詞彙關係在中文單字脈絡促發效果中所扮演的角色

Abstract

在漢字單字促發認知作業中操弄促發字與目標字間脈絡關係是探討單字字彙辨識歷程的重要研究手段，其中語義、聯想與詞彙三種脈絡關係備受重視。而詞彙關係做為一種重要脈絡關係，是否涉入字彙辨識的自動化歷程，或是僅做為策略性歷程中預期線索的角色，仍尚待進一步探討。為此，本研究使用單字促發典範配合字彙判斷作業以及唸名作業進行三個實驗，在盡可能排除語義與聯想關係影響的前提下，操弄詞彙關係有無、詞頻高低、促發字衍生詞詞群大小以及促發字位於首字之衍生詞詞群大小，觀察詞彙關係在不同刺激呈現間距下所得促發效果型態。當刺激呈現間距越短，所得促發效果越可能來自於自動化歷程的貢獻；而當刺激呈現間距較長時，則可透過兩個詞群變項做為偵測預期策略的指標，觀察詞彙關係是否做為預期線索涉入策略性歷程。研究結果顯示，不管是字彙判斷作業或是唸名作業，詞彙關係均在刺激呈現間距僅50毫秒時即可發現促發效果，顯示詞彙關係自動化地涉入字彙辨識歷程。而在刺激呈現間距於150毫秒時，詞群大小不同會使得促發效果反應型態產生差異，顯示詞彙關係在可使用預期策略的實驗條件下確實可能扮演預期線索的角色。

Manipulating different contexts in priming tasks in Chinese characters is an important research method to explore the Chinese character recognition process. Forming semantic relation, associative relation, or a two-character word between a prime and a target character are three important ways to explore the effect of the contexts in priming tasks. To solve the issue of whether word relation is involved in automatic character recognition process or is only strategically used as a clue to predict target characters, the present research used lexical decision task and naming task with a single-character priming paradigm in three experiments. Word relation, word frequency, neighborhood size of the prime character, and neighborhood size of the prime character as the first constituent in a word were manipulated at different stimulus onset asynchronies (SOA) to observe the patterns of priming effects when semantic and associative relations were excluded. If priming effect can be observed at a short SOA, the effect is more likely to be from the automatic recognition process. At a long SOA, the two neighborhood size factors could be used as detectors to check whether word relation is involved in strategic processes. The results showed that the word relation yielded priming effects in both tasks at an SOA of 50 milliseconds. It evidenced that word relation is consistently involved in the automatic Chinese character recognition process. The results also showed that different neighborhood sizes rendered different patterns of priming effects at an SOA of 150 milliseconds. This demonstrated that word relation indeed could be used as a clue to predict target characters when participants have enough time to process prime characters.