Previous studies have found that accentuation plays a very important role in spoken language comprehension. However, the specific mechanisms of accentuation affecting spoken language processing remain a matter of debate. One account is that accentuation is related to the focus distribution of the sentence, and the other is that accentuation influences speech processing by modulating the allocation of selective attention. The purpose of the current study is to explore （1） how accentuation influences the allocation of temporally selective attention during on-line spoken language comprehension. In Chinese, different lexical tones＇ pitch patterns are different when being accented. So we want to explore （2） how the effect of accentuation on attention allocation changes with lexical tones.
To study temporally selective attention, EEG technique and attention probe paradigm were used in the present study. A linguistic attention probe ＇ba＇ was presented concurrently with the critical word in the sentence. The materials used in the present study were isolated Chinese spoken sentences, i.e., sentences presented without explicitly provided discourse-level contextual information. All the critical words were single Chinese words. The stimuli were presented through E-prime 1.1. The critical word in the carrier sentence was either accented （Accent） or de-accented （DeAccent）. Meanwhile, its lexical tone was either tone three （Tone Three） or not （not Tone Three）（tone one or tone four）. Thirdly, the critical word in every sentence was added with a linguistic attention probe （a pure tone ＂ba＂ lasting 50ms） or not. Subjects were instructed to listen to the sentences carefully but to ignore the probe. The brain electrical activity was recorded from 64 electrodes mounted on the participants＇ heads with Neuroscan 4.3 software （NeuroScan, Hemdon, VA, USA）.
Firstly, to estimate whether N1 effect was elicited by the attention probe, we analyzed the eight original ERP waveforms and found that, relative to the critical words without probe, the critical words with probe elicited a larger N1 effect. Secondly, we examined how accentuation and lexical tone modulated the N1 effect through the difference of waveforms （the condition with probe subtracts the condition without probe） in the four experimental conditions （Accent-Tone Three, Accent-not Tone Three, DeAccent-Tone Three, DeAccent-not Tone Three）. The results showed that, relative to the DeAccent condition, the Accent condition elicited an earlier N 1 effect [F（1, 18） = 7.75, p 〈 001 ], and the N 1 effect of accentuation was not changed with lexical tones. Furthermore, to examine how accentuation and lexical tone influenced the later stage of semantic processing, the original ERP waveforms elicited by the critical word that was not added with an attention probe was analyzed. The results showed that, in the time window of N400, accented critical words elicited a larger N400 than de-accented critical words [F（1, 18） = 27.84, p 〈 .001 ]. No significant interaction between accentuation and lexical tone was found. These results indicated that, during on-line speech processing, accentuation can modulate temporally selective attention （the information that was accented can attract more attention） and consequently influence the depth of subsequent semantic processing. The effect of accentuation on attention allocation was not affected by lexical tones. Combining the results of time window of N l and N400, we speculate that accentuation can influence speech processing by modulating listeners＇ selective attention.