Speech perception concerns the process of perceiving phonological units from running and variable acoustic signals. It has been indicated that the perception of phonetic categories experiences several different levels of analysis, including the level of processing simple acoustic feature, the level of integrating complex acoustic patterns, and the level of phonetic category perception. While few experiments examine the time courses of these process stages, which are the critical evidence for the existence of these analysis levels. The present study investigated the time course of feature analysis and integration in the perception of Chinese vowel category. Applying priming paradigm, three experiments were administered, and the spectral similarity and the ISI (Interstimulus interval) between the priming sounds and the target sounds were varied. In experiment 1, the priming sounds were two sine-wave tones, with frequency equal to the frequency of the second formant (F2) of the target vowel /a/ and /i/, respectively. Four ISIs were selected: 50ms, 100ms, 200ms, and 300ms. Participants were asked to identify vowels as quickly as they can by pressing buttons. The results found that the priming effect was significant only for the 50ms ISI condition: the identification of vowels in the matching conditions (the frequency of priming tone was equal to the F2 frequency of vowel) was faster than that in the non-matching conditions (the frequencies of the tone and the F2 of vowel were different). In experiment 2, the priming sounds were two tone complexes, which were composed of two tones at the frequencies of the first two formants of the vowel /i/ and /a/ respectively. The ISIs and the experimental procedures were similar to those in experiment 1. The results showed that the priming effect was significant for both 50ms and 100ms ISI conditions: the identification of vowels was faster when the frequencies of the priming sounds and the vowels were matched. In experiment 3, the priming sounds were the onset 100ms segments of the target vowels, the ISIs and experiment procedures were the same as previous experiments. The results found that the priming effect was significant for all ISI conditions. The results of three experiments indicated that with the increasing similarity of spectra between priming sounds and target sounds (from tone, complex, to vowel primes), the duration of priming effect was lengthened (from 50ms to 300ms). The results not only provided evidence for the existence of several early levels of analysis, but also revealed the time course of these processing stages.