烏爾都語者對華語噝音的感知與產出

Perception and Production of Mandarin Sibilants in Urdu-speaking Learners

對於以烏爾都語為第一語言或優勢語言的學習者而言，華語噝音的對比是一大挑戰，特別是在區辨和產出發音部位不同的塞擦音與擦音（如捲舌音與齦顎音）上。為了解釋這些學習模式並為華語發音教學提供參考，本文以毆知同化模型為理論基礎，結合實驗語音學方法，探討以烏爾都語為母語或優勢語之華語學習者在毆知與產出之間的關係。研究採用ABX任務、華烏輔音相似度評分任務、華語仿說任務與烏爾都語產出任務。本文分析了15位具烏爾都語背景的華語學習者數據，其中多數為零基礎，且整體參與者的華語程度皆相當有限（平均值為2.05/7）。ABX任務的結果顯示，受試者能穩定地區辨華語噝音對比，表明其對目標音之間的聲學差異具有敏毆度。然而，單憑聽辨表現並不能掲示這些語音對比在學習者音韻與語音系统中的表徵。因此，本研究透過相似度評比任務，檢視了華語噝音是如何在知覺上對應至烏爾都語的範疇。結果顯示，學習者呈現出「單類別同化模式」：華語捲舌音與齦顎音被對應至母語硬顎音範疇。這種毆知同化模式亦反映在學習者的產出中。華語仿說數據的聲學分析顯示，學習者在發音方式（送氙對立）上呈現出烏爾都語的正遷移。然而，發音部位卻受毆知同化的制約：學習者的捲舌音與齦顎音於頻譜重心出現合併現象，與烏爾都語中產出的硬顎音範疇相似。綜合而言，烏爾都語背景之學習者的發音困難源自於「範疇層级的同化」，而非聲學聽辨能力不足。本文建議，華語語音教學方法應著眼於音韻與語音範疇之重構，而非僅僅提升其知覺敏毆度。

For learners whose first or dominant language is Urdu, Mandarin sibilant contrasts pose persistent challenges, particularly in distinguishing and producing Mandarin affricates and fricatives that differ in place of articulation, such as retroflex and alveolo-palatal categories. To account for these learning patterns and to inform Mandarin pronunciation instruction, this study adopts the Perceptual Assimilation Model (PAM) and experimental phonetic methods to examine the relationship between perception and production in Urdu-speaking learners of Mandarin. An ABX task, a Mandarin-Urdu consonant similarity-rating task, a Mandarin imitation task, and an Urdu production task were conducted. Data from fifteen Urdu-speaking learners of Mandarin were analyzed, most of whom were naïve and all of whom showed limited proficiency (mean = 2.05/7). Results from the ABX task indicate that learners could reliably discriminate Mandarin sibilant contrasts, indicating sensitivity to acoustic differences among the target sounds. However, discrimination performance alone does not reveal how these contrasts are represented in the learners' phonological/phonetic system. The similarity-rating task therefore examined how Mandarin sibilants were perceptually mapped onto Urdu categories. In this task, learners demonstrated a single-category assimilation pattern: Mandarin retroflex and alveolo-palatal sibilants were mapped onto a single palatal category in Urdu. This perceptual assimilation pattern was also reflected in learners' production. Acoustic analyses of Mandarin imitation data show positive transfer from Urdu for manner of articulation (aspiration contrasts). However, place of articulation was constrained by perceptual assimilation: learners' retroflex and alveolo-palatal sounds merged in spectral center of gravity, resembling palatal sibilants produced in Urdu. These findings indicate that pronunciation difficulties among Urdu-speaking learners stem from category-level assimilation rather than insufficient acoustic discrimination, suggesting that L2 pronunciation instructional approaches should target phonological/phonetic category restructuring rather than perceptual sensitivity alone.