suggests that detailed second-language phonological representations are required before phonological overlap can be exploited, a behavioral signature with direct implications for how the bilingual brain weights sublexical codes across the cortical language network.
Topic Areas: Multilingualism,