語境效應於漢語偏向型同音異形詞歧義解困歷程之研究

本研究欲探討語境限制在漢語偏向型同音異形詞的解歧歷程中的效應。透過兩個跨模式唸名作業，我們欲了解以下研究問題：（1）詞義相對頻率在中性語境下是否能影響詞義的啟動？（2）句內語境限制是否能影響漢語偏向型同音異形詞解困的初始階段？（3）加入全局語境是否能對歧義解困的初始階段造成影響？本研究的結果顯示，詞義相對頻率對同音異形詞不同詞義的啟動無明顯效應。在中性語境內，我們觀察到同音異形詞的主次要意義唸名時間差距無顯著差異。其次，強的句內語境限制能單獨啟動合適詞義，並且抑制不合適詞義的啟動。在沒有語境支持的情況下，即使是偏向型同音異形詞的主要意義仍然無法提取。第三，提供篇章限制對於正確詞義的啟動並沒有顯著的促進效果，反而造成了整體唸名時間的延緩。本研究的結果不支持語境早期效應的模組假說，亦不支持語境能選擇性提取詞義的重新排序模型。我們的結果較符合互動假說下的語境敏感模型（e.g., Vu et al., 2000）。在此模型中，強度大的語境能夠單獨提取其所支持的歧義詞義。然而，我們的結果亦顯示，當句內語境線索充足時，增加全局語境限制並無明顯額外效應。這可能是由於全局語境無法在詞彙的初始階段起作用，或因原有的句內語境線索已足夠，導致全局線索無法達到加乘效應。

關鍵字

歧義解困；偏向型同音異形詞；語境效應；句內語境；全局語境

並列摘要

The present study aims at exploring the effects of contextual constraints on the resolution of Mandarin biased heterographic homophones. Through two cross-modal naming tasks, we purported to answer the following research questions: (1) Does relative frequency affect meaning activation in a neutral context? (2) Do local contextual constraints influence meaning activation at the early stage of biased heterographic homophone resolution? (3) Does the inclusion of a global context have an early effect during ambiguity resolution? The results indicated that relative meaning frequency does not affect meaning activation. In the neutral context, naming times for both the dominant and the subordinate meanings did not differ significantly. Secondly, strong local contextual constraints could lead to selective activation of the correct meaning, while inhibiting the activation of improper meanings. Without contextual support, even the dominant meaning of a biased heterographic homophone could not be successfully activated. Thirdly, providing discourse constraints did not facilitate activation of the correct meaning, but rather led to an overall latency in naming times. The results of the study disconfirm the Modular Hypothesis, which opposes an early effect of contextual constraints. Additionally, the results cannot be explained by the Reordered Access Model, which argues against selective access of the correct meaning by contextual information. Our results are most compatible with the Context Sensitive Model (e.g., Vu et al., 2000). According to this model, a strong context can activate the single meaning it supports. Nevertheless, our results suggest that global context does not yield additional effects when local constraints are sufficient. This may be caused by the fact that global context does not have an early effect at all, or that global constraints do not exert extra effects when there are enough constraints in the local context.