Social anxiety disorder (SAD) is a common and highly distressing disorder (e.g., Fehm et al., 2005; Kennedy et al., 2009; Perry et al., 2016) that is characterized by an excessive fear of negative evaluation by others in social or performance situations (American Psychiatric Association, 2013, 2022). Until recently, it has been predominantly conceptualized as a condition of ‘distorted or biased information processing’ (Clark & Wells, 1995; Hofmann, 2007; Rapee & Heimberg, 1997) based on the idea that individuals with high degrees of social anxiety (SAs) have a tendency to see social danger where in fact there is none. This tendency is thought to cumulate and eventually maintain SAD (Beard & Amir, 2008; Heeren et al., 2012; Schmidt et al., 2009).

Yet, there is growing evidence that SAs are truly evaluated in a negative way. Heerey and Kring (2007) found that interactions with SAs were characterized by, e.g., nervous fidgeting, or poor reciprocity of smiling (Heerey & Kring, 2007). As a consequence, conversations with SAs evoked increased negative affect in interaction partners (Alden & Wallace, 1995; Creed & Funder, 1998; Meleshko & Alden, 1993; Voncken & Bögels, 2008).

While the above-mentioned behaviors are more of a deliberate nature, the question remains whether SAs show disruptions in subtle, more automatic behaviors in social interaction. Research by Asher et al. (2020) showed that non-verbal behavioral synchrony increased in non-anxious couples during conversation while it decreased in anxious/non-anxious dyads. As behavioral synchrony is associated with increased positive affect (Tschacher et al., 2014) and increased sense of rapport (Miles et al., 2009), a lack of syncing with others may actually lead to the effect that SAs fear most.

In a different line of research, Givon-Benjio and Okon-Singer (2020) reported that increased degrees of social anxiety were related to a preference for larger interpersonal distance (IPD) and a tendency to underestimate interpersonal distances to strangers when compared to friends (see Givon-Benjio et al., 2020 for replication in individuals diagnosed with SAD).

While keeping more IPD to others may be a means to downregulate anxiety, breaches of an SA’s personal space by others should increase anxiety (Perry et al., 2013). Accordingly, Wieser et al. (2010) showed that high social anxiety was associated with avoiding the gaze of a virtual male agent at 1.5m IPD, while the gazes of male agents were avoided irrespective of degree of anxiety at 0.5m. SAs also showed avoidance in the form of subtle backward head-movements irrespective of IPD (Wieser et al., 2010). In addition, closer IPD and direct gaze were associated with heart rate deceleration. These last findings are particularly intriguing as decelerations have been linked to orienting and freezing responses in the past, while accelerations is thought to indicate fearful or phobic responding (Klorman et al., 1977; Ruiz-Padial et al., 2005; for critical discussion, see Barry & Maltzman, 1985).

It is only since this millennium, that researchers have started to investigate human freezing behavior as reflected by reduced body sway in response to threat (Azevedo et al., 2005; Hagenaars et al. (2014a); Roelofs, 2017; Volchan et al., 2017). It is proposed that a common response to imminent (unavoidable) threat is to ‘freeze’, before flight or fight responses are triggered (Blanchard, 2017; Bracha, 2004; Fanselow, 2022; Fanselow et al., 1987; Hagenaars et al., 2014a). Eilam (2005), however, argued that freeze and flight responses are mutually exclusive but may, nevertheless, alternate quickly, dependent on the situation. In the same line, Hagenaars et al. (2014a), suggested that when freeze/fight/flight responding had evolved for survival purposes, it must be dynamic and flexible rather than being characterized by chronological order and lack of overlap. Accordingly, Fanselow (2022) postulates that, under more, context, type and direction of approaching threat determines the ‘adaptive defense threshold’ and thus the timing, kind and magnitude of responding.

A few studies investigated freezing-like behavior in response to social threat. Roelofs et al. (2010a) and Noordewier et al. (2020) placed healthy participants on a stabilometric force platform and/or measured their heart-rates while passively viewing neutral, angry, and happy facial expressions. Viewing angry faces was associated with freezing-related behavior, e.g., a reduction in body sway. Moreover, reduced body sway was associated with increased state anxiety. In patients with SAD, Levitan et al. (2012), found reduced body sway irrespective of stimulus type (neutral objects, social threat, generic threat). Niermann et al. (2017), investigated freezing behavior after a physiological and social stress induction (Smeets et al., 2012) and found associations between diminished recovery from freezing and increased internalizing symptoms (such as, e.g., social anxiety).

In sum, there is reason to assume that SAs show distinct postural behaviors in social interactions, and in particular when their interpersonal space is breached. Yet, research assessing postural changes in human beings has seldomly looked at the impact of a trait such as social anxiety. It is particularly remarkable that all studies used static stimuli while threatening situations are typically of a dynamic nature. To our knowledge, postural balance fluctuations have not, yet, been investigated when the cause of threat is gradually approaching.

In this proof-of-principle study we therefore investigated at what distances between participant and an approaching research assistant (RA), avoidance behaviors and/or freezing would occur and which role social anxiety would play herein. We expected increasingly marked backward swaying of the participant when the RA would come closer. Such impulsive backward swaying would indicate embodied negative evaluation (Bargh & Chartrand, 1999; Niedenthal et al., 2005). In addition, we expected that freezing responses, indicated by reduced body sway, would increase with a decreasing distance between RA and participant. Finally, we expected that freezing and simultaneous avoidance would occur earlier (at a greater distances) with higher degrees of social anxiety. The stance of the RA (moving vs still) was controlled for.

We investigated if decreasing interpersonal distance was related to subtle avoidance (backward leaning) and increased freezing-like behavior (reduced body sway), as well as the role of social anxiety here in. To test this, we used a stabilometric platform to measure objective movements accurately and applied it to a dynamic stranger-approach paradigm in a proof-of-principle study.

Against expectations, the distance between the approaching RA and the participant had no general effect on avoidant body posture or freezing-like behavior. Typically, one would have expected that an approaching, unfamiliar person would make participants retreat, at least when their personal space was breached. This was indeed partially demonstrated by Wieser et al. (2010). In a VR setting, participants made backward head movements and avoided eye gaze of approaching male digital agents at an uncomfortable 50 cm distance. However, they used fixed distances where the agents stopped (0.5 and 1.5 m) instead of dynamic distance changes as used in our study and in our study gaze avoidance was undermined by instructing participant to look the RA in the eyes. Yet, our study could have shown similar results as backward head movements should be readily reflected in our sensitive measure of body posture.

Unexpectedly, social anxiety did not have any influence on avoidant body posture when the interpersonal distance decreased. This is not in line with the results of Wieser et al. (2010) and Rinck et al. (2010). The embodied cognition theory suggest that automatic evaluative/emotional responses towards environmental stimuli are readily reflected in associated impulsive bodily responding (and vice versa: van Dantzig et al., 2009). Thus, higher degrees of social anxiety should have increased the salience of the experimental situation as social and potentially threatening, which should have been reflected in automatic bodily avoidance. Numerous studies using prototypical approach-avoidance tasks based on joystick-movements towards social stimuli, found such automatic avoidance behavior towards angry but not neutral faces in highly socially anxious individuals (e.g., Heuer et al., 2007; Lange et al., 2008; Roelofs et al., 2010b). It is possible that particularly angry faces as representation of prototypical threat evoke threat-related avoidance responses in SAs while approaching, neutral looking RAs might be merely unpleasant but inapt to evoke threat responding (Lange et al., 2014) as observed in our study.

Body sway, on the other hand, was influenced by participants’ degree of social anxiety interacting with the distance of the RA. Interestingly, the pattern was opposite to what was expected: Participants with high levels of social anxiety showed increased bodily movement at 260cm, and between 200cm and 100cm distance, with a peak at 180cm. These results seem to contradict numerous previous studies that found reduced movement in response to threat in unselected samples (e.g., Azevedo et al., 2005; Hagenaars et al., 2014b; Roelofs et al., 2010a). Studies with participants showing characteristics related to social anxiety, however, reported more contradictory results. For example, while insecure attachment was associated with increased freezing-like responding (Niermann et al., 2015), internalizing symptoms were not (Niermann et al., 2017). Two studies found no effects of stimulus valence but instead general reductions of body sway throughout the experiment (social anxiety disorder: Levitan et al., 2012; panic disorder: Lopes et al., 2009), and one study found reduced body sway for blocks with painful stimuli, but not modulated by social context (Karos et al., 2020). Importantly, there is also evidence for a lack of freezing behavior in psychopathology. Stoffels et al. (2017), for example, reported reduced heart rate (typically associated with freezing) in response to aversive pictures in healthy controls but not in patients with borderline personality disorder. In the same line, Fragkaki et al. (2017) discovered attenuated freezing responses in patients with posttraumatic stress disorder. Finally, Hagenaars et al. (2015) found that heart rate reductions (indicative of freezing) were absent when participants were prepared for the aversive stimuli but not when being unprepared. Warning participants in the current study about the (very intimate) approach of the RA might have prevented the evaluation of the situation as threatening and may have undermined typical threat related responding. This is also compatible with Fanselow’s (2022) idea that context and type of threat can influence the timing, type and degree of defensive responding. Taken together with Trower and Gilbert's (1989) evolutionary approach that social anxiety is not so much determined by imminent threat of death but more by survival related resources, freezing may be less likely in the approaching stranger context than other behavior that signals discomfort or unease.

Note that higher degrees of state anxiety at baseline, irrespective of social anxiety, were associated with decreased body sway/freezing, indicating that our measure was sensible enough to pick up freezing-like behavior when participants are stressed. This is in line with previous work from, e.g., Hagenaars et al. (2014b) or Roelofs et al. (2010a). It might be that particularly state assessments of subjective experiences of stress/fear are more directly linked to physiological and automatic behavioral responding. Our assessment of SA, however, is more based on anxiety traits that may not necessarily be related to the stress levels in this particular situation. True is, that, in general, state stress and trait SA are positively correlated, but the predictability of our specific procedure may have evoked an attitude often observed with high functioning SAs: knowing the job, getting it done, with nervous fidgeting and restless movements indicating discomfort (Heerey & Kring, 2007; Voncken et al., 2008, 2012) but no pronounced fear reaction. In line with this notion, research has indicated that SA is a heterogeneous condition based, under more, on distorted interpretations and evaluations of environmental cues and social situations (Clark, 2001; Hofmann, 2007; Rapee & Heimberg, 1997). As indicated above, Fanselow (2022) as well as Trower and Gilbert (1989) acknowledge that decreases in adaptive defense thresholds are highly related to subjective evaluations and interpretations of the context. It is therefore thinkable that high degrees of social anxiety are generally related to increased stress in a situation, but that it is primarily the degree of SA in these healthy participants that allows for evaluating the situation as unpleasant but not threatening which then could determine their subtle behavior accordingly. It is important to note that the social anxiety levels of this particular sample were unexpectedly higher than is typical of a student sample. If the range between a score of 30–60 on the LSAS marks mild to moderate degrees of social anxiety then about 54% of the participants fall in that range. About 10% score 60 or higher which might be indicative of a clinically relevant degree of social fears (Mennin et al., 2002; Rytwinski et al., 2009).

In sum, increases of body sway may be more indicative of general unease with or nervousness in a situation rather than reflecting the expectation of eminent threat. For our understanding of social anxiety, this could mean that SAs show subtle signs of discomfort in a social situation. Once noticed by interaction partners they may feel uncomfortable themselves and may start disliking the situation, thereby fulfilling the SAs ‘prophecy’ of being disliked by others (Asher et al., 2020; Tschacher et al., 2014).

Remarkably, we found differences in body sway between high and low socially anxious individuals when the RAs were at a moderate distance. That is, RA-distance related changes in body sway were in the form an inverted U-shape. Especially on the uncomfortably short distances, one would have expected anxiety related effects to be strongest. This may have been a floor effect: The unnaturally close interpersonal distances are likely to be perceived as highly unpleasant for everyone, irrespective of social fears (Bailey et al., 1972; Dosey & Meisels, 1969; Hayduk, 1983). But exaggerated threat interpretations and resulting anxiety responses in social anxiety become primarily apparent in ambiguous situations and not necessarily in situations that actually are highly uncomfortable or threatening. In addition, subtle signs of discomfort and unease, rather than obvious ones, would manifest in a distance range where normal social interactions take place and not at a far distance, when there is clearly no social threat.

Irrespective of these results, a few limitations need to be considered. As indicated above, the predictability of what participants could expect may have diminished the experience of threat and related behaviors. Luo et al. (2019) found indeed that task instructions can readily attenuate automatic responding to emotional stimuli. In addition, our participants were, of course, free to leave the situation whenever unbearable while freezing-like behavior may be more likely to occur without (knowledge of) escape options (compare: Clark, 1993; Sanderson et al., 1989). For example, Buss et al. (2004) found that children responded with more freezing to an approaching stranger when they were restrained in a high-chair instead of freely playing. Finally, while our setup was considerably more dynamic than, e.g., that of Wieser et al. (2010), approaching participants in steps rather than in a continuous motion in combination with distance judgements may still be suboptimal and decrease ecological validity. Future research should delineate specific responses to dynamic social interactions and distinguish social threat from social discomfort herein to provide the most ecologically validity when measuring subtle behavior. Although not consistently supported in SA research (e.g., Cheng et al., 2022), the addition of physiological measures to the current study setup may help to further the understanding of subtle behaviors in social interactions.

To conclude, with our new experimental setup, we found that high degrees of social anxiety were associated with deviations from subtle ‘default’ behaviors seen as normal in social interaction: SAs showed increased movement when their own personal space was intruded. Future research may explore whether such behavior elicits feelings of discomfort in others, and if so, whether that would lead to true negative evaluation of SAs. Treatments may target these subtle behaviors to break this vicious circle. If psychotherapy, “primarily tackles faulty cognitions while there is a considerable probability that the fear of negative evaluation is justified, chances of substantial and sustained recovery are undermined.” (Lange et al., 2014, p. 360).