Do listeners hear /u/ or /i/ when presented with the stimuli represented by the spectrum above? Formant frequencies are appropriate for /u/ while spectral tilt is intermediate between that for /i/ and /u/. However, amplitude is very low and the peak is not prominent relative to the surrounding harmonics. Second formant peak could either be masked by that of or it may not have sufficient local spectral prominence to be audible.
Schematized spectra of stimuli at the (a) amplitude and (b) spectral tilt endpoints for both (c) full and (d) incomplete spectrum stimuli. Note that the (d) incomplete-spectrum /i/ has a peak around 900 Hz, consisting of harmonics with the same levels as those in the full-spectrum /i/.
Total /i/ responses to (a) full- and (b) incomplete-spectrum stimuli as a function of amplitude and amplitude (which is a linear function of spectral tilt). Amplitude measurements were calculated at formant frequencies via Fourier transform of the stimuli. Maximum possible number of responses is 112 for each condition. Lines give the 30%, 50%, and 70% crossovers in responses as determined via cubic interpolation.2
Spectra of stimuli used in Exp. 2. (a) and (b) have a peak at 900 Hz, while in (b) and (d), the peak has been excised.
Total hits to (a) full- and (b) incomplete-spectrum stimuli in Exp. 2 as a function of both amplitude and spectral tilt . The left-most column of responses is missing from (a) as there was no peak at 900 Hz in the unmodified stimulus (i.e., /i/ endpoint). Maximum possible number of responses is 48 for each condition. Lines give the 30%, 50%, and 70% crossovers in responses as determined by cubic interpolation.
Total number of subjects for whom hit rate was greater than the proportion of /u/ responses for (a) full- and (b) incomplete-spectrum stimuli in Exp. 2—i.e., the number of subjects who detected the peak at 900 Hz in the detection task more frequently than they labeled the corresponding stimulus as /u/ in the identification task as a function of both peak level and spectral tilt . Ties were recorded as 0.5. The left-most column of responses is missing from (a) as there was no 900-Hz peak in the unmodified stimulus. Maximum possible is 12 subjects for each condition, while the expected value, given the null hypothesis, is 6. The contour line delineates the minimum value (9) with which the null hypothesis would be rejected at the level of significance using a sign test. The opposite hypothesis—that listeners were less likely to detect the peak at 900 Hz than to identify the corresponding stimulus as /u/—was not significant for any stimulus (maximum value of 3).
Article metrics loading...
Full text loading...