Frequency discrimination of stylized synthetic vowels with two formants

Abstract

Just-noticeable differences (jnd's) in the formant frequencies of synthetic two-formant "vowels" were measured for normal hearing subjects. The jnd's were examined for a change in only the first or the second formant, and for a combined change of both formants. For the combined change two quantitative relations between the formant frequencies were used; one with equal relative changes in both formants, and one with a double relative change for the first formant. Formant frequencies were 500, 550, and 600 Hz in the first, and 2000, 2050, and 2100 Hz in the second formant region. Both formants had either shallow or steep slopes. For the fundamental frequency of the complexes we used 100 and 200 Hz. For the single-formant changes, a "natural," and a random-phase relation were used between the individual components of the complexes. These results were compared to jnd's for a Gaussian noise that was filtered with the same spectral envelopes as the harmonic complexes. For the combined formant changes only the natural phase relation was used. A three-interval, three-alternative forced-choice task was used. All measurements were performed with roving stimulus level. For the single formant changes, the phase relations had no effect on the results. For the harmonic stimuli, jnd's were mostly smaller for the formants between two harmonics than for those at a harmonic. The results for the harmonic stimuli as well as the noise bands could be described by a model using a spectral profile comparison. For the combined formant changes smaller jnd's were found than for the single changes. These jnd's could be explained by combining measures of the perceived differences from the two separately changed formants. In this combination these measures were summed as independent variables.