Acoustic cues in production of English vowels by Hijazi Arabic learners

Abstract

Dynamic cuesin vowels—in particular, vowel inherent spectral change (VISC)—have been shown to play a major role in vowel identification in various languages and to provide important information not only for diphthongal vowels but also for monophthongal vowels. VISC is also found to provide a more detailed view and understanding of vowels. This understanding is crucial for predicting difficulty that second language (L2) learners face when learning L2 vowels. However, this type of investigation has been lacking, with the majority of first language (L1) studies on vowel dynamics being restricted to English and with the majority of research into L2 acquisition focusing on a static approach and measuring the first two formants (F1 and F2) at a monophthong vowel’s midpoint and drawing conclusions about L2 attainment on this basis. With the above background in mind, the aim of this study was to discover how or whether the dynamic properties of vowels in an L2 are acquired by learners. To do so, the study investigated the production patterns of Standard Southern British English (SSBE) vowels by Hijazi Arabic L2 (HA L2) learners compared with their L1 Hijazi Arabic (HA) patterns and to those of native English (NE) speakers using static (vowel midpoint) and dynamic (offset, slope, and direction models) cues. The thesis evaluated the role of static versus dynamic cues of the HA L1, HA L2, and NE production of vowels, as well as vowel duration, fundamental frequency (F0), and third formant frequency (F3) as additional cues. The SSBE vowel production patterns by the HA L2 group were examined within the context of current L2 learning models. Data were collected from three groups (20 HA L1, 20 HA L2, and 20 NE; both genders). To note, the 20 native HA L1 individuals are the same as the 20 HA L2 learners. The target vowels were examined in a word list with varied consonantal contexts. HA L1 participants produced eight HA vowels, and HA L2 and NE participants produced 10 English monophthong vowels. Results showed that dynamic cues provide insights into all three group production patterns that are not normally gleaned from static measures alone. Using quadratic discriminant analysis, the dynamic cues (particularly the three-point model) had higher classification rates for all three groups. In addition, vowel duration was found to play a significant role in the classification accuracy for HA and HA L2 participants, while F0 was the most important additional cue for accurately classifying NE vowels. While static measures for some SSBE vowels as produced by HA L2 learners showed similarities with those of NE speakers, dynamic measures revealed significant differences which ii suggest that the underlying production setting for SSBE vowels by HA L2 learners is still very much embedded in an HA mode. The thesis results are in line with dynamic approaches and highlight the importance of looking beyond static cues and beyond the first two formants for insights into what contributes to a non-native accent. Current L2 models need further revisions and modifications before they can account for dynamic patterns of speech by L2 learners, as they were not successful in predicting all L2 vowel production patterns under investigation