Context influences social cognitive judgments in paranoid individuals with schizophrenia.

We previously presented a study investigating neural activation during judgments of trustworthiness in non-paranoid individuals with schizophrenia (NP-SCZ), paranoid individuals with schizophrenia (P-SCZ), and healthy individuals (CON) (Pinkham et al., 2008). During functional neuroimaging, participants viewed photos of faces and made dichotomous decisions of trustworthiness (i.e. trustworthy or untrustworthy) for each image. Analysis of these behavioral responses, provided in the original publication, demonstrated that PSCZ, as compared to the other groups, rated significantly more faces as untrustworthy. However, individuals were also asked to make trustworthiness judgments about the same stimuli immediately following the MRI session. These additional ratings utilized the Likertscale from −3 (not at all trustworthy) to +3 (very trustworthy) implemented in the original Trustworthiness/Approachability Task developed by Adolphs et al. (1998). Our goal was to obtain ratings that would allow for a parametric examination of neural responses that increased or decreased according to varying levels of trustworthiness. While not our primary motivation, these ratings also permit examination of the potential role of testing context on behavioral responding by comparing ratings provided in the scanner to those provided out of the scanner. Effects of context on reaction time have previously been observed in imaging studies utilizing cognitive paradigms (Barch et al., 2001), but to our knowledge, no studies have examined this issue in regard to social cognitive tasks or distinctions between paranoid and non-paranoid individuals with schizophrenia. To assess the effects of context, out-of-scanner data were first dichotomized by recoding all stimuli with negative ratings as untrustworthy and all stimuli with positive ratings as trustworthy. Stimuli receiving ratings of zero were omitted. The percent of faces rated as trustworthy was then calculated for both in-scanner and out-ofscanner responses, and these values were entered into a repeated measures ANOVA with context (in-scanner vs. out-of-scanner) as the within-subject variable and group (P-SCZ vs. NP-SCZ vs. CON) as the between-subject variable. Analyses revealed a main effect of context; across groups more faces were rated as trustworthy outside the scanner (F(1, 33)=8.99, p=.005). In contrast to our previous results, the main effect of group was not significant indicating no overall difference between ratings provided by each group. However, a significant context by group interaction (F(2, 33)=5.62, p=.008) emerged demonstrating that only P-SCZ showed a significant effect of context. Follow-up t-tests indicated that CON and NP-SCZ rated the same percentage of faces as trustworthy both inand out-of-scanner (CON: t(11)=1.18, p=.26; NP-SCZ: t(11)=.12, p=.91) whereas P-SCZ rated significantly more faces as trustworthy outside the scanner (t(11)=5.02, pb .001; Fig. 1).