The Recognition of Spoken Mono-Morphemic Compounds in Chinese

單語素複合詞於口語中文之辨識歷程

本研究旨在探討單語素複合詞於口語中文中的詞彙存取歷程，並進一步了解文字構型於口語字詞辦識歷程中所扮演的角色。傳統漢語語言學定義複合詞為包含二個以上的國字字串，其國字是否為詞素乃非必要條件，依其定義，中文單語素雙字詞即屬於複合詞的一種；而單語素複合詞又可依其文字構型上的完整性，亦即內部國字間的緊密度，再分為連綿詞與音譯詞兩類，前者之內部國字只可與出現在該詞的相鄰國字共同出現，例如'蟑螂'，後者之內部國字還可與其它國字出現於其它字詞當中，例如'漢堡'。為了解此文字構型完整性的差異是否影響口語字詞辦識之歷程，本研究控制了目標詞的詞頻、音節頻率、跨音節預測度、群組個數並操弄其組成國字間的緊密度，分別進行了口語詞彙判斷作業（實驗一）、複誦作業（實驗二）、閱讀詞彙判斷作業（實驗三）與跨模組詞彙判斷作業（實驗四）。實驗一結果發現連綿詞所需的判斷時間較音譯詞來得短，證實了於口語字詞辦識過程裡文字構型完整性效果的存在。實驗二結果發現連綿詞與音譯詞的複誦反應時間沒有顯著差異，由於複誦過程所反應的是音韻處理過程，因此實驗一所發現的效果並非音韻上的差異所造成，同時也可推測該效果應出現於後詞彙加工階段。實驗三結果發現連綿詞與音譯詞的詞彙判斷時間沒有顯著差異，說明了實驗一所發現的效果並非來自於文字構型視覺上的激發，而是跨字預測度的資訊所造成，因為在實驗四不同類型的視覺干擾下，仍可發現與實驗一相同的效果。本研究並討論此效果的成因與其在中文口語詞彙實驗裡的意涵。

This paper explores the auditory lexical access of mono-morphemic compounds in Chinese as a way of understanding the role of orthography in the recognition of spoken words. In traditional Chinese linguistics, a compound is a word written with two or more characters whether or not they are morphemic. A mono-morphemic compound may either be a binding word, written with characters that only appear in this one word, or a non-binding word, written with characters that are chosen for their pronunciation but that also appear in other words. Our goal was to determine if this purely orthographic difference affects auditory lexical access by conducting a series of four experiments with materials matched by whole-word frequency, syllable frequency, cross-syllable predictability, cohort size, and acoustic duration, but differing in binding. An auditory lexical decision task (LDT) found an orthographic effect: binding words were recognized more quickly than non-binding words. However, this effect disappeared in an auditory repetition and in a visual LDT with the same materials, implying that the orthographic effect during auditory lexical access was localized to the decision component and involved the influence of cross-character predictability without the activation of orthographic representations. This claim was further confirmed by overall faster recognition of spoken binding words in a cross-modal LDT with different types of visual interference. The theoretical and practical consequences of these findings are discussed.