Exploring the effects of sexual prejudice on dyadic interactions through an automated analysis of nonverbal behaviours

Nonverbal behaviours (NVB) are a fundamental part of the communication process: especially indicative of individuals’ inner states such as attitudes and motivations, NVBs can deeply shape the perceived quality of the interaction. Despite their practical importance and theoretical value, NVBs in intergroup interactions (i.e. intergroup nonverbal behaviours; INVB) are an understudied topic. So far, they have been mainly investigated within interethnic contexts (i.e., White and Black people) and by employing invasive or time-consuming procedures, mainly involving subjective evaluations of video-recorded interactions by external coders. The present work aimed at extending previous literature by exploring NVB and its relationship with prejudice within gay/straight dyadic interactions, a relevant but still partially unexplored intergroup context within this field of research. Differently from ethnicity, sexual orientation is less identifiable and cannot be ascertained from visible markers such as the skin colour, but requires self-disclosure. Further and most importantly, we assessed patterns of NVBs through an RGB-depth camera – the Microsoft Kinect V.2 Sensor – that allowed us to obtain exact quantitative measures of body movements in a fully automatic and continuous way. In doing so, we conducted three experimental studies in which heterosexual participants (total N = 284) were first administered measures of implicit bias and explicit prejudice towards gay men (Study 1 & 3) or lesbians (Study 2), and then asked to interact with a gay (vs. straight; Study 1 & 3) or lesbian (vs. straight) confederate (Study 2), whose sexual orientation was manipulated (Studies 1 & 2) or disclosed (Study 3). A fake Facebook profile, shown to the participant before the interaction, revealed the confederates’ sexual orientation. In all the studies, we considered the pattern of results on two main NVBs, one concerning proxemics (i.e., interpersonal volume between interactants) and the other concerning kinesics (i.e., amount of upper body motions). We selected these NVBs because previous research revealed that they are particularly meaningful for the comprehension of the psychological immediacy between interactants (i.e., interpersonal volume) and their comfort (or discomfort; amount of upper body motions) during a dyadic interaction. Overall, our work revealed a relevant (and unexpected) pattern of findings concerning interpersonal distance. Unlike previous literature, Study 1 revealed that high (vs. low) implicitly biased participants, instead of keeping a larger distance, tended to stay closer to the confederate presented as gay (vs. straight), especially when discussing a topic concerning the intergroup relation (i.e., the situation of the gay community in Italy) than a neutral one. This result was importantly extended in Study 3: high (vs. low) implicitly biased participants that stood closer to the gay (vs. straight) confederate revealed greater cognitive depletion (i.e., lower performance on a Stroop colour-naming task) after the conversation. This latter result suggests that, at least within gay/straight men interactions, interpersonal distance is an NVB that (high implicitly biased) people can control to manage their self-presentation, with consequent greater impairment of their cognitive resources. This main finding was not replicated in Study 2, in which we focused on dyadic interactions between heterosexual participants and lesbian women, by confirming how heterosexual people’s attitudes (and their consequent INVBs) towards this minority group is distinct from those towards gay men and, presumably, people’s gender plays a more predominant role than their implicit or explicit attitudes. Further, across our studies, we found inconsistent or non-significant results concerning the participants’ upper body motion as an outcome variable. A possible explanation for these inconsistent results could be due to the relatively coarse algorithmic index that we used for this INVB. Theoretical and methodological implications of this work are discussed in the General Discussion section, together with its limitations and indications for future research.