Search

Compared with the typical population, people with autism spectrum disorder often have difficulty recognising facial expressions of emotion. However, there also exists substantial variation in expression recognition within the typical population. It is important to understand this variation because difficulty in expression recognition is associated with negative social outcomes. One approach to understanding variation in expression recognition in the typical and autistic population is to measure ‘autistic traits’, which are personality traits relating to autism in the typical population, and their association with facial expression recognition. Autistic traits can be measured in three domains: social skills, social communication, and attention to detail. Our previous research suggests that expression recognition difficulty is primarily related to the two social domains of autistic traits, and minimally with attention to detail. This finding suggests that variation in these domains might also underlie difficulty in expression recognition in autistic people. Previous research further suggests that where associations between autistic traits and expression recognition do exist, these associations are mediated by a personality trait, alexithymia, which describes difficulty recognising one’s own emotions. This finding suggests that difficulty recognising expressions in autistic and typical people should not be fully attributed to autistic traits per se, but should be understood to substantially reflect co-occuring alexithymia. However, previous research is limited in that it has tended to use stereotypical, posed depictions of facial expressions which are often perceived as faked by observers. Evidence suggests observers respond differently to facial expressions which are perceived as posed versus those perceived as genuine. For example, genuine smiles may be associated with a stronger experience of social reward, and elicit different facial mimicry responses, in typical observers. Genuine and posed expressions also differ in physical characteristics such as symmetry. Finally, interpretation of posed expressions requires acceptance of a social façade, where interpretation of genuine expressions does not. Based on the differences between genuine and posed expressions, it is not clear whether the above findings for posed expressions would be observed in the context of naturalistic facial expression stimuli perceived as reflecting genuine emotion. We sought to investigate this by measuring associations between autistic traits, alexithymia, and facial expression recognition for both posed and naturalistic expressions in the typical population.

Trustworthiness is a primary dimension of facial first impressions (Oosterhof & Todorov, 2008; Sutherland et al., 2013) and we commonly make decisions about who to trust based on how trustworthy their face looks. While impressions of facial trustworthiness are powerful, they can be inaccurate, and so, adults also consider their social partners’ trait trustworthiness (i.e. how trustworthily they actually behave) when choosing who to trust. In fact, research using economic trust games suggest that adults’ trust in another person is usually the product of an interaction between how trustworthy that person’s face looks and how trustworthily they behave (Chang, Doll, van ’t Wout, Frank, & Sanfey, 2010; Hooper et al., 2018). However, it is not yet known whether a similar interaction exists in children’s trust in others. Therefore, this study will investigate the extent to which children’s trust behavior is influenced by their social partners’ facial trustworthiness and trait trustworthiness or whether their trust decisions are based instead on an interaction between these two trust cues, as they are in adults. We will use a repeated-interactions economic trust game where participants will indicate how much of their endowed assets they would like to invest with partners, who vary both in the trustworthiness of their facial appearance and their behavior – some partners will reciprocate participants’ investment behavior (fair behavior), whereas other partners will not (unfair behavior). This will allow us to determine whether children show adult-like impressions of trustworthiness, or whether trust impressions and behavior develop across childhood.

Romina Palermo (University of Western Australia, UWA), Linda Jeffery (UWA) and Brad Duchaine (Dartmouth) were awarded an Australian Research Council (ARC) Discovery Project award to investigate why face identification skills improve over childhood (DP140101743). Participants were 203 children aged 5 to 10 years, tested with a longitudinal design over three years. Here, we list the tests and measures that were included in the test battery, and note which papers the data has been used for.

Historically, our understanding of emotion perception has centred around the human face (Ekman, 1989, 1992b; Ekman & Oster, 1979), with a large body of research across psychology and neuroscience dedicated to understanding emotions inferred from faces (Adolphs, 2002; Barrett et al., 2019; Fernández-Dols & Russell, 2017). Accordingly, there are many databases of disembodied faces shown in isolation, on a blank background without any other context and most research has investigated how people respond to the face alone. However, when we see another person in real life, their face is a small part of our perception of them. For instance, we see how they position their body as well as the scene surrounding them, which may include other people. The fact that this information is missing is important as context plays a major role in the interpretation of faces (Aviezer et al., 2011; Barrett et al., 2011; Deffler et al., 2015; Kret et al., 2013). Fortunately, more research into the effects of stimuli external to the face (herein described as context) on inferred emotional states has emerged in recent years. The need now is to synthesise this evidence. The accumulating evidence almost invariably demonstrates that context influences our perception of emotional faces and that faces influence our perception of emotional contexts (Kret et al., 2013; Lecker et al., 2020). We refer to these as face-context interaction effects. The bidirectional nature of face-context interaction effects evidences that our visual perception of emotion is neither static (i.e., it is not determined by a specific set of cues in a face or a scene) nor is it solely reliant on a single visual category alone. Rather, the available emotional information from all visual cues is used to construct our emotion perception. As new and/or more information is made available to us, we update our prediction of (and confidence in) the likely emotional state of the person (Saxe & Houlihan, 2017). Therefore, emotion perception is contextual — our perception is constructed from the information available across the whole scene, including faces, bodies, the visual scene and other people. For the purposes of this review, context will only incorporate visual details, with other domains (e.g., auditory or learned affective processes) falling outside of its scope. Visual context is the most extensively researched modality of emotion perception which makes visual context a good starting point for a meta-analysis. In addition, visual context is comparable across the same modality to faces, limiting potential confounding variables. Faces influence the perception of emotion in bodies and visual scenes, and vice versa (Lecker et al., 2020; Reschke & Walle, 2021). However, little is known about precisely how much each influences the other, and what factors moderate the size of these effects. Investigating how much faces and context influence each other will provide a better understanding of the unique and shared contributions of each in visual emotion perception. To the best of our knowledge, there have only been two previous reviews of face-context interactions in emotion perception, one by Wieser and Brosch (2012) and another by Gendron et al. (2013). Although these reviews are highly cited, they are narrative reviews that provide a general summary of the literature at the time, without using a systematic search nor explicit and stringent inclusion/exclusion criteria. In addition, by nature of being narrative reviews, they also do not statistically evidence the relationship between faces and context. The proposed review will address these gaps by: • Updating the evidence to include the past decade. Our initial title search indicates this update will more than double the research pool. Nearly 60% of all PsycInfo titles that met our search criteria (Appendix A) were published post-2012. • Extracting data for meta-analysis. Extracting data from each publication and calculating effect sizes will allow for more accurate estimates of the true effects of face-context interactions. • Coding for variables that are theorised to moderate the size of face-context interaction effects. This review will also provide evidence for current debates in emotion perception theory. At present, there is considerable theoretical debate about whether emotion perception is best explained by the common view or constructionist theory. The long standing common view of emotion perception suggests that facial expressions provide universal and standard signals which allow us to interpret and understand the emotion another person is experiencing (Ekman, 1989, 1992b). In contrast, the constructionist view of emotion posits that our perception of emotion is predictive. That is, the emotion we perceive is a prediction of the most likely emotion from an array of various possible emotions based on the available information and our previous experience (Barrett, 2017). The constructionist view seems to align most with face-context interactions. As more information is available, that information is incorporated into our working perception of the emotional experience of others. If the face and context information match emotionally, there would be fewer feasible emotional states and so we would likely be more accurate and quicker in making our judgement. If the two information sources do not match emotionally, there would be more feasible emotional states, and we would therefore be less accurate and slower in our judgement (Barrett, 2017). Investigating face-context interactions provides an opportunity to better understand the relationship between faces and context, as well as to explore if their interaction provides evidence in support of the constructionist view of emotion perception.

Trustworthiness is a primary dimension of facial first impressions and we are able to form an impression about how trustworthy someone is after just 33 ms of exposure. Our first impressions can strongly influence social interaction, and can even impact financial and court decisions. These face-based trust impressions are thought to emerge early – even seven month old infants are sensitive to variations in facial trustworthiness, and trust impressions mature across childhood, reflecting adult-like patterns by 10 to 13 years. Recently, Swe et al., (2020) showed that electrophysical cortical activity is triggered by variation in visually presented facial trustworthiness cues in adults, using the fast periodic visual stimulation (FPVS) electroencephalogram (EEG) technique. FPVS is an objective measure that provides high signal to noise ratio data and allows trust impressions to be examined implicitly and non-verbally. These characteristics make it a valuable method of examining early neural activity underlying children’s trust impressions from faces. Therefore, this study will investigate whether eight to nine year old children can percieve trustworthiness implicitly by examining their neural discrimination response to facial trustworthiness. Specifically, we will use an oddball paradigm where trustworthy faces will be shown at a base frequency of 6 Hz, followed by an oddball untrustworthy face at 1 Hz (and vice versa for the trustworthy oddball face trials; following Swe et al 2020). If children’s visual cortex is sensitive to the change in facial trustworthiness, there will be an increase in amplitude of neural activity at the oddball frequency (1 Hz), and its harmonics. We will include both an upright and inverted face stimuli condition to control for low-level information. We will also collect adults’ neural discrimination responses to facial trustworthiness, and compare child and adult groups’ responses to examine any developmental trends.