Base-Line Measure for Within-Group Comparisons
For the purpose of investigating the treatment effects separately, base-line measures of BFNE-II and LSAS-SR at Time 2 were obtained by comparing them with the initial measurements at recruitment (Time 1) 14 days earlier. Four t tests per condition showed that the mean scores on all measurements did not differ between Time 1 and Time 2 (the Imagery condition: BFNE-II, t(13) = 1.125, p = .28; LSAS-SR, t(13) = − .581, p = .571 and the Exposure condition: BFNE-II, t(12) = 1.696, p = .116; LSAS-SR, t(13) = − .581, p = .571). Thus the participants did not improve without treatment over 2 weeks.
Effects of the Imagery and Exposure Treatments on Social Anxiety
In order to investigate the effects of the two interventions, 2 (Imagery; Exposure) × 3 (Time 2, Time 4 and Time 5) repeated measures ANOVA analyses were performed for the outcome measures BFNE-II and LSAS-SR. Pairwise t test analyses were conducted in cases in which significant interactions were indicated and effect sizes (Cohen’s d) were calculated. The results are shown in Table 1.
Table 1 Means and standard deviations on outcome measures before and after imagery rescriptinga, In-Vivo Exposure b, and at Follow-up, p values and Cohen’s d The ANOVAs revealed no significant interactions between the two treatments on any of the outcome measures. This shows that the measures did not differ between conditions. However, significant single main effects were found for both conditions over time: BFNE-II, F (2, 20) = 12.832, p = .000, η2 = .379, power = .990; and LSAS, F (2, 20) = 16.008, p = .001, η2 = .433, power = .968. Thus, the results showed that the treatments had similar effects and that fear of negative evaluation, levels of anxiety and avoidance reduced following the interventions.
Post-hoc analysis revealed that both treatments were associated with a reduced fear of negative evaluation at the 14-day follow-up (Imagery, t(12) = 2.204, p = .048; Exposure, t(10) = 2.462, p = .03) and at the 4-week follow-up (Imagery, t(11) = 3.362, p = .006; Exposure, t(10) = 2.785, p = .019). Also, significant reductions were found on the LSAS-SR, for the Imagery Rescripting at the 14-day follow-up, t(12) = 3.093, p = .009; and for both treatments at the 4-week follow-up (Imagery, t(11) = 3061, p = .011 and Exposure, t(10) = 2.667, p = .024).
In addition, we calculated the reliable change index (RCI; Jacobson and Truax 1991) for each individual to ensure that any difference between the baseline and post-treatment measurements was not caused by measurement error of the outcome instruments. The RCI was based on the baseline score (Time 2) and the 14-day (Time 4) or the 4-week follow-up (Time 5), and the standard error of the difference between the two scores. For the BFNE-II, Carleton et al. (2007) norm-data from a Canadian sample of undergraduate students was used (SD = 8.94 and α = .96) and for the LSAS-SR, we used American norm-data (Baker et al. 2002; SD = 25.5 and α = .95). The change is considered reliable (p < 0.05), i.e. unlikely to be the product of measurement error, if the RCI is greater than 1.96 or below − 1.96 (Jacobson and Truax 1991). When the individual has a change score exceeding these thresholds, it can be assumed that the individual has shown significant improvement or deterioration.
For BFNE-II, the proportion significantly improved participants after treatment (Time 4) were 23% in the Imagery and 36% in the Exposure condition, and at the 4-week follow up (Time 5) 50% showed improvement after imagery and 45% after exposure. For LSAS-SR, 38% improved after imagery and 64% after exposure, and at follow up the proportion that improved was 50% for imagery and 55% for Exposure. One participant showed significant deterioration after Exposure for both the LSAS-SR and the BFNE-II scores, but at follow-up the deterioration was significant only for LSAS-SR.
Social Image/Situation Threat Ratings
As the procedure for social image/situation threat ratings differed in the two conditions, these measures could not be compared between treatments. To examine within-condition changes, however, pairwise t-tests were used. Ratings were made at pre-intervention (Time 2), post-intervention (Time 3), 14-day follow-up (Time 4) and 4-week follow-up (Time 5) (see Table 2).
Table 2 Means and standard deviations on imagery and threat ratings over conditions In comparison with pre-Imagery measurements, we found significant reductions (approx. 50%) on Image Frequency at the 4-week follow-up, t(11) = 5.159, p = .000, d = .96, on Memory Distress immediately after imagery rescripting (Time 3), t(13) = 2.754, p = .016, d = .75, and after 4 weeks (Time 5), t(11) = 5.184, p = .000, d = 1.81. The strength of the Negative Belief in oneself was significantly reduced directly after Imagery compared to the pre-Imagery ratings, t(13) = 3.446, p = .004, effect size (d = .72) and was further reduced at the 4-week follow-up, t(11) = 8.094, p = .000, d = 2.19. In contrast, no significant differences were found for Imagery Distress or Vividness ratings.
All Exposure situation ratings, including Exposure Anxiety, Exposure Avoidance, strength of conviction that the feared social catastrophes could happen (Belief SC) and the negative assumptions regarding oneself as a social person (Belief SP) were significantly reduced (d = 1.06–1.71) between Time 2 (just before Exposure) and Time 5 (the 4-week follow-up). Compared to pre-intervention ratings, Exposure Anxiety was reduced at post-intervention t(11) = 3.918, p = .002, d = 1.42 and the difference remained at the 4-week follow-up, t(8) = 4.148, p = .003. The effect size was large (d = 1.71). Similarly, Exposure Avoidance was significantly reduced at post-intervention, t(11) = 6.271, p = .000, d = 1.86. However, when comparing post-intervention ratings to the 4-week follow-up, a significant increase in Exposure Avoidance was evident, t(8) = -− 3.813, p = .005, d = .76. Nevertheless, the avoidance ratings at the 4-week follow-up were still significantly lower than at the pre-intervention, t(8) = 3.082, p = .015, with a large effect size (d = 1.30). In comparison to pre-intervention (Time 2), Belief SC was significantly reduced at post-intervention (Time 3), t (11) = 2.891, p = .015, d = 1.14 and remained at Time 5 after 4 weeks t(8) = 2.669, p = .028 d = 1.19). Turning to Belief SP, a significant reduction was found at post-intervention compared to pre-intervention belief ratings t(11) = 4.967, p = .000, d = 1.54. In contrast, negative self-beliefs grew significantly stronger over time as the post-intervention beliefs were rated significantly lower than those at the 4-week follow-up, t(8) = − 1.750, p = .118, although the latter were still significantly reduced compared to the pre-intervention rating, t(8) = 3.724, p = .006, d = 1.06.