Introduction

The pursuit of beauty is common to everyone. As the old saying goes, “Don’t judge a book by its cover”, which rather reflects the attention people paid to looks. Laboratory researchers utilized economic games to examine the influence of facial attractiveness on decision-making. Among them, the ultimatum game (UG) and the dictator game (DG) are two classic games that have been frequently used1. The preference towards attractive individuals in these social transactions was defined as “beauty premium”2. For example, Solnick and Schweitzer3 confirmed the beauty premium effect in UG, showing that participants were more generous to attractive partners who received 8–12% more in anticipated earnings compared to those with unattractive faces. Lucas and Koff4 pointed out that women provided more money to attractive men in both UG and DG, regardless of their fertility status. Similarly, Pan et al.5 found that the likelihood of accepting unfair offers from attractive proposers was higher than unattractive proposers in UG. These studies proved that facial attractiveness can influence people’s economic decision-making and social preferences.

There is also a well-known proverb that says “Sound can resonate with the soul”. Previous researches have demonstrated that several vocal characteristics are correlated with attractiveness, especially vocal acoustic features6. For example, the intensity of voices is positively associated with vocal attractiveness7. Male voices with lower fundamental frequency (F0)8, lower formant frequency (fn)9 and smaller formant dispersion (Df)10 are generally perceived as more attractive, while female voices with higher F0, higher fn and wider Df are preferred by males11,12. Additional acoustic features such as Jitter (variation of pitch), Shimmer (variation of energy), and Harmonic to Noise Ratio (HNR) also affect vocal attractiveness6,10. Vocal attractiveness induces the beauty premium effect similar to facial attractiveness13. Shang and Liu14,15 used the trust game and found that participants were more willing to invest in trustees with attractive voices. Moreover, in the dictator game as third-party observers, participants had a significantly lower willingness to punish proposers with attractive voices than unattractive-voice proposers for giving unfair offers, and they also considered unfair offers proposed by individuals with attractive voices to be more reasonable16.

It is evident that both faces and voices play an indispensable role in our daily life, conveying important social information. Attractiveness is a multisensory experience including visual, auditory and olfactory domains17. Research has increasingly focused on the intertwined nature of facial and vocal attractiveness, highlighting their shared underlying factors, particularly hormonal effects18, underscoring the complexity of human attractiveness and its evolutionary significance. In men, the level of testosterone has been suggested to be linked to both facial attractiveness and vocal attractiveness19. Men with high testosterone typically exhibit more symmetrical facial structures, and deeper voices, increasing their facial and vocal attractiveness in mate selection20. Similarly, high estrogen in women is positively associated with attractive faces and voices, manifesting more symmetrical facial features and higher-pitched voice12,20. Particularly in the periovulatory phase of the menstrual cycle, women with high estrogen tend to show higher facial attractiveness to capture the attention of men4. Furthermore, some studies found that there was a positive correlation between facial attractiveness and vocal attractiveness, indicating that people believe individuals with attractive faces also possess attractive voices21,22,23. Nonetheless, other studies about the integration effect between facial and vocal attractiveness reported inconsistent results. Rezlescu et al.24 asked subjects to rate the attractiveness of targets based on the simultaneous presentation of male faces and neutral vowels (the face and voice did not have to be from the same person). They found that facial attractiveness and vocal attractiveness independently affected overall attractiveness without integration. When the attractiveness of the two channels were mismatched, people felt disappointed with the lower attractiveness channel, which led to a poorer overall impression25. Moreover, Wells et al.26 found that faces contributed proportionally more than voices to attractiveness rating of face-voice compound, suggesting that facial attractiveness may have a processing advantage over attractiveness in the auditory channel. In the field of economic decision-making, one study27 investigated the role of visual and auditory information in DG with four conditions: a baseline group (no faces and no voices), a visual group (only faces), an auditory group (only voices), and an audiovisual group (both faces and voices). The results showed that under the condition where both faces and voices were presented, female subjects allocated more money to recipients with attractive faces. However, when only faces or voices were shown, neither male or female subjects exhibited preference towards those with attractive faces. Facial attractiveness appears to have a positive effect when combined with voice. Unfortunately, this study did not manipulate vocal attractiveness.

Besides, cues expressing different levels of social interest could affect the perception of facial and vocal attractiveness. Social interest refers to the cues conveyed to others indicating interest or disinterest28. In terms of evolutionary psychology, social interest is considered as a type of perceptual signals, which indicate mating qualities and possibilities for returns to individual mating efforts29. Jones et al.30 demonstrated that social interest cues ("I really like you" vs. "I don't really like you") influenced men's preferences for women's voice pitch. Mishra et al.31 also showed that male observers' ratings of female attractiveness in subsequently presented videos were influenced by the degrees of social interest expressed in prior videos. A recent study28 manipulated social interest and vocal attractiveness and revealed higher acceptance rate for attractive voices (compared to unattractive voices) and positive social interest (compared to negative social interest) in UG. These findings provide strong evidence of indispensable role of social interest in decision-making and intricate interactions between social interest and attractiveness.

In addition, beauty premium depends on the participants’ role and bargaining situations. Voit et al.32 asked participants to make decisions in four economic games including the dictator game (DG), the trust game (TG), the ultimatum game (UG), and the prisoner's dilemma game (PD). The results revealed that, in DG and PD, participants made relatively favorable economic decisions towards opposite-sex attractive faces. In UG, men were more influenced by facial attractiveness compared to women. In TG, participants invested more money to attractive-face individuals. It can be inferred that people have different criteria when they possess different decision-making authority. Ma et al.33 also asked male subjects to act as different roles respectively in DG and UG, showing that regardless of the roles they played, the amount they allocated to female partners increased with the level of facial attractiveness. But female proposers induced a smaller "beauty premium" than female recipients.

Evolutionary theories of human mate preferences suggest that both facial attractiveness and vocal attractiveness are important social signals and closely correlated with human mate selection1,34,35,36. Interestingly, an intriguing gender difference exists in the effects of attractiveness on decision making. For example, Van Vugt and Iredale37 found that in the public good game, men contributed more in the presence of opposite-sex audience, and their public contributions increased with the level of facial attractiveness of female audiences, whereas women did not. Farrelly et al.38 also indicated that although men and women showed preferences for attractive-face individuals of the opposite sex, in the dictator game, men cooperated more with the attractive opposite sex, while women did not show such a strong gender preference. The underlying reasons for this gender difference may be that men tend to be “purer” in their considerations of potential mates and prioritize attractiveness because factors such as female fertility, age, and other physiological characteristics are closely linked to female facial attractiveness39. However, women consider a wide range of factors, including social status, resources and earning potential1,34,40, beyond just facial attractiveness when selecting a mate. In terms of vocal attractiveness, Zhang et al.10 used event-related potential (ERP) techniques and discovered that male vocal attractiveness had an advantage in early processing stage (only male attractive voices elicited larger N1 compared to unattractive voices), whereas this advantage was not apparent in later processing stage (both male and female attractive voices elicited larger late positive component (LPC) than unattractive voices). Besides, previous research41 pointed out that compared with male participants, female participants showed less tolerant to the opposite sex than the same sex in the third-party punishment, when participants were faced with less attractive proposers, also reflecting gender difference in the decision making. To eliminate gender difference, the present study aimed to investigate the influence of males’ attractiveness (including faces and voices) on females’ decision-making, taking social interest into account. Since Yuan et al.’s study28 which used female participants and male vocal stimuli demonstrated higher acceptance for positive social interest and attractive voices about unfair offers in the ultimatum game, we used the same voice materials.

To the best of our knowledge, no research has explored the effects of attractiveness and social interest on decision-making behaviors from the perspective of audiovisual integration. The present research manipulated positive and negative social interest by employing the sentences “I like you” (我喜欢你) and “I don’t like you” (我不喜欢你), uttered by attractive and unattractive voices. Additionally, dictator game (DG) and ultimatum game (UG) were used to change social situations and participants’ roles. Female participants played against male partners with different combination of facial attractiveness, vocal attractiveness and social interest. We hypothesized that beauty premium would be induced by attractive faces and voices, as revealed by previous research4,5,14,15,42. Another hypothesis was that partners expressing positive social interest would have advantage in decision-making28,30. Moreover, we anticipated an interaction between facial attractiveness, vocal attractiveness and social interest, however, this was exploratory, as no specific hypothesis were formulated based on prior research.

Methods

Participants

A prior power analysis conducted via More*Power 6.0.443 showed that 52 participants were required to observe a significant (α = 0.05) interaction at 0.80 power. The effect size ηp2 of this study was set at 0.14. Referring to previous related studies (ranging from 34 to 60 participants)5,30,33, the final sample size was 70. Seventy female students (Mage = 21.41 years, SD = 2.19) from Southeast University who are not majoring in psychology or economics were recruited. All participants had normal or corrected-to-normal visual acuity and normal hearing. None reported visual impairments (color blindness or color weakness) or mental illness. Subjects were remunerated for participation. All methods of this study were performed in accordance with the Declaration of Helsinki.

Design and materials

The present study employed a 2 (facial attractiveness: attractive, unattractive) × 2 (vocal attractiveness: attractive, unattractive) × 2 (social interest: positive, negative) within-subject design, with facial attractiveness, vocal attractiveness and social interest as within-participant variables. The dependent variable was participants’ offer /minimum acceptable offer (MAO).

278 neutral male faces were selected from SCUT-FBP5500 Database44, the CAS-PEAL-R1 Face Database45, CUHK Face Sketch Database46 and face database of our lab at Liaoning Normal University. Participants were unfamiliar with these photographs (no movies stars, musicians or other celebrities). All photographs were edited to a uniform size (4.3 × 6.1 cm, 350 × 450 pixels) and processed into grayscale by Adobe Photoshop 2022. The horizontal and vertical visual angles were 4.2° and 5.5°, respectively. Extra features (e.g., neck, clothing, etc.) were removed to minimize the additional interference. For stimulus validation and selection, 52 college students (26 females, Mage = 21.25 years, SD = 2.13) who did not participate in formal experiment were recruited in the face evaluation experiment which included three tasks: (a) the participants needed to evaluate the facial attractiveness of the faces on a 9-point Likert scale ranging from 1 (very unattractive) to 9 (very attractive). (b) they were asked to evaluate facial trustworthiness on a 9-point Likert scale ranging from 1 (very untrustworthy) to 9 (very trustworthy). (c) they needed to evaluate the pleasure, arousal and dominance of the faces on the 9-point SAM scale47,48,49 (see Fig. 1). The participants were asked to respond based on their gut feeling about these faces. The order of these evaluation tasks was counterbalanced among participants. According to the mean ratings, 32 attractive faces and 32 unattractive faces were selected for the formal experiment. The ratings of the two categories of faces were compared by independent samples T-test, wherein attractive faces were rated as more attractive compared to unattractive faces (see Table 1 for the descriptive statistics), t(30) = 21.56, p < 0.001, 95% CI [2.71, 3.28]. Ratings of trustworthiness (t(30) = 0.29, p = 0.770, 95% CI [− 0.40, 0.54]), pleasure (t(30) = − 0.21, p = 0.831, 95% CI [− 0.46, 0.37]), arousal (t(30) = − 0.24, p = 0.812, 95% CI [− 0.38, 0.30]), and dominance (t(30) = 1.14, p = 0.260, 95% CI [− 0.07, 0.24]) were not significantly different between the two groups of faces.

Figure 1
figure 1

Pleasure, arousal, and dominance ratings (SAM scale, with cartoon figures’ characteristics representing varying degrees of pleasure, arousal, and dominance.)

Full size image
Table 1 Means and standard deviations of ratings of facial attractiveness, trustworthiness, pleasure, arousal and dominance (M ± SD).
Full size table

The voice materials were collected from Yuan et al.'s research28, in accordance with the information standards of the European Voice Database in Geneva50, which consisted of two sentences: “我喜欢你” (Chinese Pinyin: “wo xi huan ni” which means “I like you”) and "我不喜欢你" (Chinese Pinyin: “wo bu xi huan ni” which means “I don’t like you”) respectively. These sentences were read aloud by 60 male native speakers of Mandarin Chinese in a neutral mood and were recorded in a quiet laboratory, by a high-quality microphone (PHILIPS, SHM1000) and Adobe Audition CS6 software. After eliminating materials with poor articulations and noises, 107 male voices were retained. The duration of the voices ranged from 800 to 1200 ms across speakers. All voices were saved as uncompressed .wav-files with a uniform intensity of 70 dB and a sampling rate of 44.1 kHz.

For stimulus validation and selection, 54 participants (29 females, Mage = 21.44 years, SD = 2.52) who did not participate in follow-up formal experiment were recruited in the voice evaluation experiment including three tasks: (a) The participants needed to evaluate the attractiveness of the voices on a 9-point Likert scale ranging from 1 (very unattractive) to 9 (very attractive). (b) They were asked to evaluate trustworthiness of the voices on a 9-point Likert scale ranging from 1 (very untrustworthy) to 9 (very trustworthy). (c) They were asked to evaluate pleasure and arousal of the voices using the 9-point SAM scale (see Fig. 1). The participants were asked to respond based on their gut feeling about these voices. The order of these evaluation tasks was counterbalanced among participants. According to the mean ratings, we used 32 attractive voices and 32 unattractive voices in the experimental trials. Among each group, 16 had positive social interest and 16 had negative social interest (see Table 2 for the descriptive statistics).

Table 2 Means and standard deviations of ratings of vocal attractiveness, trustworthiness, pleasure and arousal (M ± SD).
Full size table

A two-way ANOVA was conducted on the ratings of the voices. The independent variables included vocal attractiveness (with two levels: attractive and unattractive) and social interest (with two levels: positive and negative). Regarding the ratings of attractiveness, the main effect of vocal attractiveness was significant, F(1, 28) = 804.20, p < 0.001, ηp2 = 0.97. Neither the main effect of social interest (F(1, 28) = 0.32, p = 0.579, ηp2 = 0.01) nor the interaction between the attractiveness and social interest was significant (F(1, 28) = 0.34, p = 0.564, ηp2 = 0.01). The main effects of attractiveness and social interest, and interaction effects regarding the ratings of trustworthiness, pleasure and arousal were not significant, Fs ≤ 4.05, ps ≥ 0.054.

In addition, using Praat software and publicly available scripts51,52, we measured the acoustic parameters of the voices (see Table 3) since these parameters are the key influential factors of vocal attractiveness9,12,53,54,55. 2 (vocal attractiveness: attractive; unattractive) × 2 (social interest: positive; negative) ANOVA was performed on the acoustic parameters. Consistent with previous studies6,10, the main effects of vocal attractiveness on F0, f1 and Df were significant, F(1, 28) = 17.67, p < 0.001, ηp2 = 0.39; F(1, 28) = 9.35, p = 0.005, ηp2 = 0.25; F(1, 28) = 11.588, p = 0.002, ηp2 = 0.29. These findings demonstrated that, vocal attractiveness is strongly influenced by acoustic features. The interaction effects between social interest and vocal attractiveness, as well as the main effects of social interest, were not significant on other parameters, except for a significant interaction between vocal attractiveness and social interest on F3, F(1, 28) = 4.50, p = 0.043, ηp2 = 0.14. Further simple effect analysis revealed that none of the four simple effects between social interest and vocal attractiveness were significant, Fs ≤ 3.91, ps ≥ 0.058.

Table 3 Acoustic parameters of voices (M ± SD).
Full size table

Procedure

Stimulus presentation and data acquisition were controlled by E-Prime software version 3.0 on a Lenovo ThinkPad T14 laptop (14-inch monitor, 2440 × 1400 pixels resolution, 60 Hz refresh rate). Sennheiser HD201 headphones were used to present audio stimuli. Participants were tested individually in a sound-attenuated room. Listening volume was adjusted to a comfortable level for each participant. They first completed the decision-making task and then followed by the rating task. The viewing distance in both tasks was 55 cm. To maintain the single-blind nature of this study, participants were informed a simple cover story that the study aimed to investigate differences in monetary decision-making among individuals with diverse personalities. Besides, partners in the games were described as students from local colleges who had already made their decisions in a prior experiment, but in fact, the regularities were designed by the researchers. Considering that if the study was conducted as a genuinely incentivized experiment where participants were paid based on their decisions, it is both feasible and likely to yield more authentic behavioral responses. Drawing upon the incentive structure of previous research28,56, all participants were told that their payment for participation was ¥30 plus the payoff from two randomly selected trials in each game, but actually, they all got paid ¥35 at the end of the experiment. The deception of compensation received approval from the Ethics Committee of the Psychology Research Center at Southeast University. None of the participants reported the real aim and regularity of the experiment. At the end of the experiment, researchers provided a debriefing (including experimental remuneration) to all participants.

Our experimental tasks were adopted from Lucas and Koff’s research4. Participants were asked to make decisions in two classic experimental economics games: dictator game (DG) and ultimatum game (UG). Both games included two roles: the proposer and the recipient. In the DG, the participant acted as the proposer and determined how to split ¥10 between herself and the recipient. The recipient had no choice but to accept the allocation distributed by the proposer. The UG consisted of two tasks. In the first task, participants were asked as the proposer to split a stake of ¥10 between herself and the recipient. The participants were informed that the recipient could choose to accept or reject the proposer's allocation. If the allocation was accepted, they would receive the money as proposed, but if the allocation was rejected, both players received nothing. In the second task, the participant acted as the recipient to indicate the lowest offer they would accept from the proposer.

In order to familiarize participants with our tasks, detailed instructions and 8 practice trials were provided before the formal experiment in each task. The faces and voices used in the practice trials would not appear in the formal experimental trials, and the income in the practice trials would not be counted in participants’ final payment. In the DG (Fig. 2), at the beginning of each trial, a fixation was presented in the center of the screen for 1000 ms. Afterwards, a face and a voice of the partner were presented simultaneously for 2000 ms. After the stimuli disappeared, a no-time-limit frame which contained a line of instruction (“How much do you want to allocate to him?”) and an 11-point scale was shown. The scale had a mark of “¥0” at one end, “¥10” at the other end and “¥5” in the middle. Participants acted as proposers by clicking on the allocation amount on the scale using the mouse as quickly and accurately as possible. Finally, a blank screen appeared for 1000 ms, and then the next trial started. In the UG1 during which participants acted as proposers, trials followed the same structure as the DG task described above. In the UG2 during which participants were recipients. Trials followed the same structure as the DG task with the following changes. With the offset of face-voice pairs, participants were presented with a line of instruction (“What’s the minimum amount you would be willing to accept from him?”). The order of UG and DG was counterbalanced between participants. The order of UG1 and UG2 was also counterbalanced. Each task consisted of 2 blocks. Each face-voice pair was presented once in one block, and was presented once again in another block. There were totally 192 trials (32 face-voice pairs × 2 repetitions × 3 games) which appeared in a random sequence.

Figure 2
figure 2

Schematic representation of dictator game (DG) and ultimatum game (UG1 & UG2). At the beginning of each DG trial, a fixation appeared for 1000 ms, followed by the presentation of a face-voice pair simultaneously for 2000 ms. Next, the participant acted as the proposer and decided how much in ¥10 she wanted to allocate to the recipient. Once the decision had been made, a 1000-ms blank screen would be presented. Then, the next trial started. The experimental procedure of UG1 was the same as that of DG. In the UG2, trials followed the same structure as the DG task except the following changes. The participant as the recipient decided the minimum amount she would be willing to accept from her partner on the 11-point scale.

Full size image

After the decision-making task, participants were asked to rate the attractiveness of all faces, the attractiveness of all voices, and overall attractiveness of all face-voice pairs on a 9-point Likert scale ranging from 1 (very unattractive) to 9 (very attractive), respectively. The order of the three posttests was counterbalanced between participants.

Statistical analysis

Data were analyzed using IBM SPSS Statistics 22. The dependent variable was participants’ offer/minimum acceptable offer (MAO) (in yuan) in each game. 2 (facial attractiveness: attractive; unattractive) × 2 (vocal attractiveness: attractive; unattractive) × 2 (social interest: positive; negative) repeated measures analysis of variance (ANOVA) was performed on each dependent variable. Greenhouse–Geisser correction was used when the assumption of sphericity was not met.

Ethics declarations

The research involving human participants was reviewed and approved by the Ethics Committee of the Psychology Research Center at Southeast University, China. The participants provided their written informed consent to participate in this study.

Results

The agreements of the outcomes across repetition between two blocks for all face and voice conditions in three games were high (all Cronbach’s α ≥ 0.87). The analyses of response times and exploratory multiple regression are in the Supplementary Information file (Supplementary information).

Dictator game (participants as proposers)

The main effect of facial attractiveness was significant on the DG offer, F(1, 69) = 81.90, p < 0.001, ηp2 = 0.54, such that participants gave higher offers to attractive recipients (M = 3.35 yuan, SD = 0.17) comparing to unattractive recipients (M = 2.68 yuan, SD = 0.17). The main effect of vocal attractiveness was also significant, F(1, 69) = 14.02, p < 0.001, ηp2 = 0.17. Participants gave higher offers to recipients with attractive voices (M = 3.10 yuan, SD = 0.17) comparing to recipients with unattractive voices (M = 2.93 yuan, SD = 0.16). Moreover, the main effect of social interest was significant, F(1, 69) = 61.00, p < 0.001, ηp2 = 0.47. Participants were more generous to the recipients expressing positive social interest (M = 3.62 yuan, SD = 0.18) than the recipients expressing negative social interest (M = 2.42 yuan, SD = 0.18). In addition, the interaction between vocal attractiveness and social interest was also significant, F(1, 69) = 13.20, p = 0.001, ηp2 = 0.16.

Importantly, the 3-way interaction between facial attractiveness, vocal attractiveness and social interest was significant (as illustrated in Fig. 3), F(1, 69) = 6.50, p = 0.013, ηp2 = 0.09. Further simple simple effect analysis found that participants allocated more money to their partners with attractive voices than to their partners with unattractive voices under the condition of positive social interest, regardless of the level of facial attractiveness, Fs ≥ 5.86, ps ≤ 0.018. However, under the conditions of negative social interest, difference in offers between attractive voices and unattractive voices was not significant, regardless of the level of facial attractiveness, Fs ≤ 2.63, ps ≥ 0.110. Other simple effects were similar with the main effects described above, Fs ≥ 39.64, ps < 0.001 (Table 4).

Figure 3
figure 3

(A) The mean DG offer as a function of facial attractiveness, vocal attractiveness and social interest. (B) The mean UG1 offer as a function of vocal attractiveness and social interest. The error bars represent standard errors. ***p < 0.001.

Full size image
Table 4 Means and standard deviations for three games’ outcomes (M ± SD).
Full size table

Ultimatum Game 1 (UG1, participants as proposers)

The main effect of facial attractiveness was significant, such that participants allocated more to attractive recipients (M = 4.52 yuan, SD = 0.10) compared to unattractive recipients (M = 3.94 yuan, SD = 0.11), F(1, 69) = 61.80, p < 0.001, ηp2 = 0.47. The main effect of vocal attractiveness was significant, such that participants allocated more to the recipients with attractive voices (M = 4.30 yuan, SD = 0.10) comparing to recipients with unattractive voices (M = 4.16 yuan, SD = 0.11), F(1, 69) = 11.57, p = 0.001, ηp2 = 0.14. The main effect of social interest was also significant, such that participants allocated more to the recipients who expressed positive social interest (M = 4.68 yuan, SD = 0.11) compared to recipients who expressed negative social interest (M = 3.76 yuan, SD = 0.13), F(1, 69) = 42.90, p < 0.001, ηp2 = 0.38. The interaction between vocal attractiveness and social interest was significant (as illustrated in Fig. 3), F(1, 69) = 17.45, p < 0.001, ηp2 = 0.20. Simple effect analysis indicated that under the condition of positive social interest, the recipients with attractive voices (M = 4.80 yuan, SD = 0.11) received more money than the recipients with unattractive voices (M = 4.57 yuan, SD = 0.12), F(1, 69) = 22.14, p < 0.001, ηp2 = 0.24. However, under the condition of negative social interest, difference between attractive voices and unattractive voices was not significant, F(1, 69) = 1.39, p = 0.243, ηp2 = 0.02. Compared to the recipients who expressed negative social interest, participants were more generous to the recipient if he expressed positive social interest no matter whether the recipient had an attractive voice, Fs ≥ 35.25, ps < 0.001. Other interaction effects were not significant, Fs ≤ 1.40, ps ≥ 0.241.

Ultimatum Game 2 (UG2, participants as recipients)

In the UG2 offer, the main effect was significant only for social interest, F(1, 69) = 8.33, p = 0.005, ηp2 = 0.11, participants expected higher MAOs when faced with the proposers expressing positive social interest (M = 3.59 yuan, SD = 0.14) compared to those expressing negative social interest (M = 3.14 yuan, SD = 0.16). Other main effects or interaction effects were not significant, Fs ≤ 3.53, ps ≥ 0.065.

Post-experiment ratings of facial and vocal attractiveness

A independent samples T-test on the mean facial attractiveness ratings after the formal experiment was conducted, wherein attractive faces (M = 5.24, SD = 0.52) were rated as more attractive compared to unattractive faces (M = 2.96, SD = 0.39), t(30) = 14.14, p < 0.001, 95% CI [1.95, 2.62]. A 2 (vocal attractiveness: attractive; unattractive) × 2 (social interest: positive; negative) ANOVA on the mean vocal attractiveness ratings found a significant main effect of vocal attractiveness, F(1, 28) = 225.85, p < 0.001, ηp2 = 0.89, where attractive voices (M = 5.95, SD = 0.53) were rated as more attractive than unattractive voices (M = 3.80, SD = 0.25). No significant results were found for the main effect of social interest (F(1, 28) = 2.49, p = 0.126, ηp2 = 0.08), or the interaction between social interest and vocal attractiveness(F(1, 28) = 0.45, p = 0.510, ηp2 = 0.02). These confirms that the pre-experiment selection of stimuli was effective.

Discussion

The main purpose of this research was to examine the effects of facial attractiveness, vocal attractiveness and social interest on cross-sex decision making in the dictator game where participants acted as proposers and the ultimatum game where participants acted as proposers and recipients. The results confirmed findings of prior studies on the effects of facial or vocal attractiveness in bargaining games5,14,15,42, and extend this work by showing the influence of social interest in beauty premium. The effects of attractiveness and social interest also vary across different gaming roles.

As expected, participants as proposers allocated more money to partners with attractive faces in both DG and UG, which was in line with the well-documented beauty premium of facial attractiveness in decision-making3,5,42,57. According to evolutionary theories1,40, attractive faces symbolize high mate value, thereby prompting individuals to exhibit more generous and cooperative in social transactions. Electrophysiological research further revealed neural evidence for beauty premium. Attractive faces activated reward-related neural circuits, possessing subjective value of reward39, which makes people willing to sacrifice their own interests for the benefit of attractive people. In addition, the effect of facial attractiveness varied depending on participants’ role. When acting as proposers, participants allocated more money to attractive recipients. However, as recipients, they did not show any preference for attractive faces. This contradicts with prior research5,42 in which participants acting as recipients were more likely to accept unfair offers from attractive proposers compared to unattractive proposers. The reason might be that different roles imply different social responsibilities32,33. Unlike recipients, proposers need to consider more information in order to come up with allocation schemes for both parties. They have to weigh the gains of both parties, inevitably being influenced by social preferences such as attractiveness and social interest. However, when participants acted as recipients, they only needed to specify their MAO, meaning they may only concentrate on their own desired gains. So their considerations were not as comprehensive as those of the proposers. Besides, it may be related to the design of our experiment. In previous studies33,42,57, the offers made by proposers were predetermined by researchers, and participants only needed to make a simple judgment of acceptance or rejection. Whereas our study adopted Lucas and Koff’s experimental design4 and allowed participants to freely decide their MAO, which may have led to more complex and natural behavioral outcomes. This setting simulated more authentic decision-making contexts and obtained more realistic behaviors. Also, Lucas and Koff4 investigated the effect of facial attractiveness on females’ decisions across the menstrual cycle in DG and UG. They reported no correlation between facial attractiveness with MAO.

In terms of social interest cues, in DG and UG1, recipients who expressed positive social interest benefited more than those who expressed negative social interest, which aligns with Yuan et al.’s study28. Social interest cues are important information related to individual evolution, so people tend to allocate more cognitive resources to these threatening or beneficial information29. The "goodwill" information released by positive social interest cues can induce participants' approach motivation58, leading them to allocate more money. In other words, participants perceived that partners expressing positive social interest were more likely to respond and give positive feedback to their social effort behaviors. However, when participants took on the role of recipients in UG2, they expected partners expressing positive social interest to give them more money. These findings suggested that people had both a preference and a higher expectation for people expressing positive social interest. One possible reason may be that positive social interest implies positive emotional connotations, leading participants to expect these "good partners" to be more generous and to give them more money.

Moreover, unlike Yuan et al.28 who reported independent effect of vocal attractiveness and social interest on decision-making in UG, the preference for attractive voices in our study was restricted by social interest. Only in the condition of positive social interest, proposers tended to allocate more money to recipients with attractive voices than recipients with unattractive voices. In the condition of negative social interest, the beauty premium of voices disappeared. The possible explanation is our experimental design was different with Yuan et al.’s28. In Yuan et al.’ study28, participants were recipients who decided whether to accept the allocation of ¥10 between the proposer and the participants. The allocation ranged from 9:1 to 5:5. The highest amount participants could get was ¥5, while the lowest was ¥1. However, in the present study, the options were wider (from 0:10 to 10:0). Thus the highest amount participants could get was ¥10, while the lowest was 0. Furthermore, participants acted as both proposers and recipients, different roles may alter the interaction of vocal attractiveness and social interest. It should be noted that compared to vocal attractiveness, facial attractiveness is less influenced by social interest. This result may be primarily due to the fact that we controlled social interest through vocal stimuli, inevitably leading to a close association with vocal attractiveness.

Conclusions

In conclusion, the present research confirmed beauty premium by showing that attractiveness effects from audiovisual channels are associated with decision outcomes. Female proposers offered more money to attractive-face males than unattractive-face males. Female proposers also allocated more money to males expressing positive social interest than males expressing negative social interest, whereas female recipients expected males expressing positive social interest to offer them more money than males expressing negative social interest. Under the condition of positive social interest, female proposers allocated more money to attractive-voice males than unattractive-voice males, while this effect was absent under the condition of negative social interest. Beauty premium and social interest effects also varied across gaming situations and roles.

Limitations and future directions

There are several limitations of our study: firstly, different social interests were expressed through semantic meanings of voice stimuli without real social interaction, which restricted the generalizability of the findings. Therefore, future studies should enhance the ecological validity of studies by incorporate face-to-face interactions. Secondly, the complexity of the 2 × 2 × 2 factorial design may hinder the extraction of meaningful insights from the study, as it significantly complicates data interpretation. Future research should narrow the scope of factors to deliver a clearer message regarding impact of multi-channel attractiveness on decision-making. Furthermore, the present study deceived participants about compensation as we did not have enough funding. Future research should conduct the study as a genuinely incentivized experiment, where participants are paid based on their decisions. That is both feasible and likely to yield more authentic behavioral responses.