華語語句產生與知覺的計時特性

The Speech Timing Characteristics of Mandarin Sentence Production and Perception

本研究目的在探討華語語句計時的聲學和聽知覺韻律特性，由語句的語音聲學和聽知覺反應探討華語音節等時性的存在，考驗三種計時候選單位（空白起始單位、母音起始單位和子音起始）的等時性。實驗一使用句型結構相似的句子為材料，分析二十一位正常說話者用五種語速所說語句音段（子音、母音、空白）時長，比較三種計時單位的成對變異指數（PVI）。結果發現在各語速下以母音起始為等時的單位有最低的PVI值，尤其在慢語速時，母音起始單位最具有等時性，也發現母音時長確會受其後接子音的影響，而子音前空白或靜默則會受到前接母音的影響。為維持母音起始單位的等時性，單位內各音段具有消長互補特性。實驗二發現聽者可估計在四種語速語句中被噪音取代音節的數量，平均正確率達70%。在實驗三中參與者除了需判斷噪音段之前的音節母音時長是否拖長外，還需依照原語句的計時架構仿說語句，並填補噪音取代的音節，結果參與者判斷正確率平均達73%，且發現參與者所產生語句的母音起始單位之PVI值較其他單位為小。本研究結果支持以母音起始的類音節為華語韻律計時的基本單位，語句中言語計時單位的切分點大致是位於語句中各音節的母音起始處，推論語音的產生與知覺皆是以此單位所組成的語句韻律框架進行處理。

The behavior of speech timing plays an important role in both speech production and perception. ‘Isochrony’ is the source speech rhythm, and there suggests a timing unit with equal duration in the speech stream. Based on the isochronous timing unit, the language users can produce sentences naturally and also implement speech segmentation for the perceived speech stream. The purpose of this study is to investigate the timing isochrony in Mandarin sentence production and perception. In the experiment 1, 21 mandarin speakers were instructed to read 42 sentences at five speaking rates. These sentences had similar syntactic structures in order to test the effects of neighboring syllables. Three units were proposed for speech timing: units with rime onset (rime + pause + C), units with gap onset (gap + CV), and units with consonant onset (CV + gap). Pairwise variability index (PVI) was compared for the three rhythmic unit. The results show that Rime onset unit had the significantly lowest PVI. Furthermore, it suggests that for maintaining the isochrony of rhythmic units in sentences, the durations of the rimes are influenced by the consonant in the following syllable across the syllables, and the gaps duration are also influenced by the vowel in the preceding syllable. In the experiment 2, the listeners were asked to judge the number of syllables in a spoken sentence with a segment replaced by white noise. The results showed that the listeners could successfully predict the number of syllables in the noise segment with 70% accuracy, even though four speaking rate variation. In the experiment 3, the participants were asked to imitate the sentences that had a segment replaced by noise, and some with prolonged vowels just before the noise segments. When given the sentences in written Chinese, most participants could successfully imitate the sentence with little timing difference from the original sentences, and they also could judge the prolongation of vowels with 73% accuracy. The results also show that the unit with rime onset had the smallest PVI values at most of the speaking rate situations. These results suggest that the basic unit of Mandarin speech timing is related with the vocalic onset, where is the boundary between an initial consonant and a rime of a syllable, or between a glide and a vocalic vowel in a syllable without an initial consonant.