The aim of this study was to apply one-third-octave analysis for measuring an acoustic correlate of hypernasality in the speech of adults with a range of aetiologies (dysarthria, maxillectomy and cleft palate). Subjects included 12 speakers with hypernasality and 12 normal controls. The speech material was the vowel /i/ segmented from two Cantonese single words produced by each speaker. The results showed that speakers with hypernasality had significantly higher energy level for the one-third-octave bands centred at 630, 800 and 1000 Hz, and significantly lower amplitude for the band centred at 2500 Hz than speakers with normal resonance. These results are in general agreement with past findings about nasalization of vowels. This study showed that one-third-octave analysis has high intrajudge reliability and is applicable to the speech of adults with hypernasality due to different etiologies.