Page 30 - Fall 2011

encounters being resolved before escalation to violence. Vocalizations often play a key role in such cases, with intim- idation through vocal signaling of size believed to be a criti- cal factor (Bradbury and Vehrencamp, 2011). Results from a number of nonhuman species are consistent with this view, for instance demonstrating correlations among vocalizer body-size, vocal-tract length, and formant frequencies, as well as listeners’ sensitivity to vocalizer resonance cues (Fitch, 1997; Fitch and Fritz, 2005; Harris et al., 2006; Reby and McComb, 2003; Riede and Fitch, 1999). Using signaling to influence contest outcomes necessarily creates selection pres- sure for exaggeration, however, in this case of apparent size. In one extreme case, red deer males have been shown to lower their larynx more than 30 cm when vocalizing (Fitch and Reby, 2001). Females are indeed affected by resulting res- onance cues, but do not show this effect themselves. If voice-related intrasex competition also occurs in humans, it is thus reasonable to expect that males will be most affected. Male vocal characteristics in particular should be correlated with overall body-size, but vocalizers may also exaggerate those cues. Between age-sex classes, at least, it is clear that key vocal characteristics are significantly correlated with body-size. Both F0 and formants are lower in adults than in children (Hirano et al., 1983; Hollien et al., 1994), and in adult males than in adult females (Hillenbrand et al., 1995; Rendall et al., 2005). Human listeners are also sensitive to these differences, and use them to identify vocalizers as men, women, or children (Coleman, 1976; Owren et al., 2007). However, correlations between voice and body-size are much weaker within age-sex class—including in adult males. In fact, there may be no relationship between F0 and body-size in either males or females (Rendall et al., 2007). There is stronger evidence of a reliable correlation between vocal- tract length and body-size, but the degree of correlation is again modest, and not entirely consistent across studies. The picture is also complicated by the fact that human listeners are not very good at judging vocalizer body size (Collins, 2000; van Dommelen and Moxness, 1993; González, 2004). Furthermore, judgments tend to be based on F0 differences, which is the less-reliable cue (Rendall et al., 2007). Formants do predominate when stimuli are equated for discriminabili- ty on the two dimensions (Pisanski and Rendall, 2011), but with the caveat that naturally occurring resonance differ- ences between the sexes are significantly smaller than pitch differences. In other words, equating for discriminability means presenting formant cues that are arguably proportion- ately larger than the F0 cues. Overall, then, results concerning intrasex competition based on body-size signaling are mixed. On the one hand, it is clear that disproportionate sex differences do exist for both F0 and formants. Furthermore, F0 cues sway listener judg- ments for both male and female vocalizers, while it is specif- ically male versions that are exaggerated. A smaller, but detectable effect is also present for formants, most often in male voices. On the other hand, within-group correlations between vocal characteristics and body-size are uncertain for F0 and modest for formants. Furthermore, listeners are para- doxically more swayed by vocal pitch, which is almost entire- ly unreliable. One possible explanation for these seemingly contradic- tory outcomes is that reliable body-size cues are unneces- sary—vocalizers may instead be capitalizing on the strong, global relationship between physical size and both pitch and resonance that exists in the world at large. In other words, because a strong relationship exists between the size of an object or animal and associated pitch and resonance cues in the world at large, listeners are swayed even by unreliable vocal cues (Rendall et al., 2004). F0 cues may also be easiest to exaggerate, as the human larynx grows more or less per- pendicularly to the body axis, and can protrude from the neck without disturbing other tissue (Fitch, 2000; Fitch and Hauser, 2002). As the pharynx grows along the body axis and oral cavity length is likely constrained by mandible and tooth geometry, vocal-tract length remains more proportional to body-size as a whole. One might even argue that F0 cues have become exaggerated to the point of unreliability in human males, with formant cues differing only in being affected to a lesser degree. However, that account leaves unexplained why listeners would be differentially sensitive to the less reliable cue. A second, quite different argument is that F0 cuing is more accurate than hitherto believed. In this view, relying on height and weight differences importantly underestimates male-female differences. Specifically, human males have 60% more lean muscle-mass than females, and 80% greater mus- Loudspeakers and room acoustics carry the message Room Acoustics Software www.odeon.dk Human Voice in Evolutionary Perspective 29