Search

Skilled perception of speech involves not just recognizing the words and sentences that a talker produces, but also perceiving properties imputed to the talker, such as being a foreigner. Because being perceived as foreign can have social consequences, it is important to understand the characteristics that contribute to this percept. Foreign accent perception is often studied in relation to talker characteristics, such as age of learning the second language (L2) or age of arrival in the L2-speaking country. However, listeners generally do not have direct access to such information. In order for the perception of foreign accent to be fully understood, it must be studied in relation to physical characteristics of the speech signal.This dissertation reports a series of six experiments that elicited American English monolinguals' ratings of various properties of productions from native talkers of American English, Hindi, Korean, Mandarin, and Spanish, as well as their free classifications of talker language background. In the first four experiments, listeners heard samples of English and rated foreign accentedness or non-nativeness for each production. In the final two experiments, listeners rated non-Englishness for samples of each talker producing his or her native language (L1). Linear mixed effects regression models revealed that VOT, F1 frequency, and F2 frequency correlated with ratings of accentedness and non-nativeness for syllable- and word-length stimuli. In addition, F3 frequency and F2 tilt correlated with the ratings for syllable-length stimuli, and vowel duration with the ratings for word-length stimuli. Non-Englishness ratings for word-length stimuli were closely related to listeners' ability to recognize the stimuli as known English words. Free classification results revealed that across listeners, grouping patterns for native talkers were more consistent than for most non-native talkers. The correlates of multidimensional scaling analyses of the free classification responses were similar to the correlates of the ratings.The results of this investigation reveal, for a varied sample of non-native English, which characteristics of the speech signal may lead American English monolinguals to identify a talker as foreign. When perceiving syllable-length stimuli, listeners seem to attend to phonetic details resulting from transfer from the non-native talker's L1, while indications of the talker's L2 fluency may begin to influence perception in units as small as disyllabic words. Such information may effectively identify priorities for L2 learners of English interested in accent reduction.

Recent work has shown that preschoolers track informants’ past reliability concerning familiar information and labels, and they use this information to judge the correctness of novel information and labels they provide. But linguistic factors also sway children’s choices for social interaction, for which native-accented speakers are preferred. The present study uses the selective trust paradigm to consider how accentedness interacts with speaker reliability with native- and foreign-accented informants. The results show that speaker reliability and accentedness affect four-year-olds’ choices, but the impact of these factors differed by response type. Preschoolers preferred to ask the native-accented speaker for information, regardless of his reliability. However, in choosing which label to learn, preschoolers selected the reliable speaker’s label, regardless of accent, and correctly identified the unreliable speaker. This study provides evidence suggesting that young children separate their social biases from their objective assessment of novel information.

The issues relative to foreign accent continue to puzzle second language researchers, educators, and learners today. Although once thought to be at the root, maturational constraints have fallen short of definitively accounting for the myriad levels and rates of phonological attainment (Bialystok & Miller, 1999, p. 128). This study, a Posttest-only Control Group Design, examined how the pronunciation accuracy of adult, English language learners, as demonstrated by utterance length, was related to two input stimuli: auditory-only input and auditoryorthographic input. Utterance length and input modality were further examined with the added variables of native language, specifically Arabic and Spanish, and second language proficiency as defined by unofficial TOEFL Listening Comprehension and Reading Comprehension section scores. Results from independent t tests indicated a statistically significant difference in utterance length based on input modality (t(192) = -3.285. p = .001), while with the added variable of native language, factorial ANOVA results indicated no statistically significance difference for the population studied. In addition, multiple linear regression analyses examined input modality and second language proficiency as predictors of utterance length accuracy and revealed a statistically significant relationship (R 2 = .108, adjusted R 2 = .089, F(3, 144) = 5.805, p = .001), with 11% of the utterance length variance accounted for by these two factors predictors. Lastly, hierarchical regressions applied to two blocks of factors revealed statistical significance: (a) input modality/native language (R 2 = .069, adjusted R 2 = .048, F(2, 87) = 3.230, p = .044) and ListenComp (R 2 = .101, adjusted R 2 = .070, F(3, 86) = 3.232, p = .026), with ListenComp iv increasing the predictive power by 3%; (b) input modality/native language (R 2 = .069, adjusted R 2 = .048, F(2, 87) = 3.230, p = .044) and ReadComp (R 2 = .112, adjusted R 2 = .081, F(1, 86) = 3.629, p = .016), with ReadComp increasing the predictive power by 4%; and (c) input modality/native language (R 2 = .069, adjusted R 2 = .048, F(2, 87) = 3.230, p = .044) and ListenComp/ReadComp (R 2 = .114, adjusted R 2 = .072, F(2, 85) = 2.129, p = .035), with ListenComp/ReadComp increasing the predictive power by 4%. The implications of this research are that by considering issues relative to input modality and second language proficiency levels especially when teaching new vocabulary to adult second language learners, the potential for improved pronunciation accuracy is maximized. Furthermore, the heightened attention to the role of input modality as a cognitive factor on phonological output in second language teaching and learning may redirect the manner in which target language phonology is approached.

This paper seeks to address the issue by examining two factors that potentially affect a listener’s perception of foreign accented speech: degree of familiarity (as acquired through a work or personal environment) and amount of ESL or linguistic training. Speech samples were recorded from 18 international students from Hispanic, Asian, and Middle-eastern backgrounds and across all proficiency levels as designated by their academic English program. Six native English speakers were also recorded to serve as a basis for comparison. Listeners were drawn from two pools: people with ESL and/or linguistic training (n=42) and laypersons with no such specialist training (n=36). After completing a background questionnaire to assess familiarity with foreign accented speech, each listener rated all 24 speech samples on the dimensions of comprehensibility, degree of accent, and communicative ability. Results indicate that participants with ESL/linguistic training rate foreign accented speech more positively on all three dimensions than laypersons with no such training. Additionally, degree of familiarity with foreign accented speech is positively correlated with how participants rated the accented speech samples. a number of highly significant interactions between these and other factors including sex of the speaker, proficiency level of the speaker, and L1 family of the speaker were found as well.