A categorical perception experiment involving an identification and a discrimination task was administered to two groups of speakers of Chinese (Cantonese and Zhumadian Mandarin), in order to establish to what extent the results would reflect the phonetic form or the functional status of monosyllabic pitch contours. Each group was presented with stimuli produced by speakers of their own variety. Generally, a difference between monosyllabic pitch contours can express an intonation contrast (question vs statement) or a lexical contrast. We hypothesized that intonation contrasts are perceived gradiently, while lexical contrasts are discrete. The Zhumadian Mandarin results were in perfect agreement with our working hypothesis. Because the Zhumadian lexical tone contrasts are expressed through variation in pitch shape and the intonation contrast is expressed through variation in pitch height, the Cantonese group was recruited to test our hypothesis with a language that expresses a lexical tone difference through pitch height variation and the intonation contrast through a pitch shape difference. The Cantonese lexical tone contrasts were perceived discretely, but so was the intonation contrast. We conclude that only intonation contrasts that are expressed by means of pitch height variation have no phonological representation, while intonation contrasts expressed through pitch shape differences and all lexical contrasts do. These results confirm widespread assumptions about the tonal representations of (lexical and intonational) pitch contrasts in a way that earlier experimental findings had failed to do.