Fig 1.
Spectrograms illustrating the acoustic structure of a typical (a) male roar, (b) male scream, (c) female roar, and (d) female scream. Note the higher F0 and more chaotic spectral structure of roars than screams.
Fig 2.
Discriminant function analysis illustrating acoustic separation of voice conditions, (a) for all vocalizers, (b) for male vocalizers only, and (c) for female vocalizers only. Each data point represents the centroid of a vocal stimulus as a function of the first two discriminant variables that maximize individual separation. Larger black circles represent mean group centroids for each voice condition. The radar plot on the bottom right of panel (a) represents the loadings of the acoustic variables on the first two discriminant functions. Mean amplitude, amplitude variability, and amplitude modulation were the main factors separating voice conditions on the first function (DF1, Table A in S1 Tables). The second function (DF2, Table A in S1 Tables) relied mostly on F0 and harmonics-to-noise ratio. The pattern of separation was similar in male (b) and female (c) vocalizers.
Table 1.
Mean acoustic characteristics of male vocal stimuli.
Figures in square brackets represent standard errors.
Table 2.
Mean acoustic characteristics of female vocal stimuli.
Figures in square brackets represent standard errors.
Fig 3.
Attributed strength as a function of actual strength, when listeners rated (a) male speech stimuli, (b) male vocalizations, (c) female speech stimuli, and (d) female vocalizations. Each data point represents the mean strength rating averaged across listeners attributed to each vocalization. Blue circles represent distress stimuli, red circles represent aggressive stimuli. Open circles represent speech stimuli, closed circles represent vocalizations. R2 values for each regression line are reported in the graphs. Removing the strongest female vocalizer from our analyzes did not affect the significance of our results.
Table 3.
Strength attributions: Linear mixed model testing the effects of vocalizer sex, listener sex, stimulus context, and stimulus type on rated strength.
Fig 4.
Attributed height as a function of actual height, when listeners rated (a) male speech stimuli, (b) male vocalizations, (c) female speech stimuli, and (d) female vocalizations. Each data point represents the mean height rating averaged across listeners attributed to each vocalization. Blue circles represent distress stimuli, red circles represent aggressive stimuli. Open circles represent speech stimuli, closed circles represent vocalizations. R2 values for each regression line are reported in the graphs.
Table 4.
Height attributions: Linear mixed model testing the effects of vocalizer sex, listener sex, stimulus context, and stimulus type on rated height.
Table 5.
Strength estimation: Linear mixed models testing the effects of actual strength, stimulus context, stimulus type, and listener sex on the rated strength of females and males.
Table 6.
Height estimation: Linear mixed models testing the effects of actual height, stimulus context, stimulus type, and listener sex on the rated height of females and males.
Table 7.
Standardised linear mixed model coefficients representing the sensitivity of listeners to variation in vocalizer strength and height.
Each coefficient represents the average of listeners’ individual slopes for the relationship between actual strength/height and attributed strength/height. Significances represent whether each average slope was significantly different from zero. Separate models are reported for male and female vocalizers.