The current project tested whether changes in experiential avoidance (EA) and social anxiety disorder (SAD) symptoms bidirectionally predicted one another from session to session. Despite the increasing ubiquity of targeting EA through empirically-supported treatments (e.g., Barlow et al., 2004
; Hayes et al., 2006
; Roemer et al., 2008
), to our knowledge, this is the first study to directly test how EA and SAD symptoms may bidirectionally interrelate throughout
an intervention—a key step in establishing EA as a mechanism of SAD (Kazdin, 2009
). In partial support of hypotheses, weekly changes in SAD symptoms preceded and predicted changes in EA during the subsequent week. Surprisingly however, this effect was not bidirectional, as weekly changes in EA did not precede and predict changes in SAD symptoms during the subsequent week. Critically, these findings accounted for the overall trajectory of change in both variables (i.e., EA and SAD) and both variables’ preceding measurement, thereby eliminating two influential sources of potential covariation (i.e., important alternative explanations for these variables’ interrelations).
Our novel findings beg the question, why might decreases in SAD symptoms precede and predict less avoidance of internal states? The measure of SAD used in the present study, the SPIN, may provide important clues. Factor analyses have identified three factors captured by the SPIN: fear of negative evaluation, fear of physical symptoms (i.e., akin to anxiety sensitivity), and fear of uncertainty (i.e., akin to intolerance of uncertainty) in social situations (Campbell-Sills et al., 2015
). In other words, our findings suggest that reductions in fears of negative evaluation and uncertainty, as well as reductions in anxiety sensitivity, prospectively lead to reductions in self-reported EA.
Surprisingly, there is a dearth of research testing whether fears of negative evaluation, physical sensations, or uncertainty predict greater or less use of EA in individuals with SAD symptoms. However, indirect evidence suggests that each may predict EA in the context of the other disorders, such as symptoms of eating disorders (Espel-Huynh et al., 2019
) and generalized anxiety disorder (Akbari & Khanipour, 2018
). Thus, building on the present study, a meaningful follow-up could explore how these constructs (i.e., fears of negative evaluation, physical symptoms, and uncertainty) work together to predict more or less use of EA for individuals with SAD symptoms. For example, prior research suggests that individuals with SAD symptoms fear that physical symptoms of anxiety will lead to negative evaluation (e.g., due to increased shaking, sweating, and blushing), thereby increasing the possibility of rejection (Kashdan & Steger, 2006
). Thus, we speculate that a natural, though often unhelpful, response is to use EA to try to eliminate physical and emotional indicators of anxiety, eliminating the possibility of rejection. However, if an individual with SAD symptoms blushes, but does not fear negative evaluation or can tolerate the possibility of being rejected, then there may be little reason to fear internal and physical sensations related to anxiety. In turn, there may be little need to use EA.
Contrary to predictions based on theoretical and empirical precedent, there was no evidence from this study that change in EA preceded and predicted change in SAD symptoms. This is the first study, to our knowledge, that followed recommendations to test session-by-session changes in both
constructs to evaluate EA as a treatment mechanism of SAD symptoms (Kazdin, 2009
). This finding was particularly notable, given that many empirically-supported treatments focus on directly targeting EA to reduce symptoms of psychopathology, including anxiety disorders (e.g., Barlow et al., 2004
; Hayes et al., 2006
; Roemer et al., 2008
). This finding was also surprising in light of evidence that EA mediates treatment outcomes for SAD (Niles et al., 2014
) and unidirectionally or bidirectionally predicts anxiety symptoms and stress (Moroz & Dunkley, 2019
; Spinhoven et al., 2014
; Wenze et al., 2018
There are a few possible explanations for our divergent pattern of findings. First, the present analyses controlled for the overall trajectory of change in both variables (i.e., EA and SAD) and both variables’ preceding measurement, thus teasing apart sources of apparent change that may have influenced findings from past studies. In controlling for these sources, it is possible that our findings more accurately capture the relations between weekly fluctuations in SAD symptoms and EA, independent of longer-term trajectories or regression to the mean.
Second, none of the previous studies focused on the relations between EA and SAD symptoms within a weekly timeframe. Thus, when considering longer-term shifts over the entire course of therapy (e.g., 12 to 20 weeks), changes in EA may be an important mediator of changes in SAD symptoms (e.g., Niles et al., 2014
). However, patients might not experience the long-term benefits of accepting their emotions until they have practiced for several weeks, learned to embrace the discomfort of emotions, or witnessed the benefit of emotional acceptance in their lives. If supported empirically with future research, mapping this longer time course may have important clinical utility. For example, a therapist might tell their patient that accepting difficult emotions will lead to lower SAD symptoms with practice, but the patient might not experience this change in symptoms for several weeks or even months—a caveat that could help manage patients’ expectations about the pace of recovery.
The findings from the present study must be considered in the context of a few key limitations. First, experts in structural equation modeling and BLCSM hold that sample sizes approaching 100 to 200 are generally preferable (Kline, 2011
). That said, some have proposed that, “Smaller sample sizes are needed when the distributions of continuous outcome variables are normal in shape and their associations with one another are all linear” (Kline, 2011
, p. 12), conditions that were met in the present study. Thus, we were approaching adequate sample size, but our promising findings should be replicated with a larger sample. Second, although our sample was selected for elevated SAD symptoms and EA, our findings may not generalize to individuals diagnosed with SAD or individuals seeking treatment for SAD. Third, because we used a pilot intervention without a control group, we cannot draw firm conclusions about the intervention’s efficacy or the extent to which the findings reflect naturalistic versus manipulated interrelations between EA and SAD symptoms. Fourth, we used the BEAQ to evaluate changes in EA, because it has been recommended over other widely-used measures due its excellent construct, convergent, and discriminant validity (Rochefort et al., 2018
; Tyndall et al., 2019
). However, the BEAQ has been criticized for 1) using items related to behavioral (vs. internal) avoidance (Tyndall et al., 2019
), and 2) demonstrating poor fit in certain samples (e.g., treatment-seeking veterans; Byllesby et al., 2020
Despite these limitations, the present study is the first, to our knowledge, to test the bidirectional dynamic interrelations between EA and SAD symptoms as they unfold from week to week. Our findings suggest that changes in SAD symptoms precede and predict changes in EA, but not vice versa. This meaningful finding advances our understanding of how these constructs temporally interrelate throughout treatment. Moreover, they suggest that altering facets of SAD, such as fears of negative evaluation, physical sensations, and uncertainty, may have important downstream implications for reducing EA, representing an important area for future research. Critically, our findings also suggest that changes in EA do not necessarily lead to changes in SAD symptoms the following week. Given that many empirically-supported treatments target EA, this study highlights a need for future work to continue evaluating whether EA is indeed a mechanism of SAD symptom change.
Thank you to Dr. April Smith, Dr. Neil Brigden, and the ACE and SCOUT Labs for their invaluable feedback and support across the evolution of this project. Thank you to the research assistants for their hard work, accountability, and creativity in collecting our data (in alphabetical order): Chris Allen, Michael Archiable, Elise Ashford, Savannah Bliese, Alex Bronston, Robert Floyd, Christi Miley, Emma Kleymeyer, Yasamean Tadayon, and Isabelle Webber. This work was supported by funding provided to the first author from the Association of Behavioral and Cognitive Therapies, the Miami University Graduate School, and the Miami University Department of Psychology.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.