Classical Chinese poems have strict regulations on the acoustic pattern of each syllable and are semantically meaningless. Using such poems, this study characterized the temporal order of tone and vowel processing using event-related potentials (ERPs). The target syllable of the poem was either correct or deviated from the correct syllable at tone, vowel or both levels. Vowel violation elicited a negative effect between 300 and 500 ms regardless of the tone correctness, while tone violation elicited a positive effect between 600 and 1000 ms. The results suggest that the vowel information was available earlier than the tone information. Moreover, there was an interaction between the effect of vowel and tone violations between 600 and 1000 ms, showing that the vowel violation produced a positive effect only when the tone was correct. This indicates that vowel and tone processing interacts in the later processing stage, which involves both error detection and reanalysis of the spoken input. Implications of the present results for models of speech perception are discussed.