Formant continuity and auditory scene analysis: The effect of vowel formant manipulations on the perception of synthetic nasal consonants

Abstract

The perception of a synthesized nasal /m/ changes to /n/ as the frequency of the second formant of a preceding vowel is increased when there are no transitions between the vowel and nasal; but /m/ is heard consistently when 20-ms transitions are introduced. A possible explanation is provided by auditory scene analysis: Formants of the vowel and nasal that are contiguous and close in frequency may be grouped using principles of similarity and good continuity into a single perceptual stream. Further experiments found a robust change in percept for prototype /n/ as well as /m/. Transitions from the vowel that conflict with the formant structure of the consonant also cause a similar change in percept for both nasal prototypes. The proximity of the second formant of the vowel to a formant in the nasal prototype is therefore unnecessary for the change in percept to occur; the presence of the vowel formant near to one of two target frequencies at the boundary with the nasal seems to be sufficient to determine the nasal percept and takes precedence over the structure of the nasal prototype. Thus, these results do not show strong evidence for auditory scene analysis applied to formants.