There are consistent patterns that could be considered a non-verbal signal of guilt in humans: Guilt was most closely associated with frowning and neck touching

Are there non-verbal signals of guilt? Eglantine Julle-Danière, Jamie Whitehouse, Alexander Mielke, Aldert Vrij, Erik Gustafsson, Jérôme Micheletta, Bridget M. Waller. PLoS, April 24, 2020. https://doi.org/10.1371/journal.pone.0231756

Abstract: Guilt is a complex emotion with a potentially important social function of stimulating cooperative behaviours towards and from others, but whether the feeling of guilt is associated with a recognisable pattern of nonverbal behaviour is unknown. We examined the production and perception of guilt in two different studies, with a total of 238 participants with various places of origin. Guilt was induced experimentally, eliciting patterns of movement that were associated with both the participants’ self-reported feelings of guilt and judges’ impressions of their guilt. Guilt was most closely associated with frowning and neck touching. While there were differences between self-reported guilt and perception of guilt the findings suggest that there are consistent patterns that could be considered a non-verbal signal of guilt in humans.

Discussion

This study aimed to identify which facial movements were perceived as guilt when guilt was induced in a laboratory experiment. We found that judges gave a higher rating of guilt in videos where people were seen frowning (AU4 Brow Lowerer) and touching their neck (Neck Touching). We used instances when judges reported seeing guilt to create 1s-window of interest and conduct our analysis only on those time windows of guilt. Doing this, we identified facial movements reliably associated with the perceived expression of guilty. Judges reported other emotions at the same time as guilt in only 14% of the guilt pinpoints. Moreover, pinpoints of guilt revealed specific facial movements that were not present in control videos. This made us fairly confident that the facial expressions identified were associated with the experience (perception) of guilt.

General discussion

In two studies, we aimed to identify facial movements and behavioural displays associated with the experience of guilt in humans. In the first study, we examined the production of guilt using a novel induction technique. In the second study, we examined whether others perceived guilt from the face of those experiencing guilt. We used an extensive, bottom-up coding scheme to identify facial patterns associated with the experience (production and perception) of guilt as part of a dynamic sequence of behaviour, combined with a robust bootstrapping method to analyse our data.

We found a positive relationship between the level of self-reported guilt and the extent this individual was judged as feeling guilty by others. This supports the idea that guilt could have evolved as an observable phenomenon with a potential communicative social function. The patterns identified in this experiment showed some consistency between what people do when feeling guilty and what people see when identifying guilt. Our first study showed that guilt was associated with frowning, lip stretching and neck touching [AU4 Brow Lowerer, AU20 Lips Stretch; 59], as well as looking towards the right (AU52 Head Right, AU62 Eyes Right), which was probably an artefact of the position of the computer. Our second study showed that the identification of guilt in others was associated with frowning, eyes widening, and neck touching [AU4 Brow Lowerer, AU5 Upper Lid Raiser, AU10 Upper Lip Raiser; 59], as well as looking down and sideways (AU54 Head Down, AU61 Eyes Left, AU62 Eyes Right, AU64 Eyes Down), another potential artefact due to the experimental set-up. Thus, it seems that in this study, guilt was associated with a non-verbal pattern of frowning and neck touching.

Using a bottom-up methodology allowed us not only to approach our question without any a priori assumptions regarding the results, but it also increased the likelihood that the movements identified in our studies (AU4, AU20, and neck touch) are associated with the experience of guilt and no other secondary moral emotion. Indeed, the “guilt” pinpoints identified by the judges (Study 2) were mainly instances of identification of guilt alone, with only 14% of the total number of guilt pinpoints associated with more than one emotion (see S1 Study of Table 2). This allowed us to focus our analysis on facial movements associated with the experience of guilt only. Moreover, even though guilt is often mistaken for embarrassment or shame, the embarrassed display has been characterised by the joint production of gaze down, controlled smiles, head turns, gaze shifts, face touches [44], and the occasional blushing [90]; and the typical face of shame was described with head and gaze movements down [43–45]. None of the movements we found associated with the expression of guilt were associated with those of other negative self-conscious emotions. During the AU selection process, most facial movements associated with either embarrassment or shame were discarded from further analysis, with the only exception of face touching. Face touch can emphasise embarrassment displays, but it is not necessary for the identification of embarrassment [44]. A previous study suggested a link between blushing and admission of guilt [91]; combining FACS analysis with thermal imaging techniques might have revealed changes in facial temperature in guilty participants, which could be unconsciously used by observers in their judgments.

This bottom-up methodology also diverges from previous research examining the facial display of guilt, which is why we may have found a more concrete candidate for the display of guilt. One notable previous study used a literature-based conceptualisation of the experience of guilt to present three candidates’ displays to their participants [8]. In that study, using a top-down approach, the participants were presented with displays selected based on previous literature, which associated the experience of guilt with the experience of self-contempt, sympathy, and pain. The authors tested whether their conceptualisation of guilt accurately described a facial display associated with the experience of the emotion. The results were not conclusive as the candidates’ displays were more often associated with emotions other than guilt [8]. A more recent study associated the experience of guilty feeling with increased skin conductance and gaze avoidance [92]. We did not find gaze avoidance (i.e. actively avoiding to look in another person’s direction) to be part of the facial signal of guilt, even though participants in the guilt condition looked down and around more than participants in the control condition. Yet, this could be due to our experimental design: participants in the guilt condition might have been looking down at the laptop more than people in the control condition. It is thus unclear in our design whether guilty participants avoided eye-contact or focused on an object associated to their wrongdoing (the laptop could be incriminated for the deletion of data on the USB stick, removing the fault from them).

Both the production and perception of guilt was associated with self-directed behaviour (i.e., scratching, neck or face touching), which are often classified as displacement behaviours, and are defined as a group of behaviours that appear irrelevant to the situation in which they are displayed, but can gain communicative value over time [61]. The production of such behaviours has been shown to increase in stressful, negative, situations [93,94]. Self-directed behaviours may be used when individuals try to distance and protect themselves from an unpleasant situation, acting as a short-term diversion of attention, which could, in turn, reduce the negative feeling associated to the situation at hand [93,95,96]. Self-directed behaviour could thus help regulate the level of stress associated with emotionally challenging situations [94], such as the guilt induction experienced by our participants in Study 1. Indeed, some studies have shown that self-directed behaviours are common in situations such as embarrassment [44], discomfort [20], and anxiety and guilt [97], which focussed on hand movements and found a correlation between the production of self-directed behaviours (i.e., scratching) and anxiety and guilt feelings. In our study, we found that the experience of guilt was associated with self-directed behaviours (neck touching), which appears to be in line with previous research. However, the production of self-directed behaviours could be due to the experimental design: participants were seated at a table, in front of a computer. However, the setup is unlikely to have elicited those movements, as participants in the control condition, also seated at a computer, did not display as many self-directed behaviours.

More recent conceptualisations of emotional experiences [27,28,35–37] argue for a less universal and omnipotent link between the experience of an emotion and behavioural outcomes. In an emotional context, multiple systems will be triggered (e.g., cognitive processes, physiological systems, motor expressions; [35]), leading to multiple behavioural outcomes (e.g. facial signals), one of which might be used by observers when responding to the situation [35]. As such, an individual feeling guilty might produce multiple facial signals, one of which will be more strongly associated with the subjective, constructed, feeling of guilt (e.g., frown, lips stretch and neck touching); an observer might perceive those facial signals and rely mainly on specific ones to interpret the emotional state of the guilty individual (e.g., frown and neck touching).

It is important to remain cautious in the interpretation of our data. We need to acknowledge that if neck touching was present more in association with feelings of guilt, only 12.5%of the individuals displayed neck touching. Self-directed behaviour, however, were displayed in over 64% of the individuals during the guilt induction. Even though few participants displayed neck touching, our results showed it is a significant signal of guilt. We need to consider the possibility that by reducing our dataset to 1-second windows, we could have excluded non-verbal signals important for the onset of the experience of guilt. By focussing on the apexes of the expressions, we might have lost secondary signals contributing to the reliable identification of guilty signals. Our results provide preliminary information regarding the non-verbal signals exhibited more in association with guilty feelings. A follow-up study, using a reduced ethogram focussing on the movements identified here could allow to reach a better agreement score between coders and thus increase the K’s alpha and the validity of our results [67,68]. We also need to consider the fact that providing contextual information might have influenced the judges in their decisions. To assess the impact of context, we conducted a follow-up study comparing the judgements made with and without contextual information provided [98]. Our judgement study also presents some linguistic limitations. Even if there are differences in the appraisal and behavioural outcomes between shame and guilt, it has been previously shown that English speaker use “guilt” and “shame” interchangeably [99]. To overcome this conceptual barrier, we conducted another judgement study, without providing contextual information [98,100]. We hope to gauge how the expression of guilt is perceived when no verbal/written content needs to be understood first. Moreover, to compare various judgement methodologies [emotion words vs action tendencies vs dimensions; 101], we conducted another follow-up study to help us have a better understanding of how people conceptualise the facial expression produced when experiencing guilt, using different types of words and classification methodologies [forced choice vs free labelling vs dimensions; 100]. This way, we hoped to introduce more variability in the emotional judgements, looking at patterns of mislabelling of guilty displays.

These are the first studies to look at the genuine expression of guilt and the perception of secondary emotion using spontaneous dynamic stimuli. Judges had to rely on genuine, dynamically presented facial expressions to recognise and rate emotions. They were exploratory studies, using simple analysis and focussing on the behavioural signals associated with a guilt-inducing situation. We have however collected more extensive data; now that we identified a facial signal associated with the experience of guilt, more in-depth analysis (such as a lens modelling [35]) would be an interesting step to further break down the mechanisms associated with guilt.

Our experiments support a drive towards a new scientific culture, studying facial expressions using novel approaches removed from the dichotomous debate about nature vs nurture [73,102]. Previous research extensively looked at the behavioural consequences of guilty feelings: it can promote directed action towards those who have been wronged [4], it can reduce prejudice behaviours [13] and increase generosity [6]. We focussed on the first reactions people have when realising they did something wrong and the guilty feelings emerge; we were able to identify reliable candidates characterising the experience of self-reported guilt. Building on this, we conducted a study to investigate guilty people’s propensity to repair the relationship, as well as the impact of a facial expression on the person wronged, i.e. the victim, reaction [103]. Together, our results suggest that guilt is expressed on the face and communicates the experience of guilt to others through a signal.