SOCIAL ANXIETY AND THREAT
PERCEPTION
AN EVENT-RELATED POTENTIAL
STUDY
Master Degree Project in Cognitive Neuroscience
One year Advanced level 30 ECTS
Spring term 2020
Adithi Sutradhar
Supervisor: Oskar MacGregor
Examiner: Antti Revonsuo
1. Introduction ... 4
2. Background ... 6
2.1 Facial width-to-height ratio ... 6
2.1.1 FWHR and masculine facial features ... 7
2.1.2 The FWHR and intra-sexual competition ... 8
2.1.3 FWHR as a cue of threat ... 9
2.2 Social anxiety and cognitive explanations of threat perception ... 11
2.2.1 Social anxiety and perception of angry faces as threatening ... 13
2.3. The Electrophysiology of Facial Processing ... 14
2.3.1 Late Positive Potentials (LPP) ... 14
2.3.2 Social anxiety and electrophysiology ... 15
2.4 The Current Study Hypothesis ...16
3 Material and Methods ... 18
3.1 Participants ... 18
3.2 Stimuli ... 18
3.3 Study Design and Procedure ...19
3.4 Data Collection ... 20
3.5 Data processing ... 21
4. Analysis and Results ... 23
4.1 Experiment A... 23
4.1.1 Descriptive Statistics... 23
4.1.2 Subjective Facial Ratings with different facial morphological conditions ... 24
4.1.3 Electrophysiological Results ... 24
4.2 Experiment B ... 27
4.2.1 Descriptive Statistics... 27
4.2.2 Subjective facial ratings with high and low FWHR and emotion ... 27
4.2.3. Electrophysiological results ... 28
5. Discussion ... 31
5.1 LPP and Threat perception ... 32
5.2 The sample and effect size ... 32
5.3 Enhanced processing of the eye region in high socially anxious individuals? ... 33
5.4 Do high FWHR angry faces resemble social cues of threat? ... 34
5.5 Limitations ... 35
6. Conclusion ... 36
References ... 38
3
Abstract
The late positive potential (LPP) is an event-related potential (ERP) component associated with increased affective processing which seems to strongly respond to threats and to be sensitive to emotional faces. Some studies indicate that the LPP is modulated by anxiety symptoms, while others fail to find support for these observations. The facial width-to-height ratio (FWHR) is a facial-masculinity metric that refers to cheekbone width, divided by upper facial height (top of the lip to between the brows). Consequently, FWHR has by some researchers been proposed to serve as a cue of threat. For example, high FWHR and diverse emotional faces (e.g., angry faces) are perceived as more threatening than low FWHR faces.
Individuals with social anxiety are thought to be biased towards the threat. The literature has indicated that high FWHR faces in combination with angry facial expression can elicit larger LPPs compared to low FWHR and neutral faces. The current experiment investigated subjective ratings in addition to the LPP in response to high and low FWHR faces in combination with an angry and neutral expression, to examine how different facial morphology and affective cues influence the perception of threat to individuals with high social anxiety.
This data, in combination, suggests that high FWHR is a salient threat-related social stimulus that might have a firm influence on the perception of other peoples’ faces. Initial results do not support a significant relationship between increased LPP modulation in individuals with high social anxiety compared to individuals with low social anxiety. However, it opens up for discussion regarding how social anxiety should be approached in future LPP research.
Keywords: late positive potential, FWHR, social anxiety, threat perception
4
1. Introduction
Facial expression is the easiest way to externally anticipate emotional content in an individual. Thus, faces with different features are spontaneously interpreted as reflecting emotional, motivational, and social tendencies (Cloutier, Mason, & Macrae, 2005; DeBruine, 2005). What is more, facial expression conveys vital information on the emotional states of interaction with partners (Folster, Hess, & Werheid, 2014). Research from cognitive neuroscience and psychology shows that the human visual system is extremely responsive to facial cues and quickly processes cues such as identity, gender, age, and emotional expression, and monitors human-to-human interactions (Schupp et al., 2004). One of the factors influencing human facial processing that has received a lot of attention recently is the so-called facial width-to-height ratio (FWHR), a facial-masculinity metric that refers to cheekbone width, divided by upper facial height (top of the lip to between the brows), and associated with different behavioral traits. Pieces of evidence show that males with a high FWHR tend to behave more aggressively and dominantly and that this facial metric, as a cue of threat, certainly guides behavior (Carré & McCormick, 2008).
Multiple lines of research suggest that angry faces are detected more quickly than friendly faces among both neutral and emotional distracters (Schupp et al., 2004). Importantly, angry expressions – characterized by frowning brows, staring eyes, and a shut mouth with tense lips (Ekman & Oster, 1979) – seem to have a particularly strong effect on individuals. Observing an angry face has greater chance to inducing fear response in participants having social anxiety. Evidence from some studies (Broadbent & Broadbent, 1988; Eysenck, MacLeod, &
Mathews, 1987) shows that anxious individuals have an attentional bias towards threat-related stimuli in the environment. The level of internal emotional expression when responding to external emotional stimuli can be reflected in the late positive potential (LPP), which is a time-locked component that can be found using an event-related potential (ERP) with electroencephalographic equipment (EEG).(Hajcak et al., 2010).
This thesis sets out to further investigate the effects of social anxiety on the perception of different morphological and affective threat cues. The background section provides a comprehensive overview of this effect. First off, the FWHR will be introduced from an evolutionary perspective and linked to anger and threat detection. The influence of social
5 anxiety on threat perception, as well as other relevant perspectives, will be discussed within the context of so-called attentional bias, and threat perception will be related to an event- related potential (ERP) component known as the late positive potential (LPP). In the current study, the aim was to determine the potential interaction of FWHR and social anxiety. More specifically, the study investigated how socially anxious individuals interpreted high FWHRs as compared to low FWHRs. This was done by focusing on the LPP, a centro-parietal positive potential occurring around 400 ms post-stimulus, and continuing for up to several seconds, and thought to reflect affective processing. For experiment A it was hypothesized that if high FWHR faces are more emotionally salient, they would elicit higher LPP amplitude, and be subjectively rated as more threatening than low FWHR faces. For experiment B it was hypothesized that high FWHR faces in combination with angry expressions would instinctively be rated as threatening faces. To test these assumptions, participants were divided into two groups. In experiment A, 21 participants and experiment B, 25 participants were tested; first, subjective ratings of the level of threat perceived in the facial stimuli were collected, and second, each subject had their electroencephalograph (EEG) signal recorded during the perception of facial stimuli, to analyze the LPP responses to the different faces.
This thesis describes the study in question, by first giving a theoretical background to the research question, then providing an overview of its methods and results, before finally discussing its ramifications.
6
2. Background
This chapter will cover the initial theoretical aspects of the experiment. It will include a general explanation of faces and face recognition as well as research on the facial width-to- height ratio (FWHR). After this, theories of social anxiety will be outlined, and then social anxiety as a factor of threat perception will be described. Finally, two electrophysiological experiments related to facial processing, particularly focussing on the LPP, will be discussed concerning the current research question.
2.1 Facial width-to-height ratio
The facial width-to-height ratio (FWHR; bizygomatic width divided by upper-face height) is a sexually dimorphic characteristic of the face. The FWHR has been demonstrated to affect how a face is perceived, especially male faces (more than female faces). Initially, the metric was described by Weston et al. (2007). Considering the physical differences between the sexes, male body sizes are overall larger than female bodies including the face. More specifically, male faces develop a wider structure than females. The growth of sexually dimorphic facial features during puberty in males is driven by testosterone, resulting in the growth of the jaw, cheekbones, brow ridges, the length from the brows to the bottom of the nose, and facial hair (Enlow & Hans, 1996; Verdonck et al., 1999). High FWHR is perceived as more masculine and has been associated with higher levels of testosterone in puberty and adulthood (Gangestad, Thornhill, & Garver-Apgar, 2005; Lefevre, Lewis, Perrett, & Penke, 2013). Research indicates that this hormone is especially involved with cranial growth during puberty in males. A longitudinal study on boys with delayed puberty conducted by Verdonck et al. (1999) found a significant association when they administered testosterone to the subjects for a year. Initially, the facial measurements of the subjects with delayed puberty were smaller when compared to equally aged controls and cranial growth occurred with the administration of testosterone in the subjects. In sum, several bodies of research evidence suggest that the FWHR is a metric thoroughly compatible with male faces (e.g. Carré et al., 2009; Geniole et al., 2012). The FWHR is hypothesized to exist in males as a signal for dominance, aggression, and trustworthiness (e.g. Carré et al., 2009; Geniole et al.,2012;
7 Stirrat & Perrett, 2010). Therefore it is plausible that sensitivity towards FWHR has been selected over time (Carré, & McCormick, 2008; Carré, McCormick, & Mondloch, 2009;
Haselhuhn, Ormiston, & Wong, 2015). Male faces with high FWHR are perceived as more threatening and dominant by observers (Geniole et al., 2015). Thus, if high FWHR is a signal of aggressive and dominant behavior in an individual, it can be interpreted as a cue of threat, which means that FWHR is a significant signal for individuals to react to (Geniole et al., 2015).
2.1.1 FWHR and masculine facial features
The FWHR was first mentioned in Weston et al. (2007) as a sexually dimorphic facial feature that was greater in males than females and autonomous of changes in body size.Interestingly, Geniole et al. (2015) found that FWHR predicted body size which suggests that the behavioral tendencies and perceptual judgments related to FWHR could be related to body size after all. In earlier studies, connections have been found between men’s fWHR and direct aggressive behavior (e.g., Carré, & Mccormick, 2008) as well as indirect indications of aggression, e.g., proving untrustworthy in an economic game (Stirrat & Perrett, 2010). In line with Carré & Mccormick (2008), faces that were judged as more aggressive had a broader bizygomatic width and shorter distance between eyes and mouth, indicating a higher FWHR in aggressive faces. However, other facial features such as more prominent eyebrows, horizontally narrower eyes, larger nose, broader chin, as well as lower fatty deposits, especially on the chin and cheeks also correlated with perceived aggressiveness. Few facial features including high FWHR as well as a wider jawline and heavy brows are considered to be typically masculine traits (e.g. Gangestad et al., 2005). Windhager et al. (2011) found that men that were judged more dominant and masculine by women had a wider and more prominent lower jaw as well as smaller eyes, nose, and lips compared to the average man.
Social interaction and judgment occur primarily in the upper part of the face and eyes, which is why the FWHR may be one of the immediate metrics that are perceived and appraised (Geniole et al., 2015). Findings from research illustrate that in perceiving other peoples’
faces, both humans and other primates tend to direct their attention towards eye movements on medial features, e.g., the eyes, nose, and mouth (Althoff & Cohen, 1999 ). The study conducted by Dotsch and Todorov (2012) investigated that facial areas were used when people make judgments of dominance and trustworthiness of others. When participants made judgments of dominance and submissiveness, the facial areas used to make these judgments
8 included the regions around the eyes, on the brows, and the outline of the face, including the chin and jaw-line. Considering that patterns of eye movements are fundamentally related to subsequent recognition, the study conducted by Henderson et al. (2003) found that recognition memory for faces was significantly impaired when participants were prevented from generously making saccades to observe facial features. One of the potential mechanisms for this link is that eye movements may permit the encoding of relations among facial features and permit precise feature details to be encoded. It seems that eye movements are crucial for face recognition. Changes in the eye scanning behavior during the initial observation of faces have an impact on face recognition memory.
2.1.2 The FWHR and intra-sexual competition
From an evolutionary perspective of sexual selection, the development of some specific features, e.g., facial features, is advantageous for accessing mates (Darwin, 1871). This may have lead to individuals developing features attractive to the opposite sex (intersexual selection) or same-sex peers in the competition for access to the opposite sex (intrasexual selection) (Darwin, 1871). Some other research findings also suggest male facial structures evolved not as features to attract mates but to intimidate and compete against same-sex competitors (Geniole et al., 2015). It is thought to have been primarily developed as an intra- sexual competition mechanism for male dominance, to better be able to compete against other members of the same sex for reproductive success (Geniole et al., 2015). Regarding inter- sexual selection, the relationship to the FWHR is less clear. Studies have evaluated the possible relationship between the FWHR in males – or masculinity in general – and attractiveness with mixed results. Some studies have found no relationship between FWHR and attractiveness (Geniole et al., 2012). Additionally, other studies have reported negative correlations between attractiveness ratings and high-FWHR pictures. A meta-analysis by Geniole and colleagues found to be a stronger correlation in experiments where the higher proportion of participants were female raters. Females rated high FWHRs as being more unattractive than did males, even though the effect was true for both sexes.This result provides further support for the FWHR’s possible function as an intra-sexual competition mechanism primarily, rather than for attracting mates (Geniole et al., 2015).
Further research has found that men with greater FWHR behave as more aggressive and dominant, and describe themselves as more aggressive than men with smaller FWHR
9 (Geniole et al. 2015). Men with higher FWHR are perceived to be less intelligent and friendly, and more aggressive (Hehman, Leitner, & Gaertner, 2013), untrustworthy (Stirrat&
Perrett, 2010), and aggressive (Carré, McCormick, & Mondloch, 2009; Geniole, Keyes, Mondloch, Carré, & McCormick, 2012). Regardless, in recent literature, FWHR has been rated as more threatening and indicating dominant behavior within human males, and FWHR also operates as an evolved cueing system of intra-sexual threat and overall aggressiveness in men (Geniole et al., 2015). A study conducted by Carré & McCormick (2008) investigated the FWHRs effect among participants. The results from this study first indicated that FWHR among men predicted reactive aggressive behavior as well as dominant behavior. These results were then followed up with a study among hockey players. FWHR was correlated positively with the time spent in the penalty, and the authors then argued that FWHR serves as an honest signal for aggressive tendencies among men.
2.1.3 FWHR as a cue of threat
However, the question is why high FWHR is interpreted as a cue of threat. High FWHR resembles components of facial expressions of anger and is therefore considered threatening (Eisenbruch, Lukaszewski, Simmons, Arai, & Roney, 2017). Moreover, Angry faces usually involve raising the chin and lowering the brow, which produces a configuration that both resembles high FWHR and enhances perceptions of formidability (Sell, Cosmides, & Tooby, 2014). Thus, angry facial expressions may serve to emulate higher FWHR (Carré et al., 2009). Interestingly the relationship between FWHR and threat perception has been observed in other primates as well. Brown capuchin monkeys with smaller FWHRs were found to be less assertive and dominant than those with larger FWHRs (Lefevre, Wilson, et al., 2014). In another study, FWHRs of more dominant species of macaques were found to be larger than the FWHRs of more tolerant species of macaques (Borgi & Majolo, 2016).
A problem with this view, however, is explaining the fact that observers accurately can predict a man’s propensity for aggression solely based on FWHR. On the other hand, if wider-faced men are regularly perceived as angrier than narrow-faced men due to the similarity of a high FWHR with an angry face, they might receive submissive, mistrustful, and aggressive treatment. Over time, this could potentially lead to the acquisition of more exploitative and behavioral strategies (Eisenbruch et al., 2017). This model of explanation would essentially attribute the relationship between FWHR, behavior, and observer judgments to social forces. In sum, it is difficult to know if the relationship between judgments of threat
10 and FWHR is due to an evolved mechanism in humans or simply just an effect of the resemblance of high FWHR with an angry face.
A study conducted by Hansen & Hansen (1988) demonstrates how angry faces are interpreted as a cue of threat. In the experiment, participants were tasked with detecting the presence of faces with different emotional valence in a crowd across numerous experiments. In the first experiment, participants were tasked with detecting stand-out faces in a crowd, where the faces had either happy, neutral, or angry expressions. The participants were tasked to determine if there were faces that had angry or neutral expressions if the crowd consisted of happy faces. The results from the first experiment illustrated that participants were more faster and accurate when the stand-out faces had an angry expression in both happy and neutral crowds than happy faces in angry crowds or neutral faces in angry crowds. However, neutral faces in happy crowds lead to similar effects as angry faces, as well as happy faces being identified similarly fast and accurate in both neutral and angry crowds as neutral faces in happy crowds. This suggests a bias for identified faces within crowds, with angry faces being processed faster and more accurately than other kinds of faces when unique to their crowd.
A third experiment was conducted to further investigate the effect in which the researchers investigated multiple hypotheses of angry faces being favored in face processing. The experiment set out to show that angry faces should be identified faster than happy faces regardless of crowd size. They tested the hypothesis that the time needed to identify angry faces should not be heavily influenced by the number of faces in the crowd as well as the time to identify happy faces should be more influenced by the number of faces in the crowd. Thus, the results from this study provide a potent demonstration of angry faces being privileged over other kinds of faces. To put this in an evolutionary context, note that it would be favorable for an organism to detect threats more rapidly than other kinds of stimuli and that this effect applies to social situations as well, given that angry faces serve as a cue of threat (Hansen, & Hansen, 1988).
In summary, much support has been found for FWHR being an honest cue of threat (e.g., Geniole et al., 2015) with other features such as brow height and jaw width also being known to influence threat perception. Both FWHR and angry faces have been shown to function as reliable signals of social threat. Thus, when an individual is presented with a face that either has an angry expression or a face with a high FWHR, or a combination of both, threat-related
11 mechanisms within the individual should be activated. Together with FWHR and angry expressions previously discussed in this thesis, threatening faces should result in a larger emotional response in individuals with high social anxiety rather than individuals with low social anxiety.
2.2 Social anxiety and cognitive explanations of threat perception
Social anxiety is a persistent and excessive fear of being evaluated by other people, and the avoidance of situations involving inspection and possible negative evaluation (American Psychology Association, 2013). Some research findings suggested that anxious individuals are surprisingly sensitive to threat-related stimuli in the environment and prone to information processing biased in consideration of threat-related stimulations compared to non-anxious individuals. Some evidence indicates that attentional bias towards threat plays a significant role in the etiology and maintenance of anxiety disorders. According to Rapee and Heimberg’s cognitive model (1997), individuals with social anxiety are hyper-vigilant in monitoring their external environment for signs of negative evaluation from others, which is considered as the core feature of anxiety.
To support his idea, some studies were conducted by researchers with both clinically anxious individuals (MacLeod, Mathews, & Tata, 1986; Mogg, Mathews, & Eysenck, 1992) and participants with high levels of self-reported anxiety (Broadbent & Broadbent, 1988;
Eysenck, MacLeod, & Mathews, 1987) with a series of studies using a variety of selective attention tasks to independently support the subjects’ statements. However, some researchers assume that one of the anticipating factors is anxious people might be sensitive and turn their attention towards threatening stimuli. Thus, they would possibly perceive high FWHRs as more threatening stimuli than low FWHRs. The study conducted by Bar-Haim (2010) suggested that high anxious individuals have an attentional biased tendency towards threat- related stimuli particularly reinforced by their sensitive processing system.
There are different cognitive explanations of anxiety, differing about concerning the roles they assign to biases in attention, interpretation, memory, and judgment in the etiology and maintenance of anxiety. The schema theories of anxiety describe the, (e.g., Beck, 1976; Beck
& Clark, 1997; Beck, Emery, & Greenberg, 1985) cognitive processing is guided by schemas that largely determine how information is attended to, interpreted, and remembered. Anxious individuals are thought to be biased towards the threat, and as a result, threat-related material
12 is favored at all stages of processing, which includes attention and stimulus encoding as early processes and memory and interpretation as later processes.
Kendall's (1987) theory is one that interrelates the relationship of childhood anxiety, pathological fear, and anxiety in the later stage arising from the chronic overactivity of schemas around themes of danger and death. Thus these activated schemas are predicted to focus on processing resources on threat-related perception and information. In Kendall’s theory, cognitive distortion is a key feature, and these distortions pertain to cognitive processes that are biased or erroneous and therefore yield dysfunctional and maladaptive thoughts and behaviors (Daleiden & Vasey, 1997).
The unified theory of anxiety (e.g. Eysenck, 1997) suggests that individuals with anxiety selectively attend to threatening cues: individuals with social anxiety carefully observe the facial expressions of others for the negative evaluation of signs, disapproval, or rejection and associated such faces in the social environment of subjects (attentional bias).
Recent models of anxiety emphasized that high levels of anxiety are precipitated by some indicators such as allocation of time towards attention bias and threat perception (e.g., Eysenck, 1992; Mathews & Mackintosh, 1998; Mogg & Bradley, 2008). Some researchers have proposed that during early and automatic stages of processing anxious individuals tend to direct their attention toward threat whereas during more strategic and later stages of processing, they tend to direct their attention away from the threat (Williams et al.,1988) and others (Amir, Foa, & Coles, 1998; Mogg et al., 2008).
Likewise, pieces of evidence from some research groups have demonstrated that anxiety has a partial impact on initial detection of threat but a prolonged effect to keep attention on the source of threat (Fox, Russo, Bowles, & Dutton, 2001; Yiend & Mathews, 2001). A study conducted by Muris et al. (2000) indicates that participants express early detection of threat and a high level of negative feelings and cognitions. That is, they have proposed that a delay in disengaging from threat stimuli might be the primary attentional difference between anxious and non-anxious individuals. However, at the empirical level of studies, threat-related biases in anxiety are somewhat confusing, as there are also contradictory findings, and there are no fully satisfactory explanations of the cognitive mechanism of threat perception and anxiety.
13 2.2.1 Social anxiety and perception of angry faces as threatening
Research findings have attested that selective attention towards threatening facial expressions precipitates and maintains symptoms of social anxiety. To explain the cognitive component of anxiety, threat bias is considered as a core element and associated with prolonged clinical anxiety disorders and their treatments (Cisler & Koster, 2010). One of the recent studies conducted by Yoon et al. (2014) applied Signal Detection Theory (SDT) to obtain discrimination measures and separate perceptual sensitivity from response criterion. This theory clarifies and regulates the question of whether anxious individuals have an advantage at identifying threats in truly threat-related expressions (high sensitivity to anger), or whether they are simply more likely to respond that, regardless of the actual emotion, all faces look threatening. However, Yoon et al. (2014) reported that social anxiety was related to both greater sensitivity to mild angry expressions and a response bias towards labeling other expressions as angry. Following this idea, Gutiérrez-García et al. (2017) also found the relevant correlation between attentional bias and angry faces. In their study, participants categorized the expressions, where social anxiety was associated with enhanced detection of anger and disgust at low-intensity levels, relative to non-anxious controls. The reason for this behavior may be the nature of social anxiety as specified above, with fear of negative evaluation as a feature. A negative evaluation is more likely to come from people showing anger and disgust, whereas the other expressions do not convey a direct threat. The special sensitivity to anger and disgust is in line with prior research (anger: Bell et al., 2011; Yoon et al., 2014; disgust: Heuer et al., 2010).
Moreover detecting threatening faces and responding appropriately to them in terms of presenting signals was critical for survival during evolution, several authors have suggested that angry faces may represent generic signals of threat. Therefore, threatening faces may be fear-relevant not only for individuals with social anxiety but for all humans (Öhman, 1986).
However, according to this suggestion, threatening faces should be detected more rapidly and more efficiently than neutral stimuli (Öhman, Lundqvist, & Esteves, 2001). In support of this theory, generalized social phobia and non-anxious controls both detected angry faces faster than happy ones in a neutral crowd in a face-in-the-crowd paradigm; however, the generalized social phobic showed even faster reactions to angry faces than the controls (Gilboa- Schechtman, Foa, & Amir, 1998).
14 Based on these studies, it is not settled how threat perception distortions are associated with different types of anxiety symptoms, but it is clear that individuals with social anxiety are prone to judging social stimuli as threatening, Likewise, these behaviors preserve and upgrade social fear which can develop avoidance behavior in later stages in development. However, some evidence found that socially anxious individuals interpreted images of ambiguous social scenarios more negatively or less positively than non-anxious ones (Mobini, Reynolds, &
Mackintosh, 2012; Morrison & Heimberg, 2013).
2.3. The Electrophysiology of Facial Processing
EEG (electroencephalography) is a non-invasive procedure where the summed activity of large populations of cortical neurons is recorded through the scalp using scalp electrodes (Luck & Kappenman, 2011). Generally, EEG is a very useful tool in assessing emotional processing in anxiety (MacNamara, Kappenman, Black, Bress, & Hajcak, 2011). However, because EEG is a rather rough measure of brain activity, it cannot be used to measure specific neural processes on a fine-grained level (Luck, 2011). While the term ”EEG” mainly refers to the continuous observation of larger waveforms, the ERP (event-related brain potential) methodology lets the researcher focus on specific, time-locked epochs within the recorded EEG signal. The most important advantage of the ERP technique is that it reflects event- related neural activity more directly by enhancing the signal-to-noise ratio (SNR). This enables the researcher to study changes in amplitude, brought on by a stimulus, with millisecond precision concerning an event of interest, such as the onset of a target stimulus.
For this reason, ERPs are especially useful for studying neural responses to facial stimuli in participants in a non-invasive, time-sensitive, and reliable way. Neural activity is increasingly used in addition to behavioral measures in studies of anxiety and attention biases toward threatening stimuli.
2.3.1 Late Positive Potentials (LPP)
The late positive potential (LPP) is an event-related-potential (ERP) component that reflects facilitated attention to emotional stimuli. (Schupp et al., 2000) Particularly, it is thought that emotionally arousing (both pleasant and unpleasant) stimuli elicit greater LPPs (Schupp et al., 2004). It develops at around 300-400 ms after stimulus onset over centro-parietal areas of the brain and may last for several seconds, depending on the duration of the stimuli (Cuthbert,
15 Schupp, Bradley, Birbaumer, & Lang, 2000). Moreover, emotional or neutral stimuli like pictures, faces, or words (Cuthbert et al., 2000; Schupp et al., 2000, 2004) with high evolutionary relevance (e.g., threat) are found to be linked to increased LPP amplitudes (Schupp, Flaisch, Stockburger, & Junghöfer, 2007). The LPP has been persistently shown to be sensitive to emotional scenes. Its amplitude is also enhanced for emotional facial expressions (Schupp et al., 2004) compared to neutral faces. Further, emotional modulation of the LPP is remarkably stable and does not appear to habituate away over time (Codispoti, Ferrarib, & Bradley, 2006. Moreover, the evidence from previous literature suggests that a larger LPP amplitude would be elicited during the processing of angry compared to happy or neutral faces in general (Duval et al., 2013). The larger LPP magnitude appears during the processing of prototypical angry faces compared to prototypical happy and neutral faces ( Schupp et al., 2004). Thus, in the present study, the LPP is of particular interest because of its sensitivity to emotional content (Cuthbert, Schupp, Bradley, Birbaumer, & Lang, 2000).
The notion of angry faces being treated differently than other kinds of faces is further backed up by electrophysiological data (Schupp et al., 2004). A study conducted by Schupp et al., (2004) investigated the differences between threatening, neutral and friendly faces. In the piloting stages of this study, angry faces were rated with higher emotional arousal within the subjects and higher unpleasantness than other faces. The results from the EEG demonstrated that angry faces augmented the late positive potential (LPP) in contrast to neutral and friendly faces. The link between angry faces inducing higher emotional arousal as well as augmenting the LPP is consistent with previous literature. The LPP has also been illustrated to be increased for stimuli higher in emotional significance (Cuthbert et al., 2000). Additionally, in early ERPs studies, processing of affective stimuli repeatedly showed larger parietal late positive potentials (LPPs) when subjects were processing emotionally arousing (pleasant and unpleasant) stimuli as compared to neutral ones (Cuthbert, Schupp, Bradley, Birbaumer, &
Lang, 2000). Thus, if angry faces are more unpleasant and arousing for individuals with social anxiety than for healthy controls, this should be reflected in enlarged LPPs in response to angry faces.
16 2.3.2 Social anxiety and electrophysiology
The ERP experiment conducted by Moser et al. (2013) found that late, positive ERPs were related to social anxiety during the processing of emotional faces, while several other studies have reported negative results. None of the researchers Erika et al. (2016) or Mühlberger et al. (2008) found any effects involving the LPP on emotional faces in clinical or sub clinical participants. The study conducted by Mühlberger et al. (2008) found that it is difficult to distinguish between photographs of real people and artificially created faces, which confounded their results. There is as yet no significant evidence that the P3 (another late positive ERP component) plays a major role in the processing of threatening faces in social anxiety. Irrespective of whether the late ERPs truly signal, to some extent, elaborate frontal processes such as memory or judgment, these inconsistencies mirror behavioral and self- report data showing that individuals with social anxiety, in general, do not differ from controls in interpretation and judgment of emotional faces.
The literature shows significant evidence that anxious people tend to allocate more attention to threatening stimuli than to neutral stimuli. But due to limited work on the topic, there exist no strong predictions about how differences in FWHR are associated with anxiety and threat perception. If FWHR provides a cue of threat, as proposed by some researchers (Geniole et al., 2015) it is possible that the LPP could be affected, perhaps signaling an observer’s subjective feelings of threat. In the present study, an experiment will be carried out to measure directly whether there is any significant relationship between high FWHR as a hint of threat perception and high social anxiety.
2.4 The Current Study Hypothesis
The study aims to investigate if individuals with social anxiety show a stronger reaction when they are exposed to high FWHR faces if they considered those faces as threatening. To enable easier comparisons, participants will be divided into two groups based on their Social Interaction Anxiety Scores (SIAS). Cues of threat will be presented to participants as different faces that differ along various dimensions believed to affect the LPP, described in further detail below. In social facial processing, high FWHR is considered a threatening factor.
Therefore I expect to see a typical threat reaction to high FWHR. So far, there are no results in the literature that support a strong link between typical threat responses to high FHWR
17 faces and high levels of social anxiety. One of the scant results that exist so far is that individuals with social anxiety tend to have stronger reactions to threatening stimuli.
Therefore, if it is so that FWHR is a threatening factor in social facial processing, we can hypothesize that individuals with social anxiety would have stronger reactions to high-FWHR faces, compared to individuals without social anxiety. The present experiment tests this hypothesis by investigating emotion-related electrophysiological responses, in particular the LPP, which is commonly used to measure the emotional salience of stimuli. If this hypothesis turns out to be true then this would provide evidence that a high FWHR is a salient threat- related social stimulus with a strong influence on the perception of other peoples’ faces. If individuals with high social anxiety appeared to show no stronger reactions to high FWHR faces, then this would suggest that the FWHR is not a reliable predictor of perceiving other individuals as threatening.
18
3 Material and Methods
3.1 Participants
In the present study, we recruited 46 participants (after artifact rejection); 21 for experiment A (7 female and 14 male) and 25 for experiment B (11 female, 14 male). Initially, the total number of participants was 58 (30 for experiment A and 28 for experiment B). The participants were recruited primarily through word of mouth at the university and at the international student residency. The participants were current students of the University of Skövde and were all fluent in English. All participants reported being free of neurological or psychiatric disorders, having no dyslexia as well as normal or corrected-to-normal vision, and being right-handed. They had no record of epilepsy and were not on any type of neuropsychiatric medications. All participants gave their informed consent following the Declaration of Helsinki before participation.
3.2 Stimuli
For both experiments, stimuli were presented on a 23-inch screen with a 1920 x 1080 pixel resolution (HP Compaq LA2306x). The stimuli, which were also 1920 x 1080 in size, were shown for 1500 ms, following a 400 ms fixation point, and followed by a blank screen with a jittered latency of 300-500 ms, for a total stimulus onset asynchrony of 2200-2400 ms. The facial stimuli were made in FaceGen Modeller Core v3.18 (Singular Inversions). The faces were then placed on a grey background (RGB index: 152, 152, 152), as shown in Figure 3.1.
The study consisted of eight different male facial images in both experiments. For experiment A images with different facial morphologies, varying across three factors (brow height, cheek width, jaw width) with two levels of each factor (weak and strong), and in experiment B with all possible combinations of neutral versus angry faces, young versus old faces, and narrow versus broad faces. Note that, for experiment A the jaw width condition and experiment B the age condition was not analyzed in this thesis, but was instead used in other students’ theses.
19 .
Figure1. FaceGen Modeller faces. A) Low FWHR faces, with a neutral expression (left) and an angry facial expression (right). B) High FWHR faces, with a neutral expression (left) and an angry facial expression
3.3 Study Design and Procedure
Upon arrival, participants were greeted and then asked to read the study description and provide their informed consent. They were asked to choose a playlist/artist/album on Spotify to play for the duration of the experiment if they had not already provided specifics before arriving. After providing consent, the participants were instructed to sit down in the experimental room, and head measurements were conducted. Then they were seated in a sound-attenuated and dimly lit EEG lab with an approximate eye-screen distance of 105 cm.
This first section (Section A) for both experiments included three questionnaires and a face- rating task. The order of the questionnaires, and the questions of each questionnaire, was randomized between participants. The procedure of the main experiment began after the participants’ responses to the questionnaires were recorded. Of the three questionnaires, two were used for other students’ theses, while I focus on the Social Interaction Anxiety Scale (SIAS; Mattick, & Clarke, 1998) for anxiety level measurements of the participants. The SIAS consists of 20 items that are rated on a 5-point Likert scale ranging from 1 (“not at all characteristic of me”) to 5 (“extremely characteristic of me”). The original scale ranges from 0-4. I changed the scale range so that the version used here ranges from 1-5, in order to make it compatible with the questionnaires used for other students' theses. Since this change does
20 not affect the order of the data, and we only used the data to perform a simple median split, this change was unproblematic. Items are self-statements describing reactions to social interactions as individuals or groups. The SIAS is scored by summing the ratings (after reversing the 3 positively worded items).
The facial rating task included all eight faces, which were rated on a 9-point Likert scale about threat rate (1 = least threatening; 9 = most threatening). The faces for the face-rating task were presented for 1500 ms. The experimenters gelled the electrodes with conductive gel while participants were occupied with the questionnaires. The participant answered the questions using the numbers on a keyboard.
When the participants were finished with the first section, new instructions were presented to prepare them for the second section (Section B) for both experiments. This is considered the main experiment and consisted of 12 blocks, where each block consisted of the eight faces shown eight times each in a pseudo-randomized order, with no faces being shown twice in a row, for 64 trials in total per block. After each block, there was a short break. After the sixth block, participants were given a longer break. In addition to the 64 trial faces, there was a one-in-eight chance at the start of each trial that the participants would see one of the eight faces but with closed eyes. They were instructed to respond as quickly as possible to these closed-eye faces by pressing a button on a hand controller. The purpose of this reaction task was to keep subjects alert throughout the experiment. The results of the closed-eye trials were not counted as part of the 64 trials per block and were excluded from further analysis.
3.4 Data Collection
In the present study, brain activity was recorded using 17 active Ag/AgCl electrodes. 13 of these were placed in a stretchable cap (g.GAMMAcap) and positioned according to the International 10/20 Placement System at the following locations: AF3, AF4, Fz, FC3, FC4, Cz, CP1, CP2, CPz, Pz, P5, P6, and Oz. The other four electrodes were placed at the right and left mastoid (for subsequent offline re-referencing) and at the external canthus and suborbital of the right eye (to capture ocular movements). Electrodes were online-referenced to the right mastoid, and FPz served as ground. The active electrode impedances were transformed by the system to output impedances of about 1 kOhm.
21 In the experiment, the EEG data were acquired in MATLAB v8.5.1.281278 (MathWorks, 2019) with a g.USBamp amplifier (g.tec). It was sampled at 256 Hz, and filtered online with an eighth-order Butterworth low pass filter with a half-power (-3dB) cutoff at 60 Hz, by an internal digital signal processor within the amplifier.
3.5 Data processing
The data processing offline analysis was performed using the toolboxes EEGLAB v13.6.5b (Delorme, & Makeig, 2004) and ERPLAB v7.0 (Lopez-Calderon, & Luck, 2014) in MATLAB. Continuous EEG data were re-referenced to the average of the mastoids and filtered with a 180th-order stopband notch filter at 50 Hz, to remove line noise.
As a pre-processing step for removing artifacts by Independent Component Analysis (ICA), the data were filtered with a second-order Butterworth bandpass filter with a half-power (- 3dB) cutoff at 1 and 30 Hz (the higher highpass filter settings are recommended for ICA analysis). The EEG data were then segmented into epochs of 1900 ms, with a 400 ms pre- stimulus baseline and 1500 ms post-stimulus. Epochs exceeding three standard deviations above the joint electrode probability activity limits were rejected, after which ICA was run.
The Multiple Artifact Rejection Algorithm (MARA) was used to automatically identify ICA components reflecting artifacts (Winkler, Haufe, & Tangermann, 2011).
The ICA weights were then transferred back onto the pre-processed, unepoched data (that had only been subjected to the notch filter), and the relevant MARA-detected components were removed from this data. Subsequently, these data were filtered with a second-order Butterworth highpass filter with a half-power (-3dB) cutoff at 0.1 Hz, and, as before, segmented into epochs of 1900 ms, with 400 ms pre-stimulus baseline and 1500 ms post- stimulus. Step-wise artifact rejection was performed in ERPLAB 7.0 (all epochs containing step-like activity greater than 100 μV in a moving window of 200 ms with a step size of 20 ms were rejected). For experiment A, 9 subjects with an artifact detection rate above 20% in total (across all conditions) were rejected from further analysis. For the remaining subjects (n=21), epochs were averaged for each participant, and each experimental condition, and lowpass filtered at 30 Hz to aid visual inspection. Moreover, for experiment B, 3 subjects with an artifact detection rate above 20.0% in total (across all conditions) were rejected from further analysis. For the remaining subjects (n= 25, epochs were averaged for each participant and each experimental condition, and lowpass filtered at 30 Hz to aid visual inspection.
22 In line with a large body of prior research, which found that the LPP is most pronounced over central-parietal sites (Foti, & Hajcak, 2008; Hajcak et al., 2009; Wieser et al., 2014), the LPP was quantified across a cluster of central-parietal electrodes (Cz, CP1, CP2, CPz, Pz) as a function of the condition in two time windows following stimulus onset: 400-1000 ms, and 1000-1500 ms.
23
4. Analysis and Results
Statistical analysis was run using IBM SPSS Statistics 25 for mean amplitude data on subjects with adequate data quality. For experiment A, within-subject conditions were strong vs. weak brows, strong vs. weak cheek, and early vs. late LPP. For experiment B, within-subject conditions were high vs low FWHR, angry vs. neutral expression, and early vs. late LPP. For both experiments, between-subject factors were total anxiety scores, measured by the Social Interaction Anxiety Scale, with high vs. low anxiety scores relative to each other. All relevant conditions were subject to a mixed-model 2x2x2x2 repeated-measures ANOVA. Any significant effects were subject to independent samples t-tests as a post-hoc measure. The reasoning behind dividing the LPP into two different time-windows is twofold: to ensure comparability with other literature on the subject, and because there is reason to believe that different kinds of stimuli affect different parts of the LPP in different ways (Weinberg &
Hajcak, 2010).
4.1 Experiment A
4.1.1 Descriptive Statistics
The total number of participants (n=21) were divided into two groups based on anxiety scores, one group of high anxiety scores (n=10) and one group of low anxiety scores (n=11).
The division between the two groups was a median split, i.e. all participants with a score higher than the median were selected for the high anxiety group and all participants with a score below the median were selected for the low anxiety group. The anxiety scores differed between the two groups with the high anxiety group having a mean and standard deviation (SD) of 59.20 ± 6.9 and the low anxiety group having a mean and SD of 44.36 ± 6.6. To make sure the two groups were different enough on anxiety scores from each other, an independent t-test was conducted between the two groups. The total anxiety scores were shown to be significantly different from each other between the two groups (P < 0.001).
24 4.1.2 Subjective Facial Ratings with different facial morphological conditions
Figure 2 shows the subjective rating of faces with strong vs. weak brows and cheeks between anxiety groups of the total sample. A 2x2x2 repeated measures ANOVA (including the factors brow, cheek, and anxiety) was calculated to test for a significant difference between values. The results indicate that brows are the only factor that exhibits significant differences (F(1,19)=35.57, P < 0.01), while none of the other factors were found to be significant (P >
0.05).
Figure 2. Means of face ”threateningness” ratings for different facial morphologies: weak brows weak cheeks, weak brows strong cheeks, strong brows weak cheeks, strong brows strong cheeks, separate for the high and low social anxiety groups. Bars represent mean value +/- 1 SD.
4.1.3 Electrophysiological Results
To test the hypothesis and to determine if there were any significant differences in the different facial morphological components, first, a 2x2x2x2 repeated-measures ANOVA was conducted on the LPP (400-1500ms) for the average of the electrode cluster of Cz, CP1, CP2, CPz, Pz, with the facial variables (time, brows and cheek) and anxiety level (high vs. low
25 anxiety group) as factors. The interaction effect between the anxiety group and any of the other factors was not found to be significant (P > 0.05). However, some other effects did show significance. An almost significant interaction was found between brow and cheek (F(1,19) = 3.842, P =.06), as well as a three-way interaction between time, brows and cheek (F(1,19) = 3.741, P =.06). Moreover, for the early LPP, a significant two-way interaction was found between brow and cheek (F(1,19) = 8.215, P =.01), and one-way interaction for brows (F(1,19) = 4.663, P =.04).
26 Figure 3: Grand average event-related potential waveforms recorded from sites CPz, Pz, CP1, CP2, Cz for the high and low anxiety groups for faces with weak and strong brows and weak and strong cheeks. The time window is divided as early (400-1000 ms) and late (1000-1500 ms).
As suggested by Kujawa et al. (2015), effects from anxiety could be observed in the latter part of the LPP, but although visual inspection of the plotted waveforms between the groups revealed a noticeable difference in the LPP amplitudes between the different conditions in the early stage (see Figure 3), it was not significant and may have occurred by mere chance.
However, eyebrows are a rather subtle facial feature, so maybe the effect is more pronounced for stronger stimuli. The waveforms in Figure 3 illustrate possible effects occurring in the P3 time window of the gathered ERPs.
Next, as there were no significant differences between the high and low anxiety groups, new waveforms were created by binning all subjects across both anxiety groups together.
27 Figure 4: Grand average event-related potential waveforms recorded from sites CPz, Pz, CP1, CP2, Cz for faces with weak and strong brows, and weak and strong cheeks. The time window is divided as early (400-1000 ms) and late (1000-1500 ms).
4.2 Experiment B
4.2.1 Descriptive Statistics
The total number of participants (n=25) were divided into two groups based on anxiety scores, one group of high anxiety scores (n=14) and one group of low anxiety scores (n=11), as in experiment A. Again, a median split was used to define the group boundary. The anxiety scores differed between the two groups with the high anxiety group having a mean and standard deviation (SD) of 59.23 ± 6.3 and the low anxiety group having a mean and SD of 41.50 ± 6.02. An independent t-test between the two groups showed that the total anxiety scores were significantly different from each other (P < 0.001).
4.2.2 Subjective facial ratings with high and low FWHR and emotion
The subjective ratings of the faces with different facial morphology and emotion are illustrated in Figure 4. The figure summarizes the descriptive information of the ratings of the faces as high and low FWHR, the emotion between anxiety groups of the total sample. A 2x2x2 repeated-measures ANOVA was calculated to check for significant differences between values. The results indicated that only a three-way interaction between emotions, FWHR, and anxiety exhibited significant differences (F(1,23)=7.88, P=0.01), while none of the other factors were found to be significant (P > 0.05).
28 Figure 5: Means of face ”threateningness” ratings for low FWHR neutral, high FWHR neutral, low FWHR angry, high FWHR angry, separate for the high and low social anxiety groups. Bars represent mean values +/- 1 SD.
4.2.3. Electrophysiological results
To determine if there were any significant differences in the different facial morphological components and emotion first, a 2x2x2x2 repeated-measures ANOVA was conducted on the LPP (400-1500ms) for the average of the electrode cluster of Cz, CP1, CP2, CPz, Pz, with the variables time, FWHR, emotion, and anxiety level (high vs. low anxiety group) as factors.
The interaction effects found to be significant were a one-way interaction emotion (F(1,22))=
4.368, P=.04), time (F(1,22)) = 72.63, P=.00); two-way interactions between FWHR and emotion (F(1, 22) = 5.293, P =.03) and emotion and time (F (1,22)= 14.56, P=.001. The only three-way interaction found to be significant was the same interaction between FWHR, emotion, and time (F(1,22) = 12.54, P =.002)
29 Figure 6: Grand average event-related potential waveforms recorded from sites CPz, Pz, CP1, CP2, Cz across anxiety groups for faces with low FWHR (angry and neutral) faces and high strong FWHR (angry and neutral) faces. The time window is divided as early (400-1000ms) and late (1000-1500ms).
Again, as there were no significant differences between the high and low anxiety groups, new waveforms have been created by putting all the subjects together.
30 Figure 7: Grand average event-related potential waveforms recorded from sites CPz, Pz, CP1, CP2, Cz for faces with low FWHR (angry and neutral) faces and high FWHR (angry and neutral) faces. The time window is divided as early (400-1000ms) and late (1000-1500ms).
31
5. Discussion
The purpose of the present study was to investigate whether higher social anxiety in individuals results in stronger reactions when exposed to a threatening face. This has been done by using the LPP as a proxy for emotional arousal. As a threat-inducing variable, faces with different morphology, operationalized as high and low FWHR, have been presented as cues of threat. In the present study, no statistically significant increase in LPP amplitude could be observed in individuals with high social anxiety compared to individuals with low social anxiety. However, the literature reviewed above suggests that anxiety should have a significant influence on LPP (MacNamara & Hajcak, 201; Kujawa et al., 2015). It is difficult to say whether the null finding reported here was merely due to the small sample size.
Possibly, using computer-generated artificial faces with their rather poor ecological validity as stimuli, lead to small overall emotional and corresponding electrophysiological reactions.
Within the context of the present study which did show significance were brow height and emotional expression. In the experiment, only brow height had a significant effect on facial ratings and LLP magnitude and producing the greatest effects. In experiment B concerning the variables FWHR and anger, which presumably should serve as a cue of threat, only the angry and neutral expressions showed a significant main effect. If FWHR is a cue of threat, then these results indicate that expression seems to have a more robust effect than FWHR. A significant interaction effect for FWHR was found between time and expression, as well as a three-way interaction between time, FWHR, and expression. Given that angry faces have been shown to be a reliable way of inducing a cue of threat across multiple studies (Fox et al., 2000; Hansen, & Hansen, 1988; Oosterhof, & Todorov, 2008; Schupp et al., 2004), there is the further possibility that, while the FWHR is a cue of threat, an angry high FWHR individual is more threatening than either an angry low FWHR individual or a neutral high FWHR individual.
32 5.1 LPP and Threat perception
Given that in the present study the facial ratings correlate nicely with the magnitude of the LPP in the 400-800ms window, the results support that LPP indeed indicates threat perception. The LPP reaction was found to be significant in the earlier of the two time windows, following similar studies investigating the early and late LPP, suggesting that differences between a neutral and a threatening condition are more likely to occur in the early LPP time window rather than the late (Weinberg, & Hajcak, 2010). Concerning the LPP, a time window from 400-800 ms is well within the boundaries of previous research related to threatening content, e.g., 350-750 ms (Schupp et al., 2000), 400-600 ms (Schupp, Öhman, et al., 2004) has been used. On the other hand, some other researchers have used window 400- 600 ms to measure how threat-related content affects the P300 (Schupp et al., 2007).
However, the LPP in the late window (800-1500 ms) together with the uncertain separation of the LPP with the P300 component (Hajcak et al., 2010), makes it hard to relate the results to perceptual and underlying neurological mechanisms. Therefore, it is preferable to not observe the early LPP, as a sole component to provoke sustained positive reaction to stimuli.
Moreover, given that the stimuli used in the current study were of neutral faces, and that the low and high levels for brow did not differ greatly when compared to angry and neutral faces in previous studies (e.g., Schupp, Öhman, et al., 2004), a less sustained LPP of lower magnitude would be expected in the current study. The stimuli used in this experiment thus can safely be thought to have had the anticipated effect on participants within the context of threatening stimuli evoking differences in the early time window rather than the late.
5.2 The sample and effect size
There are numerous potential explanations as to why the initial hypothesis (individuals with social anxiety would have stronger reactions to high-FWHR faces, compared to individuals without social anxiety) of this study was not supported by the data. The most obvious explanation as to why no statistical significance was reached in the analysis is because of how the sample was put together. There is reason to believe that the high social anxiety group in this experiment might not have been a true high anxiety group, since the group-defining criterion was a simple median split. The scores of SIAS are somewhat higher than would be expected based on the normative data of Heimberg, Mueller, Holt, Hope, and Liebowitz (1992), who found a mean of 19.9 on the SIAS in a community sample. Heimberg et al.
33 defined the clinical cutoff for social phobia as being equal to or greater than 34 on the SIAS (using the scoring range 0-4), but the use of the SIAS total score cutoff may be slightly more strict for students than for clinical clients (e.g., in the present study, a high social anxious group having mean 59.20 and a low social anxious group having mean 44.36). Evidence shows that patients with social anxiety scored higher on the SIAS than undergraduates and community controls (Heimberg et al., 1992; Holt, Heimberg, & Hope, 1992; Mattick & Clark, 1998; Rapee, Brown, Antony, & Barlow, 1992). Consequently, the sample might not have been the best possible for using anxiety as a factor; as the students were not clinically diagnosed as anxious, the entire sample might be considered as a low anxious population.
An additional explanation is that the real effect size could be much smaller than expected by the hypothesis. Should the effect on the LPP be very small, it need not matter that the sample might be considered a low anxiety sample; rather, in this case, a higher number of participants would be required to achieve the statistical power needed for the effect to show any significance. Additionally, the influence of facial morphology might not be sufficiently effective to cause measurably large changes to LPP amplitude, unlike showing faces with different emotional expressions, e.g., angry and neutral faces.
5.3 Enhanced processing of the eye region in high socially anxious individuals?
In the present study, some significant measures have indicated how different facial morphology and affective cues influence the perception of threat. For experiment A, brows are found to be the only factor that leads to significant differences for all participants regardless of which anxiety group they belonged to, while no significant differences were found for other factors, e.g., cheeks. The current study failed to observe any signs of higher- rated perceived threat in faces with relatively wide cheekbones, meaning high FWHR (Geniole et al., 2015). For both anxiety groups, the “weak cheeks strong brows” condition has a high reaction in relatively earlier stages of the LPP. Most studies on FWHR incorporated brow height in their measure of the upper facial height. Part of the variance seen in these studies was likely driven by brow height. The study conducted by Stirrat and Perrett (2010) found that faces with lower FWHR were judged as more trustworthy than faces with higher FWHR despite not incorporating brow height, Interestingly, they presented the pictures simultaneously in pairs to their participants, who then had to choose which face they thought seemed most trustworthy. This comparison of the faces might have intensify the effects of the
34 manipulated FWHR as the faces are seen in relation to each other. Moreover, a study conducted by LoBue and Larson (2010) also suggests that biases for angry face features, such as the downward “V” shape of the eyebrows, have been found to generate attentional bias for threatening stimuli in both child and adult populations. Moreover, study results by Schmitz et al. (2012) demonstrated that early posterior negativity (EPN) amplitudes in socially anxious individuals were enhanced to eye pairs regardless of gaze direction. One possible interpretation of this result could be that High Social Anxiety participants show a generally enhanced processing of the eye-region independent of gaze direction. The EPN to angry faces was recently found to be increased by high state social anxiety, triggered by the anticipation of a social stressor (Wieser et al., 2008).
5.4 Do high FWHR angry faces resemble social cues of threat?
For experiment B the effect of emotion and FWHR did turn out to be significant, confirming the validity of the experiment. Faces with high FWHR (angry expression) elicited higher LPP amplitudes for both the high and low anxiety group. This is not surprising, since an angry face has been demonstrated to be a reliable way of inducing a cue of threat across numerous studies (Fox et al., 2000; Hansen, & Hansen, 1988; Oosterhof, & Todorov, 2008; Schupp et al., 2004) and an angry high FWHR individual plausibly is more threatening than an angry low FWHR individual. The results show that angry and high FWHR together elicit a larger effect on the LPP. It remains unclear which of these factors is causing the effect here. A previous study has suggested that a facial structure subtly resembling anger, such as the FWHR, might elicit a cue of threat due to its likeness to angry faces (e.g. Carré et al., 2009;
Geniole & McCormick, 2015). Facial features that resemble those of angry faces might also cue strength and fighting ability and therefore be threatening (Sell et al., 2014). In the present study, the possible cue of threat elicited from a neutral, high FWHR could only be measured from subjective threat ratings, not from EEG recordings. However neutral faces with angry face features e.g., low brows (Langner et al., 2010), could be perceived as the person having aggressive intentions and therefore be threatening (Said et al., 2009). There is a strong support from the finding that faces categorized as neutral (Engell, Haxby, & Todorov, 2007) were still rated in agreement to have more or less threatening personality traits (Said et al., 2009).
Furthermore, the small increase in the subjective threat ratings for the neutral, high FWHR face is difficult to analyze for several reasons. Firstly, the ratings came from a relatively small
35 sample. Secondly, participants witnessed and rated each face only once. Moreover, the angry facial expressions elicited a much higher threat rating, regardless of FWHR.
5.5 Limitations
There are further possible limitations to the current study also worth noting. For instance, the faces presented in this experiment to the participants were artificially generated using a computer program. As per the need for the study, these artificial faces were created to make certain that the only relevant variables that were changing between the faces were those we wanted to: i.e., brow, cheek, jaw or FWHR, age, and emotion. This might result in a lack of authenticity of the stimuli. Previous studies have more commonly utilized photographs of real people when inducing a cue of threat in participants (Fox et al., 2000; Hansen & Hansen, 1988; Schupp et al., 2004). This might be yet another reason why the effect hypothesized in this thesis did not attain statistical significance, as it might have been mitigated by non- authentic cues of threat.
As mentioned above the low sample size in the present study reduces the validity of the results. Previous research regarding ERP measures has found significant results using sample sizes between 20 and 28 (Cuthbert et al., 2000; Schupp, Junghöfer, et al., 2004; Schupp, Öhman, et al., 2004). Therefore, the sample size of the present study is at the small end of the spectrum, but might still be adequate to generate enough power.
The present study had lacked a proper homogeneous representation of the sexes with a significant bias towards male participants. There is no certain evidence of whether or not there could be any potential differences between sexes and their automated perception of social threat. If this is a factor then this would interpret that anxiety is a strong modifying factor for adult males than adult females.
During the experiment and EEG recording, participants were allowed to listen to the music of their choice, which might conceivably constitute confounds regarding the results. On the upside, music might have a beneficial effect on tiredness or boredom. Moreover, Luck (2014) argues that the potential confounds introduced by boredom and tiredness that arise when participating in an EEG-experiment are larger than the possible confounds introduced when allowing participants to listen to music.
