To clarify the relationship between the perceptual diagnoses of hypernasality and its physical correlates, we conducted three experiments using the Japanese vowel /i/. We performed a spectral analysis of hypernasality on five cleft palate patients and six velum resection patients to investigate the spectral features related to hypernasality (experiment 1). Psychoacoustic experiments were conducted using various spectrally modified vowels to investigate the relationships between the spectral features and auditory perceptions of hypernasality (experiment 2). We analyzed the estimated spectral envelopes using an articulatory speech synthesis model to clarify the relationship between the spectral features and the velopharyngeal opening (experiment 3). Experiments 1 and 2 suggested four spectral features as the characteristics of hypernasality: (1) broadening of the F1 bandwidth, (2) an additional peak at around 1 kHz [P1], (3)^* a decrease in the magnitude of F2, and (4)^* a dip between F2 and F3 [D2]. In particular, the simultaneous modification of the decrease in the magnitudes of F2 and D2 is a very important cue for the auditory perception of hypernasality. Experiment 3 showed that the dips caused by nasal coupling resulted in the spectral modifications. In addition, the dip regions appeared to spread from the low-frequency region (around F1) to the high-frequency region (above F2), as the velopharyngeal opening enlarged. This suggests that the decrease in the magnitude of F2 and the dip between F2 and F3 [D2] are related to excessive acoustic coupling of the nasal cavity.^* New findings.