Study Overview
The present study stems from a larger project focused on the nature and neurobiology of social affiliative deficits in psychosis (R01-MH110462)14, 32-34. Participants completed two assessments: a baseline clinical session and a two-phase laboratory session. At the baseline clinical session, eligibility was confirmed; participants provided informed written consent; and demographic, diagnostic, symptom, and other self-report data were acquired. Participants were instructed to abstain from taking sedatives/benzodiazepines for at least 12 hours prior to the MRI assessment. None of these individuals disclosed concerns or exhibited noteworthy withdrawal or rebound effects. Latency between the two sessions was <2 weeks (M=6.5 days, SD=2.9). During the two-phase laboratory session, participants completed (a) the SAET outside the scanner, and (b) the social and monetary incentive delay (SID/MID) paradigms inside the scanner (Figure 1) as well as additional tasks not reported here14. Following the last scan, participants were debriefed and compensated. Procedures were approved by the University of Maryland, Baltimore Institutional Review Board.
Participants
Recruitment. To capture a broad spectrum of motivation and pleasure deficits, maximizing range and statistical power, a mixed transdiagnostic adult sample—including both clinical and community participants—was recruited31. A modest number of psychiatrically healthy community participants was included (19.4%; Table 1) to ensure that the full range of affiliative function was captured31. Clinical participants were recruited from outpatient community mental health clinics in the Washington, DC-Baltimore metropolitan region. Community participants were recruited via online advertisements (e.g., Craigslist).
Enrollment Criteria. General inclusion criteria included 18-60 years of age, English fluency, and normal or corrected-to-normal vision, and consent to be videotaped during study participation. General exclusion criteria included moderate or severe substance use disorder in the past 6 months or mild substance use disorder in the past month, indexed by the Structured Clinical Interview for DSM-5 Research Version (SCID-5-RV)35; standard MRI contraindications; lifetime neurological, developmental, or cognitive disorder, indexed by medical history or cognitive testing; or a lifetime history of serious head injury. Clinical inclusion criteria included a lifetime psychotic disorder (Table 1), clinical stability (i.e., no inpatient hospitalizations in the past 3 months and no changes in psychoactive medication in the past month), indexed by medical history. Community inclusion criteria included absence of current psychiatric diagnoses or medication (past 6 months), and absence of lifetime psychotic, mood, or personality disorder, indexed by SCID-5 and self-report.
Final Sample. A total of 120 participants completed the baseline clinical assessment. Of these, 12 did not attend the neuroimaging session due to psychiatric hospitalization (n=1), study withdrawal (n=10), or inclement weather (n=1). The remaining 108 participants included a mixture of clinical (n=87) and community (n=21) participants. Of these, 39 participants were excluded from analyses due to study withdrawal (n=8), MRI safety concerns (n=5), poor fit in the scanner (n=3), technical problems (n=8), excessive movement (n=1), or inadequate task compliance (hit-rate <10% during any scan; n=14), yielding a final sample of 69 individuals (Table 1). Examining those who are included in imaging analyses (n=69) and those who were excluded (n=39) indicated no significant differences in gender, age, or education (all ps > .05) and no symptom differences as measured by the BPRS and CAINS scales (all ps > .05).
Clinical Assessments
Diagnoses. Diagnoses were determined using the SCID-5. Assessments were conducted by well-trained Master’s level interviewers supervised by doctoral-level clinical psychologists.
Clinician-Rated Symptoms. The Clinical Assessment Interview for Negative Symptoms (CAINS; Horan et al., 2011; Kring et al., 2013) is a well-established 13-item interview that indexes deficits in Motivation and Pleasure (MAP; 9 items; e.g., amotivation, asociality, and anhedonia; α=0.80) and Expression (4 items; e.g., affective flattening and alogia; α=0.87) (Supplementary Table S1). The CAINS has been successfully deployed in a variety of clinical and non-clinical populations9, 14, 27, 29, 36, 37. For hypothesis testing, MAP served as the primary index of social amotivation and anhedonia. The expanded Brief Psychiatric Rating Scale (BPRS) is a 24-item interview that was used to index Positive Symptoms (8 items; α=0.69), Depression/Anxiety (4 items; α=0.74), and Agitation (6 items; α=0.53)38, 39.
Self-Reported Social Function. The 7-item Interpersonal Relationships scale from the Specific Levels of Functioning (SLOF) instrument was used to index interpersonal functioning (α=0.89) (Supplementary Table S1)40, 41. Consistent with prior studies, in the current sample more severe MAP symptoms were associated with poorer interpersonal functioning (r = -.56, p < .001).
Social Affiliation Enhancement Task (SAET)
The Social Affiliation Enhancement Task (SAET) encompasses a validated suite of procedures for cultivating social rapport, trust, and affiliation (Figure 1a) (for details, see Refs. 10, 14). Prior work by our group demonstrates that the SAET significantly enhances affiliative feelings, perceived closeness, and willingness to interact with the partner14, consistent with work using similar paradigms17.
Social and Monetary Incentive Delay (SID/MID) fMRI Paradigms
Overview and Procedures. As shown in Figure 1b, paralell incentive-delay paradigms were used to probe neural reactivity to social and monetary reward42, 43. Both paradigms took the form of balanced 3-condition (Reward Level: High, Low, None) randomized, event-related, repeated-measures designs (paradigm order counterbalanced; 2 scans/paradigm; 22 trials/condition/scan). General task structure, timing, and procedures were identical across paradigms. Because we did not harbor a strong a priori hypothesis about the impact of MAP symptoms on the anticipation-versus-presentation of social reward, trial timing was optimized via simulations to maximize the detection of global differences in reward sensitivity, while remaining mindful of participant burden and tolerability (variance inflation factors < 2.55). Participants were completely informed about the task structure and contingencies prior to scanning. They were instructed that the goal of both paradigms was to maximize reward receipt and that this was contingent on the speed of their response to a briefly presented visual target. Responses were made using the first digit of the dominant hand and an MRI-compatible response-pad (MRA, Washington, PA). To maintain a comparable level of difficulty across paradigms, trials, and participants, the response-time threshold (signaled by the duration of the target presentation) was adaptively adjusted on a trial-by-trial basis (±25-ms; target hit-rate: 66%). Too-slow responses (‘misses’) triggered the presentation of the No-Reward audiovisual clips (Figure 1c). No-Reward clips were presented on all No-Reward trials, irrespective of response time (hit/miss). Prior to scanning, participants practiced abbreviated versions of the paradigms and staff provided feedback as necessary to ensure participant comprehension. Stimulus presentation and behavioral data acquisition was controlled using Presentation (version 19.0, Neurobehavioral Systems, Berkeley, CA). Hit-rate was matched across paradigms, t(68)=0.85, p=0.40 (SID: M=64.8%, SD=0.08; MID: M=65.5%, SD=0.05) and unrelated to the severity of MAP symptoms, |r|<0.08, p>0.51.
SID Outcomes. Prior neuroimaging studies of social reward in psychosis have relied on static photographs of positive facial expressions posed by unfamiliar adult models27-29. Here we capitalized on naturalistic audiovisual clips of the experimental partner from the SAET, enhancing ecological validity and translational relevance (Figure 1). Building on preclinical work in university students44, this approach enabled us to manipulate the intensity of nonverbal (facial expressions and gestures), paralinguistic (vocal intonation), and verbal (praise) indicators of social reward expressed by a psychologically meaningful social partner. High-Reward clips featured large open-mouth smiles, thumbs-up gestures, and verbal feedback indicative of exceptional performance (Amazing!, Awesome!, Fabulous!, Fantastic!, Spectacular!) and expressed in an ebullient manner (Figure 1c). Low-Reward clips featured small closed-mouth smiles and verbal feedback indicative of good performance (Decent, That was cool, That was fine, That was nice, That was neat), expressed in a mildly positive manner. No-Reward clips were devoid of facial expressions and gestures; instead, the partner simply instructed the participant to prepare for the next trial (Continue, Get ready, Keep going, Next one, Proceed) in a neutral monotone.
MID Outcomes. As shown in Figure 1c, High-Reward audiovisual clips featured 10 coins falling into a bowl, Low-Reward clips featured 4 coins falling in a bowl, and No-Reward trials featured confetti falling into a bowl. In addition to the audiovisual clips, successful performance of the High- and Low-Reward MID trials was incentivized by $1.00 and $0.20, respectively, in monetary compensation. On average, participants earned $32.59 (SD=1.25).
MRI Data Acquisition
MRI data were acquired using a Siemens Magnetom TIM Trio 3 Tesla scanner (32-channel head-coil). Foam inserts were used to mitigate potential motion artifact. To further mitigate motion artifact, for the final 14 participants, a strip of medical tape was positioned just above the forehead, providing tactile feedback45. Sagittal T1-weighted anatomical images were acquired using a magnetization prepared rapid acquisition gradient echo (MPRAGE) sequence (TR=2,400 ms; TE=2.01 ms; inversion=1,060 ms; flip=8°; slice thickness=0.8 mm; in-plane=0.8 mm2; matrix=300×320; field-of-view=240×256). A T2-weighted image was collected co-planar to the T1-weighted image (TR=3,200 ms; TE=564 ms; flip=120°). To enhance resolution, a multi-band sequence was used to collect oblique-axial echo planar imaging (EPI) volumes (acceleration=6; TR=1,250 ms; TE=39.4 ms; flip =36.4°; slice thickness=2.2 mm, number slices=66; in-plane=2.1875 mm2; matrix=96×96; 355 volumes × 4 scans). Images were collected in the oblique axial plane (approximately −20° relative to the AC-PC plane) to minimize potential susceptibility artifacts. The scanner automatically discarded 7 volumes prior to the first recorded volume. To enable fieldmap correction, two oblique-axial spin echo (SE) images were collected in each of two opposing phase-encoding directions (rostral-to-caudal/caudal-to-rostral) co-planar to the functional volumes (TR=7,220 ms; TE=73 ms). Respiration and pulse were acquired using a respiration belt and photo-plethysmograph affixed to the first digit of the non-dominant hand. Participants were continuously monitored using an MRI-compatible eye-tracker (Eyelink 1000; SR Research, Ottawa, Ontario, Canada) and the AFNI real-time motion plugin46. Eye-tracking data were not recorded.
MRI Data Processing Pipeline
Methods were optimized to minimize spatial normalization error and other potential sources of noise, and are similar to those detailed in other recent reports by our group14, 47, 48. Data were visually inspected before and after processing for quality assurance. All participants provided 4 usable scans.
Anatomical Data. T1-weighted images were inhomogeneity corrected using N449 and filtered using ANTS DenoiseImage50. Brains were extracted using BEaST51 and brain-extracted-and-normalized reference-brains52. Brain-extracted T1 images were normalized to a version of the brain-extracted 1-mm T1-weighted MNI152 (version 6) template modified to remove extracerebral tissue53. Normalization was performed using the diffeomorphic approach implemented in SyN (version 2.3.4)50. T2-weighted images were rigidly co-registered with the corresponding T1 prior to normalization. The brain-extraction mask from the T1 was then applied. Tissue priors were unwarped to native space using the inverse of the diffeomorphic transformation54. Brain-extracted T1 and T2 images were segmented—using native-space priors generated in FAST (version 6.0.4)55—for subsequent use in T1-EPI co-registration (see below).
Fieldmap Data. SE images and topup were used to create fieldmaps. Fieldmaps were converted to radians, median-filtered, and smoothed (2-mm). The average of the motion- and distortion-corrected SE images was inhomogeneity corrected using N4 and masked to remove extracerebral voxels using 3dSkullStrip (version 20.2.14).
Functional Data. EPI files were de-spiked using 3dDespike, slice-time corrected to TR-center using 3dTshift, and motion corrected to the first volume using ANTS (12-parameter affine). Transformations were saved in ITK-compatible format for subsequent use56. The first volume was extracted and inhomogeneity corrected for EPI-T1 co-registration. The reference EPI volume was simultaneously co-registered with the corresponding T1-weighted image in native space and corrected for geometric distortions using boundary-based registration55. This step incorporated the previously created fieldmap, undistorted SE, T1, white matter (WM) image, and masks. To minimize potential normalization error, reference EPI volumes were spatially normalized to the MNI template using SyN, intensity standardized, and averaged to create a study-specific EPI template57-59. Normalized EPI reference volumes were then normalized to the study-specific. To minimize incidental spatial blurring, the operations necessary to transform each EPI volume from native space to the reference EPI, from the reference EPI to the T1, from the T1 to the MNI template, and from the MNI template to the study-specific EPI template were concatenated and applied to the processed EPI data in a single step. Normalized EPI data were resampled (2 mm3) using fifth-order b-splines and spatially smoothed (6-mm) using 3DblurInMask.
fMRI Data Modeling
General Approach. For each participant, first-level modeling was performed using general linear models (GLMs) implemented in SPM12 (version 7771), using the default autoregressive model and temporal band-pass filter set to the hemodynamic response function (HRF) and 128 s60. Consistent with past work14, 47, 48, nuisance variates included volume-to-volume displacement and its derivative, motion (6 standard parameters, global volume-to-volume displacement, and temporal derivatives), cerebrospinal fluid (CSF) signal, instantaneous pulse and respiration signals, and ICA-derived nuisance signals (e.g., global motion)61. Volumes with excessive volume-to-volume displacement (>0.66 mm) were censored. The inter-trial interval served as the implicit baseline.
Hypothesis Testing. For hypothesis testing purposes, reward signals were modeled using variable-duration rectangular (‘box-car’) regressors that spanned the entire trial, separately for each combination of reward level (High, Low, None) and outcome (Hit/Miss) (Figure 1b). Regressors were convolved with a canonical hemodynamic response function (HRF) and its temporal derivative.
Reward Anticipation and Presentation. To explore the relevance of finer differences in neural reward signaling, we separately modeled the anticipation and presentation phases of the trial using delta functions time-locked to the onset of the cue and outcome, respectively, for each combination of reward level and outcome (Figure 1b). Although our incentive-delay paradigms were not originally optimized for this modeling approach, collinearity proved acceptable (variance inflation factors <3.36)62. Regressors were convolved with a canonical HRF.
Analytic Strategy
Overview. Analyses were implemented in SPSS (version 27.0.1; IBM, Armonk, NY), SPM1260, and in-house MATLAB code (version 9.14.0.2239454; The MathWorks, Natick, MA). Diagnostic procedures and data visualizations were used to confirm that test assumptions were satisfied63 and key conclusions remained unchanged using robust regression (not reported)64. Some figures were created using created using R (version 4.0.2)65, Rstudio (version 1.2.1335)66, ggplot2 (version 3.4.1)67, and MRIcron (version 1.0.20190902)68. Clusters and peaks were labeled using the Harvard–Oxford atlas69-71, supplemented by descriptions of the orbitofrontal cortex, the ventral striatum, and its two major divisions: the core and shell (Supplementary Figure S1)72-77.
Confirmatory Testing. Whole-brain voxelwise (‘second-level’) repeated-measures (‘random effects’) general linear models (GLMs) were used to confirm that the SID and MID tasks robustly engaged the ventral striatum, as indexed by the cardinal High-versus-No-Reward contrast for hit trials. Significance was assessed using p<0.05, whole-brain familywise error (FWE) corrected for cluster extent, and a cluster-defining threshold of p<0.00178.
Hypothesis Testing. The overarching goal of this study was to test the hypothesis that blunted ventral striatum reactivity to social incentives is associated with more severe clinician-rated MAP symptoms. To do so, we used a standard voxelwise regression, with mean-centered CAINS MAP as the predictor, mean-centered biological sex and age as nuisance variates, and the High-versus-No-Reward contrast as the outcome (Figure 1d), consistent with prior work26, 28. Significance was assessed using p<0.05, FWE corrected for the volume of the anatomically defined ventral striatum (Figure 1d)79. The same approach was used to probe potential associations with ventral striatum reactivity to monetary incentives.
Specificity Analyses. When a significant association was observed, a voxelwise multiple regression was used to test whether ventral striatum reactivity to that incentive (e.g., social) continued to explain significant variance in MAP symptoms when statistically controlling for mean-centered reactivity to the other incentive (e.g., monetary), sex, and age (p<0.05, ventral striatum FWE corrected). For a similar voxelwise-covariate approach, see Ref. 80. Follow-up analyses also allowed us to test whether MAP symptoms explain significant variance in ventral striatum reward signaling, over-and-above mean-centered affective flattening/alogia, positive symptoms, depression/anxiety, agitation, and binary diagnostic status (i.e., case-versus-control; p<0.05, ventral striatum FWE corrected).
Secondary Analyses. The same general approach was used to determine the relevance of disaggregating striatal responses to the anticipation-versus-presentation of reward (see above for fMRI modeling details). Here again, when a significant association was detected, voxelwise multiple regression was used to test whether ventral striatum reactivity to that phase of the trial (e.g., anticipation) continued to explain significant variance in MAP symptoms when statistically controlling for mean-centered reactivity to the other phase (e.g., presentation), sex, and age (p<0.05, ventral striatum FWE corrected). For a similar approach, see Ref. 81.
Exploratory Analyses. Voxelwise regressions were used to explore potential associations between ventral striatum reward signaling and self-reported interpersonal functioning (SLOF; p<0.05, ventral striatum FWE corrected), and to assess associations between MAP symptoms and reward signaling in other, less intensively scrutinized brain regions (p<0.05, whole-brain FWE corrected).