Martijn van Teffelen

Interpretation bias modification for hostility 141 6 State hostility and general psychiatric symptom measures To test the hypothesis that CBM-I leads to larger reductions in state hostility and general psychiatric symptoms than AC, four separate mixed regression models were run with cognitive aspects of hostility (AQ-H), state anger (STAXI-2S), self-reported aggression (FOAS) and general psychiatric symptoms (K10) as dependent variables and time, condition and condition by time as independent variables. Findings are shown in Table 3. Results showed that CBM-I led to a lesser reduction in general psychiatric symptoms (K10) ( d = -0.17, 95%CI = -0.34 to 0.00) from pre to post intervention compared to AC. Results on self-reported state aspects of hostility (AQ-H, STAXI-2S, and FOAS) revealed a time effect in which all variables reduced from pre- to post intervention ( p s < .001). None of the interaction effects were significant ( p s > .054). Trait hostility measures To test the prediction that CBM-I results in larger reductions in hostility traits than AC, four separate mixed regression models were run with overall hostility (PID5-H), cognitive aspects of hostility (AQ-H) and self-reported aggression traits (FOA) as dependent variables and time, condition and condition by time as independent variables. Results are shown in Table 3. For the self-reported hostility measures (PID-5H, AQ-H, and FOA) findings revealed significant time effects ( p ’s < .007), but no significant interaction effects ( p s > .18). Next, a Poisson regression model was run with behavioral aggression (VDT) at post-intervention as dependent variables and condition and behavioral aggression at baseline as independent variables. Results showed that CBM-I resulted in greater reductions in behavioral aggression from pre- to post intervention ( B = -.28, SE = .08, d = -0.29, 95% CI = -.47 to -.10) compared to AC (see Figure 4). Last, given that men express aggression more physically than women (Björkqvist et al., 1992) we explored CBM-I effects on behavioral aggression for men. Results showed that the effect of CBM-I increased ( d = -0.41, 95% CI = -1.10 to -0.20) compared to AC. Carry-over effects on working alliance in subsequent psychotherapy To test the hypothesis that CBM-I leads to beneficial carry-over effects on working alliance in subsequent psychotherapy we explored the data of n = 17 ( n = 8 for CBM-I and n = 9 for AC) participants who engaged in psychotherapy after the experiment. First, an independent samples t-test was run on working alliance as dependent variable. Results demonstrated that the two conditions did not differ significantly in terms of working alliance in psychotherapy, t = 1.45, p = .168. Then correlations were calculated to examine relations between change scores on the outcome variables and working alliance. Results showed that none of the outcome- change scores significantly related to working alliance ( p s > .069). However, when working alliance was regressed on post-test VDT scores, correcting for that the indicated usefulness per condition (i.e., interaction) did not depend on hostility level ( OR = .32, p = .171).