| Summary: | 博士 === 高雄師範大學 === 英語學系 === 100 === Phonological differences related to the /l/-/r/ liquid contrast in English and the phonotactics of the native language are likely to influence the English pronunciation of Taiwanese learners of English as a Foreign Language (EFL). This research is an attempt to analyze (1) the error patterns that emerge from TEFL learners’ liquid production; and (2) the theoretical framework that can account for the error data.
The main purposes of this dissertation are: (1) to detect the linguistic behavior of TEFL learners’ liquid acquisition, (2) to identify the effects that can trigger production errors of English liquids for university-level TEFL learners, (3) to contribute to the research on interlanguage phonological development.
Two production tasks were conducted to test the performance of sixty university-level learners. The design of the target words mainly was based on the distribution of English liquids in the four positions for later comparison, analysis and discussion. Following the error analysis, the acoustic measures were to establish the articulatory contours for /l/ and /r/ of NS norm and TEFL production, as well as to elicit the articulatory characteristics of production errors, according to two acoustic cues: formant values and formant transitions.
The major findings of this dissertation are reported as follows. First, the data-driven approach used to resolve the first research question was to assess the error patterns. The error rate shows that the hierarchy of difficulty was ranked as /l/ > /r/. The error patterns in the four positions are summarized below: (1) The word-initial position: There was a tendency to substitute /l/ and /r/: /l/ was realized as /r/, while /r/ was realized as /l/. (2) The word-final position: The error rates indicate a preference for word-final /l/ deletion and the replacement of /l/ with /r/. The word-final /r/ data indicates a tendency for /r/ deletion. (3) The word-initial cluster: The word-initial /l/ cluster errors mainly involved insertions, substitutions, and deletions, whereas the word-initial /r/ cluster errors mainly involved deletion and substitution. Notably, both /l/ and /r/ in word-initial clusters were realized as /l/ most frequently. (4) The word-final cluster: Deletion was the most common error in the word-final and word-final clusters.
Second, an examination of the formant values shows that articulatory difficulty was related to the movements of two articulators: the lip and the tongue. Considerate correspondence between the NS norm and the TEFL production was shown in the contour of English liquids with acoustic measures. Three important findings for formant values: (1) The over-exaggeration in the formant values (F2, F3, and F3-F2,) of the TEFL subjects indicated the articulatory differences between L1 and L2. (2) The mean F3 value of the TEFL word-initial and word-final /r/ fell within the range of the F3 of the NS /l/, showing why the /l/ and /r/ segments were not contrastive in TEFL production. Moreover, the formant transitions gave the curves for articulation configuration. A comparison between the results of NS and TEFL transition modes show certain acoustic deviations in the TEFL interlanguage production. The degree of backness of the vowel affected the TEFL liquid production, since this required more articulatory modifications to change the tongue position.
Following acoustic measures, the examinations described the articulatory characteristics of production errors. The distance between the F3 and F2 reflected the emphasis on the position of the tip of the tongue. The lower the value of F3, the more foreign-accented the /l/ articulation was. The higher was the value of F3, the more foreign-accented the initial /r/ was articulated. The F3-F2 value in a relatively wider range indicates that TEFL /r/ was articulated in a more mouth opening position. The degree of retroflection influenced categorical contractiveness.
Fourth, resolving the second research question involved examining two theoretical frameworks: the similarity effect in the Speech Learning Model (SLM), and the markedness effect in the Markedness Differential Hypothesis (MDH). To summarize, it was found that the SLM analysis could account for the data better: (1) native language similarity suppressed the markedness effect on liquid production; (2) the learnability of “unfamiliar” /r/ overrode “similar” /l/, if more meaningful input was provided; and (3) the systematic error distribution in the four different positions could be explained. Most importantly, our discussions could outline the hierarchy of learning path, that is, /r/ > /l/. The SLM model confirmed the interaction between phonetic similarity and the articulatory effect. Third, both learning experiences and production difficulty were emphasized.
|
|---|
