In the present study, the time course of tonal and vowel information processing of the spoken words in Mandarin Chinese was investigated using a delayed-response paradigm. Idiomatic materials, providing semantically highly constraining contexts, were utilized. Besides being presented normally, the terminal monosyllabic words in idioms were manipulated with Tonal, Vowel, or Triple violations (i.e. with consonantal, tonal, and vowel mismatches). Event-related potential results showed that all three violations elicited larger widespread negativities in comparison with the Control condition, with the Triple violation effect starting first from 150 ms, then the Vowel violation, and the Tonal violation being the latest. The different starting times of the violation effects suggest that the access of tonal information is slower than that of vowel information, even though the lexical tones are very important in distinguishing the meaning of Chinese words.