中文認字前得音來自於低空間頻率？

Prelexical Phonology by Low Spatial Frequency Channels?

在認出中文字之前語音訊息是否已經處理一直是中文研究的重要爭論之一。由於中文並無如拼音文字般明確的形音對應原則，如何在認字之前得音為此爭議的關鍵環節。針對此點，Chua（1999）認為字音與字形分別由低空間頻率與高空間頻率負責，二者互相獨立；而由於低空間頻率處理的速度快於高空間頻率，因此字音的消息可以藉由低空間頻率所提供的整體字形輪廓率先取得，而待高空間頻率消息到達時方完成字形的辨認。為檢驗此假說，本研究採用成語判斷作業，先呈現四字成語中的三字，接著出現形似音同、形似音異、形異音同、形異音異等四組假目標字。當假目標字包含所有頻段訊息時，同音字的正確率低於非同音字（實驗一）。但當假目標字只包含低空間頻率時，此同音字效果減弱或消失（實驗二）。當假目標字不包含低空間頻率時，同音效果重現（實驗三），而且在去除低與高空間頻率，只保留中間頻段的實驗情況中仍有此同音效果（實驗四）。形似假目標字的正確率比形異假目標字來得低，且此字形效果不因空間頻段訊息的不同而有差異，在四個實驗中都持續穩定存在。研究結果反駁中文認字前得音來自於低空間頻率的觀點。

Whether prelexical phonology exists in a logo- graphic script such as Chinese has long been debated. In contrast to English, there is no explicit grapheme-phoneme correspondence rule for Chinese characters, which makes the idea of prelexical phonology seemingly implausible. To provide a logical basis for prelexical phonology in Chinese character recognition, Chua (1999) proposed two independent routes for Chinese character recognition: one through phonology that is carried by low spatial frequencies, and the other through a lexicon that is carried by high spatial frequencies. The low frequency channels can be activated earlier than the high frequency ones, so the global shape of a character provides the phonological information even before the full identification of the character is possible. In this study, a verification paradigm using four- character idioms was adopted to test Chua's low- spatial-frequency mediated phonology model. Three characters of each four-character idiom were presented first, followed by four kinds of foils designed factorially of orthography (similar, dissimilar) x homophone (yes, no). The accuracy in rejecting orthographically similar or homo- phonic foils was found lower than their controls when the foils were unfiltered (Experiment 1). The homophone effect was not consistently found when the foils contained only low spatial frequency information (Experiment 2); it was found when low spatial frequency information was absent (Experiment 3); and it was also found when only a middle range of spatial frequencies was intact (Experiment 4). In contrast, the orthography effect was found consistently in all 4 experiments. These results argue against the low-spatial-frequency mediated phonology view.