Consonant gemination as a phonological feature plays a major role in the Sinhala language. The absence of true gemination in English causes perception problems for native English speakers when attempting to distinguish minimal pairs in Sinhala created by gemination. This study examined whether native English learners' difficulties in perceiving consonant gemination in Sinhala could be reduced by creating phonological awareness of it through formal training. Four native Sinhala speakers were asked to record thirty-two Sinhala minimal pairs. These recordings were used to set up the audio test instruments. Twenty-four native English speakers participated in the pre-test, a teaching/training session, and the post-test. The pre-test consisted of an AX discrimination task, where the subjects heard two stimuli (A and X) and had to decide if those two stimuli were the same or not. The 20-minute teaching/training session, which was the study's primary independent variable, was given to educate the participants about the gemination contrast in Sinhala. The same audio test was repeated as the post-test. The data analysis included descriptive statistics and a t-test for dependent samples through SPSS statistics version 20. The findings of the study showed a significant difference between the pre-test and the post-test. The data also revealed the teaching/training session to have a high level of effectiveness regarding gemination contrast. Perception of the gemination contrast had increased in the post-test, while perception of words without this contrast had decreased in accuracy, possibly as a result of hypercorrection.

本篇論文旨在從優選理論(Optimality Theory)來探討日語中的英語外來語。所研究的問題有四個，包括英語詞彙日語化過程中的語音代換(segmental substitutions)與音韻調整(phonological modifications)的現象，英語輔音韻尾促音化(consonant gemination)的現象，以及外來語截短(loanword truncation)現象的研究。在語音代換方面，基於過去以對比分析(Contrastive Analysis)的研究只重視英日語間音與音的對應，然共通語法(Universal Grammar)為當代語言學研究的主要目標，因此本文以優選理論這一針對共通語法所提出的當代音韻學理論中的制約(constraints)排行與違反的概念來重新分析。在本文中，語音代換主要是起因於共通語法下的音段標顯性制約(segmental markedness constraints)、極大性制約(MAX-IO)、及特徵一致性制約(IDENT[F] constraints)在日語語法中排行的結果。在音韻調整方面，日語主要是以元音添加(vowel epenthesis)來處理英語韻尾輔音及輔音群。過去的分析將元音添加視為是一種規則(rules)運作的結果，然這種分析仍需要制約的概念。有鑑於此，本文純以制約的排行與違反來分析這個現象。在本文中，音韻調整及元音添加主要是起因於共通語法下的一組音節結構完好性制約(syllable wel-formedness constraints)與極大性、依循性制約(DEP-IO)在日語中排行的結果所致。在英語韻尾輔音促音化方面，由於許多過去的研究將此一現象歸因於是模仿英語原字，特別是其閉音節結構的結果。根據此一看法，本文主張此一現象乃是起因於與韻尾輔音相關的音拍一致性制約(IDENT[m])因模仿英語原字這一目的而獲得優先被滿足權。也就是當其排行高於反孳生輔音制約(anti-gemination constraints)時，就會導致促音化的現象。在外來語的截短現象方面，過去學者的分析引發不少的問題，如Suzuki(1995)根據優選理論所提出的分析卻與Ito(1990)的統計結果相矛盾等。有鑑於此，本文作者亦進行外來語截短詞包括長度、音節類型、截短方向及重音等的統計。根據這一統計結果，本文將截短現象分成「最小截短」與「非最小截短」。兩者各由其相關的制約排行所控制。此外，本文並指出日語外來語的「最小截短」能驗證emergence of the unmarked (McCarthy and Prince 1994, 1995b, 1997)這一主張。 / This thesis aims to study English loanwords in Japanese within the framework of Optimality Theory (Prince and Smolensky 1993, McCarthy and Prince 1993ab). It deals with four issues: the segmental substitutions and phonological modifications in loanword nativization, consonant gemination, and loanword truncation. The segmental substitutions are attributed mainly to the Japanese-specific ranking of the UG segmental markedness constraints, together with the higher ranked MAX-IO. The choices of substitute sounds are mainly determined by the rankings of lDENT[F] constraints in Japanese. The phonological modifications are attributed mainly to the Japanese-specific ranking of the UG syllable well-formedness constraints. In addition, vowel epenthesis as a way of modification is prompted by the dominance of MAX-10 over DEP-IO. As for consonant gemination, it is not motivated by any native requirements in Japanese, and thus results from a conscious effort to imitate English source forms by having lDENT[μ] acquire a priority to be respected. That is, when it dominates the anti-gemination constraints, consonant gemination takes place. Finally, on the basis of the statistical results, the present study distinguishes two kinds of truncation: the minimal and non-minimal truncations. In the minimal truncation, FTBIN and NONFIN(σ) are responsible for the accented LL and HL output patterns while in the non-minimal truncation WDBIN and MAX-BT lead to the unaccented four-mora outputs as well as LLL outputs.