從聽辨認知分析現代漢語音譯借詞

The Correspondence in Mandarin Loanword Phonology: A Perception-Map Account

本篇研究是以聽辨認知模式（Perception-map model）建構的制約排序層級（constraint ranking），分析現代漢語音譯（phonetically-translated）、借詞（loanword）、對應（correspondence）的處理。借詞是不同的語言接觸，經由移借產生的新興詞彙，一般而言，語音轉譯為首要處理策略。在借詞對應調整過程中，深受借入語和借出語的音韻系統力量影響，在借詞譯音轉換對應過程中，聽辨認知的感受是很重要的關鍵。本篇研究現代漢語自英語詞彙，經由音譯產生的借詞，發現輔音串在詞首、詞中及詞尾的位置，會有不同的對應處理策略。再深入輔音串類別分析，發現輔音串中塞音的保留或刪除在詞彙不同的位置，有不同的處理策略；而如舌尖擦音/s/，則是強烈傾向保留轉譯而不刪除。研究中所提出聽辨認知建構的制約排序層級，並非專為借詞處理的，是存在於普遍語法中。

This study tries to provide a perceptual account to analyze the correspondent relationship in Mandarin Chinese loanwords. When different languages contact， loanwords would occur. It seems that loanword forms， the correspondence between borrowing language and native language， are influenced by the native phonological systems， and by the perception. In fact， the P-map model assumes perceptual knowledge of speakers and listeners are the key for the nice communication. When the borrowing forms are phonetically translated， the correspondence of loanwords formed may be due to avoid the similarity in the contrast; may be due to strengthen the contrast.In this paper aims on the process of the consonants in word-initial， word medial， word final positions. Though observing the data， there is an asymmetry for the consonants in different positions. From the perceptual of view， the prominence unit such as onset is a salient unit. Thus， the consonants in the initial (onset) position tend to preserve， but the consonants in the final (coda) position may be deleted. In Optimality theory framework， the preference is explored by constraint ranking. Especially， it is obvious that the faithfulness constraints such as ONSET-FAITHFULNESS are higher-ranked than the markedness constraints. The deletion， or the epenthesis of vowels， or the adjustment of segments to break the consonantal cluster may be due to avoid the perceptual similarity or confusibility in coda position， or due to strengthen the contrast in onset position. And it is explored through the ranking of the relevant constraints such as *COMPLEX.