Martijn van Teffelen

Interpretation bias modification for hostility 131 6 STUDY 2 Methods Participants Participantswere sampled in twoways. First, we recruited in twooutpatient andone inpatient mental healthcare facility in the Netherlands (i.e., METggz, Mondriaan – Radix and U-Center). Participants were screened for eligibility while on the waitlist for treatment. Second, participants were sampled in the local community using an advertisement in local news media asking for ‘people with a short fuse.’ Inclusion criteria were age between 18 and 60 years, mastery of the Dutch language, basic computer skills and a score above 1.27 on the hostility scale of the Personality Inventory for DSM-5 (PID-5H) (Van der Heijden et al., 2014). This cutoff equals one standard deviation above the mean in both a Danish (a comparable population to the Netherlands) community as well as a clinical sample (Bach et al., 2016). Exclusion criteria were being on a waitlist for treatment shorter than four weeks 2 , IQ estimate below 80, observed psychotic/manic symptoms during intake, suicidality and not having access to a computer. A participant flow diagram is presented in Figure 1. In the pre-registration we determined sample size in the following way; to detect an effect of d = 0.81 (Hallion & Ruscio, 2011) with α = .05 and β = .10, anticipating a 25% drop-out, N 1 = N 2 = 10.5 * 2/0.81 2 / .75 ≈ 43 participants are needed per condition. However, if we would have used the obtained effect size in S1 ( d = 0.51) 3 to detect an effect with α = .05 and β = .20 we would have needed a minimum of n = 61 participants per condition. In total, N = 135 people entered the study, of whom n = 37 were lost to follow-up. Sample characteristics are shown in Table 1. Intervention arms The intervention arms were similar to Study 1. Materials Interpretation bias. To measure interpretation bias we used the SIP-AEQ, similar to Study 1. In addition, we administered the Word Sentence Association Paradigm-Hostility (WSAP-H) scale (Dillon et al., 2016). In the WSAP-H, participants were presented with sixteen hostile and sixteen neutral word-sentence pairs and are asked to indicated how well each word is related to the sentence on a 6-point Likert scale ranging from 1 (not related at all) to 6 (very related). Mean scores on the hostile and neutral word-sentence pairs are a measure for respectively hostile and benign interpretation bias. In the present study, we split up the WSAP-H in two parts to enable measurement pre- and post-intervention. The instrument shows good internal consistency (in this study α ’s range from .70 to .72) and adequate discriminant validity (Dillon et al., 2016). 2 Our CBM-I intervention took place over the course of four weeks. To avoid confounding effects of other therapies, people with a waiting time shorter than four weeks were excluded. 3 We thank one of the reviewers of this paper for the valuable suggestion to include this in our manuscript.