Vocal fundamental and formant frequencies affect perceptions of speaker cooperativeness

In recent years, the perception of social traits in faces and voices has received much attention. Facial and vocal masculinity are linked to perceptions of trustworthiness; however, while feminine faces are generally considered to be trustworthy, vocal trustworthiness is associated with masculinized vocal features. Vocal traits such as pitch and formants have previously been associated with perceived social traits such as trustworthiness and dominance, but the link between these measurements and perceptions of cooperativeness have yet to be examined. In Experiment 1, cooperativeness ratings of male and female voices were examined against four vocal measurements: fundamental frequency (F0), pitch variation (F0−SD), formant dispersion (Df), and formant position (Pf). Feminine pitch traits (F0 and F0−SD) and masculine formant traits (Df and Pf) were associated with higher cooperativeness ratings. In Experiment 2, manipulated voices with feminized F0 were found to be more cooperative than voices with masculinized F0, among both male and female speakers, confirming our results from Experiment 1. Feminine pitch qualities may indicate an individual who is friendly and non-threatening, while masculine formant qualities may reflect an individual that is socially dominant or prestigious, and the perception of these associated traits may influence the perceived cooperativeness of the speakers.

Similarly, voices also elicit personality attributions. Voice pitch (F 0 ) influences the perception of personality traits such as truthfulness, persuasiveness, nervousness, and friendliness (Apple, Streeter, & Krauss, 1979;Kramer, 1977). More recently, voice pitch has been associated with perceptions of trustworthiness (Klofstad, Anderson, & Peters, 2012;Tigue et al., 2012) and sexual infidelity (O'Connor, Re, & Feinberg, 2011). Voice pitch is sexually dimorphic in humans, and lower pitch in men is commonly associated with masculinity and attractiveness (Feinberg, 2008). As with masculine facial traits, masculine vocal traits are also associated with negative traits, such as physical dominance and threat potential (Puts, Apicella, Cárdenas, & Cardenas, 2012;Wolff & Puts, 2010). However, whether feminine voices or masculine voices are associated with prosocial personality traits remains somewhat unclear. While feminine faces are generally found to be more trustworthy than masculine faces, a number of studies have shown that masculine voices are more trustworthy than feminine voices (Apple et al., 1979;Klofstad et al., 2012;Tigue et al., 2012;Todorov, Baron, & Oosterhof, 2008). This seemingly contradictory pattern of results suggests that further examination of the factors influencing perceptions of vocal prosociality is warranted.

Trustworthiness
Voice pitch has been associated with perceptions of trustworthiness. Voices with low F 0 are considered more truthful and trustworthy than voices with high F 0 , in both male and female voices (Apple et al., 1979;Klofstad et al., 2012;Tigue et al., 2012). Masculine (i.e. low-pitch) voices are considered attractive in men, and feminine (i.e. highpitch) voices are considered attractive in women (Feinberg, 2008;Feinberg, Jones, DeBruine, et al., 2005). Attractiveness is often associated with positive personality attributions via a "halo" effect (Eagly, Ashmore, Makhijani, & Longo, 1991;Zuckerman, Miyake, & Elkin, 1995). The halo effect may explain why low-pitched male voices are considered trustworthy (Klofstad et al., 2012;Tigue et al., 2012); however, masculinized pitch also makes men seem likely to engage in sexual infidelity (O'Connor et al., 2011), which is not in line with a straightforward halo effect.
These seemingly paradoxical findings suggest that perceptions of prosociality may have a more complex link with vocal masculinity. Additionally, Tigue et al. (2012) found that lower-pitched voices were considered more trustworthy than their higher-pitched counterparts in voices of both sexes, suggesting that trustworthiness may not be exclusively related to vocal attractiveness, at least among female speakers. Rather, there may be a generalized effect of vocal masculinity being considered trustworthy in voices of both men and women. Additionally, Klofstad et al. (2012), Tigue et al. (2012) and O'Connor et al. (2011) each used manipulated versions of stimuli (raised and lowered F 0 ). Because listeners chose between very masculine and very feminine male voices, as opposed to measuring impressions based on normal variation in a naturalistic sample, this may have led to choosing masculine voices because the feminized voices sounded too high-pitched in direct comparison with masculinized voices. Furthermore, these studies did not examine acoustic traits other than F 0 .
F 0 variation is another vocal attribute that may affect the perception of prosociality. In contrast to jitter or F 0 tremor, which are perceived as voice roughness, F 0 variation is captured by measuring the standard deviation in voice pitch throughout an utterance. As such, the pitch variation (F 0 −SD) captures the amount of within-utterance variation in pitch, and low values of F 0 −SD are perceived as monotony. A high variation in F 0 (F 0 −SD) is considered a pleasant vocal attribute (Apple et al., 1979;Scherer, 1974), and its presence in both play behaviour in non-human primates and in human child-directed speech suggests that it may be used as a signal of affiliation (Goedeking, 1988;Trainor, Austin, & Desjardins, 2000). Variation in F 0 may, then, also be related to perceptions of prosociality. Formant measures (formant dispersion, D f , and formant position, P f ) may also influence listeners' attributions of prosociality, due to their relationships with dominance and intrasexual competition (Puts et al., 2012;Puts, Hodges-Simeon, Cárdenas, Gaulin, & Hodges, 2007).

Dominance
The link between low voice pitch and trustworthiness is a surprising one, given the association between masculinity, anti-social behaviour, and dominance (Mazur & Booth, 1998). Masculinesounding male voices are considered to be cues to dominance, which could aid intra-sexual competition (Puts et al., 2012(Puts et al., , 2007Wolff & Puts, 2010), and low F 0 is associated with dominance both cross-culturally in humans and within nonhuman species (Morton, 1977;Ohala, 1983Ohala, , 1984. Thus, speakers who wish to sound submissive may affect higher-pitched vocalizations, with the goal of sounding small and nonthreatening (Ohala, 1984). Low F 0 , D f and P f are related to body size, and it has been suggested that these traits serve as cues to threat potential (Feinberg, Jones, Little, Burt, & Perrett, 2005;Puts et al., 2012Puts et al., , 2007. Speakers with naturally higher measurements of these vocal traits may be perceived as submissive, which could give listeners impression of being naturally more prosocial and cooperative. A high F 0 may be related to increased perceptions of submissiveness because nervousness (such as that brought about by lying or fear) has an impact on vocal fold tension. An autonomic nervous response via vagus nerve stimulation tightens the vocal folds, which increases F 0 (Charous, Kempster, Manders, & Ristanovic, 2001). F 0 variation may also be influenced by emotional arousal, and may reveal emotional traits of the speaker, such as whether they feel confident or threatened (Banse & Scherer, 1996;Hodges-Simeon, Gaulin, & Puts, 2011); a low F 0 variation has been suggested as a means of intimidation (Hodges-Simeon, Gaulin, & Puts, 2010).
Because low F 0 and F 0 −SD are related to threat potential and intimidation, and because high measures of these traits may be related to nervousness and fear, individuals with naturally higher pitch and pitch variation may, accordingly, be perceived as submissive, which in turn could positively influence perceptions of prosociality. Low measures of these traits may negatively influence ratings of cooperativeness, as dominant individuals may use threat or physical strength to get their way, while cooperation requires working in tandem to a common, mutually beneficial end. Thus, voices that sound masculine and dominant may be considered attractive, or even trustworthy, but an inverse relationship between masculinity and cooperativeness may be expected because masculine individuals may behave in a more selfish way or be less likely to acquiesce to the needs of others (Booth & Osgood, 1993;Dabbs & Morris, 1990).
Listeners' own dominance may additionally influence the way they attribute prosociality to others.  finding that dominant men are less sensitive to facial dominance cues in other men lends support to the idea that social trait attribution may be modulated in part by the individual differences of the listeners. This is also supported by research showing that taller (i.e. more dominant) men are less sensitive to dominance cues in masculinized faces and voices than shorter men (Watkins, Fraccaro, et al., 2010); however, this study also found that height was not associated with selfrated dominance, nor was self-rated dominance associated with dominance attributions in faces and voices. A possible explanation for the differing results presented in the two aforementioned studies may be in the way dominance was measured.  measured dominance as a personality trait using an 11-item questionnaire (Goldberg, 1999), while Watkins, Fraccaro, et al. (2010) utilized a single scaled question about the participants' dominance, which may be more reflective of the participants' conceptions of their own physical dominance rather than capturing dominant personality characteristics. Research by Wolff and Puts (2010) did not find that selfrated physical dominance, physical aggressiveness, or morphometric measures of strength predicted dominance attributions of others; however, the measures taken by these researchers focus on traits that reflect physical formidability rather than dominant personality traits such as those measured by . It may thus be reasonable to suspect that individual differences in dominance as a personality characteristic may interact with the way social traits are perceived in the others.

The present research
In the present study, we examined ratings of cooperativeness for male and female voices based on a naturalistic sample (Experiment 1). Here, we examined measurements of pitch (F 0 ) and pitch variation (F 0 −SD), as well as two measures of formants (formant dispersion, D f , and formant position, P f ). In Experiment 2, we examined the effect of manipulated F 0 on ratings of cooperativeness. In both Experiment 1 and Experiment 2, we additionally measured the dominance of the subjects who rated the stimuli, in order to determine whether this factor affected how cooperative they found the voices of others.

EXPERIMENT 1 Method
Stimuli Stimulus donors were 16 men (ages 18-30, mean age 20.4 years, SD =2.73) and 16 women (ages 18-23, mean age 19.4 years, SD =1.46). All were undergraduate psychology students at the University of Stirling. Recordings were obtained using an Audio-Technica AT-4041 microphone with a cardioid pickup pattern, at a distance of approximately 65 cm using a preamp (M-Audio Audiobuddy). Audio was recorded directly to hard disk as .wma files using Windows Movie Maker v.2.1.4027.0 (Microsoft Inc., 2004), with a 48kHz sampling rate and 16-bit quantization. The room was quiet and partially soundproofed with 1.5 ′′ -thick sound-dampening foam. Participants were recorded while reading a scripted text. This text was selected because of its neutrality of content (see Cowan & Little, 2013). For the purposes of this experiment, 5 seconds of speech was extracted from this scripted recording: "October frequently brings the first frost of the season over the greater part of the UK." Extraction was completed using Audacity Team (2012). We excluded participants whose first language was not English, and those who exhibited difficulties reading from a script (e.g. omitting words, stuttering, long pauses, or repeating words). Additionally, participants over the age of 30 were excluded from our stimulus set so that perceived age would not play a role in participants' ratings (Linville & Fisher, 1985;Mulac & Giles, 1996).
For analysis and playback, audio files were converted to single-channel .mp3 at 320kbps/48kHz using Switch v.2.04. All voice measurements were obtained using Praat v.5.3.03 (Boersma & Weenink, 2013). F 0 was measured using Praat's autocorrelation algorithm. Pitch was searched for 65-300Hz for male voices, and 100-600Hz for female voices, according to the manufacturer's recommendations (Boersma & Weenink, 2013). Measurements of the first four formants were taken (F 1 -F 4 ) using Linear Predictive Coding with the BURG algorithm, using 10 poles and pre-emphasis. Maximum frequencies were set at 5500Hz for female voices and 5000Hz for male voices-again per manufacturer recommendations. These formant measures were used to calculate both formant dispersion (D f , see Fitch, 1997), which is the average distance between the four formants in Hz, and formant position (P f , see Puts et al., 2012), which is obtained by assigning each formant a z-score and taking the mean of these four standardized measures. F 0 −SD (the within-utterance standard deviation of F 0 ) was also recorded. All measurements were obtained using voiced segments of speech only. These four measurements were chosen in order to capture vocal masculinity, due to the sexual dimorphism exhibited by each of these measures. Additionally, these four measures have all been related to attributions of social traits in previous studies.

Subjects
Participants (N = 79) were psychology undergraduates at the University of Stirling: females (n = 57) were aged 18-35 (M = 19.56 years, SD =2.8); males (n = 22) were aged 18-30 (M = 19.95 years, SD =2.5 years). All took part in the study to fulfil a course requirement. All phases of this experiment were approved by the University of Stirling Ethics Committee.

Procedure
Following Havlicek, Roberts, and Flegr (2005) and , participants completed the 11-item dominance subscale of the IPIP (Goldberg, 1999, ipip.ori.org), resulting in a range of scores from 18 to 43, with a mean score of 31.9 (SD = 5.2). This questionnaire was administered prior to stimuli exposure. Male and female voices were presented in separate blocks and randomized within each block. Participants were asked to rate the voices for how cooperative they thought the person sounded. For the purposes of this study, we defined cooperativeness as a measure of how likely you think a person might be to work with you towards a mutually beneficial goal-e.g. writing a presentation or contributing to group work. In these situations, cooperative people will do their fair share of the work required. A person who is uncooperative is not likely to contribute their fair share of work or resources, but will still enjoy the rewards of effort provided by others.
This definition stresses mutual benefit cooperation and highlights the possible existence of defectors/ free-riders. Voices were rated on a 7-point Likert scale, with 1 indicating low cooperativeness and 7 indicating high cooperativeness.

Data treatment
For ANOVA analyses, the 16 voice stimuli for each sex were placed into high-or low-F 0 groups (separated into two quantiles of the 8 highest and 8 lowest F 0 in the sample). This median split was performed in order to maximize statistical power due to the low number of voices sampled (N = 16 for each sex) and to make the results more comparable to experiments that use manipulated stimuli. This same method was used to create high and low quantiles based on the other traits measured (F 0 −SD, D f , P f ). Voices that fell into one high group did not necessarily fall into high groups on the basis of other measurements-for example, voices with high F 0 were not entirely the same as voices with high D f , etc. While this method of collapsing the data into two separate groups is not without its disadvantages, similar methods have been usefully applied by previous researchers to examine differences between groups based on high and low measurements of other traits (e.g. Cowan & Little, 2013;Little, Burt, Penton-Voak, & Perrett, 2001;Penton-Voak & Chen, 2004;Penton-Voak et al., 2003;Stanton, Liening, & Schultheiss, 2011). Linear mixed effects models were performed in R (R Core Team, 2014), with the lmerTest package (Kuznetsova, Brockhoff, & Christensen, 2014).

Results
For tests using the stimulus as the unit of analysis, mean cooperativeness ratings were calculated by averaging the ratings of all listeners. Mean ratings were also calculated separately for male and female listeners in order to determine whether male and female listeners use different cues for perceiving cooperativeness. For tests using the listener as the unit of analysis, we calculated each participant's mean cooperativeness rating given to all stimuli, and also calculated separate mean ratings of male and female voices. Additionally, we calculated the rating given by each participant to high-F 0 and low-F 0 voices separately; the same method was used to calculate ratings based on high/low F 0 −SD, D f , and P f .
The mean acoustic measurements of all parameters are similar to the averages of those examined in previous research, barring mean female voice F 0 , which is lower than the population-level average. While the minimum and mean F 0 are lower, the upper limit is on par with those measured by other researchers (e.g. Feinberg, DeBruine, Jones, & Perrett, 2008;Puts et al., 2012), which indicates that we have simply captured a wider range of female F 0 , and not an unrepresentative sample. Based on the types of analyses we use, there are no reasons to suspect that these lowerthan-average female voices should elicit a pattern of results that would differ in directionality from a more restricted stimulus set, as we have captured a wider range of F 0 than is typically utilized.

Vocal measurements & cooperativeness ratings
All four vocal measurements obtained (F 0 , F 0 −SD, D f , P f ) were sexually dimorphic, with all measures significantly lower for male voices than for female voices (independent-samples t-tests; all t . 5.24, all p , .001) (see Table 1). None of the vocal measurements obtained revealed significant correlations with cooperativeness ratings (Table 2); however, many of the correlation coefficients are notable: F 0 was positively but non-significantly related to cooperativeness for both male and female voices; when male voices were rated by other men, the correlation approached significance, r (16) = .48, p = .059. P f was negatively related to cooperativeness ratings for both male and female voices; when female voices were rated by other women, this negative correlation also approached significance, r = −.48, p = .059 (see Figure 1).
A 2 × 2 ANOVA (sex of voice; high/low F 0 ; sex of listener as a between-subjects factor) revealed a significant main effect of F 0 only, F(1, 77) = 34.36, p , .001, with high-pitched voices being found more cooperative than low-pitched voices, in both male and female stimuli, when rated by both male and female listeners (see Table 3). There was also a significant interaction between speaker sex and F 0 , F(1, 77) = 6.33, p = .01. No other significant effects or interactions were observed (all F , 3.19, all p . .08). Repeating the analysis separately for male and female voices revealed that the main effect of F 0 on cooperativeness ratings was stronger for female voices, F(1, 77) = 38.16, p , .001, than for male voices, F (1, 77) = 5.41, p = .02 (see Figure 2a). For the measure F 0 −SD, the 2 × 2 ANOVA (sex of voice; high/low F 0 −SD; sex of listener as a between-subjects factor), a significant main effect of F 0 −SD was revealed, F(1, 77) = 48.00, p , .001, indicating that a high F 0 −SD was found more cooperative than low F 0 −SD in both male and female voices. Similarly, a 2 × 2 ANOVA for D f (sex of voice; high/low D f ; sex of listener as a between-subjects factor) revealed a significant main effect of D f , F(1, 77) = 34.85, p , .001, with low D f being found more cooperative in voices of both men and women. The 2 × 2 ANOVA for P f (sex of voice; high/low P f ; sex of listener as a between-subjects factor) returned no significant main effect of P f , F(1, 77) = 0.03, p = .87. A significant interaction between P f and sex of listener was observed, F(1, 77) = 12.23, p , .001, as well as a three-way interaction between P f , sex of listener, and sex of voice, F(1, 77) = 5.25, p = .03. Analysing male and female stimuli separately revealed a main effect for female voices, F (1,77) = 5.40, p = .02, such that voices with a higher P f received higher ratings of cooperativeness; however, among male voices, those with a lower P f were rated as more cooperative.

Mixed effects models
We performed additional analyses of what vocal traits influence ratings of cooperativeness using linear mixed effects models. Random effects in all of the models were the listener and voice stimulus. For male listeners of male stimuli, our model was significant, χ 2 (4) = 16.38, p = .003. There were significant fixed effects of F 0 , F = 8.29, p = .004, and listener dominance, F = 11.08, p , .001, on voice cooperativeness ratings. There was also a significant interaction between F 0 and listener dominance, F = 9.41, p = .002, and a near-significant effect of D f , F = 4.01, p = .07. The model shows that men with higher voice pitch and low formant dispersion were rated by other men as more cooperative, and that male listeners who were low in dominance gave higher cooperativeness ratings than did high-dominance listeners. The interaction between F 0 and dominance suggests that high-dominance men found high F 0 to be cooperative, while low-dominance men favoured voices that were low in F 0 .
Our model for female listeners was also significant, χ 2 (6) = 32.39, p , .001. Women rated male stimuli as cooperative based on three fixed factors. F 0 was positively associated with cooperativeness, F = 15.10, p , .001. A high F 0 −SD was also associated with cooperativeness ratings, F = 5.41, p = .04. There was also a significant effect of women's age on cooperativeness ratings, such that older women gave higher cooperativeness ratings, F = 17.42, p , .001. We also observed a significant interaction between listener age and speaker F 0 , F = 13.81, p , .001, indicating that older women found low-pitched voices cooperative, while younger women found high-pitched voices cooperative. A non-significant interaction between speaker F 0 and F 0 −SD, F = 3.61, p = .08, was also present. Although not a significant factor within the model, the inclusion of this interaction term significantly improved the overall model, χ 2 (1) = 4.15, p = .04. The direction of the interaction indicates that a high pitch variation positively influenced cooperativeness ratings, particularly when voices were low in pitch; the positive effect of pitch variation was less pronounced for voices that were high in F 0 . Voices with a low pitch and also a low pitch variation were found the least cooperative, while voices with a low pitch and high pitch variation were found the most cooperative.
For male listeners of female stimuli, there was a significant fixed effect listener dominance on cooperativeness ratings, F = 8.82, p = .003, and a non-significant main effect of F 0 , F = 3.01, p = .08, indicating that men who were high in dominance found women to be less cooperative than did low-dominance men, and high F 0 was Note: Voices were categorized by high or low F 0 , F 0 −SD, D f and P f for ANOVA analyses.
generally found more cooperative than low F 0 . As with men listening to male voices, a similar significant interaction between dominance and F 0 was found, F = 5.00, p = .03. This interaction suggests that high-dominance men found highpitched female voices to be more cooperative, and low-dominance men favoured lower-pitched female voices. Our overall model was significant, χ 2 (3) = 13.67, p = .003. For women listening to the voices of other women, we found a significant model with two fixed effects, χ 2 (2) = 9.46, p = .009, indicating that higher cooperativeness ratings were predicted by both a high F 0 , F = 5.05, p = .04, and a high F 0 −SD, F = 4.35, p = .057. While the significance value for F 0 −SD approached significance within the model, the inclusion of the term did significantly improve the overall model, χ 2 (1) = 4.57, p = .03. In these four models presented, the addition of further factors, including vocal measurements and listener age and dominance, and interactions between these, did not significantly improve the models beyond the results presented.
Among male listeners, scores on the dominance questionnaire were negatively correlated with mean cooperativeness ratings, such that low dominance was related to higher ratings of cooperativeness in

EXPERIMENT 2
The results from Experiment 1 indicated that F 0 and F 0 −SD were strongly linked to cooperativeness judgements. F 0 has been manipulated in numerous experiments by other researchers examining subjective traits such as attractiveness, dominance, and trustworthiness (e.g. Feinberg, Jones, Little, et al., 2005;Puts et al., 2007;Tigue et al., 2012). Thus, we further examine the relationship between F 0 and cooperation by repeating Experiment 1 using a stimulus set consisting of voices with manipulated F 0 .

Stimuli
Eight male voices (ages 19-30 years, M = 21.0 years, SD = 3.74) and 8 female voices (ages 18-23 years, M = 19.6 years, SD = 1.92) were randomly selected from the stimuli used in Experiment 1. Pitch manipulations were made using Praat v.5.3.56 (Boersma & Weenink, 2013). Using Praat's pitch-synchronous overlap add (PSOLA) method, each voice was manipulated in Hz by +0.5 equivalent rectangular bandwidths (ERBs), which is perceptually equivalent to a manipulation of +20 Hz (Traunmüller & Traunmüller, 1990). This created a raised and lowered version of each voice, resulting in a total of 16 male and 16 female voices. The PSOLA method alters the pitch of the voice, while leaving other aspects (e.g. formants) unchanged. Numerous other experiments have successfully used the PSOLA method in experiments examining perceived attractiveness, dominance, and trustworthiness (e.g. Feinberg et al., 2006;Jones et al., 2010;Klofstad et al., 2012;Puts, 2005;Tigue et al., 2012;Vukovic et al., 2008), allowing this experiment to be directly comparable to a large amount of previously published literature. Amplitude was scaled to create a constant presentation volume using RMS (root-mean-squared) method.

Participants
Participants (N = 101) were psychology undergraduates at the University of Stirling. Females (n = 70) were aged 16-40 years (M = 20.3 years, SD 4.66 years); males (n = 31) were aged 17-49 years (M = 20.7 years, SD 5.81 years). All took part in the study to fulfil a course requirement.

Procedure
Apart from the stimuli, the procedure for Experiment 2 was identical to Experiment 1. Stimuli were again presented in separate blocks of male/female voices. Each block consisted of 8 voices, which had been both raised and lowered in F 0 , resulting in 16 voice stimuli per block. Within each block, the order of presentation was randomized.

Results
As in Experiment 1, we calculated each participant's mean cooperativeness rating given to all stimuli, and also calculated separate mean ratings of male and female voices. Additionally, we calculated each participant's mean rating of high-F 0 and low-F 0 voices for both male and female voices separately.
Listeners rated both male and female voices that had been raised in pitch as significantly more cooperative than voices that had been lowered Age was positively correlated with cooperativeness ratings for female raters, r(70) = .26, p = .028. When further examined by sex of speaker, the correlation remained for male voices, r(70) = .27, p = .023, but not for female voices, r (70) = .20, p = .104, though the directionality of the effect is the same. There was no significant correlation found between age and cooperativeness ratings by male listeners, r(31) = −.18, p = .33. No significant correlations between dominance scores and cooperativeness ratings were found among male listeners [rating women: r (31) = −.08, p = .66; rating men: r(31) = .03, p = .86].
Using ANCOVA, we investigated whether selfmeasures of dominance were related to listeners' sensitivity to masculinity cues when judging the prosociality of others (within-subjects factor: mean cooperativeness rating [masculinized, feminized]; between-subjects factor: sex of listener; covariates: age, dominance score). No significant effect of listener dominance was found for men listening to voices of other men, F(1, 67) = 0.17, p = .68, suggesting that listeners' own dominance did not affect men's sensitivity to dominance cues of other men while assessing cooperativeness. We did observe a non-significant interaction between listener age and masculinity cues, F(1, 67) = 3.46, p = .074, which suggests that older men may have been more sensitive to dominance cues than were younger men, and rated masculinized voices as less cooperative than did younger men. There was no effect of age or dominance on men's sensitivity to cues of female masculinity (all F , 1.30, all p . .26). We also found no significant effects of age or dominance on women's sensitivity to masculinity cues when assessing the cooperativeness of other women (all F , 2.69, all p . .11), or of other men (all F , 0.92, all p . .34).
While we found no relationship between dominance scores and cooperativeness judgements among female listeners in Experiment 1, we did observe a significant negative correlation in Experiment 2, irrespective of F 0 manipulation. Here, dominance was negatively correlated with cooperativeness ratings of women rating female voices, r(70) = −.26, p = .03, but not when women rated male voices, r(70) = −.08, p = .52.

GENERAL DISCUSSION
Results from Experiment 1 indicated that feminine pitch traits (high F 0 and high F 0 −SD) and masculine formant traits (low D f and P f ) were considered more cooperative-sounding in male voices than those with masculine pitch traits and feminine formant traits. In female voices, feminine pitch traits (high F 0 and high F 0 −SD) were also considered more cooperative than masculine pitch traits. A masculine D f was also considered more cooperative than a feminine D f , while femininity in P f was found more cooperative for female speakers.
Experiment 2 confirmed our findings from Experiment 1 regarding a positive association between F 0 and cooperativeness ratings, with feminized voice pitch being found more cooperative in the voices of both men and women. Individual differences of the listeners also influenced cooperativeness ratings. Among female listeners, age was positively correlated with cooperativeness ratings given to other women in both Experiment 1 and Experiment 2. In Experiment 1 only, dominance in male listeners was negatively correlated with cooperativeness ratings given to women, and men who were low in dominance found male voices with masculine F 0 cooperative, while high-dominance men found feminine F 0 cooperative.

What makes a voice sound cooperative?
Both Experiment 1 and Experiment 2 illustrate that a high F 0 is associated with perceptions of cooperativeness. Male listeners also displayed a tendency to rate voices with a high F 0 −SD as more cooperative than voices with a low F 0 −SD. Additionally, voices with low formant measures were rated as more cooperative than voices with high formant measures (both D f and P f for men, and D f for women); however, our mixed model analyses confirmed this relationship between a low D f and increased ratings of cooperativeness for men listening to other men's voices only. It is important to note that while D f is a measure of the spacing between the formants, P f is a measurement of the mean frequency of the formants. So, while cooperative male voices had little space between the formants (a low D f ), those formants also have a low mean measured value (a low P f ), according to our ANOVA results. For women's voices, a low D f was considered cooperative, as well as voices with a high mean measured formant value (a high P f ).
Feminine pitch and pitch variation, combined with masculine formants, appear generally to be vocal traits that influence perceptions of cooperativeness. Our results regarding the relationship between high pitch traits (F 0 and F 0 −SD) and ratings of cooperativeness were the most clear, and our mixed models also support the relationship between masculine D f and ratings of cooperativeness. While our results concerning P f were not conclusive based on the mixed effects model presented in Experiment 1, the relationships we uncovered using the median-split technique are intriguing, especially considering the differing directionality for male and female voices, and this is worth examining further. Future research may wish to examine more thoroughly the relative importance of pitch and formant traits.

Pitch and pitch variation
High voice F 0 positively influenced perceptions of a speaker's cooperativeness in voices of both sexes, in both a naturalistic (Experiment 1) and a manipulated sample (Experiment 2). In both studies, a higher voice pitch elicited increased ratings of cooperativeness in male and female voices. Low F 0 is associated with masculinity (Feinberg, 2008;Pisanski, Mishra, & Rendall, 2012;Pisanski & Rendall, 2011;Puts et al., 2012) and is related to high testosterone levels among men (Dabbs & Mallinger, 1999). Given that men with high testosterone are prone to numerous antisocial and risktaking behaviours (Apicella et al., 2008;Archer, Birring, & Wu, 1998;Booth & Osgood, 1993;Coates & Herbert, 2008;Dabbs & Morris, 1990;Mazur & Booth, 1998;O'Connor, Archer, & Wu, 2004;Rowe, Maughan, Worthman, Costello, & Angold, 2004;Stanton et al., 2011;Studer, Aylwin, & Reddon, 2005), the perception of masculinity may suggest a general air of uncooperativeness or lack of prosociality. This finding underscores the importance of F 0 as a male intrasexual signal, with low F 0 indicating physical dominance (Hodges-Simeon et al., 2010;Vukovic et al., 2011;Wolff & Puts, 2010). Given that physically dominant persons may use their physical strength to get what they want, and less dominant individuals may be less likely to physically challenge others, persons with lower perceived masculinity and dominance may be considered desirable as potential cooperators. Additionally, vocal femininity may be associated with a certain degree of compliance, which may also be captured by the construct of "cooperativeness".
Voices with high F 0 −SD (i.e. more dynamic, less monotone voices) were rated as more cooperative than voices with low F 0 −SD. While some of the apparent effect of F 0 −SD in our ANOVA results may be attributed to its correlations with other measured traits, our mixed model analyses present a more fine-tuned picture of how this trait stands alone as a main effect, particularly for female listeners, for whom F 0 −SD was positively related to cooperativeness ratings of both male and female voices. Hodges-Simeon et al. (2010) found that low F 0 −SD predicted higher ratings of physical dominance, which may be an undesirable trait in potential cooperators. F 0 −SD is sexually-dimorphic, with high F 0 −SD being a feminine trait. This trait in men, then, may sound friendlier and less dominant than monotone voices. The fact that we also found high F 0 −SD to be associated with cooperativeness in female voices (for female listeners only) lends further credence to F 0 −SD's inverse relationship with perceived dominance in voices of both sexes. Our results support the suggestion that variation in F 0 may serve to elicit and maintain positive emotional states in the listener (Traunmüller, Eriksson, & Traunmüller, 1995). Hodges-Simeon et al. (2010) liken F 0 variation to a "smile", positing high F 0 −SD as a submissive social gesture. Our data support such an association.
We also found an interaction between pitch and pitch variation in male voices, when judged by female listeners. Men's voices with a high pitch variation were considered more cooperative generally, but also influenced cooperativeness ratings to a greater degree when the voices were also low in F 0 . The positive effect of pitch variation on cooperativeness ratings was less extreme for voices that had a high F 0 , suggesting that negative perceptions of a low F 0 may be ameliorated if the speaker also has a high pitch variation.

Formants
Low D f is generally considered a masculine trait (Feinberg, Jones, Little, et al., 2005;Wolff & Puts, 2010) and is associated with a larger body size (Feinberg, Jones, Little, et al., 2005) due to the allometric relationship between body height and vocal tract length (Evans, Neave, & Wakelin, 2006;Fitch, 1997). While we may not expect a masculine trait to be associated with cooperativeness, it has previously been suggested that formants have a greater effect on dominance judgements than on judgements of masculinity (Hodges-Simeon et al., 2010;Pisanski & Rendall, 2011), suggesting a link between the formants and perceptions of social traits. If the formants are a somewhat reliable indicator of speaker size or height, listeners may find taller individuals to be more cooperative, possibly because height could confer prestige. While voices with low D f are considered dominant, Puts et al. (2007) suggest that this could also be related to social dominance or prestige. This generalized association between low D f and cooperativeness may be linked to the social benefits of height. Taller individuals have greater social status (Cavelaars et al., 2000;Power, Manor, & Li, 2002), higher levels of education (Power et al., 2002;Silventoinen, Lahelma, & Rahkonen, 1999), and greater earnings than shorter individuals (Judge & Cable, 2004;Loh, 1993), which may make them more favourable as potential cooperators. It could be that listeners are attuned to cues of social status, while male listeners are more attuned to specific cues of physical dominance. Indeed, while both male and female listeners seem to be influenced by D f , male listeners alone seem to be influenced by P f . Although this relationship was not confirmed in our mixed model results, the relationship is nonetheless an intriguing one. P f is associated with masculinity and dominance (Puts et al., 2012) and was sexually dimorphic in our data. It is also negatively related to height, weight, arm strength, and physical aggression, suggesting that this trait may be generally indicative of threat potential (Puts et al., 2012). Why a trait thus linked to threat potential would be positively associated with cooperativeness is unclear, though it may be that male listeners also associate this trait with social dominance, while female listeners may not respond to the trait in purely social terms.
Further research may give more attention to the inter-relationships between dominance, prestige, and prosociality. The median split used in Experiment 1 to divide our stimuli into groups based on high and low measures of vocal traits does unfortunately contribute to some loss in variation among these traits and is dependent upon the voices in our particular stimulus set. The nature of the data required that differing median splits be made for each of the four measured vocal traits, such that different vocal stimuli fell on either side of this split dependent upon the trait under analysis. This calls attention to the variability and complexity of vocal characteristics and serves as testament to their respective importance when examining perceptions based on these traits.

Individual differences
While we found no effects of listener age or selfmeasures of dominance to be associated with sensitivity to dominant vocal cues in Experiment 2, we did observe generalized effects of age and dominance on mean ratings of cooperativeness given across both studies. Women's age was positively correlated with cooperativeness ratings of male voices in both Experiment 1 and Experiment 2, and our mixed models in Experiment 1 confirm that older women tended to give higher cooperativeness ratings to men's voices than did younger women. We also found that women's age was associated with their sensitivity to F 0 as a cue to cooperativeness in male voices, such that older women found low-pitched voices cooperative, while younger women tended to favour higherpitched male voices. Age can be judged by listening to vocal stimuli with reasonable accuracy (Mulac & Giles, 1996;Ramig, Scherer, & Titze, 1984), and older women generally prefer older men as potential mates (Buss, 1989;Buss & Schmitt, 1993;Mathes, Brennan, Haugen, & Rice, 1985). With age, women may become more confident and socially able and may, thus, find young men less intimidating (the mean ages of stimulus donors in Experiment 1 and Experiment 2 were 20.4 years and 21.0 years, respectively, while the age range of female listeners was 18-35 years and 16-40 years, respectively). Women's ratings of young female voices did not produce the same pattern of results, which seem to be unaffected by the age of the listener. There was no effect of age among male listeners in either Experiment 1 or Experiment 2.
Self-rated dominance appeared to play a role in how male listeners attribute vocal cooperativeness. Low-dominance men gave higher mean cooperativeness ratings to male and female voices than high-dominance men in Experiment 1. The dominance of the listener also interacted with men's sensitivity to F 0 as a cue to cooperativeness. Lowdominance men found low-pitched male and female voices more cooperative, and high-dominance men tended to favour high-pitched male and female voices. It may be that dominant men feel more socially favourable towards voices that exhibit signs of a small physical stature and low masculinity, as these voices may not pose any threats to the listener's own perceived dominance status.
While this result was not replicated in Experiment 2, we found a similar effect of dominance on women's cooperativeness ratings of female voices (which we did not find in Experiment 1). Like men rating other men in Experiment 1, high dominance in women was associated with lower overall cooperativeness ratings given to the voices of other women.  demonstrated that dominance (measured as a personality characteristic) influences the way men perceived dominance in other men, and it may also affect the way listeners perceive other social traits, such as cooperativeness. Dominant personality traits may "interfere" with the way both men and women perceive cooperativeness, such that those who are dominant may tend to view same-sex individuals as uncooperative and favour low masculinity in a cooperative partner, possibly due to an enhanced sense of intra-sexual competition among high-dominance individuals. Additionally, low-dominance individuals may have a generally more positive view of others, as they are less likely to engage in aggressive and antisocial behaviour (Ehrenkranz, Bliss, & Sheard, 1974;Rowe et al., 2004).

Conclusion
While we expected vocal qualities indicating smaller, shorter feminine individuals to be found more cooperative than those indicating larger, taller, more masculine individuals, our results here were mixed. Feminine F 0 and F 0 −SD positively influenced cooperativeness ratings. F 0 ′ s link to perceptions of speaker masculinity is reflected herehigh-pitched and dynamic voices sounded more cooperative than low-pitched and monotone voices. Formants also appeared to play a role in cooperativeness judgements, with more masculine formant measures (D f and P f ) being found more cooperative than feminine formants in male voices. These seemingly dichotomous results further illustrate the relative importance of F 0 and the formants on person perception and the perception of social traits.
Self-rated dominance had some effect on how male and female listeners perceived vocal cooperativeness. We also found that high-dominance men seemed to be more sensitive to F 0 , as they found high-pitched voices of male and female speakers as cooperative, while low-dominance men found low-pitched voices to be more cooperative. However, the results regarding the negative link between dominance and the attribution of prosociality in others that we found in Experiment 1 were not replicated in Experiment 2. We also found that older women tended to rate lowpitched male voices as cooperative, while younger women favoured higher-pitched male voices, perhaps because masculine voices may sound intimidating to younger women. Further experiments may usefully examine the individual differences of listeners and how these affect their perceptions of social traits.
In summary, our results demonstrated that different vocal traits can work in synchrony to create complex interpersonal judgements. While we found that pitch alone was a consistent factor influencing listeners' cooperativeness ratings of both male and female voices, we also found that pitch variation and formants also play an important role in the perception of cooperativeness.