The protocol included a descending chromatic scale over a perfect fifth (8 semitones) from ranges selected to show how the singer’s formant might show characteristic and variable patterns among and within subjects. In a single subject the higher formants are expected to vary little, given a limited F0 range and a constant vowel, and the evenly distributed harmonics over the duration of the maneuver (8 seconds) assure that the long time average spectra (LTAS) show characteristic levels and frequencies of the singer’s formant for the given range. (The spectrograms appear fuzzy because a broader bandwidth is used for LTAS.)
The figure above compares LTAS of two sopranos singing the maneuver on the vowel [a] in middle register (here D5 to G4). The contrast between the two with respect to the singer’s formant is evident. The upper display (subject 2) shows F3 and F5 about 1400Hz apart, with a weaker F4 approximately half way between them. In the lower display (subject 3) F4 and F5 are brought lower and closer to one another. Curiously enough, this greater compactness of the singer’s formant does not result here in a higher level with respect to the first formant.
The figure above compares spectra of two vowels, [E] (above) and [a], in the middle range (D4-flat to G3-flat) of a tenor (subject 6) with an unusually strong singer’s formant. Although the singer’s formant is often considered to be independent of the vowel, note the marked differences between the the two: relative to the level of the first formant (here the first peak) the singer’s formant is 8 dB higher in the back vowel [a], where it is clearly articulated in a way that clusters F3, F4, and F5 into a single effective formant, whereas the [E] shows an apparent gap of about 500 Hz between the independent peaks of F3 and F4. Not surprisingly, the relative level of the singer’s formant in the tenor voice easily exceeds that of the two soprano voices.