Language has been considered a prototype of a complex, well-learned, multi-tiered intellectual activity that reflects the fundamental architecture of cognition (Carpenter, Miyake, & Just, 1995). Language comprehension involves the integration of semantic, syntactic, and pragmatic sources of knowledge. Much recent psycholinguistic work has assumed that syntactic and semantic interpretations of a sentence are produced by separate cognitive processes that can work serially or in parallel (Fodor, 1983). The mechanisms and time course of the integration process are a major issue in psycholinguistic research (Gunter, Stowe, & Mulder, 1997). However, there is still controversy about which model (serial model or parallel model) would be applicable. In this article, we introduced and compared several language studies which used event-related brain potential (ERP) measures, focused on time course of sentence processing, and reported divergent findings. Although these different findings might be results of using different methods (e.g. using different grammatical violation), tasks, or language, it is not yet clear where the divergence has come from. We discussed the advantages of ERP measures for disentangling these problems, and investigating the temporal structures of processes involved in language comprehension.