Imagery Rescripting Versus Cognitive Restructuring for Social Anxiety: Treatment Effects and Working Mechanisms

Participants

Highly socially anxious individuals were recruited via advertisements on university campus and social media. To be included, participants had to score ≥ 30 (clinical cut-off) on the German version of the Social Interaction Anxiety Scale (SIAS; Stangier et al., 1999). Results of a sample-size calculation (two-tailed, α = .05, power = .80, run with G*Power 3.1; Faul et al., 2007) with medium to large effect sizes (d = .70; Morina et al., 2017) showed that a sample size of 76 was required to detect significant differences between active treatments (ImRs + CR) versus NIC.

During the first session, eligible participants were administered the Mini International Neuropsychiatric Interview (M.I.N.I. 5.0.0; Sheehan et al., 1998; German version: Ackenheil et al., 1999) to screen for exclusion criteria: (1) current diagnosis of Major Depressive Disorder, (2) current and/or lifetime diagnosis of Posttraumatic Stress Disorder/Psychotic Disorder/Bipolar Disorder, (3) Substance Dependence during the past 12 months, (4) acute suicidal tendencies. Further exclusion criteria were: (5) age < 18 or > 35 years, (6) current psychological treatment, (7) pregnancy, (8) severe physical illness. The restricted age range was applied to obtain a more homogenous sample regarding age. Participants had to meet the following inclusion criteria: (1) negative mental self-image(s) in feared social situations, (2) aversive social experience related to the image, and (3) maladaptive self-belief (see Section "Imagery Interview").

A total of 96 participants attended Session 1 of whom 16 had to be excluded (n = 10 current/lifetime diagnosis of mental disorders specified above; n = 4 no negative mental self-image; n = 2 no maladaptive self-belief). Three participants did not attend the follow-up session, leaving a final sample of 77 participants (81% female; age: M = 22.46, SD = 3.88). All participants gave written informed consent and were reimbursed by receiving partial course credit or 20€. The study was approved by the Research Ethics Committee of the Faculty of Psychology and Educational Sciences at LMU Munich.

Clinical Interviews

The M.I.N.I. (Sheehan et al., 1998; German version: Ackenheil et al., 1999) was administered to assess current diagnoses according to DSM-IV (American Psychiatric Association [APA], 2000). Additionally, the SAD module of the Structured Clinical Interview for DSM-IV (SCID-I; First et al., 2002; German version: Wittchen et al., 1997) was administered.

Imagery Interview

The Imagery Interview was based on the Waterloo Images and Memories Interview (WIMI; Moscovitch et al., 2011) and on the interview used by Norton and Abbott (2016). The semi-structured interview assessed negative self-imagery, aversive memories, and maladaptive self-beliefs. Participants were asked to define their most anxiety-provoking social situation and to imagine themselves being in such a situation. They were instructed to become aware of whether there was a mental image that comes to their mind in this kind of situation and to describe the mental image in detail. Participants were then asked when they first felt the way they did in the image and to visualize and describe the respective event. This was used to determine whether there was an early aversive memory related to the mental image. In order to specify the idiosyncratic self-belief derived from the negative mental image and the aversive memory, participants were asked: “What do the image and the memory tell about you as a person?”. Participants were instructed to summarize the meaning in form of a short statement.

Speech Task

In order to measure reactions to a social stressor, participants were asked to give a 3 min video-recorded impromptu speech (Norton & Abbott, 2016) on a given political topic in both sessions (the order of two topics was counterbalanced).

Symptom Measures

The 20-item SIAS (Mattick & Clarke, 1998; German version: Stangier et al., 1999) was used to assess social interaction anxiety during the past seven days on a 5-point scale (0 = not at all to 4 = extremely). The 12-item Brief Fear of Negative Evaluation Scale-Revised (BFNE-R; Carleton et al., 2006; German version: Reichenberger et al., 2016) was administered to measure fear of negative evaluation by others on a 5-point scale (1 = not at all characteristic of me to 5 = extremely characteristic of me). In order to test for baseline group differences in depressive symptoms, the Patient Health Questionnaire-9 Item (PHQ-9; Krönke et al., 2001; German version: Löwe et al., 2002) was administered.

Speech Task Measures

In order to verify the relevance of the speech task as a stressor we asked participants to indicate how anxious they had felt or would have felt when giving a speech/presentation during the last week (0 = not at all anxious to 3 = extremely anxious). The Probability and Consequences Questionnaire (PCQ; Rapee & Abbott, 2007) asks participants to rate their appraisal of the likelihood (7 items) and cost (7 items) of negative evaluation of their speech on a 5-point scale (0 = not at all likely/bad to 4 = extremely likely/bad). Subjectively experienced levels of distress were assessed using Subjective Units of Distress (SUD, 0 = not at all distressed to 100 = extremely distressed). Self-assessment manikins (SAM; Bradley & Lang, 1994) were used to assess self-reported physiological arousal (1 = very calm to 9 = very aroused).

Measures of Underlying Mechanisms

Interventions

Procedure

The study comprised two sessions, which were one week apart. Two experimenters carried out different parts of the procedure so that the speech task and intervention were not administered by the same experimenter. During Session 1, Experimenter 1 administered the clinical interviews and baseline measurements (t0: sociodemographic data, SIAS, BFNE-R, public speaking anxiety, SUIS, ERQ), followed by pre-speech measures (SUD, SAM, PCQ) and the speech task. Experimenter 2 then conducted the Imagery Interview and administered pre-treatment questionnaires (t1: intellectual and emotional belief, PANAS-X). Then, participants were randomly allocated to ImRs (n = 25), CR (n = 27), or NIC (n = 25). The allocation sequence was computer-generated and Experimenter 2 was blinded until the beginning of the interventions, Experimenter 1 was blinded during the entire study. Immediately after the interventions or the waiting period, participants completed post-treatment measures (t2: intellectual and emotional belief, PANAS-X). During Session 2, which took place one week later, Experimenter 1 administered the follow-up questionnaire (t3: SIAS, BFNE-R, intellectual and emotional belief) and the second speech task, again including speech task measures administered prior to the speech task (SUD, SAM, PCQ). Finally, participants were fully debriefed.

Statistical Analyses

A series of 2(Time) x 3(Condition) repeated measures ANOVAs were carried out for social anxiety symptoms (t0; t3), for speech task measures (pre-speech1; pre-speech2), and for positive and negative emotions (t1; t2). To follow up significant interactions, planned contrasts on change scores were conducted (ImRs+CR vs. NIC; ImRs vs. CR). Effects on intellectual and emotional self-beliefs were tested with 3(Time) x 3(Condition) repeated measures ANOVAs. Significant interactions were followed up using planned contrasts (ImRs+CR vs. NIC; ImRs vs. CR). For ImRs, Pearson correlations were computed between mechanisms and symptomatic change. A significance level of α = .05 (two-tailed) was used for all analyses. Partial eta squared ( ${\mathrm{\eta }}_{\mathrm{p}}^2$) or Cohen’s d were used as effect sizes.

Participant Characteristics and Baseline Comparisons

No significant baseline differences between conditions emerged (see Table 1). Mean age at time of the aversive event was 12.86 years (SD = 4.55; range 3-27), with significant differences between groups¹ (ImRs: M = 13.88, SD = 4.90; CR: M = 13.76, SD = 4.60; NIC: M = 10.88, SD = 3.55), F(2, 74) = 3.78, p = .027.

Social Anxiety Symptoms

Social Interaction Anxiety

For SIAS scores (see Figure 1), there was no main effect of Condition, F(2, 74) = 1.97, p = .147, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .05, but a significant effect of Time, F(1, 74) = 17.94, p < .001, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .20, and a significant interaction, F(2, 74) = 3.22, p = .046, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .08. Planned contrasts revealed no difference between the active treatment groups compared to NIC in reducing social interaction anxiety, t(74) = 1.05, p = .298, d = 0.26. However, CR led to stronger decreases than ImRs, t(74) = 2.29, p = .025, d = 0.64.

Click to enlarge

Figure 1

Effects of ImRs vs. CR vs. NIC on (a) Social Interaction Anxiety (SIAS), and (b) Fear of Negative Evaluation (BFNE-R)

Note. Error Bars Represent SEM.

Speech Task Measures

For both subscales of the PCQ², there were significant main effects of Time, Fs(1, 71) > 9.74, ps < .003, ${\mathrm{\eta }}_{\mathrm{p}}^2$s ≥ .12, but no significant interactions, Fs(2, 71) < 2.28, ps > .110, ${\mathrm{\eta }}_{\mathrm{p}}^2$s ≤ .06. The main effect of Condition was significant for probability, F(2, 71) = 3.13, p = .050, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .08, but not for cost of negative evaluation, F(2, 71) = 1.13, p = .330, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .03. ImRs and CR did not yield significantly greater reductions of appraisals of negative evaluation than NIC (see Table 2).

For distress (SUD), a significant effect of Time emerged, F(1, 70) = 17.41, p < .001, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .20, but neither the main effect of Condition nor the interaction were significant, Fs(2, 70) < 2.12, ps > .128, ${\mathrm{\eta }}_{\mathrm{p}}^2$s < .06 (see Table 2).

Results for arousal (SAM) revealed a significant effect of Time, F(1, 72) = 11.35, p = .001, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .14, but neither a significant main effect of Condition nor a significant interaction, Fs(2, 72) < 1.05, p > .354, ${\mathrm{\eta }}_{\mathrm{p}}^2$ < .03 (see Table 2).

Mechanisms

Effects on Social Anxiety Symptoms

Contrary to hypothesis, we found that one session of cognitive restructuring (CR) is more effective than one session of imagery rescripting (ImRs) and no intervention control (NIC) in reducing social interaction anxiety. No significant differences between groups emerged for fear of negative evaluation. When confronted with the speech task, participants in all conditions demonstrated equal reductions in distress, arousal, and negative appraisals suggesting that if CR and ImRs are administered as very brief interventions no beneficial effects emerge over and above mere exposure to the speech. The speech task represents a strength of the study, but our findings suggest that the speech task may be susceptible to exposure effects, thereby reducing its ability to capture between-group differences in anxiety across time. Taken together, we could not replicate previous findings regarding the effects of the interventions on responses to a social stressor (Norton & Abbott, 2016). Our findings support previous evidence that one session of CR exerts positive effects on social anxiety symptoms (e.g., Norton & Abbott, 2016; Shikatani et al., 2014). Contrary to expectations, we were not able to replicate earlier findings on the benefits of stand-alone ImRs (Nilsson et al., 2012; Norton & Abbott, 2016; Reimer & Moscovitch, 2015) on social anxiety symptoms. This result is surprising given the similarities between studies (i.e., one session of ImRs, no cognitive preparation); however, a sub-clinical sample was included in our study whereas participants were diagnosed with SAD in previous research (Nilsson et al., 2012; Norton & Abbott, 2016; Reimer & Moscovitch, 2015). Although the severity of self-reported interaction anxiety in our study was comparable to previous studies (ø40 [this study]; ø37 [Nilsson et al, 2012]; ø44 [Norton & Abbott, 2016]), the low rate of diagnoses in the present sample could indicate that the impairment caused by the social anxiety symptoms was not sufficient to fulfill diagnostic criteria and that participants are able to cope with their negative mental images.

As our ImRs procedure closely followed the procedure of Norton and Abbott (2016), it seems rather unlikely that procedural differences explain the inconsistent findings. Alternatively, ImRs as used in this study might need to be optimized. First, ImRs might not have been optimally delivered (e.g., insufficient reactivation of emotions or the hotspot; short duration of ImRs [ø 22min in the present study]). Second, we do not know to what extent participants were able to put themselves in their younger self´s perspective. Third, in order to ensure internal validity we used a highly standardized ImRs protocol whereas other studies administered ImRs in a more individualized way and with a more active therapist/ experimenter (e.g., Norton & Abbott, 2016). Fourth, participants were instructed to introduce changes themselves in the present study. Finally, as dysfunctional self-beliefs were not explicitly addressed during ImRs it cannot be ruled out that the rescripting did not show a good enough match with the dysfunctional self-beliefs in the sense of providing corrective information and experiences to modify this belief. This may provide another explanation why ImRs was not associated with long-term effects in our study. Therefore, as the ImRs protocol used in the present study represents only one specific implementation of ImRs, it is conceivable that other versions of ImRs might have yielded more stable effects. For example, in accordance with the protocol by Wild and Clark (2011), a combination of ImRs with CR (Lee & Kwon, 2013; Wild et al., 2008) might yield more stable treatment effects. Different ImRs techniques have been applied in both research and clinical practice; however, it remains an open question how ImRs is best realized (e.g., with or without cognitive preparation; active vs. passive role of patient/therapist), therefore, future research is clearly needed to identify the most effective implementation of ImRs.

Mechanisms Underlying Imagery Rescripting

In line with our hypothesis and with previous evidence (Holmes & Mathews, 2010), a single session of ImRs led to stronger increases of positive emotions than CR and NIC. In contrast, negative emotions significantly decreased across time with no differences between conditions. ImRs and CR more strongly reduced maladaptive intellectual and emotional beliefs from pre- to post-intervention compared to NIC, but only for CR reductions remained stable across time. In ImRs, neither changes in positive emotions nor in emotional beliefs correlated with symptomatic outcomes.

Although our results indicate that brief ImRs led to beneficial (short-term) effects, it remains to be tested whether the aforementioned mechanisms play a role in producing symptomatic change, as ImRs did not yield improvements on symptom measures in the present study. Moreover, our results challenge the notion that emotionally anchored reappraisal is a mechanism specific to ImRs. In fact, brief CR seems to be more effective in targeting maladaptive emotional beliefs in the longer-term, counter to the theoretical idea that cognitive treatment strategies primarily change intellectual meaning levels (i.e., propositional level). However, after a single session of CR mean levels of emotional beliefs were still high at follow-up and more systematic research is needed to test whether emotional beliefs can be further reduced with multiple treatment sessions.

Limitations

ImRs and CR were delivered as very brief interventions within a non-therapeutic setting. Thus, the interventions deviate from treatment as used in clinical practice limiting its generalizability. However, laboratory-based studies in healthy or subclinical samples are a valuable means to investigate mechanisms involved in psychological treatments under highly controlled and standardized conditions (e.g., Van Den Hout et al., 2017). Although we inquired about the meaning of the mental image, we did not assess how distressing and how relevant the image was regarding participants´ social anxiety symptoms. The distress/impairment caused by the image should be inquired in future studies as it is conceivable that only the modification of distressing images might be associated with long-term effects on social anxiety symptoms. Moreover, it remains unclear whether participants adhered to the ImRs instructions and how distressed they were during ImRs as distress during ImRs was not assessed. Therefore, we cannot verify the correct implementation of ImRs and that emotional activation was sufficient. Emotional beliefs were rated on a one-item VAS, which might reduce reliability.

Conclusion

The present study compared the effects of ImRs vs. CR as stand-alone single-session interventions in socially anxious individuals and aimed to examine mechanisms underlying symptomatic change. Results indicate that a single session of CR effectively reduces social anxiety symptoms. The present study raises the question how ImRs for socially anxious individuals should optimally be implemented in order to yield symptomatic change. We propose that more individualized ImRs protocols, higher treatment intensity, cognitive preparation, and/or directly targeting dysfunctional self-beliefs might be necessary to yield therapeutic effects.