| Speech sounds vary across different conditions and subjects; nevertheless, listeners perceive the phonemes without difficulties. Categorical perception (CP) occurs when listeners map the varying speech sounds into discrete phonemic categories. In CP, to discriminate a pair of stimuli that cross a category boundary is much easier than those that lie within the same category, even though both pairs are separated by an equal physical difference. CP is one of the important properties essential for speech perception.;Pitch contour or its acoustic correlate, fundamental frequency (F0), distinguishes lexical meanings in tone languages. Two topics on CP of lexical tones were studied in the thesis: (1) the factors influencing CP, and (2) the temporal process of CP. These two topics were investigated through both behavioral and event-related-potential (ERP) methods on Cantonese and Mandarin tones.;Four factors were studied. They were (1) intrinsic acoustic properties of pitch contours by comparison between continua of level tones and contour tones; (2) positions of target syllables relative to context (without contextual sentence, at the beginning and at the end of the contextual sentence); (3) language backgrounds by comparison between listeners with different tone experiences; and (4) carrier syllables (real word, non word, and nonspeech). Three temporal stages were studied in the same experimental paradigm. They were (1) the preattentive stage investigated through the mismatch negativity (MMN); (2) the attentive stage investigated through the P300; and (3) the overt response stage investigated through the hit rate data.;All these four factors influence the degree of CP. In the discussion, both general auditory processing and language specific processing are suggested to be responsible for the various types of exhibition of CP, although they have different weights for different factors. Different patterns of CP were also observed in three temporal stages due to different weights of these two types of processing. In summary, a multistage model which includes both general auditory processing and language specific processing is proposed to explain the CP of lexical tones. This model improves previous models by proposing that the weights of these two types of processing in speech perception depend on the types of factors, and the temporal processing stages.;Finally, for the first time in the literature, the thesis also reported that even though a tone contrast (i.e., level vs. rising) is present in both tone systems, the same contrast is perceived differently by the two groups of subjects by virtue of their different language experiences. |
