Abstract: The results of previous psycho-phonetic studies indicate that Mandarin native speakers discriminate tone 1 and tone 2 categorically, while they discriminate tone 2 and tone 3 continuously. Tone 3 in Mandarin is phonologically described as a low tone, while the results of perceptual experiments suggest that Mandarin speakers tend not to identify a low level tone as tone 3. This study aims to answer the following questions: 1) Does register have an effect on the perception of tone 2? 2) Is it true that there is no category boundary between tone 2 and tone 3? 3) Do Mandarin speakers perceive a low-rising tone in the same way as they perceive a falling-rising tone? 4) What do the results of perceptual experiments indicate about the phonological features of tone 2 and tone 3 in Mandarin?
Three psycho-acoustical experiments were conducted. Experiment Ⅰ aimed to investigate how Mandarin speakers discriminate tone 1 and tone 2, using rising tone continuum which was created by changing the onset frequency step by step. Experiment Ⅱ aimed to investigate how Mandarin speakers discriminate tone 2 and tone 3, using falling-rising tone continuum which was created by changing the ending frequency and the turning point position of pitch curve step by step. Experiment Ⅲ aimed to investigate how Mandarin speakers perceive level-rising tone, using level-rising tone continuum which was also created by changing the ending frequency and the turning point position of pitch curve step by step. The subjects were 11 Mandarin native speakers. The subject tasks of the three experiments were all identification, i.e., to identify tone category for each tone they heard.
The results of experiment Ⅰ showed that Mandarin speakers also discriminated tone 1 and tone 2 when a relative scale, i.e. semitone, was used in step design of the tone continuum, and that when the register was relatively high, a lager rising range was needed to identify tone 2. The results of experiment Ⅱ demonstrated that 1) the main effects of the turning point position and of the ending pitch of a falling-rising tone on identification of tone 2 and tone 3 were both significant, and the interaction of the two factors was also significant; 2) under certain conditions of final pitch and of turning point position there existed a boundary between tone 2 and tone 3, although the boundary was not significant under all conditions. The results of experiment Ⅲ showed that both the turning point position and the ending pitch of a falling-rising tone had effects on identifying tone 2 and tone 3. However, the rate of identification of tone 3 was always below 20 percent.
In conclusion, the results of this study indicate that 1) identification of tone 2 does not need a large rising range when the tone register is not high; 2) discrimination of tone 2 and tone 3 is quasi-categorical; 3) the pattern of pitch curve plays an important role in identification of tone 2 and tone 3, i.e., the falling feature is significant for tone 3; 4) there may exist both a high range and a low range in the Mandarin tone system , and tone 2 and tone 3 belong to the low tone range.

Key words: tone 2, tone 3, perception, phonological feature