Pairs of Cantonese sentences were constructed which differed in only a mono-syllable character in each case. The two syllables involved could either differ in their vowels or their tones, or both. Subjects heard recordings of one of each pair of sentences and were asked to identify the sentence. It was found that difference in segmental phonemes provided more effective perceptual clues than difference in tonal phonemes, and segmental phonemes alone appeared to be not significantly different from segmental phonemes plus tonal phonemes. Obtained results were not related to the possible response bias due to relative naturalness/familiarity of particular constituent sentences of the pairs. Some possible explanations are discussed.