Contrary to previous research findings, our findings suggest that third-party observers cannot reliably detect attraction in others

Can third-party observers detect attraction in others based on subtle nonverbal cues? Iliana Samara, Tom S. Roth, Milica Nikolic, Eliska Prochazkova & Mariska E. Kret. Current Psychology, Apr 8 2022. https://link.springer.com/article/10.1007/s12144-022-02927-0

Abstract: In a series of three studies, we examined whether third-party observers can detect attraction in others based on subtle nonverbal cues. We employed video segments of dates collected from a speed-dating experiment, in which daters went on a brief (approx. 4 min) blind-date and indicated whether they would like to go on another date with their brief interaction partner or not. We asked participants to view these stimuli and indicate whether or not each couple member is attracted to their partner. Our results show that participants could not reliably detect attraction, and this ability was not influenced by the age of the observer, video segment location (beginning or middle of the date), video duration, or general emotion recognition capacity. Contrary to previous research findings, our findings suggest that third-party observers cannot reliably detect attraction in others. However, there was one exception: Recognition rose above chance level when the daters were both interested in their partners compared to when they were not interested.

General discussion

In a series of three experiments, we found no strong evidence supporting the notion that people can reliably detect attraction or its absence in thin video slices of people on a date based on nonverbal subtle emotional cues. However, we found that accuracy was increased based on whether the person presented in the video was attracted to their partner. Specifically, we found that the third-party observers were more accurate in detecting attraction when the daters were attracted to their partners than detecting the absence of attraction when the daters indicated not being attracted to their partner. In addition, recognizing attraction was not influenced by age or length of the stimuli presented.

In accordance with previous findings (e.g., Place et al., 2009), we found that people cannot reliably detect attraction from initial interactions. Given that previous findings have emphasized the importance of subtle nonverbal cues in communicating attraction (e.g., Eibl-Eiblsfeldt, 1989; Keltner & Buswell, 1997), one might question whether the observed low accuracy in detecting attraction might be the result of a low frequency of occurrence of behaviours associated with attraction. In other words, was there sufficient information present in the stimuli themselves that the participants might have picked up? Indeed, we only found minor numerical differences in behaviours associated with attraction (e.g., coyness, genuine smiles) in the First Impression 3-s videos (see Supplemental Material). Thus, the observed low accuracy might result from the low frequency of behaviour occurrence. Nonetheless, our findings replicate previous research (e.g., Place et al., 2009) and further support the notion that people cannot reliably detect attraction when viewing others in the initial phases of their interaction.

Our findings do not provide support for the notion that third-party observers can detect attraction when viewing segments from later phases of a date, which contrasts with previous research (Place et al., 2009). In all experiments, participants performed near chance level independent of the length of the segment (3, 6, or 9 s) or the phase of the interaction (first impression or verbal interaction). Our analyses (see Supplementary Material) of the coded behaviours illustrate that daters that were attracted to their partner exhibited behaviours associated with attraction for a longer duration compared to daters that were not interested in their partner (in videos taken from the middle of the speed date). This finding suggests that the observed low accuracy is not due to the low frequency of behaviour occurrence. Instead, it might be more probable that people cannot detect attraction as third-party observers using thin video slices even when the signs of attraction are there.

It may be advantageous for humans to mask what they feel in certain situations, and they often use their cognitive resources to do so (Kret, 2015). This masking might render interpreting nonverbal cues more complex and thus, lead to confusion and awkward social encounters (Abbey, 1982; Abbey & Melby, 1986) when the expressions of the sender are misinterpreted (Burgoon et al., 2002; Grammer, 1990). These factors may be a source of error in people involved in a one-on-one interaction (i.e., a date), given that the high-intensity motivational environment might decrease accurate emotion detection (Maner et al., 2005; Prochazkova et al., 2021).

It has been speculated that the ability to detect attraction in others has an adaptive function, allowing people to collect more information to guide their mating choices (see Simao & Todd, 2002). However, a more parsimonious explanation would be that the ability to detect attraction as a third-party observer is merely a by-product of detecting attraction when faced with a potential mate, which would undoubtedly be a beneficial quality for anyone navigating their romantic environment. However, previous research consistently demonstrates that people cannot detect attraction in others and instead project their interest to a given partner (Lee et al., 2020; Samara, Roth, & Kret, 2020; see also Prochazkova et al., 2021). Thus, it remains possible that people cannot detect attraction above chance level.

Emotions can be efficiently detected from facial expressions (Ekman, 1992). Previous research has shown that basic emotions, such as disgust, fear, and happiness, can be recognized in scenes within 200 ms (Righart & de Gelder, 2008). This effect suggests that detection and recognition of emotional expressions likely rely on quick facial expression processing (see also Meeren, van Heijnsbergen, & de Gelder, 2005, for similar findings on the interaction between facial expressions and body language). Here, we examined whether attraction can be detected as efficiently as other emotions. Given our null findings, we cannot conclude whether indeed attraction can be detected as efficiently as other emotions based on three experiments. Future research should help elucidate how easily and accurately complex emotions like attraction are perceived and processed.

In all experiments, we consistently found that people are likely to detect attraction when the person observed is indeed exhibiting such signals. Indeed, even though attraction cannot be expressed with a single behaviour (Moore, 1985), people likely have experience in decoding such cues and are thus more likely to detect them efficiently. This is further corroborated by our consistent replication of this effect in initial encounters as well as later in the interactions irrespective of video length (3, 6, and 9 s). Date members that were attracted to their partner likely illustrated affiliation more clearly (e.g., see Grammer et al., 1999). In contrast, disinterested partners might have opted to display rejection more subtly (or perhaps not at all), making it more challenging to interpret. However, it should be noted that we did not find robust differences in attraction cues between daters that were interested in their partner compared to daters that were not in the 3-s stimuli, even though a robust difference was found for coy smiles in the 9-s stimuli. An alternative explanation for the finding is that participants were more likely to detect attraction when indeed, participants had a general propensity to respond positively rather than negatively (see Supplemental Material). This could be due to expectancy effects, given that participants were informed that these video segments are from a blind date study. Future research should further investigate the role of expectancy effects in the ability of third-party observers to detect attraction.

This finding directly contrasts with previous research (Hall et al., 2015 Experiment 2). In their study, the authors asked participants to view 1-min segments of others on a date and indicate whether they thought the person on the video was flirting with their partner. Given that the people that report feeling attracted to their partner are also more likely to report flirting (Hall et al., 2015; Experiment 1), this is a reliable indicator of detecting attraction. Furthermore, their results suggest that participants were more accurate in detecting attraction when the person depicted was not flirting than when they were flirting. The authors suggest that these findings could be due to a) the implicit risk of openly displaying interest in another, which would have rendered any flirting difficult to decode, and b) that the probability of flirting in zero-acquaintance settings is relatively low (e.g., Abbey, 1982; Saal et al., 1989); therefore, people might not be familiar with flirting expressions in such settings. We disagree with both of these interpretations. Flirting, in general, is quite ambiguous, as flirting cues are also easily confused with friendliness (Farris et al., 2008; Moore, 2010). Furthermore, previous research has documented several flirting signals in first time-encounters, such as self-grooming (McCormick, Perper, & Jones, 1983), suggesting that these are signals typically exhibited in such situations. Crucially, in a previous study (Prochazkova et al., 2021), it was found that almost half (44%) of the participants reported that they would be interested in going on another date with their partner rendering the reduced-likelihood interpretation unlikely. In short, we consistently show that attraction is detected above the chance level when it is indeed there.

Based on the Perception–Action Model of Empathy (PAM; de Waal & Preston, 2017), we expected that participants with more experience with romantic interactions (i.e., adults) would be more accurate in detecting attraction than participants with less experience with romantic interactions (i.e., children). However, in Experiment 2, we found no substantial differences between adults and children, suggesting that children’s lower accuracy in detecting attraction in Experiment 1 was likely due to cognitive overload.

One limitation that should be discussed is the fact that our responses were coded in a binary way. This approach was necessary to calculate accuracy based on the responses of the study conducted by Prochazkova et al. (2021), where responses were also coded binary. It could be argued that this approach reduced the variation that would otherwise be shown if responses were coded in a continuous way. This is indeed possible, even though it should be noted that using a scale for attraction and a binary response for another date has been shown to correlate highly (Roth et al., 2021a, 2021b). Nonetheless, future studies using speed-dating paradigms could also employ a continuous response regarding attraction and willingness to go on another date, which can then be used in studies employing third-party observers. In this manner, a more nuanced accuracy scale can be calculated.

In conclusion, here we demonstrate that people might not reliably detect when others are attracted to their partner and when not. Furthermore, we showed that the overall accuracy in detecting attraction is not influenced by age, or the phase of the interaction observed. The only factor that reliably influenced accuracy is whether attraction is indeed present.