中文摘要：

书面语言阅读过程中，关于读者能否提取预视词汇的语义信息，以往的研究仍有争议。鉴于边界技术在研究该问题时具有无法严格控制预视词呈现时间等一些局限性，近年来有研究者利用ERPs技术结合伴侧RSVP范式对拼音语言中的语义预视效应进行研究。本研究首次将这一方法应用到中文句子阅读中，并对实验材料进行了更严格的控制，保证关键屏（目标名词位于右侧副中央凹处）内容在实验条件间完全相同，避免以往类似研究中关键词自身属性信息对实验结果产生干扰。ERPs数据分析发现，在关键屏呈现后，语义违背相比合理条件诱发了更大的N400成分，而关键后屏（目标名词位于中央凹处）呈现后，两种条件问的N400成分无显著差异。该结果支持中文句子阅读中预视词汇语义信息可以得到提取，并参与随后的语义整合过程。

英文摘要：

Previous studies have confirmed that readers can obtain sub-lexical information （e.g., orthographic or phonological） from the parafovea during reading. However, no consensus has been reached as to whether semantic information can also be extracted from the parafoveal word. A possible explanation for this controversy may be the limitations of the boundary paradigm in eye-tracking reading experiments, which have been the main methodology used in previous studies. For instance, the duration of parafoveal word viewing cannot be held constant across conditions in the boundary paradigm. Thus, in this study, an ERPs paradigm using rapid serial visual presentation （RSVP） with flankers was used to study extraction of semantic information from the parafoveal word in the reading of Chinese sentences. A total of 84 sets of Chinese sentences were constructed for this study. Unlike previous studies, in each set of sentences, the two-character verbs were manipulated, yielding a congruent or incongruent context with the target single-character noun that appeared 4 or 5 characters after the verbs. Thus the critical trials （containing the target noun as the right flanker） were the same across congruent and incongruent conditions. Each sentence was presented character by character at fixation, flanked 2~ bilaterally by the preceding character （n-l） on the left and the next character （n＋l） on the right. Each viewing consisted of three characters presented on the screen for 100 ms with an inter-stimuli interval （ISI） of 400 ms. Twenty-four participants＇ ERPs were recorded individually with 40 tin electrodes （10-20 System） in a sound-attenuating, electrically shielded booth. The EEG and EOG signals were sampled at 500 Hz and filtered digitally offline with a 0.02 to 30 Hz band pass. Epochs of interest were selected time-locked to the onset of the critical trials with a 200 ms pre-onset baseline window and a 1000 ms window after onset. Mean ERPs amplitudes from 300-450 ms and 800-950 ms were computed f