Learning from in-group and out-group models induces separative effects on human mate copying

Abstract

Mate copying is a social learning process in which individuals gather public information about potential mates by observing models’ choices. Previous studies have reported that individual attributes of female models affect mate copying, yet little is known about whether and how the group attributes of models influence mate copying. In the current behavioral and functional magnetic resonance imaging studies, female participants were asked to rate their willingness to choose the depicted males as potential romantic partners before and after observing in-group or out-group female models accepting, rejecting or being undecided (baseline) about the males. Results showed that participants changed their ratings to align with the models’ acceptance or rejection choices. Compared to rejection copying, the effect of acceptance copying was stronger and regulated by in- and out-group models, manifesting a discounting copying effect when learning from out-group models. At the neural level, for acceptance copying, stronger temporoparietal junction (TPJ) activity and connectivity between TPJ and anterior medial prefrontal cortex (amPFC) were observed when female models belonged to out-group members; meanwhile, the functional connection of TPJ and amPFC positively predicted the rating changes when learning from out-group models. The results indicated that participants might need more resources to infer out-group members’ intentions to overcome the in-group bias during acceptance copying.

Introduction

Learning from others’ choices and outcomes is an efficient way to acquire knowledge and skills (Kang et al., 2021). In the field of mate choice, although sexual selection based on genetic preferences is widespread when choosing a mate (Fisher, 1930; Majerus, 1986; Andersson and Iwasa, 1996; Mead and Arnold, 2004; Andersson and Simmons, 2006; Varela et al., 2018), evidence from both non-human animals (Dugatkin and Godin, 1992; Freed-Brown and White, 2009; Kniel et al., 2015; Dagaeff et al., 2016; Danchin et al., 2018; Gierszewski et al., 2018) and humans (Jones et al., 2007; Hill and Buss, 2008; Little et al., 2008; Place et al., 2010; Zhuang et al., 2016, 2017, 2021) shows that observing other agents’ (the same-sex models) mate choices has a substantial impact on an agent’s own choice. For example, after observing other women showing interest in (acceptance) depicted men, female participants increased their willingness to choose the men as romantic partners and vice versa (Zhuang et al., 2021). In another study, female participants rated male individuals who were pictured with other females as more desirable than males who were pictured alone or with other males (Hill and Buss, 2008). Such a non-independent process in which individuals gather public information about potential mates by observing models’ choices has been broadly known as ‘mate copying’.

Mate copying has been considered adaptive with saving individuals’ time and energy spent on mate choice and reducing the risk of incorrect decisions (Wade and Pruett-Jones, 1990; Danchin et al., 2004; Wagner and Danchin, 2010). Individuals obtain public information by observing models’ attitudes towards potential mates; in addition, the attributes of same-sex models can affect individuals’ learning from models’ choices. Previous studies have suggested that there is a ‘selection bias’ in mate copying, which is related to the individual attributes of models, such as attractiveness (Sigall and Landy, 1973; Waynforth, 2007; Yorzinski et al., 2010; Little et al., 2011), character (Chu, 2012) and age (Dugatkin and Godin, 1993). For instance, compared to those paired with unattractive women, men paired with attractive women’s faces were later rated as more attractive than they had been previously (Waynforth, 2007), which might be because the preferences of attractive women should carry more weight for other females (Chu, 2012). Indeed, for females, the depicted male has been chosen by an attractive female, which can at least convey this message: the depicted male has some hidden positive qualities, which are unobserved directly from physical appearances, such as high social status, substantial financial resources or desire to invest in children (Rodeheffer et al., 2016). These results suggest that individuals are more willing to ‘copy’ the models who can provide more beneficial information to guide their mate choice. However, existing studies have mainly focused on the models’ individual attributes (e.g. attractiveness); it is still an open question whether the group attributes of models, such as belonging to a in-group or out-group, affect mate copying.

Mate copying has been suggested to be a strategic use of public information via social learning (Richerson and Boyd, 2005; Little et al., 2008; Kavaliers et al., 2016; Santos et al., 2017; Gierszewski et al., 2018; Varela et al., 2018). Intergroup context is one of the non-negligible regulatory factors in social learning (Golkar et al., 2015; Howard et al., 2015). Research has shown that people are more likely to change their beliefs, behaviors or attitudes to conform to in-group members (Mackie et al., 1992; Mackie and Wright, 2003; Cialdini and Goldstein, 2004; Lin et al., 2018; Kang et al., 2021). For example, Lin et al. (2018) conducted a study in which participants rated a series of images in different emotional contexts. The same images were then presented to participants with ratings from either in-group or out-group members, and the participants were asked to re-rate the images. The results revealed that participants were more likely to change their ratings to align with in-group members than with out-group members. Keeping in line with in-group members is essential for the individual, such as contributing to maintaining positive self-esteem (Tajfel and Turner, 1986), obtaining social approval from in-group members (Alvaro and Crano, 1997; Cialdini and Goldstein, 2004) and improving access to or protect resources conducive to survival or reproduction (Correll and Park, 2005). While existing research has investigated the impact of in-group and out-group members on behaviors, attitudes and emotional experiences, whether and how these members influence social learning in the field of mate choice remains unclear.

The present study aimed to explore the behavioral and neural mechanisms underlying human mate copying in an intergroup context. Previous research has shown two types of strategies in mate copying: copying the acceptance of a mate and copying the rejection of a mate. These two strategies have been suggested to might be involved in different mechanisms. Evidence from behavioral results observed that participants exhibit copying effects differently under acceptance and rejection conditions (Place et al., 2010; Chu, 2012; Deng and Zheng, 2015; Zhuang et al., 2016, 2021). For instance, Zhuang et al. (2021) found that female participants showed a stronger copying effect for acceptance than for rejection, while similarly, Nöbel et al. (2022) reported that female fruit flies copied the acceptance of a mate but not the rejection. Further, by exploring the neural mechanism of mate copying, research has reported that compared to the rejection copying, participants activated more brain regions in acceptance copying. These regions include the prefrontal cortical regions in the dorsal medial prefrontal cortex (dmPFC), anterior medial prefrontal cortex (amPFC) and dorsal lateral prefrontal cortex, temporoparietal junction (TPJ), precuneus, inferior frontal gyrus, inferior parietal lobule, anterior cingulate cortex, striatum, the supramarginal gyrus, insula and some visual areas (Zhuang et al., 2016, 2021). In the present study, we aimed to investigate whether and how in- and out-group models affect participants’ choices in acceptance and rejection copying, both at behavioral and neural levels. According to the in-group bias and previous empirical evidence, we hypothesized that participants would exhibit a greater propensity to alter their choice of a potential mate when exposed to female in-group models as opposed to out-group models, particularly in the context of acceptance copying.

As mentioned above, individuals guide their mate choice based on the attributes of models and attitudes toward potential mates, which involves the processing of public information, such as how and when the public information will be most useful (Richerson and Boyd, 2005; Jones et al., 2007; Little et al., 2008; Bowers et al., 2012; Kavaliers et al., 2016), and inferring the reasons for the models’ choices, such as whether the chosen men might have a higher partner value (Chu, 2012). This procedure seems to involve a processing of mentalizing (Frith et al., 2003), also known as the theory of mind (ToM). A large number of neuroimaging studies have consistently reported the TPJ and medial prefrontal cortex (mPFC, especially dmPFC and amPFC) as the core areas in social mentalizing (Amodio and Frith, 2006; Frith and Frith, 2006; Van Overwalle, 2009; Van Overwalle and Baetens, 2009; Mars et al., 2012; Amft et al., 2015; Molenberghs et al., 2016; Schurz et al., 2021). In fact, it has been found that these brain regions of TPJ and mPFC are involved in the processing of mate copying (Zhuang et al., 2016, 2021).

Moreover, within the framework of intergroup social influence, it has been observed that brain regions involved in mentalizing, such as the TPJ and mPFC, are recruited in different ways. Prior studies found more spontaneous mentalizing involved in following in-groups when the tasks referred to a direct perception of characteristics or emotions (Adams et al., 2010; Lin et al., 2018). On the contrary, a study on prosocial decision-making revealed that Chinese participants demonstrated increased activation in brain regions associated with mentalizing when contributing to out-group members (Telzer et al., 2015). Compared to direct perception of in- or out-groups, making prosocial decisions about intergroup members seems more difficult. Particularly, deciding to help out-group members may need to carefully contemplate the costs and benefits (Stürmer et al., 2006) and overcome the social distance from out-groups, which requires more mentalizing effort and resources when facing out-group members; whereas prosocial decision-making about in-group members may be a more automatic process (Telzer et al., 2015). Similarly, in the process of mate copying, copying models’ choices is beneficial to avoid the risk of incorrect mate decisions and wasting of time and energy (Danchin et al., 2004; Wagner and Danchin, 2010). Due to the in-group bias, identifying with and following in-group members’ choices might be easier and more automatic for participants. However, when referring to the out-group models’ choices to adjust their mate choices, participants might have to overcome the unfamiliar, infrequent or novel (Merritt et al., 2021) of out-group models and take their perspective. Therefore, the process of copying out-group models might be more difficult and effortful to infer the intentions of out-group members. Given that, we hypothesized that compared to out-group members, participants copied in-group models’ choices more automatically, while more resources were needed when participants copied out-group models than in-groups, which might involve more brain regions related to mentalizing, such as TPJ and mPFC, especially to make an acceptance decision. Moreover, based on previous studies suggesting that TPJ may be functionally tightly associated with mPFC (Qin et al., 2020), we also explored the functional connectivity between these two brain areas.

In order to test these questions, the present study conducted a behavioral experiment and a task-related functional magnetic resonance imaging (fMRI) experiment respectively. The task utilized in the study was similar to that used in previous research by Zhuang et al., 2016; 2021. The behavioral results in the two experiments aimed to observe participants’ tendency to ‘copy’ the in-group and out-group models under different conditions, while the fMRI results focused on the neural mechanisms underlying this process. Specifically, we examined the roles of brain regions related to mentalizing, such as the TPJ and mPFC, and explored their functional connectivity in the process of mate copying.

Study1

Methods

Participants

We recruited 27 Chinese female students (mean age 20.48 ± SD 1.91) from the university community. A sensitivity analysis was conducted to compute required effect size based on our sample of 27 participants. Considering that the repeated-measures analyses of variance (rmANOVAs) was used to detect the effect of intergroup members and different conditions, partial η² was calculated to indicate the effect size. At the level of α = 0.05, the sample of 27 participants provided 95% power to detect effects as large as partial η² = 0.25.

All participants were healthy with normal or corrected-to-normal vision, had no history of psychological or neurological, were right-handed, reported to be heterosexual and had no significant negative events during the week before the experiment. The study protocols were approved by the Ethics Committee of East China Normal University. All participants signed informed consent before the experiment and were debriefed and paid for after the experiment.

Stimuli

Stimuli were colored photographs of men and women sourced with permission from the local university and websites allowing free to use of images. The photographic style employed conforms to a standardized set of guidelines, which stipulate the use of high-quality images featuring youthful visages with direct eye contact and a natural facial expression. Moreover, the photographs are devoid of any decorative items such as headgear, hair ornaments or jewelry. The individuals featured in the photographs are not celebrities or well-known personalities. Given that the study participants were exclusively of Chinese origin, we selected Chinese individuals to serve as the in-group models, while employing Caucasian individuals to serve as the out-group models. All photographs were cropped at the neck and adjusted to a size of 200 × 266 pixels using Adobe Photoshop (Version CS5). The brightness was adjusted across the images, and the background was kept a constant white (RGB: red = 255, green = 255, blue = 255). To obtain the final target images and in-group model images, 15 females (mean age 24.27 ± SD 1.44) rated the attractiveness of 250 in-group-male and 250 in-group-female faces using a Likert scale (1 = very unattractive, 9 = very attractive). Average attractiveness scores for the male and female faces were 4.15 ± SD 1.26 and 4.41 ± SD 1.32, respectively. These ratings were used to select 120 photographs of males with relatively average to medium (range: 4.15 + 1.26 ∼ 4.15–1.26) attractiveness ratings (to exclude highly and lowly attractive males for whom choice copying might be less relevant) and 60 photographs of females with relatively highest attractiveness ratings (mean scores 6.02 ± SD 0.37; to maximize the chances that the females depicted were perceived as models worth copying) for the main experiment. We obtained 60 photos of the out-group female models using the same method. Another 15 female participants (mean age 24 ± SD1.37) rated the attractiveness of 180 out-group-female faces. The average attractiveness score was 4.55 ± 1.32, and the highest 60 faces were selected for the out-group model (mean scores 6.07 ± SD 0.50).

The final photographs were then edited in Adobe Photoshop to create 120 couple images, which showed both a male and a female face together (800 × 600 pixels). The single photographs to create these couple images were chosen randomly and arranged side by side against a white background. Half of the 120 images were male faces and in-group female faces, and the other half were male faces and out-group female faces. In addition, to emphasize the in- and out-group attributes, we placed national flags above the faces of female models, in which the Chinese flag represented the in-group and the American flag represented the out-group (Lin et al., 2018). The side on which the female face was displayed was counterbalanced across stimuli. The couple images were divided randomly into three conditions according to the implied decision of the female character displayed in the image: ‘Accept’, ‘Reject’, and ‘Undecided’. ‘Undecided’ served as a baseline condition in which no decision was communicated. The word cues were added at the bottom of the images communicating the decision. The word cues were in simplified Chinese and read ‘女对男感兴趣’(The woman is interested in the man) for the Accept condition, ‘女对男不感兴趣’(The woman is not interested in the man) for the Reject condition and ‘女对男未表达意愿’(The woman has not yet decided whether she is interested in the man) in the Undecided condition. Participants were informed that the females’ attitudes towards the depicted male in the couple image (i.e. interested, not interested or undecided) were the female models’ preference for choosing this male as a potential romantic partner.

Procedure

Each participant completed two task phases. In the first phase, in each trial, they were shown one of the 120 male photographs with a rating scale below the image for 4000 ms. All images were randomly shown once. Participants were asked to rate how likely they were to consider each of these men as a potential romantic partner using a 1–9 Likert scale (1 = not at all, 9 = very much). A temporal jitter of 500 ∼ 2000 ms was introduced randomly between trials (Figure 1A).

Fig. 1.

Experimental design. (A) In the first phase, 120 individual male facial photographs were presented, and participants were asked to rate whether they would consider each of these men as a potential romantic partner, using a 1–9 Likert scale. (B) In the second task phase, couple images were presented. In these 120 couple images, sixty of models were Chinese female faces with Chinese flags (In-group), and the others were foreign female faces with American flags (Out-group). These images were accompanied by statements regarding whether the depicted female would consider the male as a romantic partner (conditions: accept, reject and undecided). Finally, the same male face was shown again, and participants provided a second rating of whether they would consider each of these men as a potential romantic partner, using a 1–9 Likert scale. Photograph samples were used for display. The change in ratings served as an indicator for a change of mind in their mating decision following the observation of how the depicted female considered the male.

The second phase of the experiment was of primary interest for our analyses and comprised 120 trials. Each condition contained 20 trials. In each trial, one of the 120 couple images, each containing one of the male faces from the first task phase, together with the acceptance statement of the displayed female (either Accept, Reject, Undecided) was shown for 6000 ms randomly, and the participants were asked to observe them. This was followed by another jitter delay of either 500 ∼2000 ms (randomly drawn). After that, only the male face from the respective couple image was presented again, and participants were asked to rate their willingness to choose him as a romantic partner as in the previous section, using the same Likert scale. This image was presented for 4000 ms, followed by another 500 ∼2000 ms jitter, randomly drawn as before. (Figure 1B). After finishing the task, all participants confirmed that they had never seen the depicted people before.

Data analysis

We calculated the change in rating scores from pre- to post-observation of the depicted females’ decisions. To avoid the effects of the extreme values, we eliminated the change in rating scores beyond ± 3 SD from the means within each participant. The total number of extreme trials across all participants accounted for 0.56% of the total trials. We also analyzed all the original results, which were consistent with the results after eliminating. A 2 (Intergroup members: In-group, Out-group) × 3 (Conditions: Accept, Reject and Undecided) rmANOVAs on the change of rating scores was used to analyze the data.

Results

The 2 (Intergroup members: In-group, Out-group) × 3 (Conditions: Accept, Reject and Undecided) rmANOVA for the change in attractiveness (CiA: as computed by the difference in rating scores from pre- to post-observation) showed an insignificant main effect of intergroup members, F _{(1, 26)} = 0.30, P = 0.59, ηp2 = 0.01, and a significant main effect of conditions, F _{(2, 52)} = 28.54, P < 0.001, ηp2 = 0.52. Post hoc tests with Bonferroni correction confirmed that positive CiA scores for Accept trials (M_{CiA_accept} = 0.88, SD_{CiA_accept} = 0.83), meaning that participants rated the depicted males higher after exposure to the female model, and this was significantly higher compared to undecided trials (M_{CiA_undecided} = 0.20, SD_{CiA_undecided} = 0.72), P = 0.001. The reverse was found for reject trials (M_{CiA_reject} = -0.26, SD_{CiA_reject} = 0.83), in which the negative scores indicated a change toward more negative attitudes towards the depicted males after observing the female model, and this was significantly lower compared to undecided trials, P < 0.001.

Moreover, we found a significant interaction effect between intergroup members and conditions, F _{(2, 52)} = 3.51, P = 0.037, η_p² = 0.12. Further simple effect analysis showed that in the accept condition, the CiA scores in the in-group condition (M _{CiA_IA} = 1.02, SD _{CiA_IA} = 0.90) were significantly positive than that in the out-group condition (M _{CiA_OA} = 0.75, SD _{CiA_OA} = 0.75), F_{(1, 26)} = 8.28, P = 0.008, η_p² = 0.24, meaning that after exposure to the in-group models, participants showed stronger copying effect than exposure to the out-groups. However, in the reject and undecided conditions, the CiA scores were not significantly different between the in-group models and the out-group models (P > 0.23) (see Figure 2).

Fig. 2.

Mean change in rating scores. Mean of the change in attractiveness (CiA) rating scores from pre- to post-exposure to the models’ choices for each condition. IA = in-group accept; OA = out-group accept; IR = in-group reject; OR = out-group reject; IU = in-group undecided; OU = out-group undecided.

Discussion

In the present study, by adding the factor of intergroup attributes of models, we still found that female participants adjusted their choices of the depicted male in line with the observed information of the same-sex models, which was consistent with previous results (Jones et al., 2007; Hill and Buss, 2008; Little et al., 2008; Place et al., 2010; Zhuang et al., 2016, 2017, 2021). Specifically, participants increased the rating of the depicted males when female models were interested in the male; on the contrary, participants decreased the rating when the female models were not interested in the depicted males. From the absolute value of the acceptance and rejection changing scores, we found that the learning effect in acceptance copying was stronger than that in rejection copying. More importantly, in the current study, we investigated the group attributes of the same-sex model and found that the regulation of acceptance and rejection copying by in- and out-group members was separated. For acceptance copying, compared to learning from in-groups, the copying effect was discounted when learning from out-group models. The regulation of group attributes was not significant in rejection copying. Consistent with previous studies (Place et al., 2010; Chu, 2012; Deng and Zheng, 2015; Zhuang et al., 2016, 2021), our results demonstrate the difference between acceptance and rejection copying at the behavioral level.

Study 2

Methods

Participants

Twenty-seven female students (mean age 21.04 ± SD 2.24) who had not participated in Study 1 were recruited from the university community. All participants were healthy, had no history of psychological or neurological, were right-handed, with normal or corrected-to-normal vision, were reported to be heterosexual and had no significant negative events during the week before the experiment. The study protocols were approved by the Ethics Committee of East China Normal University. All participants signed informed consent before the experiment and were debriefed and paid for after the experiment. One of the 27 participants quit the experiment due to the MRI scanning; eventually, we had 26 valid participants (mean age 21.08 ± SD 2.25).

Stimuli and procedure

We used the same stimuli as in Study 1. Each participant completed the two task phases described in Study 1 in the MRI scanner. After finishing the task, all participants confirmed that they had never seen the depicted people before. The stimuli were presented via the in vivo Esys system for fMRI (Gainesville, FL).

Imaging data acquisition

Structural images were acquired on a 3.0 Tesla Siemens Trio Tim scanner (Erlangen, Germany) with a 32-channel head coil at the fMRI laboratory of East China Normal University. T1-weighted sagittal structural images were acquired with the following parameters: repetition time (TR) = 2530 ms, echo time (TE) = 2.98 ms, field of view (FOV) = 256 mm and 192 slices to cover the whole brain. During the experiment, functional images were obtained using a gradient echo-planar imaging sequence (TR = 2000 ms, TE = 30 ms, FOV = 256 mm, slice thickness = 3 mm, matrix size = 64 × 64, slice gap = 0.3 mm, using 35 slices oriented in parallel to the anterior and posterior commissure, covering the whole brain).

Behavioral data analysis

The change in rating scores from pre- to post-observation of the depicted females’ decisions was calculated, and extreme trials were eliminated as described in Study 1. The total number of extreme trials across all participants accounted for 0.42% of the total trials. We also analyzed all the original results, which were consistent with the results after eliminating. A 2 (Intergroup members: In-group, Out-group) × 3 (Conditions: Accept, Reject, Undecided) rmANOVA on the change of rating scores was used to analyze the data.

Functional MRI data analysis.

Functional MRI data were preprocessed and analyzed using the Statistical Parametric Mapping (SPM) software, version 12 (The Wellcome Department of Imaging Neuroscience, London, UK). We performed a slice-timing correction first. The functional images were then realigned to the first image to correct for inter-scan head movements. One participant had to be excluded due to excessive head movement (translation > 3 mm and rotation > 3°). Next, each participant’s T1-weighted, three dimensional structural image was co-registered to the mean EPI image generated after realignment. The co-registered structural image was then segmented into gray matter, white matter and cerebrospinal fluid components with a unified segmentation algorithm. The functional images after the realignment procedure were spatially normalized to MNI space (resampled to 3 × 3 ×3 mm³) using on the normalization parameters estimated during unified segmentation. They were then spatially smoothed with an 8 mm full-width half-maximum Gaussian kernel.

We only analyzed fMRI data from the choice observation period for the couple images (Zhuang et al., 2021). Statistical parametric mapping of local brain activation was computed at the single-subject level using a standard general linear model. We modeled each stimulus presentation to their conditions, including IA (In-group-Accept), IR (In-group-Reject), IU (In-group-Undecided), OA (Out-group-Accept), OR (Out-group-Reject), OU (Out-group-Undecided) and convolved it with a canonical hemodynamic response function (HRF) and its time derivatives. Finally, the six movement parameters were included as covariates of no-interest. High-pass temporal filtering with a cut-off of 128 s was applied to remove low-frequency signal components. In the first-level analysis, simple main effects were computed for each participant for the conditions mentioned above by applying a’1 0ʹ contrast.

Based on the different behavioral results between acceptance and rejection copying (Place et al., 2010; Chu, 2012; Deng and Zheng, 2015; Zhuang et al., 2016, 2021), we constructed two distinct models of acceptance copying and rejection copying in the group analysis. For acceptance copying model, the four first-level individual contrast images (IA, IU, OA and OU) were fed into a 2 (Intergroup members: In-group and Out-group) × 2 (Conditions: Accept and Undecided) factorial design using a random-effects model (flexible factorial ANOVA in SPM12), while for rejection copying model, the 2 (Intergroup members: In-group and Out-group) × 2 (Conditions: Reject and Undecided) random-effects factorial design was built by IR, IU, OR and OU. The main effects of acceptance and rejection mate copying were defined respectively by using T-contrast (Accept—Undecided) and (Reject—Undecided), as well as the reverse contrasts (Undecided—Accept) and (Undecided—Reject). For acceptance copying, we analyzed the interaction effects between intergroup members and conditions using [(IA—IU)—(OA -OU)] and the reverse contrast, while for rejection copying, we conducted the interaction effects using [(IR—IU)—(OR—OU)] and the reverse contrast. In addition, during second-level analyses, a conjunction analysis using the conjunction null hypothesis (Nichols et al., 2005) was conducted first with the (Accept—Undecided) and (Reject—Undecided) contrasts to explore the common brain regions activated in human mate copying.

These analyses were performed for the whole brain, and a cluster-level threshold of P < 0.05 (FWE corrected) in combination with a voxel-level threshold of P < 0.001 (uncorrected) was used. All the activities were identified with the Anatomical Automated Labeling atlas (Tzourio-Mazoyer et al., 2002) in the Montreal Neurological Institute (MNI) space. The MarsBaR toolbox 0.44 in SPM12 (Brett et al., 2002) was used to extract beta values from the activated brain regions.

Functional MRI data analysis.

To further explore the neural connection between TPJ and mPFC (dmPFC and amPFC) in acceptance copying and rejection copying, we conducted a generalized psychophysiological interaction analysis (gPPI; McLaren et al., 2012). The TPJ (MNI: 48, −60, 39) that activated in the interaction contrast was used as the seed of interest. At the individual level, the seed region was defined as a 6 mm sphere around the above-mentioned coordinate, and the average time course was extracted for use in the gPPI analysis. The deconvolved time series from the seed region were extracted for each participant to create the physiological variable, and condition onset times were separately convolved with the canonical HRF for each condition, creating the psychological regressors. Interaction terms (gPPIs) were computed by multiplying physiological and psychological variables. Activity within the seed region was regressed on a voxel-wise basis against the interaction, with the physiological and psychological variables serving as regressors of interest (Hart et al., 2017). Individual gPPI contrast images of (IA—IU), (OA—OU), (IR—IU) and (OR—OU) were built by ‘1 −1’. Then we extracted TPJ’s gPPI parameter estimates of connectivity from dmPFC and amPFC by building a 6 mm sphere around the coordinates using the MarsBaR toolbox 0.44 in SPM12 (Brett et al., 2002) for each participant and each individual gPPI contrast. The dmPFC (MNI: −3, 54, 33) was chosen based on the conjunction analysis, while the right amPFC(MNI: 4, 52, 4) was chosen from Zhuang et al.’s (2021) mate copying paper. The beta value of TPJ—dmPFC and TPJ—amPFC were subjected to paired t-test between (IA—IU) and (OA—OU), (IR—IU) and (OR—OU), respectively.

Correlations between functional connectivity and CiA

To investigate the effect of TPJ-dmPFC and TPJ-amPFC functional connectivity on copying behavior, we conducted correlation analyses between CiA and functional connectivity of TPJ—dmPFC and TPJ—amPFC at (IA—IU), (OA—OU), (IR—IU) and (OR—OU) conditions.

Results

Behavioral results

The 2 (Intergroup members: In-group, Out-group) × 3 (Conditions: Accept, Reject, Undecided) rmANOVA for the CiA showed an insignificant main effect of group membership, F_{(1, 25)} = 2.64, P = 0.12, ηp2 = 0.096, and a significant main effect of conditions, F_{(2, 50)} = 26.09, P < 0.001, ηp2 = 0.51. Post hoc tests with Bonferroni correction confirmed that positive CiA scores for accept trials (M_{CiA_accept} = 0.59, SD_{CiA_accept} = 0.80) were significantly higher compared to undecided trials (M_{CiA_undecided} = 0.03, SD_{CiA_undecided} = 0.56), P < 0.001. The reverse was found for reject trials (M_{CiA_reject} = -0.23, SD_{CiA_reject} = 0.46), and this was significantly lower compared to undecided trials, P = 0.011.

Moreover, significant interaction between intergroup members and conditions was found, F_{(2, 50)} = 3.86, P = 0.028, ηp2 = 0.13. Further simple effect analysis showed that the CiA scores in the in-group (M_{CiA_IA} = 0.74, SD_{CiA_IA} = 0.82) were significantly positive than those in the out-group (M_{CiA_OA} = 0.45, SD_{CiA_OA} = 0.76), F_{(1, 25)} = 5.56, P = 0.026, ηp2 = 0.18. However, in the reject and undecided conditions, the CiA scores were not significantly different between the in-group female models and the out-group female models (P > 0.16) (see Figure 3).

Fig. 3.

Mean change in rating scores. Mean of the change in attractiveness (CiA) rating scores from pre- to post-exposure to the models’ choices for each condition. IA = in-group accept; OA = out-group accept; IR = in-group reject; OR = out-group reject; IU = in-group undecided; OU = out-group undecided.

fMRI results

Neural activation of acceptance copying

Main effects.

The main effects of acceptance copying computed by the (Accept—Undecided) contrast revealed significant activations in bilateral dmPFC, bilateral TPJ, left dorsal lateral prefrontal cortex, right supramarginal gyrus, right inferior frontal gyrus, right insula lobe (see Table 1 for details and the reverse contrast activation).

Table 1.

Whole brain analysis for accept choices

					MNI
k	Regions of maxima peak	BA	T	H	x	y	z
Accept—Undecided
1697	Supramarginal gyrus	8	7.22	R	6	21	60
	Superior frontal gyrus	9	7.06	L	−15	42	45
	dmPFC	9	6.77	L	−9	39	51
		9	5.8	R	15	42	45
	Middle frontal gyrus	9	5.95	L	−45	24	42
632	Middle temporal gyrus	21	6.11	L	−57	−18	−12
456	TPJ	39	7.4	L	−54	−63	33
285	Inferior frontal gyrus (p. orbitalis)	38	5.65	R	42	24	−12
	Insula lobe	48	4.64	R	30	18	−15
217	Precuneus	7	5.71	L	−6	−63	36
137	Cerebellum		5.52	R	33	−75	−39
100	TPJ	39	4.41	R	57	−60	33
84	Cerebellum		4.71	L	−27	−78	−39
Undecided—Accept
1483	Calcarine gyrus	17	6.51	L	−6	−93	6
79	Precentral gyrus	4	4.09	L	−45	−15	63

Results are reported as cluster significant after FWE correction at P < 0.05 for multiple comparisons; k = cluster size, MNI = Montreal Neurological Institute. L, left; R, right; H, hemisphere; BA, Brodmann area.

Interaction effects

Analysis of interactions between intergroup members and acceptance copying conditions computed by the contrast [(IA—IU)—(OA—OU)] and the reverse contrast [(OA—OU)—(IA—IU)]. More active regions in the right TPJ and left precuneus were shown in the contrast of [(OA—OU)—(IA—IU)]. We extracted the beta values from the peak coordinate point of right TPJ (MNI: 48, −60, 39) and left precuneus (MNI: −6, −60, 42) by building a 6 mm sphere. Paired t-tests on beta values were further performed to analyze the interaction. The results showed that the activity in the right TPJ was significantly higher in the out-group condition compared to the in-group condition for acceptance trials, t₍₂₄₎ = 2.74, P = 0.011, Cohen’s d = 0.55 (M_{beta_OA} = 0.16, SD_{CiA_OA} = 0.75; M_{beta_IA} = -0.09, SD_{beta_IA} = 0.84), while there was no significant difference between out-group and in-group for undecided trials, t₍₂₄₎ = -0.89, P = 0.384, Cohen’s d = -0.18 (M_{beta_OU} = -0.12, SD_{CiA_OU} = 0.68; M_{beta_IU} = -0.07, SD_{beta_IU} = 0.77). For the left precuneus, we also found the activity was significantly higher in the out-group condition compared to the in-group condition for acceptance trials, t₍₂₄₎ = 3.66, P = 0.001, Cohen’s d = 0.73 (M_{beta_OA} = 0.02, SD_{CiA_OA} = 0.59; M_{beta_IA} = -0.21, SD_{beta_IA} = 0.58), while there was no significant difference between out-group and in-group for undecided trials, t₍₂₄₎ = -1.18, P = 0.252, Cohen’s d = -0.24 (M_{beta_OU} = -0.27, SD_{CiA_OU} = 0.59; M_{beta_IU} = -0.20, SD_{beta_IU} = 0.60). No significant brain region was found in the contrast of [(IA—IU)—(OA—OU)] (Table 2 and Figure 4).

Table 2.

Whole brain analysis for interaction effect of intergroup members and accept choices

					MNI
k	Regions of maxima peak	BA	T	H	x	y	z
(OA—OU)—(IA—IU)
171	Middle occipital gyrus	19	4.37	L	−33	−72	36
86	TPJ	39	4.17	R	48	−60	39
84	Precuneus		4.27	L	−6	−60	42
(IA—IU)—(OA—OU)

No significant brain regions.

Results are reported as cluster significant after FWE correction at P < 0.05 for multiple comparisons; k = cluster size, MNI = Montreal Neurological Institute. L, left; R, right; H, hemisphere; BA, Brodmann area; IA = in-group accept; OA = out-group accept; IU = in-group undecided; OU = out-group undecided.

Fig. 4.

Whole-brain fMRI results for interaction effects between intergroup members and acceptance copying. Displayed are (A) right hemisphere and (C) left hemisphere views of the whole-brain results, highlighting the TPJ (A) and precuneus (C). (B) Percentage signal change of activity in the TPJ related to the same contrast for the right hemisphere, and (D) for the precuneus. IA = in-group accept; OA = out-group accept; IU = in-group undecided; OU = out-group undecided.

Neural activation of rejection copying

Main effects

The main effects of rejection copying computed by the (Reject—Undecided) contrast revealed significant activation in left dmPFC (MNI: 0 54 33, 117 voxels). The right calcarine gyrus, right superior frontal gyrus and left precuneus were found in the reverse contrast.

Interaction effects.

The contrast [(IR—IU)—(OR—OU)] and the reverse contrast [(OR—OU)—(IR—IU)] were used to analyze the interactions between intergroup members and the rejection condition. However, no significant brain region was observed in these two contrasts.

Conjunction analysis of acceptance copying and rejection copying

The conjunction analysis using the (Accept—Undecided) and (Reject—Undecided) contrasts revealed significant activations in left dmPFC (MNI: −3 54 33, 105 voxels). To further explore the specific activation of dmPFC in accept, reject and undecided conditions, we extracted the beta values from the peak coordinate point (MNI: −3 54 33). The beta values were submitted to a rmANOVA with the within-subject factor conditions (accept, reject and undecided). Results indicated a significant effect, F_{(2, 48)} = 10.98, P < 0.001, ηp2 = 0.31. Post hoc tests with Bonferroni correction confirmed that dmPFC beta value for accept and reject trials (M_{beta_accept} = 0.26, SD_{beta_accept} = 0.90; M_{beta_reject} = 0.15, SD_{beta_reject} = 1.20) was significantly higher compared to undecided trials (M_{beta_undecided} = -0.32, SD_{beta_undecided} = 1.01), P < 0.001 and P = 0.003, respectively. No significant difference between accept trials and reject trials, P = 1.00 (see Figure 5).

Fig. 5.

Conjunction results for acceptance copying and rejection copying. Displayed are (A) left hemisphere views of the conjunction results, highlighting the dmPFC. (B) Percentage signal change of activity in the dmPFC related to the same contrast for the left hemisphere.

In addition, we contrasted accept trials and reject trials directly. Regions of bilateral superior frontal gyrus, right supramarginal gyrus, bilateral dorsal lateral prefrontal cortex, bilateral TPJ, left superior parietal lobule, left inferior parietal lobule, right thalamus and left inferior frontal gyrus activated stronger in the acceptance copying than rejection copying. In contrast, only the left post-central gyrus was significantly more activated for the reverse comparison Reject > Accept.

Functional connectivity results.

The paired t-test revealed a significant difference in the TPJ (MNI: 48–60 39) and amPFC (MNI: 4 52 4) functional connectivity, t₍₂₄₎ = -2.35, P = 0.027 between (IA—IU) and (OA—OU). No significant difference was found between (IR—IU) and (OR—OU), P = 0.22. The connectivity of TPJ and amPFC was significantly stronger in the (OA—OU) (M_{rightTPJ-right amPFC} = 0.19, SD_{rightTPJ-right amPFC} = 0.28) than that in the (IA—IU) condition (M_{rightTPJ-right amPFC} = -0.04, SD_{rightTPJ-right amPFC} = 0.36) (see Figure 6B). For the TPJ and dmPFC (MNI: −3 54 33), there were no significant functional connectivity effects (P ≥ 0.23).

Fig. 6.

Functional connectivity of right TPJ and right amPFC. Displayed are (A) right hemisphere views of the interaction results, highlighting the TPJ and amPFC. (B) The strength of functional connectivity between right TPJ and right amPFC. (C) Positive correlation between the functional connectivity of right TPJ—right amPFC and mean CiA scores in (OA—OU). IA = in-group accept; OA = out-group accept; IU = in-group undecided; OU = out-group undecided.

Correlations between functional connectivity and CiA.

In the (OA—OU) condition, functional connectivity between the TPJ and amPFC correlated positively with CiA, r = 0.37, P = 0.035 (Figure 6C). There were no significant correlations between functional connectivity and CiA for other functional connectivity or conditions.

Discussion

The behavioral results in the fMRI study were consistent with Study1. Through a whole-brain analysis, we explored the relevant brain regions involved in acceptance and rejection copying at the neural level. Compared to rejection copying, acceptance copying activated more brain regions typically linked to social information processing and understanding others (dmPFC, TPJ), the mirror neuron system (inferior frontal gyrus), insula and dorsal lateral prefrontal cortex, which was consistent with previous studies (Zhuang et al., 2016, 2021). In contrast, only the dmPFC activated in rejection copying. Further, the conjunction analysis of acceptance and rejection copying showed significant activation in the dmPFC, and the extracted beta values showed no difference between accept and reject trials.

Importantly, we found significant activation of right TPJ in the interaction effect between intergroup members and acceptance copying. Further analysis of the extracted beta values found that the strength of activation in the right TPJ was significantly higher under out-group models compared to in-group models. Through exploring the functional connectivity between the TPJ and mPFC (amPFC and dmPFC), we found that during acceptance copying, compared with exposure to in-group female models, the functional connectivity of the right TPJ with amPFC was stronger when participants were exposed to out-group female models compared to in-group female models, while the difference was not significant in rejection copying. Meanwhile, we found that in acceptance copying, when the female models were out-group members, the functional connectivity between the TPJ and amPFC was significantly and positively correlated with the change in ratings. We did not find significant functional connectivity results between TPJ and dmPFC. These results might imply that dmPFC may be more involved in the processing of social information in mate copying. In addition, TPJ and amPFC are more involved in modulating acceptance copying by in- and out-group models, which suggests that when copying out-group members’ choices, participants demand more resources to infer the intentions of out-group members in acceptance copying.

General discussion

In the present reported behavioral and fMRI studies, we explored the influence of group attributes on female mate copying. Behaviorally, we validated the mate copying effect in female participants once again. Consistent with the previous results (Zhuang et al., 2016, 2021), participants increased their ratings of potential mates after observing models’ acceptance attitudes, and vice versa, indicating that females gathered public information by observing the choices of same-sex models to adjust their choices. Interestingly, we found that the in- and out-group attributes of models modulated acceptance and rejection copying separately. Previous studies reported that acceptance copying and rejection copying might refer to different mechanisms (Place et al., 2010; Chu, 2012; Deng and Zheng, 2015; Zhuang et al., 2016, 2021), and the current studies provided more evidence for the separation of acceptance and rejection copying by exploring the modulation of in- and out-group models from behavioral and neural levels.

Specifically, our results at the behavioral level revealed that when models displayed interest in a potential mate (acceptance copying), the copying effect was discounted for out-group models compared to in-group models. This finding aligns with prior research demonstrating that individuals are more likely to align their beliefs, behaviors and attitudes with their in-group rather than their out-group (Mackie and Wright, 2003; Cialdini and Goldstein, 2004; Lin et al., 2018; Wei et al., 2023). It is plausible that the background, culture, and societal norms of in-group models are more similar to those of the participants, making the choices of in-group models more representative and conducive to what participants themselves would choose. Consequently, participants exhibited a stronger copying effect after observing the choices of in-group models. In contrast, existing research on mate copying has predominantly focused on the role of acceptance cues, with limited attention given to the phenomenon of rejection copying. A recent study by Zhuang et al. (2021) has demonstrated weaker behavioral copying under rejection copying compared to acceptance copying. In the current study, we have also found a robust but weaker rejection copying effect at the behavioral level compared to acceptance copying. Furthermore, the current study has revealed that at the behavioral level, rejection copying is not influenced by in-group and out-group models, which indicates that rejection copying may be insensitive to the attributes information of models. These results suggest that participants also show interest in the rejection information to copy and are even directly influenced by it, but may not care about other scenario information, such as the attributes information of models, when coping rejection attitudes. Taken together, the present study suggests that the affiliation of the model plays a significant role in the acceptance copying, but not in the rejection copying.

Neurophysiologically, consistent with previous studies (Zhuang et al., 2016, 2021), we found that compared to rejection copying, acceptance copying activated more brain regions typically linked to social information processing and understanding others (dmPFC, TPJ), mirror neuron system (inferior frontal gyrus), insula, and dorsal lateral prefrontal cortex, which means that acceptance and rejection copying referred to two different mechanisms, and acceptance copying might involve more cognitive processing than rejection copying. In addition, in the current study, using conjunction analysis, we further observed that the dmPFC was involved in both acceptance copying and rejection copying, suggesting that the dmPFC played a potentially critical or fundamental role in mate copying processing. dmPFC has been considered a core area of social information processing (Petrides and Pandya, 2006, 2007; Apps and Sallet, 2017; Lockwood et al., 2018), which might be related to the processing of social information (e.g. how and when the public information will be most helpful) in acceptance and rejection copying.

Corresponding with the behavioral results, our fMRI study found that brain activation differences between in-group and out-group models were only observed during acceptance copying. Specifically, we observed significant activation of the TPJ in the interaction effect between group attributes and acceptance copying, while there was no significant activation in the interaction of rejection copying. More detailed, during acceptance copying, we found that the TPJ was more strongly activated when participants learned from out-group models’ choices, compared to in-group models. In several fMRI studies, the right TPJ has been repeatedly found to be relevant for reasoning about and understanding other persons’ mental states, pointing to its outstanding role in ToM processes (Saxe et al., 2004; Saxe and Wexler, 2005). Stronger activation of TPJ under out-group models probably means that participants need to be more effortful to infer why out-group models are interested in the depicted male. Furthermore, only for acceptance copying, compared to in-group models, when learning from out-groups, the functional connection between TPJ and amPFC was stronger, and the connection between TPJ and amPFC could positively predict the mate copying strength. However, we did not find the activation of dmPFC in the interaction effect, and the functional connection between TPJ and dmPFC was not regulated by group attributes of models. Although dmPFC and amPFC have been reported to be involved in the process of mentalizing (Frith et al., 2003; Lieberman, 2007; Mar, 2011), they are distinct in their function: social processes are functionally linked to dmPFC, while self and affective processes are linked to amPFC (Lieberman et al., 2019). Meanwhile, a current meta-analysis identified that TPJ and amPFC activated in exteroceptive- and mental- self-processing, suggesting that the two regions likely serve essential functions in the differentiation and integration of self-other information (Qin et al., 2020). Taken together, the results might imply that dmPFC may be more involved in the processing of social information in mate copying, while TPJ and amPFC are more involved in integrating and inferring on self—other information according to the task.

Our results on the interaction and functional connection results are similar to those reported in a previous study on prosocial decision-making (Telzer et al., 2015). Specifically, in acceptance copying, which was deemed more important by participants (Trivers, 1972; Zhuang et al., 2016, 2021), learning from out-group models was found to be more challenging than from in-group models. This resulted in the recruitment of more brain regions associated with mentalizing (TPJ) and stronger connectivity between these regions (TPJ and amPFC). Although previous studies on intergroup social influence have reported more activities in brain regions related to empathy or mentalizing in the in-group condition compared to the out-group (Balliet et al., 2014; Cikara and Van Bavel, 2014; Lin et al., 2018), the present study and Telzer et al. (2015) study, which used two different paradigms, consistently suggest that there may be at least two patterns for processing the effects of in-group and out-group members on behaviors. One of these two patterns might involve directly observing, recognizing or learning from in- and out-group members, where participants might show stronger mentalizing or empathy towards in-group members. The other pattern appears to be more complex, such as making a helpful decision (e.g. prosocial behavior) or using out-group members as a reference to benefit oneself (e.g. mate copying); for this, participants have to consider more about the out-group members, which requires a deeper understanding and inference of their motives and actions. Although further research is needed to confirm our findings, we believe that our studies provide a new perspective on social cognition in relation to in-group and out-group members and are therefore deserving of further exploration. In addition, in the interaction effect between group attributes and acceptance copying, we found that precuneus showed stronger activation when participants were exposed to out-group models than in-groups. The activation of precuneus was consistent with TPJ and had also been considered related to the ToM (Schurz et al., 2014; Plank et al., 2022), which might mean more inferring towards out-groups in acceptance copying.

Regarding rejection copying, as previously mentioned, the neural activities were different from acceptance copying, particularly in the ToM-related brain regions, such as TPJ and precuneus. These findings are similar to the results reported by Zhuang et al. (2021). Similarly with the behavioral results, we have not observed suprathreshold brain activations or functional connections influenced by intergroup factors at the neural level. Taken together, the behavioral and neural results of our study suggest that rejection copying may be a relatively stable and scenario-independent process. This may be because that rejection often signals the end of a potential romantic relationship (Zhuang et al., 2021; Nöbel et al., 2022) and thus may not require further consideration of the possible reasons for refusal.

In our studies, we used the national flag to construct the in- and out-group model messages clearly. The national flag represents the national identity and is often used to signify group membership or group affiliation (e.g. Lin et al., 2018; Marinthe et al., 2021; Maloku et al., 2023). To examine the possibility that different attributes between Chinese and American flags (such as different colors, different emotions associated with flags, and so on), rather than in- or out-group models’ choices, impacted the participants’ attitude changes, we compared the results under the baseline condition in which female models expressed no choice attitude towards potential mates. Results did not show significant rating differences between the in-group and out-group, suggesting that the presentation of flags might not lead to the changed attitudes and should not be responsible for the differences in mate copying. Additionally, we subtracted the corresponding baseline conditions from the in-group and out-group conditions during the behavioral and neural analyses, and the resulting interactive patterns aided us in eliminating the potential effects of flags. Another interesting question to consider is whether reference to other people’s mate choice would be associated with competition or aggression. In the frame of mate copying, attitudes of competition or aggression towards models are indeed observed, yet they are found more often in male mate copying (Bierbach et al., 2011; Plath and Bierbach, 2011; Auld and Godin, 2015). It is also important to acknowledge that the aggression factor may play a role in a clear competitive scenario (DiMenichi and Tricomi, 2015; Si et al., 2022). Therefore, in future studies, it is an interesting topic to include male participants and/or manipulate competitive contexts explicitly, such as creating task scenarios where participants need to compete with models for potential mates, to further explore the potential roles of aggression and competition in mate copying. Also, in addition, some limitations of these studies should also be noted. First, due to the restrictions of geography and Coronavirus Disease 2019 (COVID-19), the participants who took part in this experiment were all Chinese females. Exploring the influence of group members on female mate copying in different countries might help us to explain this issue more comprehensively. Second, a previous study has found inconsistent performance between male and female participants in acceptance and rejection copying (Place et al., 2010). Adding male participants might be considered for comparison in future studies. Third, human mate choice copying is a complex social process, and exploring the similarities and differences between mate choice copying and other social learning, such as conformity, would make a clearer interpretation of the mechanism of mate copying in the future.

To summarize, the current results show that female participants engaged in mate copying and were more likely to copy the acceptance choice compared to the rejection choice, with a stronger copying effect in acceptance. Importantly, we also found that the regulation of acceptance copying and rejection copying by the in- and out-group models was distinct. When learning from out-group models, the copying effect was discounted compared to in-groups in acceptance copying, while there was no significant difference in rejection copying. At the neural level, we found that the dmPFC, which is involved in both acceptance copying and rejection copying, may play a fundamental role in social processing during mate copying. Specifically, for acceptance copying, we observed stronger activity in the TPJ and stronger connectivity between the TPJ and amPFC when female models belonged to out-group members. Furthermore, the functional connection between the TPJ and amPFC positively predicted the change in ratings when learning from out-group models, suggesting that more resources may be required to infer the intentions of out-groups to overcome in-group biases during acceptance copying.

Data accessibility

All data and code are available at the Open Science Framework: https://osf.io/6h85j/

Contribution statement

Designed the research: Jiajia Xie, Li Zheng, and Lin Li; conducted the research: Jiajia Xie, Peng Li, and Yuyan Wu; analyzed the data: Jiajia Xie, Li Zheng, and Yang Lu; wrote the manuscript: Jiajia Xie and Li Zheng; discussed the final manuscript: Jiajia Xie, Li Zheng, Lin Li, Jinying Zhuang, Yang Lu, Peng Li, and Yuyan Wu; all authors gave final approval for publication.

Conflict of interest

The authors declared that they had no conflict of interest with respect to their authorship or the publication of this article.