In everyday life we always try to get the metaphorical beyond the literal meanings of sentences we come across. Metaphor is a kind of frequently used language expression, also a basic way of thinking for human surviving. So how is metaphor processed? Whether the metaphorical meanings can be accessed independent from the literal meanings or they’re based on the access of the literal meanings. Different from Searle’s Hierarchical Hypothesis (1993) and Glucksberg et al.’ Parallel Hypothesis (1982), Giora’s Graded Salience Hypothesis (2003) emphasized the importance of semantic salience in metaphor comprehension, if the metaphorical meanings were salient or familiar to readers, they can be comprehended prior to literal meanings and vice versa. The current study aimed to investigate the influence of semantic familiarity of sentences on the metaphor processing using behavioral and ERPs methods and tried to resolve the theoretical disputes. Experiment 1 employed a sentential violation paradigm to investigate the behavioral effects of familiarity on metaphor processing. We selected 28 nouns as the subjects of familiar sentences, and the other 28 nouns as the subjects of novel sentences. Then we made a metaphorical and a literal sentence for each noun, and a related incorrect metaphorical or literal version. So we made 224 sentences in total, which were screened and split into 4 lists. Each list contained the same number of sentences for each condition and was presented to a participant to counterbalance the sequence effects. Participants were asked to judge the acceptability of the sentences by key pressing and the reaction time (RT) and accuracy rate (ACC) were recorded. The results showed that the RT to judge the metaphorical sentences was faster and the ACC was higher for the familiar sentences, but for the novel sentences, the RT to process the literal sentences was faster and the ACC was higher. These results showed that the familiarity could modulate the processing of metaphorical meanings, and metaphorical meanings could be accessed prior to the literal processing. To further explore which processes of sentence processing the semantic familiarity can modulate, experiment 2 used high time-resolution ERPs with the similar experimental design to examine if the familiarity would influence the deflection of N400 or P600. The results showed that the metaphorical and literal processing elicited comparable N400 and P600 effects for novel sentences, while metaphorical processing induced enhanced N400 and reduced P600 effects for familiar sentences relative to literal processing. N400 is concerned to be correlated with the efforts of semantic retrieval, analysis (Kutas & Hillyard, 1980), the enhanced N400 for novel metaphorical sentences processing means difficulty of analyzing the semantic information. P600 tends to reflect the sentential reanalysis, integration or reflection (Holcomb, 1988), so the reduced P600 in familiar metaphorical processing might suggest that participants fail to deepen the processing of metaphor when be confronted with the abnormal sentences since they have difficulty in earlier sematic processing stage. Taken together, the present study demonstrates that participants can access metaphorical meanings efficiently, when the metaphor is familiar to them. While the metaphor is novel, the processing of metaphorical meaning may be based on the access of literal meanings. These results provide clear evidence for Giora’s Graded Salience Hypothesis.