L2 English fricative production by Thai learners

Abstract

In early research on L2 (second language) phonology, researchers mainly focussed on whether L2 learners can achieve ‘target-likeness’, which relates to whether or not a sound is perceived as the intended target or whether it fits into the expected IPA category as determined by trained phonetician(s). The popular model for this focus was the contrastive analysis hypothesis (CAH) (Lado, 1957). Later research extended the focus to judgements of ‘native-likeness’, which is the extent to which the speaker’s L2 sound production has native-like qualities. Methods such as accent rating tasks and acoustic measurements have become popular over time, together with investigations of how the results correlate with external factors which are thought to influence L2 speech learning. Well-known models such as the Speech Learning Model (SLM) (Flege, 1995) and the Perceptual Assimilation Model (PAM) (Best, 1995) have been very influential in this field, but are mainly based on assumptions regarding L2 learners in a naturalistic setting. The aim of this thesis is to investigate L2 English fricative production by Thai learners of English with a combination of focus on target-likeness and native-likeness through four types of analysis: impressionistic, sound identification, accent rating, and acoustic analyses. This thesis also explores external factors which may contribute to target-likeness in L2 production which is more important than native-likeness as it helps in communication between interlocutors. The L2 fricatives are divided into those that have a counterpart in Thai (/f, s/ henceforth ‘shared’ sounds) and those that do not (/v, θ, ð, z, ʃ/, henceforth ‘non-shared’). As CAH focuses on target-likeness, it predicts that shared sounds are easy to produce; SLM, on the other hand, focuses on native-likeness and predicts that shared sounds are difficult to produce. Results from the four experiments in this study show mixed results. In terms of results from impressionistic and sound identification analyses, CAH-based hypotheses accurately predict most results, which show that shared sounds are more frequently produced in a target-like manner and more accurately identified. In terms of results from the accent rating task, SLM had to be rejected in this case, as results showed that shared fricatives were more often produced in a native-like manner, unlike non-shared fricatives. In the acoustic investigation, ii differences in the realisations of L2 shared sounds supported SLM-based hypotheses in some contexts. And although SLM-based hypotheses were disconfirmed when it came to the accent rating of L2 shared and non-shared sounds, the phonetic properties of non-shared sounds in the realisations that were deemed target-like were native-like in many contexts, suggesting some L2 attainment for non-shared sounds. Taken as a whole, these results emphasise the need to focus on both target-likeness and native-likeness in investigating L2 speech production. They also imply that L1 and L2 sound comparison is context- and task-dependent.