Dissertations, Theses, and Capstone Projects

Date of Degree


Document Type


Degree Name



Cognitive Neuroscience


Valerie Shafer

Subject Categories

Cognition and Perception | Cognitive Science | Phonetics and Phonology | Psycholinguistics and Neurolinguistics


Mismatch Negativity, Speech Perception, ERP, Bilingualism, EEG


This study examines the effect of linguistic experience on the neural processing of Voice Onset Time (VOT) in Hindi and Romance language (Spanish and Portuguese) individuals who are bilingual in English and monolingual English speakers using the event-related potential (ERP) Mismatch Negativity (MMN) response. VOT is a linguistic property that measures the time elapsed between the release of a stop consonant and the beginning of voicing, that is, vocal fold vibration of a following vowel. In a double-oddball paradigm, participants’ (n = 41) ERP were recorded while listening to speech sounds differing in VOT. The bilabial short lag stop [p] was presented as the frequent (standard) stimulus, while two different bilabial long-lag stops [ph] and two different tokens of bilabial voicing lead stop [b] were presented as infrequent (deviant) stimuli in two conditions. The stimuli were naturally produced and edited consonant-vowel (CV) syllables [pa], [ba], and [pha]. The study was carried out in two paradigms, “Easy” and “Difficult”. In the Easy condition, the difference between standard (0 ms VOT) and deviants (long lag: 92 ms VOT; voicing lead: -112 ms VOT) was more pronounced, whereas in the Difficult paradigm, the stimuli were acoustically more alike (standard: 0 ms VOT; long lag: 36 ms VOT; voicing lead: -36 ms VOT). We hypothesized that behavioral and neural discrimination would match with the native-language contrastive categories. Specifically, we predicted that Hindi listeners would show good discrimination of both the voicing lead and voicing lag stimuli compared to the short-lag stimulus [pa] because Hindi has all three contrasts. In contrast, we predicted that monolingual English speakers would show a robust negativity to the short lag- vs. long-lag contrast both in the Easy and the Difficult condition, but not for the prevoiced vs. short-lag contrast because English has no boundary between the latter contrasts. For the Romance language group, we predicted robust negativity to the voicing lead contrast but no signs of discrimination for the short-lag versus long-lag contrast because Spanish and Portuguese do not have a boundary between the short-lag and long-lag phonemes. We also predicted that these cross-linguistic differences would be greater for the acoustically more difficult compared to the easy contrasts. The results revealed that all participants showed large MMNs (and a P3a orienting response) to the stimuli in the Easy condition, but it was considerably larger for the Romance and Hindi groups than the English group for the voicing-lead case. The Romance group also showed a later MMN to the aspirated deviant compared to the Hindi and English groups. Interestingly, the Hindi and Romance listeners showed larger MMNs to the difficult deviants than the American-English listeners, including the 36-ms aspiration deviant. There was no evidence that Hindi and Romance listeners were processing the voicing contrast differently. The lack of a larger MMN to the 36-ms aspirated deviant for English compared to Romance listeners can be attributed to bilingual experience. Some of the bilingual Romance language speakers may have learned to categorize short-lag and long-lag aspirated because they had early experience with English. Considering that bi-/multi-lingualism is common in many parts of the world, it is important to further examine how age of second-language learning and amount of experience, in addition to the nature of the phonologies for the two languages influence speech processing at the early pre-attentive level indexed by MMN.