1. Introduction

The human voice is highly sexually dimorphic. Pitch, the most perceptually salient feature of human voice (Banse & Scherer, 1996), is about half as high in men as it is in women (Titze, 2000). This dimorphism is due not merely to sex differences in body size; relative to both height and body volume, voice pitch is lower in men than it is in women and prepubescent children of both sexes (Titze, 2000). Sexual selection (Darwin, 1871) is the primary evolutionary cause of sex differences, and Collins (2000) suggested that sex differences in the human voice evolved though sexual selection via female mate choice. Some studies have shown correlations between female mate preferences and male voice pitch (Collins, 2000, Oguchi & Kikuchi, 1997), whereas others have examined the effects of experimental pitch manipulation on female preferences (Feinberg et al., in press, Feinberg et al., 2005, Puts, 2005). Puts, 2005 and Feinberg, Jones, Law Smith, et al., (in press) demonstrated menstrual cycle variation in women's preferences for masculine voices. Normally-cycling women's preferences for low, masculine voices increased with conception risk (Feinberg et al., 2005, Puts, 2005), and women preferred lower male voices mainly for short-term, sexual relationships (Puts, 2005). Taken together, these findings suggest that female mate choice may have influenced the evolution of male voice.

However, another type of sexual selection, intrasexual selection via male dominance competition, may also have been an important selection pressure on the voices of ancestral males. Dominance entails access to mates and resources that is relatively unchallenged by competitors. In most animals, dominance is achieved through aggression or threats of aggression, here termed physical dominance. In humans, dominance may also be achieved through skillful leadership and persuasion (Henrich & Gil-White, 2001), hereafter called social dominance. Among nonhuman animals, low voice pitch is associated with physical dominance (Morton, 1977, Morton & Page, 1992), and in humans, voice pitch is associated with interpersonal power and deference relations (Benjamin, 1981, Benjamin, 1992, Gregory, 1994, Gregory et al., 1993).

The reasons that the acoustic features of voice may have evolved as dominance signals can be clarified by examining their proximate causes. The frequency of vocal fold vibration during phonation is called the fundamental frequency, or F0, and closely determines what is perceived as pitch. The determinants of F0 are apparent from the equation

where L is the vocal fold length, σ is the longitudinal stress on the vocal folds, and ρ is the vocal fold tissue density (Titze, 2000). Thus, voice pitch is inversely proportional to vocal fold length and directly proportional to the square root of tension on the vocal folds. Longer vocal folds with less tension on them lead to lower voice pitch. The perception of pitch is also affected by formant frequencies (Higashikawa et al., 1996, Wolfe & Ratusnik, 1988). The term pitch will thus be used hereafter to refer to the perception of voice “highness” or “lowness,” which is influenced by both fundamental and formant frequencies. Formant frequencies, and, in particular, the spacing between them (formant dispersion), are largely determined by the length of the supralaryngeal vocal tract (Fant, 1960, Fitch & Hauser, 1995).

Vocal fold length and tension and vocal tract length have direct relationships to traits associated with dominance. Vocal tract length is related to body size (Fitch & Giedd, 1999), and both fundamental and formant frequencies influence perceptions of speaker size (Collins, 2000, Smith et al., 2005). In addition, substantial evidence supports an association between vocal anatomy and androgens. Under the influence of pubertal androgens, the vocal folds (Hollien, Green, & Massey, 1994) and supralaryngeal vocal tracts (Fitch & Giedd, 1999) of human males lengthen faster than the overall rate of body growth, resulting in drops in F0, formant position, and formant dispersion. Fundamental frequency continues to correlate negatively with endogenous androgen levels in young adult men (Dabbs & Mallinger, 1999) and decreases with exogenous androgen treatment (Need, Durbridge, & Nordin, 1993). In addition to its association with low voice pitch, androgen is positively related to physical aggressiveness (Archer, 1991, Harris, 1999, Ramirez, 2003) and physical prowess (e.g., Clark & Henderson, 2003). More generally, it has been suggested that masculine traits such as low voice pitch, whose development or maintenance depends on high androgen levels, may be honest signals of a competitor's health and vigor (Folstad & Karter, 1992, Zahavi & Zahavi, 1997).

Habitual voice pitch varies across individuals, but pitch is also modulated between and within interactions. If low voice pitch is generally a reliable signal of physical dominance, then the relationship between voice pitch and dominance should be reciprocal. That is, not only should pitch affect perceptions of dominance, but dominance relationships should influence how pitch is modulated during competitive interactions. Indeed, pitch modulation is related to dominance and submissiveness across nonhuman animal species (Morton, 1977, Morton & Page, 1992). Morton (1977) suggested that lowering pitch may be analogous to piloerection in that it increases the apparent size of an animal and thus the threat it poses.

In humans, voice pitch varies with emotional state (Williams & Stevens, 1972) and actors' portrayals of emotional states (Banse & Scherer, 1996), and considerable agreement about speakers' emotional states exists among listeners (Johnson, Emde, Scherer, & Klinnert, 1986), whose judgments depend partly on voice pitch (Sobin & Alpert, 1999). In particular, pitch modulation appears to convey information about the relative emotional engagement, or activation, of a speaker (Russell, Bachorowski, & Fernandez-Dols, 2003). Emotional activation raises F0 by increasing tension on the vocal fold mucosa (σ, in Eq. (1)), mainly via contraction of the cricothyroid muscles and consequent lengthening of the vocal folds (Titze, 2000). Thus, raised F0 is associated with disparate high-activation emotions such as hot anger, elation, and panic fear, whereas lowered F0 is associated with sadness, boredom, and contempt (low-activation emotions) (Banse & Scherer, 1996, Ekman et al., 1976, Sobin & Alpert, 1999, Williams & Stevens, 1969, Wittels et al., 2002).

The positive relationship between emotional stress and voice pitch in humans may reflect homology with nonhuman animals (Ohala, 1983, Ohala, 1984). Across animal species, interaction with relatively dominant conspecifics may increase activation (nervousness), inhibiting displays of dominance and raising vocalization pitch, whereas interaction with submissive associates may induce less activation, disinhibiting dominance displays and lowering pitch.

Surprisingly, no study has examined whether perceived dominance affects speakers' voice pitch during competitive human interactions. Several studies have explored whether voice pitch affects dominance ratings made by listeners, but these studies have obtained mixed results. Aronovich (1976) found no correlation between speakers' average F0 and dominance ratings made by listeners, whereas Tusing & Dillard (2000) found a significant positive correlation between F0 and dominance ratings (i.e., lower F0 voices were rated as less dominant). However, these results are difficult to interpret because neither study was experimental, so neither can address whether any acoustic sexual dimorphism, by itself, affects perceptions of dominance. Other features of vocalizations, such as speech content, intonation, speed, and inflection, may have covaried with pitch and substantially influenced ratings.

A recent experimental study investigated the effects of voice manipulation on dominance ratings. Feinberg, Jones, Little, et al. (2005) manipulated voice pitch/masculinity by shifting both fundamental and formant frequencies. Female participants rated masculinized male and female voices as more dominant than feminized voices. However, because this experiment examined the effects of voice on females' perceptions of male dominance, it could not address whether voice pitch mediates dominance competition among males.

2. Methods

3. Results

4. Discussion

Preliminary analyses showed a marginally significant negative correlation between ratings of physical (but not social) dominance and the F0 of subjects' unmodified competitive recordings. Experimental manipulation of these same recordings, however, produced strong negative relationships between voice pitch and both types of dominance, with pitch manipulation affecting perceptions of physical dominance more than it affected perceptions of social dominance. This paper is the second (after Feinberg, Jones, Little, et al., 2005) to show that manipulating the pitch of male utterances affects perceived dominance and the first to show this effect using male listeners.

One concern that can be raised regards the potential contributions of spectral influences on dominance ratings. Both the study by Feinberg, Jones, Little, et al. (2005) and the present study simultaneously manipulated fundamental and formant frequencies and examined the effects of these manipulations on dominance ratings. Thus, from these studies, the relative contributions of F0 and formant structure cannot be teased apart. Marginally significant correlations between F0 and dominance ratings of unmanipulated voices in the present study suggest an effect of F0 on perceptions of physical dominance. However, the relationship between body size and formant structure also strongly indicates that formants affect dominance attributions, and thus, the relative contributions of F0 and formant structure warrant further investigation.

In this study, perceptions of relative physical dominance predicted how men adjusted their F0 when speaking to a competitor, suggesting that F0 modulation signals circumstantial physical dominance. Participants who rated themselves as more physically dominant than the standard competitor tended to lower their F0 when speaking to him, whereas men who rated themselves as less physically dominant tended to raise it. One plausible interpretation of this correlation is that the competitor intimidated some males more than others, and this differential intimidation affected whether males raised their F0 or lowered it. The alternative explanation — that participants gauged their relative dominance retrospectively by monitoring their own pitch shifts — is logically possible but unsatisfying, because it fails to address why some men raised and others lowered their voices in the first place.

Finally, the relationship between spontaneous voice pitch and number of sexual partners was in the predicted direction but was nonsignificant. That is, men who spoke at a lower pitch reported (nonsignificantly) more mates in the past year.

These findings highlight the salience of voice in male intrasexual competition and generally support the hypothesis that intrasexual selection contributed to human vocal sexual dimorphisms. A potential limitation of this study is that it does not investigate voice pitch and dominance in women. As noted above, Feinberg, Jones, Little, et al. (2005) found that more masculine and lower-pitched female voices were rated as more dominant among females. However, the relationship between vocal acoustic parameters (such as F0) and dominance among females need not differ from the relationship among males in order for intrasexual selection to have created vocal sex differences. It is sufficient for dominance to affect reproductive success differently in the two sexes (Andersson, 1994, p.146), a condition for which there is abundant evidence in humans (e.g., Daly & Wilson, 1988). Moreover, that fact that males, but not females, exhibit a dramatic pubertal pitch change suggests that vocal sexual dimorphism is a result of sexual selection on males rather than on females.

Although both intrasexual and intersexual selection may have affected the evolution of vocal sexual dimorphism, their relative contributions are unclear. Some evidence suggests that intrasexual selection may have been a relatively stronger influence on the evolution of male voices: A two-semitone experimental manipulation of vocal pitch significantly affects both males' ratings of a man's physical dominance and fertile-menstrual-phase females' ratings of his sexual attractiveness (Puts, 2005), but the effect size (η2) of such a two-semitone pitch change is nearly 15 times greater on physical dominance ratings than on sexual attractiveness ratings (Puts, 2004). These results parallel research on other human sexual dimorphisms, such as facial masculinity, in which masculine traits have smaller or less consistent effects on attractiveness (e.g., Johnston et al., 2001, Penton-Voak et al., 1999, Penton-Voak & Perrett, 2001) compared to their effects on dominance ratings (e.g., McArthur & Berry, 1987, Mueller & Mazur, 1997, Perrett et al., 1998).

In summary, we suggest that low voice pitch evolved in men because it increased mating success via two (nonmutually exclusive) routes. First, voice pitch increased the appearance of physical dominance, enabling some men to exclude others from resources (including mates), thus, augmenting mating success. Second, a low voice, through its association with heritable, fitness-related traits such as health and physical prowess, made men more sexually attractive to women near peak fertility in their cycles.

These results should be replicated, particularly in other (non-Western) cultures. However, given the widespread association between dominance and pitch across animal species, including nonhuman primates (Morton, 1977, Morton & Page, 1992), and the cross-cultural universality of voice pitch sexual dimorphism, masculine vocal traits such as low pitch are expected to increase perceived physical dominance among men generally, with cultural variables influencing the degree, but not the direction, of this effect.