音韻組成與中文口語感知

Abstract

本論文探討音韻組成(phonological make-ups)，包括聲母響度、鼻音韻尾以及聲調對中文口語感知的影響。使用柵欄理論進行口語感知實驗，以精確掌控受試者所能聽取之刺激項長度，進而控制其所能獲得之聲學線索(acoustic cues)。實驗刺激項為臺灣國語四個聲調為一組(quadruplet)的音節共32個，嵌入「我說___給你聽」的載句中，發音人為一位臺灣母語使用者(23歲女性，台中人)。將刺激項以50毫秒為一個單位切割成由短而長的柵欄刺激項，以連續方式呈現，每個刺激項最多分為八個柵欄(第八柵欄為整個音節)。實驗受試者為交通大學招募來的30名學生(男女各半)。實驗探討因素有聲母響度(響音及阻音)、鼻音韻尾(無鼻音、舌尖鼻音及舌根鼻音)及聲調(一聲、二聲、三聲及四聲)，依照音節獨立點(IP)及聲調獨立點(TIP)兩個面向進行探討。實驗結果可就三部分來看：一、聲母響度對IP的影響不顯著，聲母的TIP為響音<阻音，表示響音之中包含可供辨識聲調的聲學線索。二、鼻音的IP為無鼻音顯著快於舌尖鼻音和舌根鼻音，舌尖鼻音顯著快於舌根鼻音，可看出臺灣國語的三種鼻音在聽辨速度上有所差異，或許跟臺灣國語鼻音合流(nasal merge)的現象有關，而鼻音韻尾對TIP的影響不顯著。三、聲調的IP及TIP皆為二聲最慢，根據受試者各柵欄之回應可看出，二聲容易和一聲和三聲混淆；另外，對發音人的聲學分析可看出，二聲的基頻上升出現在較慢的時間點，造成其感知時間的延後。

This thesis examined the effects of the phonological make-ups, including initial consonant sonorancy, nasal codas and tones on Mandarin spoken-word recognition. The present study uses the gating paradigm to precisely control acoustic-phonetic information of the stimuli presented to the subjects. Stimuli in this experiment are eight monosyllabic quadruplets, inserted in the carrier sentence. The stimuli were recorded by a female native Taiwan Mandarin speaker. The gates were constructed with 50ms increments and were presented in a successive presentation format. A total of 30 Taiwanese native speakers (15 males and 15 females) were recruited as subjects in this study. Results showed that initial sonorancy has no significant effect on Isolation Point (IP), but sonorant has a significantly earlier Tone Isolation Point (TIP) than obstruent, which shows that there might be some acoustic cues for tone recognition in sonorants. No nasal coda has a significantly earlier IP than alveolar nasal and velar nasal, and alveolar nasal has a significantly earlier IP than velar nasal. This shows that the recognition rate of different nasal coda in Taiwan Mandarin is varied, and it might relate to the phenomenon of nasal merge in Taiwan Mandarin. Tone 2 has the latest IP and TIP, and according to the subject responses at each gate, Tone 2 tends to misidentified as Tone 1 and Tone 3. Moreover, the F0 of Tone 2 rises later, which might lead the identification of Tone 2 delay.