The McGurk effect occurs when conflicting auditory and visual speech information result in an emergent percept. The incidence of the McGurk effect is greater for speakers of English than Japanese, and in turn for speakers of Japanese than Cantonese. Sekiyama postulates that this is because speakers of tonal languages rely more upon auditory than visual information in speech perception. Here this hypothesis is tested by presenting both tonal (Cantonese) and non-tonal (English) language speakers with McGurk stimuli in which the tone on syllables either varied or remained constant across trials. Cantonese perceivers relied more upon auditory information than did Australian perceivers, but over and above this, tone variation affected the relative salience of auditory and visual information. This effect was different for the two syllables [ba] and [ga]. In auditory-visual conflict trials subjects' responses appear to depend on the effect of the tone variation on the specific auditory and visual syllables. The results are discussed in terms of the auditory and visual correlates of tone.