Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Neural, behavioural and real-life correlates of social context sensitivity and social reward learning during interpersonal interactions in the schizophrenia spectrum:

Neural, behavioural and real-life correlates of social context sensitivity and social reward... Objective: Recent findings suggest that diminished processing of positive contextual information about others during interactions may contribute to social impairment in the schizophrenia spectrum. This could be due to general social context processing deficits or specific biases against positive information. We studied the impact of positive and negative social contextual information during social interactions using functional neuroimaging and probed whether these neural mechanisms were associated with real-life social functioning in schizophrenia spectrum disorders. Methods: Patients with a schizophrenia spectrum disorder (N = 23) and controls disorder (N = 25) played three multi-round trust games during functional magnetic resonance imaging scanning, with no, positive and negative infor- mation about the counterpart’s trustworthiness, while all counterparts were programmed to behave trustworthy. The main outcome variable was the height of the shared amount in the trust game, i.e. investment, representing an indication of trust. The first investment in the game was considered to be basic trust, since no behavioural feedback was given yet. We performed region-of-interest analyses and examined the association with real-life social functioning using the experience sampling method. Results: Social contextual information had no effect on patients’ first investments, whereas controls made the lowest investment after negative and the highest investments after positive contextual information was provided. Over trials, patients decreased investments, suggesting reduced social reward learning, whereas controls increased investments in response to behavioural feedback in the negative context. Patients engaged the dorsolateral prefrontal cortex less than controls during context presentation and showed reduced activity within the caudate during repayments. In patients, lower investments were associated with more time spent alone and social exclusion and lower caudate activation was marginally significantly associated with higher perceived social exclusion. Conclusion: The failure to adapt trust to positive and negative social contexts suggests that patients have a general insensitivity to prior social information, indicating top-down processing impairments. In addition, patients show reduced Department of Clinical, Neuro and Developmental Psychology, Faculty of Behavioural and Movement Sciences, and Institute for Brain and Behaviour (IBBA) Amsterdam, Vrije Universiteit Amsterdam, Amsterdam, The Netherlands CSI Lab, Institute of Psychiatry, Psychology and Neuroscience, Department of Psychosis Studies, King’s College London, London, UK Hersencentrum Mental Health Institute, Amsterdam, The Netherlands Department of Psychology, City, University of London, London, UK Corresponding author: Esther Hanssen, Department of Clinical, Neuro and Developmental Psychology, Faculty of Behavioural and Movement Sciences, and Institute for Brain and Behaviour (IBBA) Amsterdam, Vrije Universiteit Amsterdam, Van der Boechorststraat 7, 1081 BT Amsterdam, The Netherlands. Email: emehanssen@gmail.com Australian & New Zealand Journal of Psychiatry, 56(1) 60 ANZJP Articles sensitivity to social reward, i.e. bottom-up processing deficits. Moreover, lower trust and lower neural activation were related to lower real-life social functioning. Together, these findings indicate that improving trust and social interactions in schizophrenia spectrum needs a multi-faceted approach that targets both mechanisms. Keywords Functional magnetic resonance imaging, experience sampling method, trust, schizophrenia, social context processing prefrontal cortex (dlPFC) (MacDonald et al., 2000). Introduction Second, theory of mind, i.e. the ability to infer the mental The ability to integrate social contextual information and states of others, in which medial prefrontal cortex (mPFC) behavioural feedback from others is necessary for success- and temporo-parietal junction (TPJ) are implicated as key ful social interactions (Ruz et al., 2011) and is an important regions (Carrington and Bailey, 2009; Schurz et al., 2014). foundation of trust in social relationships (Lewicki and Third, reward processing, which strongly involves the cau- Wiethoff, 2006). Individuals with a schizophrenia spectrum date nucleus (Krach et al., 2010; Sanfey, 2007). (SZ) diagnosis, hereafter referred to as patients, show a Deficits in mentalizing, social reward processing (bot- reduced ability in judging social signals (Penn et al., 2008). tom-up) and social context processing (top-down) have Previous research suggests two underlying mechanisms. been suggested to underlie lower trust, paranoia and social First, there are deficits in learning from others’ behavioural disconnection in SZ (Couture et al., 2006; Kapur et al., feedback during social interactions, suggesting problems in 2005; Velthorst et al., 2016). In SZ, prior studies have bottom-up mechanisms. Second, there are problems with reported reduced activation within the dlPFC during con- the integration of a priori contextual information in a top- text processing (Barch and Ceaser, 2012; Niendam et al., down way (Chung et al., 2010; Hooker et al., 2011). Here, 2014), within the mPFC and TPJ during mentalizing (Green we investigate how these two mechanisms of social infor- et al., 2015; Lee et al., 2004; Pinkham et al., 2008), and mation processing impact on social behaviour in real-time within the caudate during both non-social (Juckel et al., interactions, using a modified neuroeconomic trust game 2006b; Murray et al., 2008) and, more importantly, social (Fett et al., 2015). reward processing (Fett et al., 2019; Gromann et al., 2013). Studies employing the trust game (Berg et al., 1995) in This earlier work leads to the hypotheses that the dlPFC, SZ have demonstrated that patients invested lower initial mPFC, TPJ and caudate play an important role in disturbed amounts, indicating lower trust towards others (Fett et al., social decision-making and context processing in SZ. 2012; Gromann et al., 2013; Lemmers-Jansen et al., 2018). In this first-time investigation of the underlying mecha- In addition, SZ has been associated with a reduced ability nisms of disturbed trust and social interaction in SZ, we to use others’ cooperative behavioural feedback to adjust therefore probed the impact of different social contexts and trusting behaviour, i.e. bottom-up processing (Fett et al., investigated the underlying neural correlates in patients with 2012; Gromann et al., 2013). Research on the effects of SZ and healthy controls, using a modified version of an inter- social contextual information on trust in the general popu- active trust game while measuring brain activity with fMRI. lation demonstrated that trust increases in response to a Since social interactions are embedded in peoples’ daily lives trustworthy interaction partner, showing a strong impact of in a complex way, it is important to elucidate the association a priori information on trust in a top-down manner (Delgado between the neural processes underlying social interactions et al., 2005). Patients, in contrast, show a diminished sensi- and daily-life social engagement in SZ. To investigate this, tivity to such prior positive information (Fett et al., 2012). we combined fMRI and the experience sampling method The reduced ability to use bottom-up information and not (ESM), a diary method (Delespaul, 1995). Initial ESM stud- being able to overcome distrust during positive interactions ies (Kluge et al., 2018; Moran et al., 2019) started to investi- in response to such information, may explain real-life social gate how brain activation during task-based fMRI translates impairment seen in SZ (Velthorst et al., 2016) and could be to real-world functioning. This method ensures high ecologi- tackled with cognitive bias modification or other ways of cal validity because it allows for real-time monitoring of cognitive remediation. However, it has not been investi- behaviour in daily-life contexts. In patients, Moran et al. gated whether the insensitivity to social context reflects a (2019) found that greater hemodynamic signal change dur- more general processing deficit or a specific bias against ing (non-social) reward anticipation in caudate, insula and positive social information. anterior cingulate was associated with greater anticipated Three core cognitive mechanisms have been suggested pleasure and motivation for daily-life activities. to underlie trust and decision-making in social contexts We hypothesized that (1) patients would show a general (Declerck et al., 2013). First, context processing and cogni- reduced sensitivity to prior information about the counter- tive control, which are subserved by the dorsolateral part, reflected in no differences in baseline trust between Australian & New Zealand Journal of Psychiatry, 56(1) Hanssen et al. 61 conditions, i.e. first investments. Controls would increase symptom severity in the 2 weeks prior to testing in patients investments from the negative to the positive context; (2) (Kay et al., 1987). Fourteen items evaluate the severity of patients would not increase trust over trials in response to positive and negative symptoms (1 = absent to 7 = extreme). benevolent behavioural feedback, whereas controls would PANSS scores are reported in Table 1. do so; (3) patients would engage the regions-of-interest (ROIs) to a lesser extent than controls: (a) in the left dlPFC Trust game. To measure the impact of social context pro- during context presentation and investment (i.e. trusting) cessing in social interactions, we employed a modified decisions, because of its specific role in context processing; multi-round trust game (Berg et al., 1995; Gromann et al., (b) in the mPFC and right TPJ during context presentation 2013; King-Casas et al., 2005). In a multi-round classic and investment decisions, given their role in mentalizing trust game, the first player, i.e. the investor, is given an ini- mechanisms, which we expected to a greater extent in con- tial endowment of £10 and has to invest a chosen amount trols while processing prior social information compared to between zero and ten pounds. This amount is tripled and no information and (c) in the right caudate nucleus during given to the second player, i.e. the trustee. The trustee then the partner’s repayments (i.e. receipt of social reward); and decides whether and how much of the tripled amount he or (4) for patients, reduced trust and reduced activation in the she wants to give back to the investor. The chosen amount ROIs during the trust game is associated with lower daily- to invest by the investor reflects trust (given that the trustee life social functioning, i.e. more time spent alone, higher can choose not to return any money). In this study, partici- perceived social exclusion and lower perceived relation- pants played the role of the investor and played the trust ship quality, measured by ESM. game three times, with three different hypothetical counter- parts. In one game (condition), the trust game was pre- sented without prior social contextual information, i.e. a Methods classic multi-round trust game, while the other two condi- Subjects tions were modified to examine social context processing. In the negative and positive context conditions, participants Twenty-five patients with an SZ diagnosis and 26 controls first played three ‘blind’ rounds, without seeing the repay- without a personal or family history of SZ were included (for ments of the interaction partner. These rounds were imple- recruitment, see Supplement – A). Inclusion criteria were (1) mented to establish the cooperativeness of the trustee (Fett age 18–65 years, (2) good understanding of the English lan- et al., 2012). When making their blind investments, partici- guage and (3) intelligence quotient (IQ) >70. An additional pants saw the following the message: ‘Determining E.H.’s criterion for patients was an SZ diagnosis according to the initial average returns’. After these three ‘blind’ invest- International Statistical Classification of Diseases and ments, they viewed the following message on the screen Related Health Problems–Tenth Edition (ICD-10; World pointing out the average returns (more or less than invested) Health Organization (WHO), 1992), which was confirmed of the trustee during the blind investments (i.e. the social with the treating National Health Service (NHS) clinician. context): ‘On average your partner E.H. returned more/less Exclusion criteria were (1) a history of any neurological con- than you invested’. This message was shown before each ditions and (2) a diagnosis of alcohol/drug dependence trial in the trust game. Participants completed a total of 120 within 6 months. One control subject did not complete MRI trials (60 experimental and 60 control), equally divided scanning due to anxiety. Two patients were excluded from over the three conditions (no context, positive context and analyses due to excessive movement (framewise displace- negative context). Control trials were included to control ment ⩾1.5 mm in ⩾20% of the volumes per run). Therefore, for general effects of motor and visual activation elicited by analyses were performed on 23 patients and 25 controls. the task. An experimental and control trial with the respec- Forty-four participants completed the ESM measurements tive phases and timings is displayed in Figure 1. (20 patients and 24 controls). The London – Harrow Research Participants were instructed that they played with a real Ethics Committee (14/LO/0071) approved this study. human player via the Internet but were actually playing with a pre-programmed computer that behaved in the same Measures benevolent probabilistic manner in all three contexts (for algorithm, see Supplement – B). After completion of the Estimated cognitive ability. To assess an estimated cognitive trust game, participants were asked whether they thought ability, an abbreviated two-test version of the Wechsler the other players were real and trustworthy, on a 7-point Abbreviated Scale of Intelligence (WASI) was used Likert-type scale. Ratings on whether the other players (Wechsler, 1999), which consisted of the vocabulary sub- were unreal did not differ between groups (16% controls test and the matrix reasoning subtest. WASI scores are and 13% patients, p > 0.83), and were unrelated to invest- reported in Table 1. ments in the trust game (p > 0.82). Also, the ratings of trust- worthiness of the interacting partners did not differ between Symptoms. The Positive and Negative Syndrome Scale groups (p > 0.59). (PANSS) semi-structured interview was used to measure Australian & New Zealand Journal of Psychiatry, 56(1) 62 ANZJP Articles Table 1. Participant demographics and patient clinical characteristics. Controls (N = 25) Patients (N = 23) Statistic p-value Age – M (SD) 36.02 (7.34) 39.86 (9.10) β = 0.23 0.11 Gender (% male) 68.0 82.6 0.24 χ (1) = 1.36 IQ – M (SD) 116.68 (10.18) 98.30 (11.80) β = −0.64 <0.0001 Diagnoses (%) Schizophrenia 73.9 Schizoaffective disorder 17.4 Psychotic disorder 8.7 Medication (%) Atypical antipsychotics 82.6 Typical antipsychotics 13.0 None 4.4 PANSS – M (SD) Negative scale 1.75 (0.45) Positive scale 2.11 (0.86) Amotivation factor 1.92 (1.15) Diminished expression factor 2.45 (1.36) P6 (suspiciousness) 2.70 (1.23) Experience sampling (ESM) Alone time (%) 55 71 β = 0.31 0.04 Social exclusion – M (SD) 1.77 (0.94) 2.64 (0.81) β = 0.45 0.002 Relationship quality – M (SD) 5.38 (1.12) 5.18 (0.86) β = −0.10 0.53 M: mean; SD: standard deviation; IQ: intelligence quotient; PANSS: Positive and Negative Syndrome Scale; ESM: experience sampling method. Statistically significant differences are bold faced. ESM – measurement of social engagement. ESM (Palmier Participants first completed a demographic questionnaire et al., 2011), a structured diary technique, was used to mea- and several practice trials before playing the trust game in sure social engagement in daily life. The ESM device (iPod) the MRI scanner. They were told that they would receive the gave a signal to fill in the questionnaire 10 times a day, by earnings from one randomly selected round to keep them means of a pseudo-random ‘beep’ on 7 consecutive days. We motivated (between £0 and £30). For fairness reasons, all included several questions to probe real-life social function- participants received a payment of £5. After the trust game, ing in terms of social engagement, social exclusion and qual- they completed the questionnaire on their perception of the ity of social relationships. The question that was used game partners. Next, the PANSS interview was adminis- assessing social engagement was ‘Are you alone?’ (yes/no). tered. Finally, an explanation of the iPod was given to par- Perceived social exclusion was assessed when individuals ticipants and they completed one practice questionnaire. were alone using an average of the two items: ‘I feel lonely’ The morning after the first session the 7 days ESM data col- and ‘I feel excluded’ (Cronbach’s α = 0.82). The perceived lection started. All participants were contacted by phone on quality of social relationships was assessed when individuals day 2 for guidance in case of any problems or difficulties were in social company with an average of the four items: ‘I with the iPod. The second session consisted of assessment like the person(s) I am with’, ‘I feel close to them’, ‘They are of the WASI subtests. Participants handed in the iPod and dependable’ and ‘I trust them’ (Cronbach’s α = 0.90). ESM experiences were discussed. At the end of the study, each items were rated on a 7-point Likert-type scale. participant received payment (£60 + £5) for participation. Statistical analyses Procedure Behavioural analyses. Statistical analyses were performed The two sessions took place at the Institute of Psychiatry, using STATA version 14 (StataCorp, 2015). We examined Psychology and Neuroscience (IoPPN), King’s College group differences in demographics using chi-square tests London. Participants gave written informed consent before and regression analyses. Investments were analysed using the study. Australian & New Zealand Journal of Psychiatry, 56(1) Hanssen et al. 63 Figure 1. Left: investment trial with positive context (1) context cue: ‘initial repayment same or more’, (2) investment cue, (3) investment choice, made by scrolling over a horizontal bar ranging from £0 to £10, which started at £5, (4) invested amount displayed, (5) wait for game partners’ response (jittered), (6) fixation cross, (7) returned amount displayed, (8) round totals (kept and given amount added) for both players (jittered) and (9) fixation cross. Right: control trial, similar to the experimental trials, except for (a) ‘Invest’ was replaced by ‘Rest’ and (b) in the investment phase, participants had to move the cursor to the marked number. mixed effects multilevel random regression analyses TR = 7.312 ms; TE = 3.016 ms; flip angle = 11°; (MIXED) to account for repeated measurements within per- FOV = 270 mm) was acquired. sons, with (first) investment as dependent variables and with Imaging data were analysed using Statistical Parametric group (control and patient) and context (negative, no and Mapping 12. Pre-processing of the functional images con- positive) and their respective interactions as independent sisted of realign and unwarp, and coregistration to individ- variables. For investments over trials, trial number (1–20) ual anatomical images. Next, using unified segmentation, was added to the model. Mixed effects multilevel random anatomical images were segmented and normalization regression analyses (MIXED) were used to examine asso- parameters were estimated. These parameters were used to ciations between investments and ESM indices of real-life transform functional and anatomical images to a Montreal social functioning (% alone, social exclusion and quality of Neurological Institute (MNI) template. Subsequently, social relationships) across contexts. Interactions were smoothing was applied (Gaussian kernel 6 mm full-width at probed with the CONTRAST command. Analyses with a half-maximum), and the last three volumes were removed at significant group effect are additionally reported with esti- the ending of the task. mated IQ as a covariate. Additional analyses on association with symptoms are reported in Supplement – E. ROI analyses. A general linear model was used per run in which the three phases in the game were modelled as FMRI data acquisition and scanning parameters. Imaging regressors of interest for the experimental and control con- data were acquired using a 3T GE Signa Neuro-optimized dition separately (see Figure 1). The cue phase was time- magnetic resonance (MR) System at the Centre of Neuro- locked to the start of each trial (duration 2 seconds), the imaging Science of the IoPPN. Functional images were investment phase started after the investment cue (duration acquired by a T2*-weighted echo-planar imaging sequence 4 seconds) and the repayment phase started at the beginning scanning 39 axial slices of 3.0 mm thick with 0.3 mm gap. of the repayments shown and was modelled until the end of The in-plane resolution was 3.3 mm × 3.3 mm (FOV 211 × the displayed totals (duration 5–8 seconds). All other game 211), flip angle = 75°, TR = 2.00 seconds and TE = 30 ms. phases were combined into one regressor of no interest (the There were 413 volumes per run. For anatomical reference, investment cue, the invested amounts, waiting for the part- a T1-weighted image (196 slices; isotropic voxels of 1.2 mm; ners’ response and the two fixation crosses). All phases Australian & New Zealand Journal of Psychiatry, 56(1) 64 ANZJP Articles Table 2. Task performance – first investments in GBP (£). Negative context, No context, Positive context, Context mean (SD) mean (SD) mean (SD) difference Statistic p 95% CI Controls (N = 25) 5.00 (2.91) 6.28 (2.25) 7.60 (2.04) Negative < no −1.28 <0.01 [−2.24, −0.32] Positive > no <0.01 1.32 [0.36, 2.28] Patients (N = 23) 6.35 (2.90) 6.48 (2.74) 6.70 (2.93) Negative = no p = 0.85 Positive = no p = 0.75 SD: standard deviation; CI: confidence interval. Statistically significant differences are bold faced (p<0.01). were modelled using a box-car function convolved with the corrected at family-wise-error (FWE) whole-brain cluster hemodynamic response function (Friston et al., 1995). To significance threshold of p = 0.05 (see Supplement – C). correct for motion, the six realignment parameters and regressors for volumes detected as motion-corrupted, cal- Results culated by DVARS metric as implemented in FSL, version Behavioural analysis 6.00 (FMRIB’s Software Library, www.fmrib.ox.ac.uk/fsl) by FSL Motion Outliers (https://fsl.fmrib.ox.ac.uk/fsl/fsl Groups did not differ in age and gender. Patients had a wiki/FSLMotionOutliers) were included in the design lower estimated cognitive ability than controls. Percentage matrix, making the total number of regressors in the model of time spent alone was higher in patients than controls. variable for each individual, with a minimum of 10 regres- Patients felt lonelier and more excluded, but reported a sors (cue phase, investment phase, repayment phase, similar quality of their social relationships compared to regressor of no interest and six motion parameters). A high- controls (see Table 1). pass filter of 128 seconds was used. Subsequently, for each phase of interest, contrast images were created by contrast- Baseline trust: context effect and group differences in first ing a specific phase of the experimental condition with the investments. First investments were examined to establish corresponding phase in the control condition. context effects on baseline trust (Table 2). There was a sig- A priori ROI analyses were performed. Talairach coordi- nificant group-by-context interaction (χ (2) = 7.34, p = 0.02) nates were converted to MNI space (https://bioimagesuite which remained significant when estimated IQ was added web.github.io/webapp/mni2tal.html), resulting in the fol- to the model (χ (2) = 7.34, p = 0.02). Estimated IQ was not lowing ROI MNI coordinates: right caudate (17, 20, 3), significantly associated with baseline trust (p = 0.32). The right TPJ (50, −56, 27), mPFC (−3, 64, 24) (Gromann et al., context effect was only significant in controls, who made 2014) and the left dlPFC (−43, 18, 29) (MacDonald and lower investments in the negative than the no context con- Carter, 2003). ROIs were created in MarsBaR with an dition (b = −1.28, 95% confidence interval CI = [−2.23, 8-mm sphere (version 0.44; http://marsbar.sourceforge. −0.33], p = 0.009) and higher investments in the positive net). For each ROI and each subject, average signal change than the no context condition (b = 1.32, 95% CI = [0.35, (beta estimate) was extracted to investigate group and con- 2.28], p = 0.007). Patients’ first investments did not differ text effects, and to test association between the fMRI and by context (both p > 0.75). In all three contexts, first invest- the ESM data. Additional analyses on association with ments did not differ significantly between groups (all symptoms (PANSS positive scale, PANSS negative scale, p > 0.07). PANSS suspiciousness, PANSS amotivation factor and PANSS diminished expression factor; Supplement – E) and Changes in trust over trials: context effect and group differ- IQ are reported. Mixed effects multilevel random regres- ences. We examined interference of context information on sion analyses (MIXED) were used to examine associations changes in trust (i.e. investments) over time (Figure 2(a) between ROI beta estimates and ESM indices of real-life and (b)). The three-way interaction between group, context social functioning (% alone, social exclusion and quality of and trial was marginally significant (χ (2) = 5.07, p = 0.07), social relationships) across contexts. The results of the and analyses by group showed a marginally significant ROI-based analyses were Bonferroni corrected at α levels interaction of context-by-trial number in controls of 0.0125 per test (0.05/4, as tests were performed with data (χ (2) = 5.26, p = 0.07) but not patients (p = 0.21). Across from four ROIs). trials, both groups showed a context effect with significant differences between negative and no contexts, which was Exploratory analyses. We also performed exploratory whole- larger in controls than patients (controls: b = −0.64, 95% brain analyses, investigating neural activation beyond the CI = [−0.88, −0.39], p < 0.0001 and patients: b = −0.39, predefined ROIs for all three game phases. Analyses were 95% CI = [−0.67, −0.11], p = 0.006), but no differences Australian & New Zealand Journal of Psychiatry, 56(1) Hanssen et al. 65 Figure 2. Investments over trials in (a) controls and (b) patients, showing investments in the negative, no and positive context conditions. Investment phase. There were no significant group-by- between the positive and no contexts (both p > 0.41). Both context interactions (all p > 0.08), nor main effects of group groups showed a significant main effect of trial number; (all p > 0.50) or context in any of the ROIs (all p > 0.11). controls increased investments over time, while patients decreased their investments (controls: b = 0.03, 95% Repayment phase. There were no significant group-by- CI = [0.01, 0.04], p = 0.002 and patients: b = −0.04, 95% context interactions or context effects for the right caudate CI = [−0.06, −0.02], p < 0.0001). or mPFC. A significant group effect showed lower right cau- In addition, analyses by context showed that the effect of date activation in patients compared to controls (b = −0.14, trial number was most pronounced and in opposite direc- 95% CI = [−0.25, −0.03], p = 0.01, Figure 3(b)). This effect tions in patients and controls in the negative context (group- remained significant when the IQ estimate was added to by-trial number interaction (χ (2) = 21.21, p < 0.0001), the model as covariate (b = −0.16, 95% CI = [−0.30, 0.004], controls: b = 0.06, 95% CI = [0.03, 0.8], p < 0.001 and p = 0.028). The IQ estimate was not significantly related patients: b = −0.05, 95% CI = [−0.08, −0.02], p < 0.001). In to right caudate activation (p = 0.66). Higher mPFC acti- the no context condition, there was a group-by-trial number vation was found in patients than controls (b = 0.32, 95% interaction (χ (2) = 8.37, p = 0.004); no significant changes CI = [0.003, 0.62], p = 0.047). However, this result did not in investments were found controls (p = 0.51), but patients survive Bonferroni correction. There were no interaction, invested significantly less over time (b = −0.06, 95% group or context effects in any other ROI (all p > 0.16). CI = [−0.10, −0.01], p = 0.02). Groups did not differ signifi- Given that we observed a behavioural group difference cantly in the positive context condition and did not show in investments over trials, we performed exploratory analy- investment changes over trials (all p > 0.30). Additional ses probing the change in ROI activation over investment analyses between symptoms and behavioural results are trials. The results are reported in Supplement – D. reported in Supplement – E. Additional analyses between symptoms and ROI results are reported in Supplement – E. fMRI analysis Associations between investment, ROI activation and real-life ROI analyses by trust game phase social functioning in patients. We were specifically interested Cue phase. There were no significant group-by-context to examine whether lower neural activation within the left interactions in any ROIs (all p > 0.12). A significant group dlPFC and the right caudate found in patients was related to effect in the left dlPFC showed lower activation in patients the level of real-world social functioning. In addition, we than controls (b = −0.59, 95% CI = [−1.00, −0.18], p = 0.004, explored this association in the less robustly increased Figure 3(a)). The effect remained significant when the IQ mPFC signal in patients. estimate was added to the model as covariate (b = −0.71, Across contexts, higher investments were significantly 95% CI = [−1.24, −0.17], p = 0.009). The IQ estimate was associated with less time spent being alone (b = −0.02, 95% not significantly related to left dlPFC activation (p = 0.51). CI = [−0.05, −0.001], p = 0.04). A significant interaction Australian & New Zealand Journal of Psychiatry, 56(1) 66 ANZJP Articles Figure 3. Signal change (in arbitrary units) in (a) the dlPFC during the cue phase (i.e. social context presentation) and (b) the caudate during repayment (i.e. receipt of reward) in controls and patients. Error bars depict the standard errors. between context and perceived relationship quality on with real-life social functioning in SZ. Patients showed no investments (X(2) = 23.01, p < 0.0001) showed trend-level differential effect of social context on first investments, associations between higher relationship quality and higher regardless of the valence or the absence of a context, investments in the positive (b = 0.85, 95% CI = [−0.06, whereas controls showed the expected distinct context 1.78], p = 0.07) and negative (b = 0.94, 95% CI = [−0.06, effect with highest levels of trust in response to positive and 1.78], p = 0.06) conditions, but not the no context condition lowest levels of trust in response to negative social infor- (p = 0.79). There was a significant interaction between con- mation. This suggests a general insensitivity to social con- text and social exclusion on investments (X(2) = 13.19, text instead of a bias against positive social contextual p < 0.0001), indicating an association between lower per- information in patients. Patients also did not increase trust ceived social exclusion and higher investments the no con- in response to benevolent behavioural feedback whereas text condition only (b = −1.17, 95% CI = [−1.97, −0.35], controls did. Within the patient group, the findings indicate p = 0.005, other p > 0.23). an association between lower trust and lower real-world There were no significant associations between any social functioning. measure of real-life social functioning and dlPFC activa- We found overall lower activation in the left dlPFC dur- tion (all p > 0.21). A significant interaction between social ing context presentation and less engagement of the right exclusion and context emerged for the right caudate caudate nucleus during repayments in patients compared to (X(2) = 8.51, p = 0.01); lower perceived social exclusion controls. These results suggest that SZ is associated with a was marginally significantly associated with higher cau- general insensitivity to social contexts and with a reduced date activation in the positive context only (b = −0.17, 95% sensitivity to social reward. On the neural level, we also CI = [−0.36, −0.018], p = 0.07, all other p > 0.18). No sig- found an association between caudate activation and lower nificant associations were present with social relationship real-life social functioning in patients. quality or time spent alone (both p > 0.17). Higher per- ceived social relationship quality was associated with Social context effect on baseline trust higher mPFC activation (b = 0.27, 95% CI = [0.002, 0.54], p = 0.048). There were no significant associations between For baseline trust (i.e. the first investment where partner mPFC activation and social exclusion or time spent alone feedback has not yet been received), we found reduced sen- (both p > 0.22). sitivity to positive and negative social contextual informa- tion in patients compared to controls. This seems to reflect a general insensitivity to social contextual information. Discussion These results strengthen and extend previous evidence of a Using a novel, modified trust game, we examined the social context top-down processing deficit in SZ (Baez impact of social context on trust and social reward and the et al., 2013; Fett et al., 2012; Niendam et al., 2014); how- underlying neural activation patterns during real-time ever, this may not generalize to tasks concerning different social interactions. In addition, we probed the associations types of information processing. This insensitivity points to Australian & New Zealand Journal of Psychiatry, 56(1) Hanssen et al. 67 persistent a priori beliefs about other people in SZ. It is where reward processing takes place, we found context important to unravel whether this context sensitivity is a independent blunted activation in the right caudate nucleus risk factor for developing an SZ disorder or secondary to in patients compared to controls. The caudate is a highly the disorder or related factors, by examining whether this innervated by dopaminergic neurons (Björklund and insensitivity is also found in first-episode patients and indi- Dunnett, 2007). Aberrant regulation of dopamine is thought viduals at high risk for SZ. In this study, patients tended to to play a key role in SZ and reward processing (Howes and approach social interactions with similar trusting behaviour Kapur, 2009), and may account for the insensitivity to as controls, in line with findings in first-episode patients social feedback, i.e. reward. This finding adds to evidence (Fett et al., 2019), but contrasting results have also been showing deficits in social reward processing in SZ found in chronic and first-episode patients and individuals (Gromann et al., 2013; Lee et al., 2018) and may explain at clinical high risk (Fett et al., 2012; Gromann et al., 2013; why patients fail to increase trust over time. A less robust Lemmers-Jansen et al., 2018). effect was found in the mPFC; activation was higher in patients compared to controls. This was unexpected, but may be related to the role of the mPFC in reward-based Trust over time: the effects of action selection (Euston et al., 2012). We speculate that this benevolent feedback may point to a compensation mechanism for reduced As hypothesized, patients did not increase trust in response engagement of the caudate, which is tentatively supported to benevolent partner feedback, in line with previous litera- by the correlations with real-world social outcomes, which ture in patients with chronic psychosis (Fett et al., 2012; are discussed in the following section. Gromann et al., 2013). Controls, however, increased trust in the negative context and showed stable levels of trust Associations with real-life social engagement after no and positive information, suggesting that prior pos- itive beliefs about others were matched by the benevolent Previous social neuroscience studies have started to inves- partner feedback. Patients seem to have difficulties to over- tigate the link between the reward related processing in the come the given prior negative information, i.e. persistent a brain and daily life (Bakker et al., 2019; Moran et al., 2019), priori negative beliefs about others. Patients showed a ten- yet associations with real-life social functioning have not dency to reduce trust, even though subjective ratings of been considered in schizophrenia, despite yielding mean- trustworthiness of the interacting partner were similar for ingful insights in healthy controls (Bickart et al., 2011; both groups. Other studies have shown that patients show Kanai et al., 2012; Lewis et al., 2011). We found that in real reduced sensitivity to social rewards, such as smiles life, patients spent more time alone than healthy controls, in (Catalano et al., 2018). Our results strengthen the evidence line with previous work (Pinkham and Penn, 2006; Velthorst that patients have deficits in bottom-up processing of part- et al., 2016). They reported higher feelings of loneliness ner feedback, which might be due to an insensitivity to and social exclusion, but a similarly good relationship qual- social reward (Fett et al., 2012; Gromann et al., 2013). The ity as controls. On the behavioural level, higher invest- reduced ability to increase trust could explain patients’ ments were associated with higher social functioning in reduced motivation to engage in social behaviour (Krach real life. A higher level of trust may create an advantageous et al., 2010). basis for engaging in meaningful social relationships (Campellone et al., 2016). This could positively impact on social support networks, which in turn could aid in recov- Neural findings during social interactions ery (Corrigan and Phelan, 2004). At the neural level, we with contextual information found that in patients, lower activation in the caudate in the In support of our hypotheses, we found reduced left dlPFC positive context was marginally significantly associated activation in patients compared to controls during social with higher perceived social exclusion. More engagement context presentation. The dlPFC has been implicated in of this key reward area may be related to heightened experi- deficits in non-social context processing in schizophrenia ence of positive social interaction, i.e. social reward (Barch and Ceaser, 2012; Niendam et al., 2014) and is (Gromann et al., 2013). This could lead to a higher sense of viewed as key region in top-down cognitive control belonging or inclusion in social relationships. In addition, (MacDonald et al., 2000). The impairment in top-down higher activation in the mPFC was associated with higher modulation of trust in response to contextual information reported relationship quality in daily life, which would sup- may be related to reduced engagement of the dlPFC in SZ. port the role of mentalizing abilities (Schurz et al., 2014). This result that dlPFC activation was lower in all contexts However, it is possible that this finding also relates to supports a general insensitivity rather than a specific bias. higher feelings of (social) reward (Euston et al., 2012). We did not see any group differences in ROI activation dur- Future evidence from larger replication studies will be ing investments. However, during the repayment phase, needed to strengthen and support these initial results. Australian & New Zealand Journal of Psychiatry, 56(1) 68 ANZJP Articles behavioural and neural level. The findings suggest that Limitations increasing trust may facilitate social engagement in patients. The current results should be interpreted in light of the follow- Also, indices of real-life social functioning seem to be associ- ing limitations. First, this study is an initial investigation of ated with lower neural activity in reward- and context-pro- social context processing in relation to neural activation and in cessing (i.e. cognitive control) brain areas. This study suggests relation to daily-life social functioning in a relatively small that improving social interaction in SZ requires a multi-fac- sample of relatively stable patients with an SZ diagnosis. eted approach in clinical practice, which considers both bot- Specifically, our patient sample had relatively low symptom tom-up and top-down processing of social information. An levels: negative and positive scores of, respectively, 14.74 example of an intervention that incorporates these two facets (standard deviation (SD) = 6.15) and 12.26 (SD = 3.25). We is the Social Cognition and Interaction Training (SCIT) (Penn defined these scores as relatively low, since, with respect to the et al., 2007), targeting social skills bottom-up and social cog- 7-point Likert-type scale of the PANSS items and the number nition top-down, which shows promising effects on social of items in the positive and negative subscale, the item scores functioning. Clinical practice will benefit from newly devel- of 1 and 2 are, respectively, absent and minimal (Kay et al., oped treatments, building on the SCIT for instance, targeting 1987), (see also Supplement – E). Consequently, our results these facets in daily life by implementing an ecological need to be interpreted with caution when it comes to generaliz- momentary intervention by means of a smartphone-based ability and larger replication studies are warranted before any treatment. In addition, future studies are warranted to investi- firm conclusions can be made. Second, medication type has gate how this is related to fine-grained assessments of func- been found to have an effect on reward processing in SZ (Juckel tional capacity and real-world social functioning. et al., 2006a); however, the majority of participants in our sam- ple were on atypical antipsychotics (see Table 1), which are Acknowledgements thought to normalize reward processing (Nielsen et al., 2012; The authors thank the participants of the DECOP project and the Schlagenhauf et al., 2008). In addition, healthy first-degree National Institute for Health Research (NIHR) Clinical Research relatives of patients with SZ show reduced social reward pro- Organizations (CROs), as well as Farah Yakub, Marieke Helmich, cessing, without any medication confounds (Gromann et al., Marina Volguraki, Katie Wong, Mathew Harvey and Tracy Bobin 2014; Hanssen et al., 2020). Therefore, the current findings are for their support with testing and recruitment. not likely to represent an enhancement of the effects; they may even reflect an underestimation, compared to expected effects Declaration of Conflicting Interests in unmedicated patients. Studying unmedicated patients is val- The author(s) declared no potential conflicts of interest with uable, but poses a great challenge due to the clinical reality. respect to the research, authorship and/or publication of this Third, subjective self-report measures, like ESM, may raise the article. question of response accuracy or social desirable, which may be related to suspiciousness. However, ESM is a widely used Funding and well-validated method in psychiatric and schizophrenia The author(s) disclosed receipt of the following financial support samples (Delespaul, 1995; Myin-Germeys et al., 2018) and has for the research, authorship and/or publication of this article: This the advantage that it does not rely on retrospective recall work was supported by funding of the Netherlands Organization (Potheegadoo et al., 2012). Moreover, our sample showed a for Scientific Research (NWO) VENI grant (451-13-035) and a good compliance and did not report any issues related to the use NARSAD Young Investigator Grant from the Brain & Behaviour of the ESM app during the debriefing. We therefore have no Research Foundation (24138). grounds to assume that responses were inaccurate. Finally, not all participants believed they were playing with a real human ORCID iDs player; however, this was not associated with investments and Esther Hanssen https://orcid.org/0000-0002-4504-0757 adequate changes in investments in controls after context infor- Imke LJ Lemmers-Jansen https://orcid.org/0000-0001-9044- mation and behavioural feedback indicated that the experimen- tal manipulation was effective. Anne-Kathrin J Fett https://orcid.org/0000-0003-0282-273X Supplemental Material Concluding remarks Supplemental material for this article is available online. This is the first study to investigate the neural correlates of social context processing and the link with real-life social References functioning in SZ. We provide evidence that patients do not Baez S, Herrera E, Villarin L, et al. (2013) Contextual social cognition modulate their behaviour in response to social context infor- impairments in schizophrenia and bipolar disorder. PLoS ONE 8: mation (i.e. top-down processing) and positive behavioural e57664. cues of others (i.e. bottom-up processing). Our results point to Bakker JM, Goossens L, Kumar P, et al. (2019) From laboratory to a general reduced sensitivity to social information on a life: Associating brain reward processing with real-life motivated Australian & New Zealand Journal of Psychiatry, 56(1) Hanssen et al. 69 behaviour and symptoms of depression in non-help-seeking young Howes OD and Kapur S (2009) The dopamine hypothesis of schizophre- adults. Psychological Medicine 49: 2441–2451. nia: Version III – The final common pathway. Schizophrenia Bulletin Barch DM and Ceaser A (2012) Cognition in schizophrenia: Core psy- 35: 549–562. chological and neural mechanisms. Trends in Cognitive Sciences 16: Juckel G, Schlagenhauf F, Koslowski M, et al. (2006a) Dysfunction of 27–34. ventral striatal reward prediction in schizophrenic patients treated Berg J, Dickhaut J and McCabe K (1995) Trust, reciprocity, and social with typical, not atypical, neuroleptics. Psychopharmacology 187: history. Games and Economic Behavior 10: 122–142. 222–228. Bickart KC, Wright CI, Dautoff RJ, et al. (2011) Amygdala volume and Juckel G, Schlagenhauf F, Koslowski M, et al. (2006b) Dysfunction of social network size in humans. Nature Neuroscience 14: 163. ventral striatal reward prediction in schizophrenia. NeuroImage 29: Björklund A and Dunnett SB (2007) Dopamine neuron systems in the 409–416. brain: An update. Trends in Neurosciences 30: 194–202. Kanai R, Bahrami B, Duchaine B, et al. (2012) Brain structure links loneli- Campellone TR, Fisher AJ and Kring AM (2016) Using social outcomes ness to social perception. Current Biology 22: 1975–1979. to inform decision-making in schizophrenia: Relationships with Kapur S, Mizrahi R and Li M (2005) From dopamine to salience to psy- symptoms and functioning. Journal of Abnormal Psychology 125: chosis: Linking biology, pharmacology and phenomenology of psy- 310–321. chosis. Schizophrenia Research 79: 59–68. Carrington SJ and Bailey AJ (2009) Are there theory of mind regions in Kay SR, Fiszbein A and Opfer LA (1987) The Positive and Negative the brain? A review of the neuroimaging literature. Human Brain Syndrome Scale (PANSS) for schizophrenia. Schizophrenia Bulletin Mapping 30: 2313–2335. 13: 261–276. Catalano LT, Heerey EA and Gold JM (2018) The valuation of social King-Casas B, Tomlin D, Anen C, et al. (2005) Getting to know you: rewards in schizophrenia. Journal of Abnormal Psychology 127: 602. Reputation and trust in a two-person economic exchange. Science Chung YS, Mathews JR and Barch DM (2010) The effect of context 308: 78–83. processing on different aspects of social cognition in schizophrenia. Kluge A, Kirschner M, Hager OM, et al. (2018) Combining actigraphy, Schizophrenia Bulletin 37: 1048–1056. ecological momentary assessment and neuroimaging to study apathy Corrigan PW and Phelan SM (2004) Social support and recovery in people in patients with schizophrenia. Schizophrenia Research 195: 176–182. with serious mental illnesses. Community Mental Health Journal 40: Krach S, Paulus FM, Bodden M, et al. (2010) The rewarding nature of 513–523. social interactions. Frontiers in Behavioral Neuroscience 4: 22. Couture SM, Penn DL and Roberts DL (2006) The functional significance Lee J, Jimenez AM, Reavis EA, et al. (2018) Reduced neural sensitivity to of social cognition in schizophrenia: A review. Schizophrenia Bulletin social vs nonsocial reward in schizophrenia. Schizophrenia Bulletin 32: S44–S63. 45: 620–628. Declerck CH, Boone C and Emonds G (2013) When do people cooper- Lee K-H, Farrow T, Spence S, et al. (2004) Social cognition, brain net- ate? The neuroeconomics of prosocial decision making. Brain and works and schizophrenia. Psychological Medicine 34: 391–400. Cognition 81: 95–117. Lemmers-Jansen IL, Fett A-KJ, Hanssen E, et al. (2018) Learning to trust: Delespaul PA (1995) Assessing schizophrenia in daily life: The experience Social feedback normalizes trust behavior in first-episode psychosis sampling method. PhD Thesis, Maastricht University, Maastricht. and clinical high risk. Psychological Medicine 49: 780–790. Delgado MR, Frank RH and Phelps EA (2005) Perceptions of moral char- Lewicki RJ and Wiethoff C (2006) Trust, trust development, and trust acter modulate the neural systems of reward during the trust game. repair. The Handbook of Conflict Resolution: Theory and Practice Nature Neuroscience 8: 1611–1618. 2: 92–119. Euston DR, Gruber AJ and McNaughton BL (2012) The role of medial Lewis PA, Rezaie R, Brown R, et al. (2011) Ventromedial prefrontal prefrontal cortex in memory and decision making. Neuron 76: volume predicts understanding of others and social network size. 1057–1070. NeuroImage 57: 1624–1629. Fett A-KJ, Mouchlianitis E, Gromann PM, et al. (2019) The neural mech- MacDonald AW and Carter CS (2003) Event-related FMRI study of con- anisms of social reward in early psychosis. Social Cognitive and text processing in dorsolateral prefrontal cortex of patients with schiz- Affective Neuroscience 14: 861–870. ophrenia. Journal of Abnormal Psychology 112: 689–697. Fett A-KJ, Shergill SS and Krabbendam L (2015) Social neuroscience in MacDonald AW, Cohen JD, Stenger VA, et al. (2000) Dissociating the role psychiatry: Unravelling the neural mechanisms of social dysfunction. of the dorsolateral prefrontal and anterior cingulate cortex in cogni- Psychological Medicine 45: 1145–1165. tive control. Science 288: 1835–1838. Fett A-KJ, Shergill SS, Joyce DW, et al. (2012) To trust or not to trust: Moran EK, Culbreth AJ, Kandala S, et al. (2019) From neuroimaging to The dynamics of social interaction in psychosis. Brain 135: 976–984. daily functioning: A multimethod analysis of reward anticipation in Friston KJ, Frith CD, Turner R, et al. (1995) Characterizing evoked hemo- people with schizophrenia. Journal of Abnormal Psychology 128: dynamics with fMRI. NeuroImage 2: 157–165. 723–734. Green MF, Horan WP and Lee J (2015) Social cognition in schizophrenia. Murray G, Corlett P, Clark L, et al. (2008) Substantia nigra/ventral teg- Nature Reviews Neuroscience 16: 620–631. mental reward prediction error disruption in psychosis. Molecular Gromann P, Heslenfeld DJ, Fett AK, et al. (2013) Trust versus paranoia: Psychiatry 13: 239–276. Abnormal response to social reward in psychotic illness. Brain 136: Myin-Germeys I, Kasanova Z, Vaessen T, et al. (2018) Experience sam- 1968–1975. pling methodology in mental health research: New insights and tech- Gromann P, Shergill S, de Haan L, et al. (2014) Reduced brain reward nical developments. World Psychiatry 17: 123–132. response during cooperation in first-degree relatives of patients with Nielsen MO, Rostrup E, Wulff S, et al. (2012) Improvement of brain psychosis: An fMRI study. Psychological Medicine 44: 3445–3454. reward abnormalities by antipsychotic monotherapy in schizophrenia. Hanssen E, Krabbendam L, Robberegt S, et al. (2020) Social and non- Archives of General Psychiatry 69: 1195–1204. social reward learning reduced and related to a familial vulnerability Niendam TA, Lesh TA, Yoon J, et al. (2014) Impaired context processing in schizophrenia spectrum disorders. Schizophrenia Research 215: as a potential marker of psychosis risk state. Psychiatry Research: 256–262. Neuroimaging 221: 13–20. Hooker CI, Tully LM, Verosky SC, et al. (2011) Can I trust you? Negative Palmier-Claus JE, Myin-Germeys I, Barkus E, et al. (2011) Experience affective priming influences social judgments in schizophrenia. sampling research in individuals with mental illness: Reflections and Journal of Abnormal Psychology 120: 98–107. guidance. Acta Psychiatrica Scandinavica 123: 12–20. Australian & New Zealand Journal of Psychiatry, 56(1) 70 ANZJP Articles Penn DL, Roberts DL, Combs D, et al. (2007) Best practices: The Sanfey AG (2007) Social decision-making: Insights from game theory and development of the social cognition and interaction training pro- neuroscience. Science 318: 598–602. gram for schizophrenia spectrum disorders. Psychiatric Services Schlagenhauf F, Juckel G, Koslowski M, et al. (2008) Reward system acti- 58: 449–451. vation in schizophrenic patients switched from typical neuroleptics to Penn DL, Sanna LJ and Roberts DL (2008) Social cognition in schizo- olanzapine. Psychopharmacology 196: 673–684. phrenia: An overview. Schizophrenia Bulletin 34: 408–411. Schurz M, Radua J, Aichhorn M, et al. (2014) Fractionating theory of mind: Pinkham AE and Penn DL (2006) Neurocognitive and social cognitive A meta-analysis of functional brain imaging studies. Neuroscience & predictors of interpersonal skill in schizophrenia. Psychiatry Research Biobehavioral Reviews 42: 9–34. 143: 167–178. StataCorp (2015) Stata Statistical Software: Release 14. College Station, Pinkham AE, Hopfinger JB, Ruparel K, et al. (2008) An investigation of TX: StataCorp LP. the relationship between activation of a social cognitive neural net- Velthorst E, Fett A-KJ, Reichenberg A, et al. (2016) The 20-year longitu- work and social functioning. Schizophrenia Bulletin 34: 688–697. dinal trajectories of social functioning in individuals with psychotic Potheegadoo J, Cuervo-Lombard C, Berna F, et al. (2012) Distorted per- disorders. American Journal of Psychiatry 174: 1075–1085. ception of the subjective temporal distance of autobiographical events Wechsler D (1999) Wechsler Abbreviated Scale of Intelligence. New York: in patients with schizophrenia. Consciousness and Cognition 21: The Psychological Corporation; Harcout Brace & Company. 90–99. World Health Organization (WHO) (1992) The ICD-10 Classification Ruz M, Moser A and Webster K (2011) Social expectations bias decision- of Mental and Behavioural Disorders: Clinical Descriptions and making in uncertain inter-personal situations. PLoS ONE 6: e15762. Diagnostic Guidelines. Geneva: WHO. Australian & New Zealand Journal of Psychiatry, 56(1) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Australian & New Zealand Journal of Psychiatry SAGE

Neural, behavioural and real-life correlates of social context sensitivity and social reward learning during interpersonal interactions in the schizophrenia spectrum:

Download PDF
12 pages

Loading next page...
 
/lp/sage/neural-behavioural-and-real-life-correlates-of-social-context-s0Z7mzwWGk

References (112)

Publisher
SAGE
Copyright
Copyright © 2022 by The Royal Australian and New Zealand College of Psychiatrists
ISSN
0004-8674
eISSN
1440-1614
DOI
10.1177/00048674211010327
Publisher site
See Article on Publisher Site

Abstract

Objective: Recent findings suggest that diminished processing of positive contextual information about others during interactions may contribute to social impairment in the schizophrenia spectrum. This could be due to general social context processing deficits or specific biases against positive information. We studied the impact of positive and negative social contextual information during social interactions using functional neuroimaging and probed whether these neural mechanisms were associated with real-life social functioning in schizophrenia spectrum disorders. Methods: Patients with a schizophrenia spectrum disorder (N = 23) and controls disorder (N = 25) played three multi-round trust games during functional magnetic resonance imaging scanning, with no, positive and negative infor- mation about the counterpart’s trustworthiness, while all counterparts were programmed to behave trustworthy. The main outcome variable was the height of the shared amount in the trust game, i.e. investment, representing an indication of trust. The first investment in the game was considered to be basic trust, since no behavioural feedback was given yet. We performed region-of-interest analyses and examined the association with real-life social functioning using the experience sampling method. Results: Social contextual information had no effect on patients’ first investments, whereas controls made the lowest investment after negative and the highest investments after positive contextual information was provided. Over trials, patients decreased investments, suggesting reduced social reward learning, whereas controls increased investments in response to behavioural feedback in the negative context. Patients engaged the dorsolateral prefrontal cortex less than controls during context presentation and showed reduced activity within the caudate during repayments. In patients, lower investments were associated with more time spent alone and social exclusion and lower caudate activation was marginally significantly associated with higher perceived social exclusion. Conclusion: The failure to adapt trust to positive and negative social contexts suggests that patients have a general insensitivity to prior social information, indicating top-down processing impairments. In addition, patients show reduced Department of Clinical, Neuro and Developmental Psychology, Faculty of Behavioural and Movement Sciences, and Institute for Brain and Behaviour (IBBA) Amsterdam, Vrije Universiteit Amsterdam, Amsterdam, The Netherlands CSI Lab, Institute of Psychiatry, Psychology and Neuroscience, Department of Psychosis Studies, King’s College London, London, UK Hersencentrum Mental Health Institute, Amsterdam, The Netherlands Department of Psychology, City, University of London, London, UK Corresponding author: Esther Hanssen, Department of Clinical, Neuro and Developmental Psychology, Faculty of Behavioural and Movement Sciences, and Institute for Brain and Behaviour (IBBA) Amsterdam, Vrije Universiteit Amsterdam, Van der Boechorststraat 7, 1081 BT Amsterdam, The Netherlands. Email: emehanssen@gmail.com Australian & New Zealand Journal of Psychiatry, 56(1) 60 ANZJP Articles sensitivity to social reward, i.e. bottom-up processing deficits. Moreover, lower trust and lower neural activation were related to lower real-life social functioning. Together, these findings indicate that improving trust and social interactions in schizophrenia spectrum needs a multi-faceted approach that targets both mechanisms. Keywords Functional magnetic resonance imaging, experience sampling method, trust, schizophrenia, social context processing prefrontal cortex (dlPFC) (MacDonald et al., 2000). Introduction Second, theory of mind, i.e. the ability to infer the mental The ability to integrate social contextual information and states of others, in which medial prefrontal cortex (mPFC) behavioural feedback from others is necessary for success- and temporo-parietal junction (TPJ) are implicated as key ful social interactions (Ruz et al., 2011) and is an important regions (Carrington and Bailey, 2009; Schurz et al., 2014). foundation of trust in social relationships (Lewicki and Third, reward processing, which strongly involves the cau- Wiethoff, 2006). Individuals with a schizophrenia spectrum date nucleus (Krach et al., 2010; Sanfey, 2007). (SZ) diagnosis, hereafter referred to as patients, show a Deficits in mentalizing, social reward processing (bot- reduced ability in judging social signals (Penn et al., 2008). tom-up) and social context processing (top-down) have Previous research suggests two underlying mechanisms. been suggested to underlie lower trust, paranoia and social First, there are deficits in learning from others’ behavioural disconnection in SZ (Couture et al., 2006; Kapur et al., feedback during social interactions, suggesting problems in 2005; Velthorst et al., 2016). In SZ, prior studies have bottom-up mechanisms. Second, there are problems with reported reduced activation within the dlPFC during con- the integration of a priori contextual information in a top- text processing (Barch and Ceaser, 2012; Niendam et al., down way (Chung et al., 2010; Hooker et al., 2011). Here, 2014), within the mPFC and TPJ during mentalizing (Green we investigate how these two mechanisms of social infor- et al., 2015; Lee et al., 2004; Pinkham et al., 2008), and mation processing impact on social behaviour in real-time within the caudate during both non-social (Juckel et al., interactions, using a modified neuroeconomic trust game 2006b; Murray et al., 2008) and, more importantly, social (Fett et al., 2015). reward processing (Fett et al., 2019; Gromann et al., 2013). Studies employing the trust game (Berg et al., 1995) in This earlier work leads to the hypotheses that the dlPFC, SZ have demonstrated that patients invested lower initial mPFC, TPJ and caudate play an important role in disturbed amounts, indicating lower trust towards others (Fett et al., social decision-making and context processing in SZ. 2012; Gromann et al., 2013; Lemmers-Jansen et al., 2018). In this first-time investigation of the underlying mecha- In addition, SZ has been associated with a reduced ability nisms of disturbed trust and social interaction in SZ, we to use others’ cooperative behavioural feedback to adjust therefore probed the impact of different social contexts and trusting behaviour, i.e. bottom-up processing (Fett et al., investigated the underlying neural correlates in patients with 2012; Gromann et al., 2013). Research on the effects of SZ and healthy controls, using a modified version of an inter- social contextual information on trust in the general popu- active trust game while measuring brain activity with fMRI. lation demonstrated that trust increases in response to a Since social interactions are embedded in peoples’ daily lives trustworthy interaction partner, showing a strong impact of in a complex way, it is important to elucidate the association a priori information on trust in a top-down manner (Delgado between the neural processes underlying social interactions et al., 2005). Patients, in contrast, show a diminished sensi- and daily-life social engagement in SZ. To investigate this, tivity to such prior positive information (Fett et al., 2012). we combined fMRI and the experience sampling method The reduced ability to use bottom-up information and not (ESM), a diary method (Delespaul, 1995). Initial ESM stud- being able to overcome distrust during positive interactions ies (Kluge et al., 2018; Moran et al., 2019) started to investi- in response to such information, may explain real-life social gate how brain activation during task-based fMRI translates impairment seen in SZ (Velthorst et al., 2016) and could be to real-world functioning. This method ensures high ecologi- tackled with cognitive bias modification or other ways of cal validity because it allows for real-time monitoring of cognitive remediation. However, it has not been investi- behaviour in daily-life contexts. In patients, Moran et al. gated whether the insensitivity to social context reflects a (2019) found that greater hemodynamic signal change dur- more general processing deficit or a specific bias against ing (non-social) reward anticipation in caudate, insula and positive social information. anterior cingulate was associated with greater anticipated Three core cognitive mechanisms have been suggested pleasure and motivation for daily-life activities. to underlie trust and decision-making in social contexts We hypothesized that (1) patients would show a general (Declerck et al., 2013). First, context processing and cogni- reduced sensitivity to prior information about the counter- tive control, which are subserved by the dorsolateral part, reflected in no differences in baseline trust between Australian & New Zealand Journal of Psychiatry, 56(1) Hanssen et al. 61 conditions, i.e. first investments. Controls would increase symptom severity in the 2 weeks prior to testing in patients investments from the negative to the positive context; (2) (Kay et al., 1987). Fourteen items evaluate the severity of patients would not increase trust over trials in response to positive and negative symptoms (1 = absent to 7 = extreme). benevolent behavioural feedback, whereas controls would PANSS scores are reported in Table 1. do so; (3) patients would engage the regions-of-interest (ROIs) to a lesser extent than controls: (a) in the left dlPFC Trust game. To measure the impact of social context pro- during context presentation and investment (i.e. trusting) cessing in social interactions, we employed a modified decisions, because of its specific role in context processing; multi-round trust game (Berg et al., 1995; Gromann et al., (b) in the mPFC and right TPJ during context presentation 2013; King-Casas et al., 2005). In a multi-round classic and investment decisions, given their role in mentalizing trust game, the first player, i.e. the investor, is given an ini- mechanisms, which we expected to a greater extent in con- tial endowment of £10 and has to invest a chosen amount trols while processing prior social information compared to between zero and ten pounds. This amount is tripled and no information and (c) in the right caudate nucleus during given to the second player, i.e. the trustee. The trustee then the partner’s repayments (i.e. receipt of social reward); and decides whether and how much of the tripled amount he or (4) for patients, reduced trust and reduced activation in the she wants to give back to the investor. The chosen amount ROIs during the trust game is associated with lower daily- to invest by the investor reflects trust (given that the trustee life social functioning, i.e. more time spent alone, higher can choose not to return any money). In this study, partici- perceived social exclusion and lower perceived relation- pants played the role of the investor and played the trust ship quality, measured by ESM. game three times, with three different hypothetical counter- parts. In one game (condition), the trust game was pre- sented without prior social contextual information, i.e. a Methods classic multi-round trust game, while the other two condi- Subjects tions were modified to examine social context processing. In the negative and positive context conditions, participants Twenty-five patients with an SZ diagnosis and 26 controls first played three ‘blind’ rounds, without seeing the repay- without a personal or family history of SZ were included (for ments of the interaction partner. These rounds were imple- recruitment, see Supplement – A). Inclusion criteria were (1) mented to establish the cooperativeness of the trustee (Fett age 18–65 years, (2) good understanding of the English lan- et al., 2012). When making their blind investments, partici- guage and (3) intelligence quotient (IQ) >70. An additional pants saw the following the message: ‘Determining E.H.’s criterion for patients was an SZ diagnosis according to the initial average returns’. After these three ‘blind’ invest- International Statistical Classification of Diseases and ments, they viewed the following message on the screen Related Health Problems–Tenth Edition (ICD-10; World pointing out the average returns (more or less than invested) Health Organization (WHO), 1992), which was confirmed of the trustee during the blind investments (i.e. the social with the treating National Health Service (NHS) clinician. context): ‘On average your partner E.H. returned more/less Exclusion criteria were (1) a history of any neurological con- than you invested’. This message was shown before each ditions and (2) a diagnosis of alcohol/drug dependence trial in the trust game. Participants completed a total of 120 within 6 months. One control subject did not complete MRI trials (60 experimental and 60 control), equally divided scanning due to anxiety. Two patients were excluded from over the three conditions (no context, positive context and analyses due to excessive movement (framewise displace- negative context). Control trials were included to control ment ⩾1.5 mm in ⩾20% of the volumes per run). Therefore, for general effects of motor and visual activation elicited by analyses were performed on 23 patients and 25 controls. the task. An experimental and control trial with the respec- Forty-four participants completed the ESM measurements tive phases and timings is displayed in Figure 1. (20 patients and 24 controls). The London – Harrow Research Participants were instructed that they played with a real Ethics Committee (14/LO/0071) approved this study. human player via the Internet but were actually playing with a pre-programmed computer that behaved in the same Measures benevolent probabilistic manner in all three contexts (for algorithm, see Supplement – B). After completion of the Estimated cognitive ability. To assess an estimated cognitive trust game, participants were asked whether they thought ability, an abbreviated two-test version of the Wechsler the other players were real and trustworthy, on a 7-point Abbreviated Scale of Intelligence (WASI) was used Likert-type scale. Ratings on whether the other players (Wechsler, 1999), which consisted of the vocabulary sub- were unreal did not differ between groups (16% controls test and the matrix reasoning subtest. WASI scores are and 13% patients, p > 0.83), and were unrelated to invest- reported in Table 1. ments in the trust game (p > 0.82). Also, the ratings of trust- worthiness of the interacting partners did not differ between Symptoms. The Positive and Negative Syndrome Scale groups (p > 0.59). (PANSS) semi-structured interview was used to measure Australian & New Zealand Journal of Psychiatry, 56(1) 62 ANZJP Articles Table 1. Participant demographics and patient clinical characteristics. Controls (N = 25) Patients (N = 23) Statistic p-value Age – M (SD) 36.02 (7.34) 39.86 (9.10) β = 0.23 0.11 Gender (% male) 68.0 82.6 0.24 χ (1) = 1.36 IQ – M (SD) 116.68 (10.18) 98.30 (11.80) β = −0.64 <0.0001 Diagnoses (%) Schizophrenia 73.9 Schizoaffective disorder 17.4 Psychotic disorder 8.7 Medication (%) Atypical antipsychotics 82.6 Typical antipsychotics 13.0 None 4.4 PANSS – M (SD) Negative scale 1.75 (0.45) Positive scale 2.11 (0.86) Amotivation factor 1.92 (1.15) Diminished expression factor 2.45 (1.36) P6 (suspiciousness) 2.70 (1.23) Experience sampling (ESM) Alone time (%) 55 71 β = 0.31 0.04 Social exclusion – M (SD) 1.77 (0.94) 2.64 (0.81) β = 0.45 0.002 Relationship quality – M (SD) 5.38 (1.12) 5.18 (0.86) β = −0.10 0.53 M: mean; SD: standard deviation; IQ: intelligence quotient; PANSS: Positive and Negative Syndrome Scale; ESM: experience sampling method. Statistically significant differences are bold faced. ESM – measurement of social engagement. ESM (Palmier Participants first completed a demographic questionnaire et al., 2011), a structured diary technique, was used to mea- and several practice trials before playing the trust game in sure social engagement in daily life. The ESM device (iPod) the MRI scanner. They were told that they would receive the gave a signal to fill in the questionnaire 10 times a day, by earnings from one randomly selected round to keep them means of a pseudo-random ‘beep’ on 7 consecutive days. We motivated (between £0 and £30). For fairness reasons, all included several questions to probe real-life social function- participants received a payment of £5. After the trust game, ing in terms of social engagement, social exclusion and qual- they completed the questionnaire on their perception of the ity of social relationships. The question that was used game partners. Next, the PANSS interview was adminis- assessing social engagement was ‘Are you alone?’ (yes/no). tered. Finally, an explanation of the iPod was given to par- Perceived social exclusion was assessed when individuals ticipants and they completed one practice questionnaire. were alone using an average of the two items: ‘I feel lonely’ The morning after the first session the 7 days ESM data col- and ‘I feel excluded’ (Cronbach’s α = 0.82). The perceived lection started. All participants were contacted by phone on quality of social relationships was assessed when individuals day 2 for guidance in case of any problems or difficulties were in social company with an average of the four items: ‘I with the iPod. The second session consisted of assessment like the person(s) I am with’, ‘I feel close to them’, ‘They are of the WASI subtests. Participants handed in the iPod and dependable’ and ‘I trust them’ (Cronbach’s α = 0.90). ESM experiences were discussed. At the end of the study, each items were rated on a 7-point Likert-type scale. participant received payment (£60 + £5) for participation. Statistical analyses Procedure Behavioural analyses. Statistical analyses were performed The two sessions took place at the Institute of Psychiatry, using STATA version 14 (StataCorp, 2015). We examined Psychology and Neuroscience (IoPPN), King’s College group differences in demographics using chi-square tests London. Participants gave written informed consent before and regression analyses. Investments were analysed using the study. Australian & New Zealand Journal of Psychiatry, 56(1) Hanssen et al. 63 Figure 1. Left: investment trial with positive context (1) context cue: ‘initial repayment same or more’, (2) investment cue, (3) investment choice, made by scrolling over a horizontal bar ranging from £0 to £10, which started at £5, (4) invested amount displayed, (5) wait for game partners’ response (jittered), (6) fixation cross, (7) returned amount displayed, (8) round totals (kept and given amount added) for both players (jittered) and (9) fixation cross. Right: control trial, similar to the experimental trials, except for (a) ‘Invest’ was replaced by ‘Rest’ and (b) in the investment phase, participants had to move the cursor to the marked number. mixed effects multilevel random regression analyses TR = 7.312 ms; TE = 3.016 ms; flip angle = 11°; (MIXED) to account for repeated measurements within per- FOV = 270 mm) was acquired. sons, with (first) investment as dependent variables and with Imaging data were analysed using Statistical Parametric group (control and patient) and context (negative, no and Mapping 12. Pre-processing of the functional images con- positive) and their respective interactions as independent sisted of realign and unwarp, and coregistration to individ- variables. For investments over trials, trial number (1–20) ual anatomical images. Next, using unified segmentation, was added to the model. Mixed effects multilevel random anatomical images were segmented and normalization regression analyses (MIXED) were used to examine asso- parameters were estimated. These parameters were used to ciations between investments and ESM indices of real-life transform functional and anatomical images to a Montreal social functioning (% alone, social exclusion and quality of Neurological Institute (MNI) template. Subsequently, social relationships) across contexts. Interactions were smoothing was applied (Gaussian kernel 6 mm full-width at probed with the CONTRAST command. Analyses with a half-maximum), and the last three volumes were removed at significant group effect are additionally reported with esti- the ending of the task. mated IQ as a covariate. Additional analyses on association with symptoms are reported in Supplement – E. ROI analyses. A general linear model was used per run in which the three phases in the game were modelled as FMRI data acquisition and scanning parameters. Imaging regressors of interest for the experimental and control con- data were acquired using a 3T GE Signa Neuro-optimized dition separately (see Figure 1). The cue phase was time- magnetic resonance (MR) System at the Centre of Neuro- locked to the start of each trial (duration 2 seconds), the imaging Science of the IoPPN. Functional images were investment phase started after the investment cue (duration acquired by a T2*-weighted echo-planar imaging sequence 4 seconds) and the repayment phase started at the beginning scanning 39 axial slices of 3.0 mm thick with 0.3 mm gap. of the repayments shown and was modelled until the end of The in-plane resolution was 3.3 mm × 3.3 mm (FOV 211 × the displayed totals (duration 5–8 seconds). All other game 211), flip angle = 75°, TR = 2.00 seconds and TE = 30 ms. phases were combined into one regressor of no interest (the There were 413 volumes per run. For anatomical reference, investment cue, the invested amounts, waiting for the part- a T1-weighted image (196 slices; isotropic voxels of 1.2 mm; ners’ response and the two fixation crosses). All phases Australian & New Zealand Journal of Psychiatry, 56(1) 64 ANZJP Articles Table 2. Task performance – first investments in GBP (£). Negative context, No context, Positive context, Context mean (SD) mean (SD) mean (SD) difference Statistic p 95% CI Controls (N = 25) 5.00 (2.91) 6.28 (2.25) 7.60 (2.04) Negative < no −1.28 <0.01 [−2.24, −0.32] Positive > no <0.01 1.32 [0.36, 2.28] Patients (N = 23) 6.35 (2.90) 6.48 (2.74) 6.70 (2.93) Negative = no p = 0.85 Positive = no p = 0.75 SD: standard deviation; CI: confidence interval. Statistically significant differences are bold faced (p<0.01). were modelled using a box-car function convolved with the corrected at family-wise-error (FWE) whole-brain cluster hemodynamic response function (Friston et al., 1995). To significance threshold of p = 0.05 (see Supplement – C). correct for motion, the six realignment parameters and regressors for volumes detected as motion-corrupted, cal- Results culated by DVARS metric as implemented in FSL, version Behavioural analysis 6.00 (FMRIB’s Software Library, www.fmrib.ox.ac.uk/fsl) by FSL Motion Outliers (https://fsl.fmrib.ox.ac.uk/fsl/fsl Groups did not differ in age and gender. Patients had a wiki/FSLMotionOutliers) were included in the design lower estimated cognitive ability than controls. Percentage matrix, making the total number of regressors in the model of time spent alone was higher in patients than controls. variable for each individual, with a minimum of 10 regres- Patients felt lonelier and more excluded, but reported a sors (cue phase, investment phase, repayment phase, similar quality of their social relationships compared to regressor of no interest and six motion parameters). A high- controls (see Table 1). pass filter of 128 seconds was used. Subsequently, for each phase of interest, contrast images were created by contrast- Baseline trust: context effect and group differences in first ing a specific phase of the experimental condition with the investments. First investments were examined to establish corresponding phase in the control condition. context effects on baseline trust (Table 2). There was a sig- A priori ROI analyses were performed. Talairach coordi- nificant group-by-context interaction (χ (2) = 7.34, p = 0.02) nates were converted to MNI space (https://bioimagesuite which remained significant when estimated IQ was added web.github.io/webapp/mni2tal.html), resulting in the fol- to the model (χ (2) = 7.34, p = 0.02). Estimated IQ was not lowing ROI MNI coordinates: right caudate (17, 20, 3), significantly associated with baseline trust (p = 0.32). The right TPJ (50, −56, 27), mPFC (−3, 64, 24) (Gromann et al., context effect was only significant in controls, who made 2014) and the left dlPFC (−43, 18, 29) (MacDonald and lower investments in the negative than the no context con- Carter, 2003). ROIs were created in MarsBaR with an dition (b = −1.28, 95% confidence interval CI = [−2.23, 8-mm sphere (version 0.44; http://marsbar.sourceforge. −0.33], p = 0.009) and higher investments in the positive net). For each ROI and each subject, average signal change than the no context condition (b = 1.32, 95% CI = [0.35, (beta estimate) was extracted to investigate group and con- 2.28], p = 0.007). Patients’ first investments did not differ text effects, and to test association between the fMRI and by context (both p > 0.75). In all three contexts, first invest- the ESM data. Additional analyses on association with ments did not differ significantly between groups (all symptoms (PANSS positive scale, PANSS negative scale, p > 0.07). PANSS suspiciousness, PANSS amotivation factor and PANSS diminished expression factor; Supplement – E) and Changes in trust over trials: context effect and group differ- IQ are reported. Mixed effects multilevel random regres- ences. We examined interference of context information on sion analyses (MIXED) were used to examine associations changes in trust (i.e. investments) over time (Figure 2(a) between ROI beta estimates and ESM indices of real-life and (b)). The three-way interaction between group, context social functioning (% alone, social exclusion and quality of and trial was marginally significant (χ (2) = 5.07, p = 0.07), social relationships) across contexts. The results of the and analyses by group showed a marginally significant ROI-based analyses were Bonferroni corrected at α levels interaction of context-by-trial number in controls of 0.0125 per test (0.05/4, as tests were performed with data (χ (2) = 5.26, p = 0.07) but not patients (p = 0.21). Across from four ROIs). trials, both groups showed a context effect with significant differences between negative and no contexts, which was Exploratory analyses. We also performed exploratory whole- larger in controls than patients (controls: b = −0.64, 95% brain analyses, investigating neural activation beyond the CI = [−0.88, −0.39], p < 0.0001 and patients: b = −0.39, predefined ROIs for all three game phases. Analyses were 95% CI = [−0.67, −0.11], p = 0.006), but no differences Australian & New Zealand Journal of Psychiatry, 56(1) Hanssen et al. 65 Figure 2. Investments over trials in (a) controls and (b) patients, showing investments in the negative, no and positive context conditions. Investment phase. There were no significant group-by- between the positive and no contexts (both p > 0.41). Both context interactions (all p > 0.08), nor main effects of group groups showed a significant main effect of trial number; (all p > 0.50) or context in any of the ROIs (all p > 0.11). controls increased investments over time, while patients decreased their investments (controls: b = 0.03, 95% Repayment phase. There were no significant group-by- CI = [0.01, 0.04], p = 0.002 and patients: b = −0.04, 95% context interactions or context effects for the right caudate CI = [−0.06, −0.02], p < 0.0001). or mPFC. A significant group effect showed lower right cau- In addition, analyses by context showed that the effect of date activation in patients compared to controls (b = −0.14, trial number was most pronounced and in opposite direc- 95% CI = [−0.25, −0.03], p = 0.01, Figure 3(b)). This effect tions in patients and controls in the negative context (group- remained significant when the IQ estimate was added to by-trial number interaction (χ (2) = 21.21, p < 0.0001), the model as covariate (b = −0.16, 95% CI = [−0.30, 0.004], controls: b = 0.06, 95% CI = [0.03, 0.8], p < 0.001 and p = 0.028). The IQ estimate was not significantly related patients: b = −0.05, 95% CI = [−0.08, −0.02], p < 0.001). In to right caudate activation (p = 0.66). Higher mPFC acti- the no context condition, there was a group-by-trial number vation was found in patients than controls (b = 0.32, 95% interaction (χ (2) = 8.37, p = 0.004); no significant changes CI = [0.003, 0.62], p = 0.047). However, this result did not in investments were found controls (p = 0.51), but patients survive Bonferroni correction. There were no interaction, invested significantly less over time (b = −0.06, 95% group or context effects in any other ROI (all p > 0.16). CI = [−0.10, −0.01], p = 0.02). Groups did not differ signifi- Given that we observed a behavioural group difference cantly in the positive context condition and did not show in investments over trials, we performed exploratory analy- investment changes over trials (all p > 0.30). Additional ses probing the change in ROI activation over investment analyses between symptoms and behavioural results are trials. The results are reported in Supplement – D. reported in Supplement – E. Additional analyses between symptoms and ROI results are reported in Supplement – E. fMRI analysis Associations between investment, ROI activation and real-life ROI analyses by trust game phase social functioning in patients. We were specifically interested Cue phase. There were no significant group-by-context to examine whether lower neural activation within the left interactions in any ROIs (all p > 0.12). A significant group dlPFC and the right caudate found in patients was related to effect in the left dlPFC showed lower activation in patients the level of real-world social functioning. In addition, we than controls (b = −0.59, 95% CI = [−1.00, −0.18], p = 0.004, explored this association in the less robustly increased Figure 3(a)). The effect remained significant when the IQ mPFC signal in patients. estimate was added to the model as covariate (b = −0.71, Across contexts, higher investments were significantly 95% CI = [−1.24, −0.17], p = 0.009). The IQ estimate was associated with less time spent being alone (b = −0.02, 95% not significantly related to left dlPFC activation (p = 0.51). CI = [−0.05, −0.001], p = 0.04). A significant interaction Australian & New Zealand Journal of Psychiatry, 56(1) 66 ANZJP Articles Figure 3. Signal change (in arbitrary units) in (a) the dlPFC during the cue phase (i.e. social context presentation) and (b) the caudate during repayment (i.e. receipt of reward) in controls and patients. Error bars depict the standard errors. between context and perceived relationship quality on with real-life social functioning in SZ. Patients showed no investments (X(2) = 23.01, p < 0.0001) showed trend-level differential effect of social context on first investments, associations between higher relationship quality and higher regardless of the valence or the absence of a context, investments in the positive (b = 0.85, 95% CI = [−0.06, whereas controls showed the expected distinct context 1.78], p = 0.07) and negative (b = 0.94, 95% CI = [−0.06, effect with highest levels of trust in response to positive and 1.78], p = 0.06) conditions, but not the no context condition lowest levels of trust in response to negative social infor- (p = 0.79). There was a significant interaction between con- mation. This suggests a general insensitivity to social con- text and social exclusion on investments (X(2) = 13.19, text instead of a bias against positive social contextual p < 0.0001), indicating an association between lower per- information in patients. Patients also did not increase trust ceived social exclusion and higher investments the no con- in response to benevolent behavioural feedback whereas text condition only (b = −1.17, 95% CI = [−1.97, −0.35], controls did. Within the patient group, the findings indicate p = 0.005, other p > 0.23). an association between lower trust and lower real-world There were no significant associations between any social functioning. measure of real-life social functioning and dlPFC activa- We found overall lower activation in the left dlPFC dur- tion (all p > 0.21). A significant interaction between social ing context presentation and less engagement of the right exclusion and context emerged for the right caudate caudate nucleus during repayments in patients compared to (X(2) = 8.51, p = 0.01); lower perceived social exclusion controls. These results suggest that SZ is associated with a was marginally significantly associated with higher cau- general insensitivity to social contexts and with a reduced date activation in the positive context only (b = −0.17, 95% sensitivity to social reward. On the neural level, we also CI = [−0.36, −0.018], p = 0.07, all other p > 0.18). No sig- found an association between caudate activation and lower nificant associations were present with social relationship real-life social functioning in patients. quality or time spent alone (both p > 0.17). Higher per- ceived social relationship quality was associated with Social context effect on baseline trust higher mPFC activation (b = 0.27, 95% CI = [0.002, 0.54], p = 0.048). There were no significant associations between For baseline trust (i.e. the first investment where partner mPFC activation and social exclusion or time spent alone feedback has not yet been received), we found reduced sen- (both p > 0.22). sitivity to positive and negative social contextual informa- tion in patients compared to controls. This seems to reflect a general insensitivity to social contextual information. Discussion These results strengthen and extend previous evidence of a Using a novel, modified trust game, we examined the social context top-down processing deficit in SZ (Baez impact of social context on trust and social reward and the et al., 2013; Fett et al., 2012; Niendam et al., 2014); how- underlying neural activation patterns during real-time ever, this may not generalize to tasks concerning different social interactions. In addition, we probed the associations types of information processing. This insensitivity points to Australian & New Zealand Journal of Psychiatry, 56(1) Hanssen et al. 67 persistent a priori beliefs about other people in SZ. It is where reward processing takes place, we found context important to unravel whether this context sensitivity is a independent blunted activation in the right caudate nucleus risk factor for developing an SZ disorder or secondary to in patients compared to controls. The caudate is a highly the disorder or related factors, by examining whether this innervated by dopaminergic neurons (Björklund and insensitivity is also found in first-episode patients and indi- Dunnett, 2007). Aberrant regulation of dopamine is thought viduals at high risk for SZ. In this study, patients tended to to play a key role in SZ and reward processing (Howes and approach social interactions with similar trusting behaviour Kapur, 2009), and may account for the insensitivity to as controls, in line with findings in first-episode patients social feedback, i.e. reward. This finding adds to evidence (Fett et al., 2019), but contrasting results have also been showing deficits in social reward processing in SZ found in chronic and first-episode patients and individuals (Gromann et al., 2013; Lee et al., 2018) and may explain at clinical high risk (Fett et al., 2012; Gromann et al., 2013; why patients fail to increase trust over time. A less robust Lemmers-Jansen et al., 2018). effect was found in the mPFC; activation was higher in patients compared to controls. This was unexpected, but may be related to the role of the mPFC in reward-based Trust over time: the effects of action selection (Euston et al., 2012). We speculate that this benevolent feedback may point to a compensation mechanism for reduced As hypothesized, patients did not increase trust in response engagement of the caudate, which is tentatively supported to benevolent partner feedback, in line with previous litera- by the correlations with real-world social outcomes, which ture in patients with chronic psychosis (Fett et al., 2012; are discussed in the following section. Gromann et al., 2013). Controls, however, increased trust in the negative context and showed stable levels of trust Associations with real-life social engagement after no and positive information, suggesting that prior pos- itive beliefs about others were matched by the benevolent Previous social neuroscience studies have started to inves- partner feedback. Patients seem to have difficulties to over- tigate the link between the reward related processing in the come the given prior negative information, i.e. persistent a brain and daily life (Bakker et al., 2019; Moran et al., 2019), priori negative beliefs about others. Patients showed a ten- yet associations with real-life social functioning have not dency to reduce trust, even though subjective ratings of been considered in schizophrenia, despite yielding mean- trustworthiness of the interacting partner were similar for ingful insights in healthy controls (Bickart et al., 2011; both groups. Other studies have shown that patients show Kanai et al., 2012; Lewis et al., 2011). We found that in real reduced sensitivity to social rewards, such as smiles life, patients spent more time alone than healthy controls, in (Catalano et al., 2018). Our results strengthen the evidence line with previous work (Pinkham and Penn, 2006; Velthorst that patients have deficits in bottom-up processing of part- et al., 2016). They reported higher feelings of loneliness ner feedback, which might be due to an insensitivity to and social exclusion, but a similarly good relationship qual- social reward (Fett et al., 2012; Gromann et al., 2013). The ity as controls. On the behavioural level, higher invest- reduced ability to increase trust could explain patients’ ments were associated with higher social functioning in reduced motivation to engage in social behaviour (Krach real life. A higher level of trust may create an advantageous et al., 2010). basis for engaging in meaningful social relationships (Campellone et al., 2016). This could positively impact on social support networks, which in turn could aid in recov- Neural findings during social interactions ery (Corrigan and Phelan, 2004). At the neural level, we with contextual information found that in patients, lower activation in the caudate in the In support of our hypotheses, we found reduced left dlPFC positive context was marginally significantly associated activation in patients compared to controls during social with higher perceived social exclusion. More engagement context presentation. The dlPFC has been implicated in of this key reward area may be related to heightened experi- deficits in non-social context processing in schizophrenia ence of positive social interaction, i.e. social reward (Barch and Ceaser, 2012; Niendam et al., 2014) and is (Gromann et al., 2013). This could lead to a higher sense of viewed as key region in top-down cognitive control belonging or inclusion in social relationships. In addition, (MacDonald et al., 2000). The impairment in top-down higher activation in the mPFC was associated with higher modulation of trust in response to contextual information reported relationship quality in daily life, which would sup- may be related to reduced engagement of the dlPFC in SZ. port the role of mentalizing abilities (Schurz et al., 2014). This result that dlPFC activation was lower in all contexts However, it is possible that this finding also relates to supports a general insensitivity rather than a specific bias. higher feelings of (social) reward (Euston et al., 2012). We did not see any group differences in ROI activation dur- Future evidence from larger replication studies will be ing investments. However, during the repayment phase, needed to strengthen and support these initial results. Australian & New Zealand Journal of Psychiatry, 56(1) 68 ANZJP Articles behavioural and neural level. The findings suggest that Limitations increasing trust may facilitate social engagement in patients. The current results should be interpreted in light of the follow- Also, indices of real-life social functioning seem to be associ- ing limitations. First, this study is an initial investigation of ated with lower neural activity in reward- and context-pro- social context processing in relation to neural activation and in cessing (i.e. cognitive control) brain areas. This study suggests relation to daily-life social functioning in a relatively small that improving social interaction in SZ requires a multi-fac- sample of relatively stable patients with an SZ diagnosis. eted approach in clinical practice, which considers both bot- Specifically, our patient sample had relatively low symptom tom-up and top-down processing of social information. An levels: negative and positive scores of, respectively, 14.74 example of an intervention that incorporates these two facets (standard deviation (SD) = 6.15) and 12.26 (SD = 3.25). We is the Social Cognition and Interaction Training (SCIT) (Penn defined these scores as relatively low, since, with respect to the et al., 2007), targeting social skills bottom-up and social cog- 7-point Likert-type scale of the PANSS items and the number nition top-down, which shows promising effects on social of items in the positive and negative subscale, the item scores functioning. Clinical practice will benefit from newly devel- of 1 and 2 are, respectively, absent and minimal (Kay et al., oped treatments, building on the SCIT for instance, targeting 1987), (see also Supplement – E). Consequently, our results these facets in daily life by implementing an ecological need to be interpreted with caution when it comes to generaliz- momentary intervention by means of a smartphone-based ability and larger replication studies are warranted before any treatment. In addition, future studies are warranted to investi- firm conclusions can be made. Second, medication type has gate how this is related to fine-grained assessments of func- been found to have an effect on reward processing in SZ (Juckel tional capacity and real-world social functioning. et al., 2006a); however, the majority of participants in our sam- ple were on atypical antipsychotics (see Table 1), which are Acknowledgements thought to normalize reward processing (Nielsen et al., 2012; The authors thank the participants of the DECOP project and the Schlagenhauf et al., 2008). In addition, healthy first-degree National Institute for Health Research (NIHR) Clinical Research relatives of patients with SZ show reduced social reward pro- Organizations (CROs), as well as Farah Yakub, Marieke Helmich, cessing, without any medication confounds (Gromann et al., Marina Volguraki, Katie Wong, Mathew Harvey and Tracy Bobin 2014; Hanssen et al., 2020). Therefore, the current findings are for their support with testing and recruitment. not likely to represent an enhancement of the effects; they may even reflect an underestimation, compared to expected effects Declaration of Conflicting Interests in unmedicated patients. Studying unmedicated patients is val- The author(s) declared no potential conflicts of interest with uable, but poses a great challenge due to the clinical reality. respect to the research, authorship and/or publication of this Third, subjective self-report measures, like ESM, may raise the article. question of response accuracy or social desirable, which may be related to suspiciousness. However, ESM is a widely used Funding and well-validated method in psychiatric and schizophrenia The author(s) disclosed receipt of the following financial support samples (Delespaul, 1995; Myin-Germeys et al., 2018) and has for the research, authorship and/or publication of this article: This the advantage that it does not rely on retrospective recall work was supported by funding of the Netherlands Organization (Potheegadoo et al., 2012). Moreover, our sample showed a for Scientific Research (NWO) VENI grant (451-13-035) and a good compliance and did not report any issues related to the use NARSAD Young Investigator Grant from the Brain & Behaviour of the ESM app during the debriefing. We therefore have no Research Foundation (24138). grounds to assume that responses were inaccurate. Finally, not all participants believed they were playing with a real human ORCID iDs player; however, this was not associated with investments and Esther Hanssen https://orcid.org/0000-0002-4504-0757 adequate changes in investments in controls after context infor- Imke LJ Lemmers-Jansen https://orcid.org/0000-0001-9044- mation and behavioural feedback indicated that the experimen- tal manipulation was effective. Anne-Kathrin J Fett https://orcid.org/0000-0003-0282-273X Supplemental Material Concluding remarks Supplemental material for this article is available online. This is the first study to investigate the neural correlates of social context processing and the link with real-life social References functioning in SZ. We provide evidence that patients do not Baez S, Herrera E, Villarin L, et al. (2013) Contextual social cognition modulate their behaviour in response to social context infor- impairments in schizophrenia and bipolar disorder. PLoS ONE 8: mation (i.e. top-down processing) and positive behavioural e57664. cues of others (i.e. bottom-up processing). Our results point to Bakker JM, Goossens L, Kumar P, et al. (2019) From laboratory to a general reduced sensitivity to social information on a life: Associating brain reward processing with real-life motivated Australian & New Zealand Journal of Psychiatry, 56(1) Hanssen et al. 69 behaviour and symptoms of depression in non-help-seeking young Howes OD and Kapur S (2009) The dopamine hypothesis of schizophre- adults. Psychological Medicine 49: 2441–2451. nia: Version III – The final common pathway. Schizophrenia Bulletin Barch DM and Ceaser A (2012) Cognition in schizophrenia: Core psy- 35: 549–562. chological and neural mechanisms. Trends in Cognitive Sciences 16: Juckel G, Schlagenhauf F, Koslowski M, et al. (2006a) Dysfunction of 27–34. ventral striatal reward prediction in schizophrenic patients treated Berg J, Dickhaut J and McCabe K (1995) Trust, reciprocity, and social with typical, not atypical, neuroleptics. Psychopharmacology 187: history. Games and Economic Behavior 10: 122–142. 222–228. Bickart KC, Wright CI, Dautoff RJ, et al. (2011) Amygdala volume and Juckel G, Schlagenhauf F, Koslowski M, et al. (2006b) Dysfunction of social network size in humans. Nature Neuroscience 14: 163. ventral striatal reward prediction in schizophrenia. NeuroImage 29: Björklund A and Dunnett SB (2007) Dopamine neuron systems in the 409–416. brain: An update. Trends in Neurosciences 30: 194–202. Kanai R, Bahrami B, Duchaine B, et al. (2012) Brain structure links loneli- Campellone TR, Fisher AJ and Kring AM (2016) Using social outcomes ness to social perception. Current Biology 22: 1975–1979. to inform decision-making in schizophrenia: Relationships with Kapur S, Mizrahi R and Li M (2005) From dopamine to salience to psy- symptoms and functioning. Journal of Abnormal Psychology 125: chosis: Linking biology, pharmacology and phenomenology of psy- 310–321. chosis. Schizophrenia Research 79: 59–68. Carrington SJ and Bailey AJ (2009) Are there theory of mind regions in Kay SR, Fiszbein A and Opfer LA (1987) The Positive and Negative the brain? A review of the neuroimaging literature. Human Brain Syndrome Scale (PANSS) for schizophrenia. Schizophrenia Bulletin Mapping 30: 2313–2335. 13: 261–276. Catalano LT, Heerey EA and Gold JM (2018) The valuation of social King-Casas B, Tomlin D, Anen C, et al. (2005) Getting to know you: rewards in schizophrenia. Journal of Abnormal Psychology 127: 602. Reputation and trust in a two-person economic exchange. Science Chung YS, Mathews JR and Barch DM (2010) The effect of context 308: 78–83. processing on different aspects of social cognition in schizophrenia. Kluge A, Kirschner M, Hager OM, et al. (2018) Combining actigraphy, Schizophrenia Bulletin 37: 1048–1056. ecological momentary assessment and neuroimaging to study apathy Corrigan PW and Phelan SM (2004) Social support and recovery in people in patients with schizophrenia. Schizophrenia Research 195: 176–182. with serious mental illnesses. Community Mental Health Journal 40: Krach S, Paulus FM, Bodden M, et al. (2010) The rewarding nature of 513–523. social interactions. Frontiers in Behavioral Neuroscience 4: 22. Couture SM, Penn DL and Roberts DL (2006) The functional significance Lee J, Jimenez AM, Reavis EA, et al. (2018) Reduced neural sensitivity to of social cognition in schizophrenia: A review. Schizophrenia Bulletin social vs nonsocial reward in schizophrenia. Schizophrenia Bulletin 32: S44–S63. 45: 620–628. Declerck CH, Boone C and Emonds G (2013) When do people cooper- Lee K-H, Farrow T, Spence S, et al. (2004) Social cognition, brain net- ate? The neuroeconomics of prosocial decision making. Brain and works and schizophrenia. Psychological Medicine 34: 391–400. Cognition 81: 95–117. Lemmers-Jansen IL, Fett A-KJ, Hanssen E, et al. (2018) Learning to trust: Delespaul PA (1995) Assessing schizophrenia in daily life: The experience Social feedback normalizes trust behavior in first-episode psychosis sampling method. PhD Thesis, Maastricht University, Maastricht. and clinical high risk. Psychological Medicine 49: 780–790. Delgado MR, Frank RH and Phelps EA (2005) Perceptions of moral char- Lewicki RJ and Wiethoff C (2006) Trust, trust development, and trust acter modulate the neural systems of reward during the trust game. repair. The Handbook of Conflict Resolution: Theory and Practice Nature Neuroscience 8: 1611–1618. 2: 92–119. Euston DR, Gruber AJ and McNaughton BL (2012) The role of medial Lewis PA, Rezaie R, Brown R, et al. (2011) Ventromedial prefrontal prefrontal cortex in memory and decision making. Neuron 76: volume predicts understanding of others and social network size. 1057–1070. NeuroImage 57: 1624–1629. Fett A-KJ, Mouchlianitis E, Gromann PM, et al. (2019) The neural mech- MacDonald AW and Carter CS (2003) Event-related FMRI study of con- anisms of social reward in early psychosis. Social Cognitive and text processing in dorsolateral prefrontal cortex of patients with schiz- Affective Neuroscience 14: 861–870. ophrenia. Journal of Abnormal Psychology 112: 689–697. Fett A-KJ, Shergill SS and Krabbendam L (2015) Social neuroscience in MacDonald AW, Cohen JD, Stenger VA, et al. (2000) Dissociating the role psychiatry: Unravelling the neural mechanisms of social dysfunction. of the dorsolateral prefrontal and anterior cingulate cortex in cogni- Psychological Medicine 45: 1145–1165. tive control. Science 288: 1835–1838. Fett A-KJ, Shergill SS, Joyce DW, et al. (2012) To trust or not to trust: Moran EK, Culbreth AJ, Kandala S, et al. (2019) From neuroimaging to The dynamics of social interaction in psychosis. Brain 135: 976–984. daily functioning: A multimethod analysis of reward anticipation in Friston KJ, Frith CD, Turner R, et al. (1995) Characterizing evoked hemo- people with schizophrenia. Journal of Abnormal Psychology 128: dynamics with fMRI. NeuroImage 2: 157–165. 723–734. Green MF, Horan WP and Lee J (2015) Social cognition in schizophrenia. Murray G, Corlett P, Clark L, et al. (2008) Substantia nigra/ventral teg- Nature Reviews Neuroscience 16: 620–631. mental reward prediction error disruption in psychosis. Molecular Gromann P, Heslenfeld DJ, Fett AK, et al. (2013) Trust versus paranoia: Psychiatry 13: 239–276. Abnormal response to social reward in psychotic illness. Brain 136: Myin-Germeys I, Kasanova Z, Vaessen T, et al. (2018) Experience sam- 1968–1975. pling methodology in mental health research: New insights and tech- Gromann P, Shergill S, de Haan L, et al. (2014) Reduced brain reward nical developments. World Psychiatry 17: 123–132. response during cooperation in first-degree relatives of patients with Nielsen MO, Rostrup E, Wulff S, et al. (2012) Improvement of brain psychosis: An fMRI study. Psychological Medicine 44: 3445–3454. reward abnormalities by antipsychotic monotherapy in schizophrenia. Hanssen E, Krabbendam L, Robberegt S, et al. (2020) Social and non- Archives of General Psychiatry 69: 1195–1204. social reward learning reduced and related to a familial vulnerability Niendam TA, Lesh TA, Yoon J, et al. (2014) Impaired context processing in schizophrenia spectrum disorders. Schizophrenia Research 215: as a potential marker of psychosis risk state. Psychiatry Research: 256–262. Neuroimaging 221: 13–20. Hooker CI, Tully LM, Verosky SC, et al. (2011) Can I trust you? Negative Palmier-Claus JE, Myin-Germeys I, Barkus E, et al. (2011) Experience affective priming influences social judgments in schizophrenia. sampling research in individuals with mental illness: Reflections and Journal of Abnormal Psychology 120: 98–107. guidance. Acta Psychiatrica Scandinavica 123: 12–20. Australian & New Zealand Journal of Psychiatry, 56(1) 70 ANZJP Articles Penn DL, Roberts DL, Combs D, et al. (2007) Best practices: The Sanfey AG (2007) Social decision-making: Insights from game theory and development of the social cognition and interaction training pro- neuroscience. Science 318: 598–602. gram for schizophrenia spectrum disorders. Psychiatric Services Schlagenhauf F, Juckel G, Koslowski M, et al. (2008) Reward system acti- 58: 449–451. vation in schizophrenic patients switched from typical neuroleptics to Penn DL, Sanna LJ and Roberts DL (2008) Social cognition in schizo- olanzapine. Psychopharmacology 196: 673–684. phrenia: An overview. Schizophrenia Bulletin 34: 408–411. Schurz M, Radua J, Aichhorn M, et al. (2014) Fractionating theory of mind: Pinkham AE and Penn DL (2006) Neurocognitive and social cognitive A meta-analysis of functional brain imaging studies. Neuroscience & predictors of interpersonal skill in schizophrenia. Psychiatry Research Biobehavioral Reviews 42: 9–34. 143: 167–178. StataCorp (2015) Stata Statistical Software: Release 14. College Station, Pinkham AE, Hopfinger JB, Ruparel K, et al. (2008) An investigation of TX: StataCorp LP. the relationship between activation of a social cognitive neural net- Velthorst E, Fett A-KJ, Reichenberg A, et al. (2016) The 20-year longitu- work and social functioning. Schizophrenia Bulletin 34: 688–697. dinal trajectories of social functioning in individuals with psychotic Potheegadoo J, Cuervo-Lombard C, Berna F, et al. (2012) Distorted per- disorders. American Journal of Psychiatry 174: 1075–1085. ception of the subjective temporal distance of autobiographical events Wechsler D (1999) Wechsler Abbreviated Scale of Intelligence. New York: in patients with schizophrenia. Consciousness and Cognition 21: The Psychological Corporation; Harcout Brace & Company. 90–99. World Health Organization (WHO) (1992) The ICD-10 Classification Ruz M, Moser A and Webster K (2011) Social expectations bias decision- of Mental and Behavioural Disorders: Clinical Descriptions and making in uncertain inter-personal situations. PLoS ONE 6: e15762. Diagnostic Guidelines. Geneva: WHO. Australian & New Zealand Journal of Psychiatry, 56(1)

Journal

Australian & New Zealand Journal of Psychiatry – SAGE

Published: May 18, 2021

Keywords: Functional magnetic resonance imaging; experience sampling method; trust; schizophrenia; social context processing

There are no references for this article.