DOI: 10.6129/CJP. 20130608
中華心理學刊 民102，55卷，3期，289-318
Chinese Journal of Psychology 2013, Vol.55, No.3, 289-318
吳瑞屯（國立台灣大學心理學系暨研究所）；楊馥菱（國立台灣大學心理學系暨研究所）；林維駿（國立台灣大學心理學系暨研究所）
摘要
漢字的設計不同於拼音文字，導致研究者超過廿年關注於語音處理是否在漢字認字歷程中扮演影響角色的議題。本文回顧台灣過去四十年的漢字字彙辨識研究議題，包括字詞處理腦側化、辨識前語音轉錄、次字彙歷程、頻率效果等，並同時闡明，只有圍繞著字頻效果發展的研究路線可以經得起深入嚴格檢驗，也較有機會衍生後續的理論進展。歐洲的拼音文字研究則除了以往的字頻變項，近期更發展至分析語料的其他統計特性，據以為操弄變項，探討該等變項如何影響行為反應的計算認知語言學研究境界。相對於中國、香港、新加坡等地，台灣具備了歷史最久的研究經驗，也建立了最多樣的字詞頻資料庫系統，以作為探討漢字認知歷程議題的基礎。根據這些多樣的字詞頻資料庫進行語料屬性之間的分析，可以很容易發現單字字頻與單字成詞數目產生明顯共變關係，單字字頻低者，成詞數目較少，但單字字頻高者，成詞數目傾向於較大。過去的單字辨識歷程研究實驗經驗與理論，都指出字頻是影響單字辨識歷程最重要的變項。而且漢字單字辨識實驗裡得到的字頻效果影響力，比拼音文字研究得到者更大。因此，字頻效果中，是否混淆或包含了成詞數目的影響力，是本文的主要關注。如果操弄單字成詞數可以穩定觀察到單字辨識受到影響，可能可以推論詞的知識會影響單字辨識，遠出於以往研究者的意料。如果關連詞的知識網絡會影響單字的辨識，那麼詞內單字自動引發的關連詞網絡知識是否也可能影響詞的辨識？這樣的思考角度可以重新評估有關鄰群詞是否會影響詞的辨識，或詞的辨識是否涉及詞內單字處理等研究者長久以來無法解決的問題。本文更舉出一些初步的實驗結果，以闡明這樣的觀點。
關鍵詞：字彙辨識、部件處理、詞彙辨識、頻率效果、鄰群效果
Beyond Phonology Matters in Character Recognition
Jei-Tun Wu（Department of Psychology, National Taiwan University）；Fu-Ling Yang（Department of Psychology, National Taiwan University）；Wei-Chun Lin（Department of Psychology, National Taiwan University）
Abstract
The structure of Chinese characters is quite different from that of the alphabetic system. This rendered for decades whether phonology plays a role in character recognition. This article reviewed previous studies over the past forty years on issues related to character recognition, including the lateralization effect in reading, phonological recoding, sublexical processing, and frequency effect. We then argued that only researches concerning frequency effect are beneficial for advancing further studies. Based on previous explorations of word frequency, European researchers in alphabetic languages have recently started to elaborate on the possible role of other statistic properties of linguistics in word recognition. Compared with China, Hong Kong and Singapore, Taiwan is equipped with a longer research history and has cultivated the most diversified linguistic corpus systems for the research on traditional Chinese language processing. Based on computations from available data base systems, it is obvious that character frequency covariates with the number of different words embedding that character. A character with higher frequency tends to combine with other characters to constitute more multi-character words. Furthermore, previous studies on character recognition have pointed out that occurrence frequency is the most important factor, with an effect size even larger than that of word recognition from studies on alphabetic languages. Hence, the present study focused on whether the frequency effect of a character
confounds with or consists of its combinability with other characters to constitute different words. If the combinability of a character affects the recognition of that character, then we can infer that a structure of network knowledge about the inter-connection between characters and words plays a role in the recognition of characters and words. From this angle, many previous studies concerning neighborhood effect or sub-lexical character processing in Chinese word recognition would need to be re-evaluated. Some preliminary results were also provided to elucidate the authors’ proposition.
Keywords: character recognition, radical processing, word recognition, frequency effect, neighborhood size effect
