Experiments with synthesized serbocroatian tones

Abstract

The paper describes two experiments in the perception of Serbocroatian tones, using stimuli generated by means of a speech synthesizer. The first experiment tested the perception of changing fundamental frequency. It was found that in listening to a risingfalling fundamental frequency contour, listeners were unable to perceive the difference between end points of the contour; they seemed to base their judgment on the difference between the peak and the terminal fundamental frequency. The second experiment tested the perception of long rising and long falling accents, using 44 synthesized forms of the word radi. Listening tests showed that the fundamental frequency pattern on the first syllable was not sufficient for identifying the accent. The additional factor that had to be present before listeners agreed in their responses was the fundamental frequency of the second syllable. For a falling first syllable, the second syllable had to acquire a fairly low value before the judgments shifted from random to falling. For a level first syllable, the fundamental frequency of the second syllable had to be at least as high as the level of the first to evoke consistent rising responses. The same applied to stimuli with a rising first syllable. In most cases, the shift was not between consistent rising or falling responses, but between either a rising or falling and a random response. The fundamental frequency patterns thus distinguish not so much between two different accents as between the presence and absence of an identifiable accent. © 1972 S. Karger AG, Basel.