Trait paranoia shapes inter-subject synchrony in brain activity during an ambiguous social narrative

Abstract

Individuals often interpret the same event in different ways. How do personality traits modulate brain activity evoked by a complex stimulus? Here we report results from a naturalistic paradigm designed to draw out both neural and behavioral variation along a specific dimension of interest, namely paranoia. Participants listen to a narrative during functional MRI describing an ambiguous social scenario, written such that some individuals would find it highly suspicious, while others less so. Using inter-subject correlation analysis, we identify several brain areas that are differentially synchronized during listening between participants with high and low trait-level paranoia, including theory-of-mind regions. Follow-up analyses indicate that these regions are more active to mentalizing events in high-paranoia individuals. Analyzing participants' speech as they freely recall the narrative reveals semantic and syntactic features that also scale with paranoia. Results indicate that a personality trait can act as an intrinsic "prime," yielding different neural and behavioral responses to the same stimulus across individuals.

Fig. 1

Experimental protocol, distribution of trait paranoia scores, and inter-subject correlation (ISC) analysis. a Schematic of the experimental protocol. Participants came to the laboratory for an initial behavioral visit, during which they completed several computerized cognitive tasks as well as self-report psychological scales, one of which was the Green et al. Paranoid Thoughts Scale (GPTS). To minimize demand characteristics and/or priming effects, the fMRI scan visit took place approximately 1 week later. During this visit, subjects listened to an ambiguous social narrative in the scanner and then completed an extensive post-narrative battery consisting of both free-speech prompts and multiple-choice items. b Distribution of scores on the GPTS-A subscale across n = 22 participants, and median split used to stratify participants into low (≤18, blue) and high (≥19, orange) trait paranoia. c Schematic of ISC analysis. Following normalization to a standard template, the ISC of activation time courses during narrative listening was computed for each voxel (v, yellow square; enlarged relative to true voxel size for visualization purposes) for each pair of subjects (i,j), resulting in a matrix of pairwise correlation coefficients (r values). These values were then compared across paranoia groups using voxelwise linear mixed-effects models with crossed random effects to account for the non-independent structure of the correlation matrix

Fig. 2

Narrative listening evokes widespread inter-subject correlation (ISC) across the whole sample. Voxels showing significant ISC across the time course of narrative listening in all participants (n = 22). As expected, the highest ISC values were observed in auditory cortex, but several regions of association cortex in the temporal, parietal, frontal, and cingulate lobes as well as the cerebellum also showed high synchrony. Also included are three representative axial slices from the cerebellum (z coordinates in Talairach space: -38, -35, -29). Results are displayed at a voxelwise false-discovery rate (FDR) threshold of q < 0.001

Fig. 3

Trait paranoia modulates patterns of inter-subject correlation during narrative listening. a Results from a whole-brain, voxelwise contrast revealing brain regions that are more synchronized between pairs of high-paranoia participants than pairs of low-paranoia participants (contrast schematized in top panel, cf. Fig. 1c). Significant clusters were detected in the left temporal pole, two regions in the right medial prefrontal cortex (one anterior and one dorsal and posterior), and the left precuneus. No clusters were detected in the opposite direction (low > high). b Results from two whole-brain, voxelwise contrasts revealing brain regions that are more synchronized within a paranoia group than across paranoia groups. The first contrast (cool colors) revealed that left lateral occipital cortex was more synchronized within the low-paranoia group (i.e., low–low pairs) than across groups (i.e., high–low pairs; contrast schematized in top panel, cf. Fig. 1c). The second contrast (warm colors) revealed that right angular gyrus was more synchronized within the high-paranoia group (i.e., high–high pairs) than across groups. For all three contrasts, results are shown at an initial threshold of p < 0.002 with cluster correction corresponding to p < 0.05

Fig. 4

Inter-subject correlation (ISC) scales continuously with trait paranoia. Post-hoc analyses for two regions of interest (ROIs) that emerged from the dichotomized contrast between high- and low-paranoia groups (cf. Fig. 3a): left temporal pole (top row) and right medial prefrontal cortex (PFC, bottom row). a Location of ROI (left) and participant-by-participant ISC matrix (right) for the left temporal pole. Participants are ordered by increasing trait paranoia score. Each matrix element reflects the correlation between two participants’ activation time courses in the left temporal pole during narrative listening. Higher correlations are visible as one moves to the right and down along the diagonal, representing pairs of increasingly high-paranoia individuals. b Scatter plot of paranoia rank vs. median ISC value—i.e., the median of each row of the ISC matrix in a. Each dot represents a participant. Rank correlation indicates a significant monotonic relationship between trait paranoia and median ISC in left temporal pole (r_s = 0.71, p = 0.0002). c Location of ROI and participant-by-participant ISC matrix for the right medial PFC. Participants are ordered as in a. d Scatter plot of each participant’s paranoia rank vs. their median ISC value in the right medial PFC. As in b, rank correlation indicates a significant monotonic relationship between paranoia rank and median ISC (r_s = 0.63, p = 0.0016)

Fig. 5

Response to mentalizing events is stronger in high- compared to low-paranoia individuals. a Regions of interest (ROIs) for the event-related analysis. LTmpPole, left temporal pole; RmPFC, right medial prefrontal cortex; LTPJ, left temporo-parietal junction; LHeschl, left Heschl’s gyrus. b Comparison of beta coefficients for each ROI for the mentalizing-events regressor between the paranoia groups (low, blue; high, orange). Each dot represents a subject. Boxes represent the median and 25th/75th percentiles, and whiskers represent the minimum and maximum. *p = 0.01; **p < 0.007; n.s., not significant (p-values adjusted to control the false discovery rate at Q = 5%). c Comparison of beta coefficients for each ROI for the non-mentalizing-events regressor (the inverse of the mentalizing-events regressor shown in b). Each dot represents a subject. Boxes represent the median and 25th/75th percentiles, and whiskers represent the minimum and maximum. d Post-hoc continuous analysis: Beta coefficients for the mentalizing-events regressor plotted against paranoia rank (coefficient values are the same as in b). Left panel: the two ROIs in which beta coefficient was hypothesized to scale with trait paranoia (LTmpPole and RmPFC). Right panel: the two control ROIs (LTPJ and LHeschl). Correlations between paranoia rank and beta coefficient: LTmpPole, r_s = 0.57, p = 0.005; RmPFC, r_s = 0.64, p = 0.001; LTPJ, r_s = −0.04, p = 0.86, LHeschl, r_s = 0.02, p = 0.95. e Beta coefficients for the non-mentalizing-events regressor plotted against paranoia rank (coefficients are the same as in c). Left and right panels as in d. Correlations between paranoia rank and beta coefficients (all n.s.): LTmpPole, r_s = −0.28, p = 0.21; RmPFC, r_s = −0.22, p = 0.33; LTPJ, r_s = 0.085, p = 0.71; LHeschl, r_s = 0.17, p = 0.44

Fig. 6

Speech analysis reveals a signature of trait paranoia in behavioral response to the narrative. a Loadings of all semantic and syntactic categories for the first component from a partial least squares regression relating features of speech during narrative recall to trait paranoia score, sorted by strength and direction of association with paranoia (those positively related to paranoia at top in orange; those inversely related at bottom in blue). b Example sentences from participant speech transcripts containing words falling into the three of the top positive categories (affiliation, health, and anxiety) and one of the top negative categories (anger). c Rank correlations between participants’ trait-level paranoia and their self-report measures of 16 emotions following the narrative (self-report was based on a Likert scale from 1 to 5). Dotted lines represent approximate threshold for a significant correlation at p < 0.05 (uncorrected). Gray-shaded area indicates non-significance