Orchestrating Acoustic Cues to Linguistic Effect *

A most convincing way to demonstrate that an acoustic property is a cue for the listener would be to find speech events that constitute minimal pairs with respect to that property, but in nature such pairs are most unlikely. The English words rapid and rabid are a minimal pair at the level of the segmental phoneme, and are near minimal at the level of the phonetic feature, but as many as sixteen acoustic properties are candidate cues to the lexical distinction. Three properties lend themselves to simple waveform editing: the duration of the stressed vowel, the duration of the closure, and the glottal buzz versus silence of the closure signal. Listener responses to stimuli having natural values of these properties show that, with a single exception, there was no decisive effect on word identification produced by a shift in the value of anyone property. Adding glottal buzz to the /pJ closure led listeners to report the word "rabid." To transform original "rabid" to "rapid," at least two properties had to be changed to achieve any significant effect, one of these necessarily the replacement of closure buzz by silence.