Perception of prominence by Estonian and English listeners.

This study deals with the perception of prominence in speech and nonspeech signals by listeners who are native speakers of Estonian or English. The stimuli consisted of repetitions of the synthetic syllable [ba] at a constant fundamental frequency (F0) or signal-correlated noise tokens whose amplitude envelope matched that of the [ba]. The basic token was 400 msec in duration. Listeners heard sequences of four stimulus tokens separated by 120 msec of silence. One token in the sequence could be lengthened to 425, 450, 475, or 500 msec and/or increased in amplitude by 3 or 6 dB; changes in duration and amplitude were independent. The speech ([ba]) and nonspeech (noise) listening tests were run separately. The listeners' task was to indicate which of the four tokens was "most prominent". The responses showed that, for English-speaking listeners, amplitude cues overrode duration cues, while Estonian listeners were more responsive to duration cues. For both groups of listeners, smaller increments in duration and/or amplitude sufficed to produce perceived prominence in the noise condition than in the speech condition. This is interpreted as being due to the absence of F0 variation. In spoken language, F0 is a significant component of stress both in production and perception. Thus the absence of F0 variation should have greater effect on the perception of speech-like stimuli than noise stimuli. The results of the experiments show this to be the case.