Researches of metaphor make people realize that metaphor is not only a means of rhetoric but also a cognitive style to explore the world. Cognitive linguistics about metaphor focused on cognitive mechanism of metaphor. With question about whether metaphor process need more cognitive effort than literal language, people found several theoretical interpretations such as standard pragmatic model, direct access view and graded salience hypothesis (GSH). GSH suggested that the fact of salient make the difference of language comprehension but not the fact of difference between literal language and nonliteral language (such as metaphor, ironic, idiom and proverb). With the advance of event relate potential, we recorded the brain electrophysiology when processing metaphor and literal language under strong or weakly salience. As with same level of salience between metaphor and literal, the literal stimulus could evoke a large N400 component because more effort according to direct access view. However, if weakly salience stimulus evoked larger N400, we could deduct that the salience impact Chinese metaphor comprehension and support the GSH. To investigate how the salience influences metaphor comprehension, we presented a modified S1-S2 task of metaphor and literal paired word with different salience (strong/weakly). All stimuli presented with paired-words extracted from rating pool with high or weakly salience as sentences form of “A is B” (such as Time is money). Participants read two levels of salience paired-words of metaphor and literal, each level with 40 trials, and judged the semantic relation between two words (S1-S2). The brain electrical activity of S2 were recorded from 64 electrodes mounted on participant’s heads according to the international 10–20 system with Neuroscan 4.3 software (NeuroScan, Herndon, VA, USA). Reference electrode was placed on the left mastoid and average mastoid reference was derived off-line using right mastoid data. EEG was amplified (half-amplitude band-pass 0.05–100 Hz) and digitized at a sampling rate of 500 Hz. The EEG was segmented into epochs of 200 ms pre-feedback (for baseline correction) and lasted 1000 ms after its onset. Two ANOVA analyses with fact of sentence type (metaphor vs. literal) and salience level (strong vs. weakly) were conducted of mean amplitude for N400 and P600 component. With results of N400, we found significant main effect of salience level [F (1,15) = 11.737, p < .01] that weakly stimulus with