Thin-slicing study of the oxytocin receptor (OXTR) gene and the evaluation and expression of the prosocial disposition

To test these three hypotheses, a sample of naive observers watched 20-s, silenced video clips of 23 target individuals with varying rs53576 genotypes. In the video clips, targets were listening to their romantic partner disclose an experience of personal suffering; thus, the target was the listener in the conversation. This conversation was selected because displays of vulnerability are compelling elicitors of prosocial behavior (31, 32) and should provide targets with a clear opportunity to respond to another person in need.

We first tested whether target differences in rs53576 predicted the prosocial judgments observers made about them. In these analyses, the focus was on the observers and their inferences about the prosociality of the targets. Because of the nested design of the study (i.e., each observer watched 23 videos), hierarchical linear modeling was used to analyze the data (33); specifically, ratings of targets (level 1) were nested within individual observers (level 2). As predicted, and portrayed in Fig. 1, observers judged targets homozygous for the G allele (M = 4.21, SD = 1.40) as more prosocial than targets carrying an A allele (M = 3.80, SD = 1.41), b = 0.41, t(2,628) = 8.22 P < 0.001. In fact, of the 10 most trusted targets, 6 were homozygous for the G allele; of the 10 least trusted targets, 9 were carriers of the A allele. These results demonstrate that differences in rs53576 systematically predict outside observers’ judgments of the prosociality of carriers based on observations of 20 s of silent behavior.

Our next set of analyses focused on several ancillary questions. First, we tested whether the effect of target genotype was invariant across observers. That is, did observers make similar inferences about the prosociality of the target, or was there significant variation in these inferences? To test this, we modeled the effect of target genotype as random at level 2 within the hierarchical linear model. In this analysis, we did not observe significant individual differences in the association between target genotype and the observers’ prosociality ratings of the targets, χ² = 110.10, df = 115, P > 0.500. Therefore, the rated differences in prosociality between targets homozygous for the G allele and those carrying the A allele were consistent across the observers.

Next, we tested whether the sex of the targets moderated the effect of genotype on the prosociality judgments of observers. Observers tended to see a bigger difference in prosociality between the male targets who were homozygous for the G allele compared with carriers of the A allele than for the female targets, b = 0.14, t(2,626) = 1.85, P = 0.063. It should be noted, however, that the difference in prosociality judgments of the female targets homozygous for the G allele and the A allele was still highly significant, b = 0.32, t(2,626) = 5.39 P < 0.001. It is also important to note that male targets were on average judged to be less prosocial than the female targets, b = −0.27, t(2,628) = −5.38 P < 0.001. Therefore, these findings suggest that observers judge both male and female targets homozygous for the G allele to be more prosocial than those with the A allele, although this difference may be more pronounced for males, a topic worthy of future investigation.

Our second hypothesis held that targets homozygous for the G allele would display more affiliative nonverbal cues than carriers of the A allele. For these analyses, the central unit of analysis was the target because they made the behavioral displays. Thus, we used an ordinary least-squares regression approach. In line with our prediction, we found that targets homozygous for the G allele (M = 0.33, SD = 0.56) displayed more affiliative nonverbal cues than targets carrying an A allele (M = −0.25, SD = 0.56), b = 0.58, t(22) = 2.48, P = 0.022. Of the 10 targets displaying the most affiliative cues, 6 were homozygous for the G allele; of the 10 targets displaying the fewest affiliative cues, 8 were carriers of the A allele. Thus, differences in display of affiliative cues as a function of target genotype closely mirrored the observers’ prosociality ratings.

Finally, we conducted mediation analyses to ascertain whether differences in affiliative cues accounted for the link between target rs53576 genotype and observer judgments of prosociality. We used a multilevel mediation approach as outlined by Zhang, Zyphur, and Preacher (34). To calculate the 95% confidence interval (CI) for the indirect effect, we used the Monte Carlo Method for Assessing Mediation using 20,000 repetitions (35). Only the relationship between affiliative cues and prosociality ratings (χ² = 171.23, P = 0.001) varied across observers; therefore, the genotype effect was modeled as fixed at level 2 and the affiliative cues effect was modeled as random at level 2. Supporting our hypothesis and as shown in Fig. 2, the difference in prosociality judgments of targets homozygous for the G allele versus carrying an A allele became nonsignificant [b =0.05, t(2,627) = 1.00 P = 0.318] when mediated by affiliative cues, CI 95% (0.08, 0.67). In contrast, target display of nonverbal affiliative cues remained a highly significant predictor of the prosociality judgments observers made about the targets, b = 0.62, t(115) = 12.88, P < 0.001. Thus, consistent with our hypothesis, individuals who are homozygous for the G allele were judged to be more prosocial as a function of displaying more affiliative cues within the period of observation.

Cooperation among nonkin is enabled when individuals can reliably identify the prosocial intentions of other individuals based on brief observations. In line with this reasoning, studies have documented that the prosocial intent of others is highly detectable on the basis of brief observations of nonverbal behavior (21, 26–29). The present study extends this research by showing that based on witnessing only 20 s of silent behavior of the targets, naive observers attributed greater prosociality to targets homozygous for the G allele on an SNP of OXTR (rs53576), which has been found in other studies to predict prosociality (15–19), than to carriers of the A allele. Furthermore, the targets homozygous for the G allele displayed greater nonverbal affiliative cues associated with prosociality (e.g., head nods, smiles) than targets carrying the A allele, which in turn ultimately accounted for how prosocial the observers judged the targets to be.

Taken together, these findings demonstrate that an SNP (rs53576) known to predict prosocial tendencies is signaled through display behavior that leads to reliable observer inferences of the target's prosociality. Our findings shed light on how people's friendship networks can align along genetic lines; recent evidence finds that people tend to have friends that share similarity among certain genes and differences among other genes (36). The communicability of genetic information on the basis of a few brief moments of behavior provides a potential mechanism for how friendship networks develop as a function of such genetic similarities and differences. The more general implication is that even SNPs can yield systematic influences upon the social inferences and relationships important to social life.

It is important, however, to take caution in interpreting the present results. Social traits like prosociality are no doubt influenced by a multiplicity of genetic and epigenetic factors, and thus single-gene paradigms are ultimately limited in their ability to explain large portions of variability in social behavior (37). We therefore suggest that variation in rs53576 is a contributing factor to people's proclivity toward prosociality, although it is certainly not the only one. Efforts should be taken to better understand its role within the context of other factors underlying a prosocial disposition, especially given that oxytocin functions through a single receptor, unlike most neuromodulators. It is important to note that the exact functional consequences of rs53576 are presently unknown; it is located in an intron that has been implicated in transcriptional repression (38). More research is required to understand how differences in this specific location of OXTR contribute to physiological processes that underlie social functioning. Nonetheless, our results are consistent with a growing number of studies linking variation in rs53567 to social functioning across numerous social domains (15–19). It is also important to note that the present study featured a limited number of targets; thus, more work is necessary to replicate and extend the present results to a larger, more diverse sample. Finally, several studies have documented that the relationship between rs53576 and social functioning is culturally dependent, and that the distribution of genotypes varies heavily by ethnicity (39–41). In the present study, all of the targets were Caucasian. This approach allowed us to avoid potential cultural and ethnic biasing of our results, but it also limits the generalizability of the present findings. Future research should investigate whether a different pattern of results emerges for targets of other ethnicities.

Collectively, our results speak to the communicability of even slight genetic variations and the power of human intuition and inference to recognize the nonverbal signatures associated with specific genotypes. What remains, however, is the challenge of understanding the specific pathways through which genes bias behavior, especially within the context of other vital endogenous and exogenous factors shaping behavioral tendencies.

This research was supported by the University of California, Berkeley Committee on Research: Research Bridging Grant, the Metanexus Institute, the Greater Good Science Center, and the Fetzer Institute.