中文摘要：

采用知觉干扰效应的经典实验范式，首次利用ERP技术探究知觉干扰效应的内部机制。结果发现，递进条件和标准条件均诱发了P2，P3，N400和晚期正成分LPC。在P2，P3和LPC成分上，标准条件的波幅均显著大于递进条件；在N400成分上，递进条件比标准条件波幅更负，潜伏期更长。此外，对递进条件下靶字各模糊水平（Level 1-Level5）的ERPs进行比较分析表明，N400成分的波幅在各模糊水平间的差异均不显著。基本支持失匹配理论及词汇自动加工理论。

英文摘要：

The perceptual interference effect is the finding that prior exposures to a more degraded stimulus compromise the identification of a less degraded version of the stimulus later. Up to now, there are two influential theories which explain this effect： competition explanations （Bruner ＆ Potter, 1964; Luo ＆ Snodgrass, 1994） and mismatch explanations （Wang ＆ Reinitz, 2001 ）. Notably, previous studies on perceptual interference almost all focused on the behavioral experiments. However, no study exploring the internal mechanism of this effect on the neurophysiological level has been showed yet. So the aim of this study is to investigate the neurophysiologieal substrate of the perceptual interference effect and to further explore the underlying mechanism of this effect. In additon, in the field of N400 studies, the debate between the prelexical account and the postlexieal account has been very intense. Is the perceptual interference effect related to the N400？ The study of this question may be another way to resolve the debate. According to the perceptual interference effect theory, because the hypotheses compete with each other or the mismatch yields inhibition, we predict that the incremental condition will produce a more negative ERP deflection than the standard condition, and this negativity is likely to be N400. If so, this study will support the N400 lexical process theory. 270 Chinese characters were selected as the experiment materials （ 180 target items and 90 probe items）. Each stimulus was drawn in white color in the center of a 60 x 60 pixel black square window, in the font of Song/Kai Ti 119 in a Flash MX 2004 environment, and then saved as a smoothed BMP image. For targets, each image was then fragmented at blurredness from Level 1 to Level 5 with a fragmentation procedure of Luo and Snodgrass （1994）. In each trial, the participants were first presented with a fixation mark for 300 - 500 ms immediately followed by the target characters. The fragmented stimulus of each character