許多聽覺知感知實驗的研究均指出輔音間有明確的音位界線（Liberman et al. 1957; Abramson & Lisker 1970），而元音間的音位界線則有許多的重疊之處（Fry et al. 1962），且元音在元音空間是呈連續性的。可想像聲調的範疇感知界線也可能是重疊和連續的。但聲調方面的研究卻有兩種不同的結果，Abramson（1961, 1979）認為泰國話的聲調感知不是範疇性的，而王士元等人（1976）的結果卻指出華語的陰平調和陽平調間有聲調感知範疇界線。 本文延續前人的研究，以自然語音來探討華語平調和低升調間、低升調和高降調、低降調和高降調間的範疇感知界線。我們使用自然語音，改變音節的音高和音強，合成了八個刺激項。辨認測驗中，受試者聽到的是帶有刺激項的句子；二級差區分測驗（AXB）受試者聽到的則是音節。辨認測驗結果顯示在50%的交叉點有一急遽的變化，而二級差區分測驗結果顯示有一相當寬且平緩答對率尖峰（peak）。這答對率尖峰和辨認測驗所得的交叉點彼此相互對應。由於后分測驗結果答對率相當的高，且答對率尖峰不像輔音範疇感知實驗中所得的那麼陡峭，本研究結果說明了聲調感知範疇界線不像輔音的感知範疇界線那麼明確。

　　　　　It is widely agreed that speech perception is categorical in nature for consonants. In contrast, speech perception is generally considered "less" categorical for vowels, reflecting the overlap in production of vowel categories and the rather continuous structure of the vocalic spaces in the languages of the world. Conceivably, overlap and continuity could also condition the perception of linguistic tones in a similar way. The literature on tone categorization, however, is scarce and controversial (Abramson, 1979; Wang, 1976). In this study, we reopened the issue of tone categorization in Taiwan Mandarin, using three tone continua: level-rising (tone 1-tone 2), rising-high falling (tone 2-tone 4), and high falling-low falling (tone 4-tone 3). The est stimuli were derived from natural utterances by giving them F0 and amplitude contours interpolated between continuum endpoints. They were presented to native speakers of Taiwan Mandarin embedded in a frame sentence (identification tests), or as single syllables (AXB discrimination tests). The identification functions showed a steep transition at the 50% crossover point, while the discrimination functions showed rather wide and smooth peaks, roughly centered on the corresponding identification crossover locations. Because the discrimination level was high throughout the entire tone continua and the discrimination peaks were not as sharp as those typically found for consonant continua, the present results support the notion of a lesser degree of categorical perception for tones than for consonants.