「偶語易安、奇字難適」：探討中文讀者斷詞不一致之原因

Even Words Are Right, Odd Ones Are Odd: Explaining Word Segmentation Inconsistency among Chinese Readers

中文讀者在被要求對文句做意識性的斷詞時，會出現彼此不一致、甚至自己前後不一致的情形。部分學者以「中文詞不具心理真實性」作為對此不一致現象的解釋。陳振宇（1994）觀察到：讀者在進行斷詞作業時，傾向將單字詞及其相鄰的雙字詞當成一個詞（例：紅旗袍），他稱此為「單字詞過度延伸」，並認為這是造成斷詞不一致的原因之一。本研究嘗試以實驗方式驗證「單字詞過度延伸」現象。在實驗中要求受試者將短文中的每一個詞切分出來，每一篇短文有兩種版本，差別在於目標字串是一個單字詞配一個雙字詞，或是一個雙字詞配另一個雙字詞。實驗一比較單雙詞組（例：紅旗袍）及雙雙詞組（例：紅色旗袍）兩種情況，實驗二比較雙單詞組（例：說明檔）與雙雙詞組（例：說明檔案）。兩個實驗的結果發現單雙詞組或雙單詞組被判斷為一個詞的比例高於雙雙詞組，證實了受試者意識性地斷詞確有「單字詞過度延伸」的現象。此現象可能與中文語詞多音節化的特性有關。劉勰、金兆梓、劉復、呂叔湘皆有類似之主張。綜合以上之發現及主張可推論：過去研究發現斷詞不一致現象，應是因為受試者在進行外顯的語言作業時，將其隱含的中文語言知識過度延伸應用所致。據此，詞在中文讀者的心理詞典裡應是真實存在，並且是一致的，而中文閱讀的單位仍應為詞。

When asked to segment a sentence into words, Chinese readers often show substantial between-individual as well as within-individual inconsistencies. Some researchers have attributed this word segmentation inconsistency to the fact that the character is the stable linguistic unit in the Chinese mental lexicon, and the word is not. Chen (1994) observed that his subjects had a tendency to treat the combination of a monosyllabic word and a disyllabic word as a single word. He characterized this tendency as "overextension of monosyllabic words," and suggested it might be the cause of the word segmentation inconsistency. The present study attempted to confirm this observation by means of experiments. The subjects were asked to draw a line between words in several short texts. The critical part of each text consisted of two words which could be paraphrased in two ways: a monosyllabic word and a disyllabic word, or two disyllabic words. Experiment 1 compared the monosyllabe-disyllabe combination with the disyllable-disyllable combination. Experiment 2 compared the disyllable-monosyllable combination with the disyllable-disyllable combination. The results showed that the monosyllabe-disyllabe combination and the disyllable-monosyllable combination did have a higher probability. of being treated as a single word than the disylla- ble-disyllable combination. The Chinese language has evolved into a multi-syllabic language with most words consisting of two sylla-bles/characters. We think the overextension tendency in our subjects' word segmentation behavior reflected the application of their implicit knowledge about this characteristic of the language in an explicit linguistic task that required metalinguistic knowledge. Several early-days Chinese scholars have discussed this tendency before, and Jin Zhao-Zi characterized it the best: "Even words are right, odd ones are odd." In line with this view, we conclude that the word is a psychologically real and stable unit in the Chinese mental lexicon. and that the reading unit in Chinese should be word.