Table 1.
Acoustic variables (F0, Fi, ΔF in Hz) and apparent Vocal Tract Length (aVTL in cm) characterising the 4 exemplars (measured on concatenated strings of CVC words).
Figure 1.
Spectrograms of vowel “” (from “book”) created from girl exemplar 1.
Spectrogram settings: window length = .025 s, maximum number of formants, 5; maximum formant frequencies, 6000–6600 Hz. The formants (labeled F1–F4) are shifted down by 18% (A) and up by 12% (C) in comparison to the original signal (B), while all other acoustic parameters, including fundamental frequency, remain unchanged.
Figure 2.
Identification and rating scores of boys' voices along the gender continua.
Scores were averaged across listeners on voice stimuli (numbered 1–16 on the x-axis) for the boys' exemplars. The mean identification scores are plotted from 0 = male to 1 = female (left y-axis) and fitted with the logistic curve (black line). The vertical lines illustrate the location of the estimated sex boundary (where 50% of the listeners rate the stimuli as female) and the location of the prototypical boy voice stimulus (100%). The percentage of stimuli identified as female follows an S-shaped pattern along the continuum of resynthesis variants. The sex identification curve is characterised by a lower plateau for stimuli 1 to 6 (ΔFs of 1138–1267 Hz), where less than 10% of the stimuli are identified as female, indicating that stimuli variant with the lowest ΔF are mostly identified as male. The percentage of stimuli identified as female then increases gradually and linearly, and while no upper plateau is reached, average scores for stimuli 14 to 16 (ΔFs of 1474–1526 Hz) varied from 76% to 85%, indicating that boys' voices with the highest ΔF are mostly classified as female. Average gender rating scores are plotted from 1 = masculine boy (or girl) to 7 = feminine boy (or girl) (right y-axis) and fitted with a linear function (straight grey line). Mean gender ratings of male voices ranged from 1.78 (SE = .07) for the lowest ΔF variants to 5.36 (SE = .08) for the highest ΔF variants.
Figure 3.
Identification and rating scores of girls' voices along the gender continua.
Scores were averaged across listeners on voice stimuli (numbered 1–16 on the x-axis) for the girls' exemplars. The mean identification scores are plotted from 0 = male to 1 = female (left y-axis) and fitted with the logistic curve (black line). The vertical lines illustrate the location of the estimated sex boundary (where 50% of the listeners rate the stimuli as female) and the location of the prototypical boy voice stimulus (100%). The percentage of stimuli identified as female also follows an S-shaped pattern along the continuum of resynthesis variants. The sex identification curve is characterised by a lower plateau for stimuli 1 to 3 (ΔFs of 1129–1184 Hz), where between 10% and 15% of the stimuli are identified as female, indicating that stimuli variant with the lowest ΔF are mostly identified as male. The percentage of stimuli identified as female then increases gradually and linearly until it reaches an upper plateau from stimuli 12 to 16 (ΔFs of 1432–1542 Hz), with average scores varying from 92% to 95% and indicating that girl voices with the highest ΔF are mostly classified as female. Average gender rating scores are plotted from 1 = masculine boy (or girl) to 7 = feminine boy (or girl) (right y-axis) and fitted with a linear function (straight grey line). Mean gender ratings of female voices ranged from 2.33 (SE = .02) for the lowest ΔF variants to 6.10 (SE = .06) for the highest ΔF variants.