Research ArticleNEUROSCIENCE

The neural circuitry of affect-induced distortions of trust

See allHide authors and affiliations

Science Advances  13 Mar 2019:
Vol. 5, no. 3, eaau3413
DOI: 10.1126/sciadv.aau3413
  • Fig. 1 Experimental task and electrophysiological and behavioral findings.

    (A) Schematic representation of hybrid functional magnetic resonance imaging (fMRI) design, trial sequence, and timing (see Materials and Methods). Participants faced blocks of trust (human icon) and NS control (NSC; computer icon) trials in random order. During trust and NS control blocks, participants expected either strong (threat) or weak (no threat) tactile stimulation at unpredictable times. At the beginning of each block, a 750-ms visual cue followed by tactile stimulation reminded participants of the game type (trust or NS control) and stimulation intensity (weak or strong) for the current block. On each trial, participants chose how much of their endowment of 24 CHF to transfer to a stranger (trust game) or invest in an ambiguous lottery that provided a 40 to 60% probability of returning an amount greater than the investment (NS control game). (B) The threat of an aversive tactile stimulation leads to a strong increase in skin conductance responses (SCRs) in the trust game (orange; P < 0.0001) and the NS control game (blue; P < 0.0001). (C) In the threat condition (relative to the no-threat condition), participants transferred significantly less to an anonymous stranger in the trust game (orange; P < 0.005; reduction due to threat in 71% of participants) and invested less into an ambiguous lottery in the NS control game (blue; P < 0.005; reduction due to threat in 73% of participants). These results are driven by the emotional arousal induced by the threat of a shock and not by the actual experience of shocks shortly before choice (table S1). (D) In the threat condition (relative to the no-threat condition), participants made their decisions significantly faster in both the trust (orange; P < 0.005) and the control (blue; P < 0.05) game. Dot plots in all panels reflect the change for each participant in the presence of threat (threat–no threat) in mean SCRs (B), mean transfers (C), and mean response latencies (D). These plots thus show the enhancement of affective arousal and the reduction of mean transfer and response latency due to threat.

  • Fig. 2 The impact of aversive affect on trust-specific TPJ activity and connectivity.

    (A) The region of left TPJ (peak at xyz = −57, −60, 27) that is selectively involved in trust compared to the NS control task as reflected by a significant main effect of game type (shown in green; see also table S2A). Aversive affect induced by the threat of a shock reduced activation in the left TPJ (relative to no threat) significantly more during the trust game than in the NS control game (significant interaction effect; peak at xyz = −60, −54, 19; table S2B). Voxels whose activity reflects this interaction effect are shown in red. All regions survived SV FWE correction, P < 0.05 (see Materials and Methods). Threat-induced reduction of TPJ activity was observed in 78% of participants during trust decisions (downward-sloping connecting lines) and in 44% of participants during NS control decisions, as shown in (B). The parameter estimates in (B) are extracted from a sphere (6-mm radius) around individual peaks within the TPJ cluster marked in red in (A). (C) The left amygdala (peak at xyz = −28, −6, −14; see table S3A) shows significantly stronger connectivity with TPJ during trust relative to control decisions as reflected by a significant main effect of threat. This coupling is disrupted by the threat of a shock specifically during trust as compared to the NS control task (significant interaction effect; peak at xyz = −26, 0, −23; shown in red). All regions survive P < 0.05 SV FWE-corrected (see Materials and Methods). Threat-induced reduction of TPJ-amygdala connectivity was observed in 76% of participants during trust decisions (downward-sloping connecting lines) and in 44% of participants during NS control decisions, as shown in (D). The parameter estimates are extracted from a 6-mm sphere around the individual peaks within the amygdala cluster marked in yellow in (C) to visualize the specific effects of aversive affect on functional connectivity between the left TPJ and left amygdala during decisions in the trust game. Dot plots in (B) and (D) reflect individual participant mean activation in each condition and are connected to illustrate the suppression of activity due to threat for each participant.

  • Fig. 3 TPJ functional connectivity is associated with transfer rates in the trust game.

    Trust-specific functional connectivity (A, B, and D) and threat-induced breakdown of connectivity (C) between the TPJ and a network of target regions. (A, B, and D) Trust-specific associations between transfer rate and trust-related neural activity reflecting the main effect of trust are shown in green activation clusters: Connectivity between the left TPJ and its targets is positively associated with trust (orange regression lines) but not with NS control decisions (blue regression lines) in the (A) bilateral posterior STS (pSTS; left peak at xyz = −62, −52, −5; right peak at xyz = 64, −43, 4), (B) dmPFC (peak at xyz = −12, 54, 40), anterior insula (left peak at xyz = −51, 21, −6; right peak at xyz = 56, 18, 1), and amygdala (peak at xyz = 28, 2 , −20). In contrast, mean transfers (i.e., investments) during the NS control task (blue regression lines) are associated with reduced connectivity strength between TPJ and these regions. In all cases, the correlation between mean transfer and connectivity strength is stronger in the trust game compared to the NS control task (see table S4A for ROI analyses). (C) The results from the interaction contrast reflecting a trust-specific breakdown of the association between mean trust and TPJ connectivity are shown in the red activation cluster in STS (peak at xyz = 64, −43, 6; table S4B): Aversive affect causes a breakdown of the association between TPJ-pSTS connectivity and mean trust. The correlation between mean trust levels and TPJ-pSTS connectivity is stronger in the no-threat compared to the threat condition (peak at xyz = 64, −43, 6; see table S4C). Specifically, there is a positive association between TPJ-pSTS connectivity and the mean trust level when distortionary aversive affect is absent (green regression line), which is eliminated by threat (red regression line). This suggests that connectivity between the TPJ and its target region in pSTS supports general trust taking only in the absence of threat. The regression lines in (A to D) predict functional connectivity strength as a function of mean transfer levels based on an extended ordinary least squares model that estimates both the slope of the relationship between mean transfers and functional connectivity in the NS control task and the increase in this relationship in the trust task (relative to the NS control task). For this purpose, we extracted data from 6-mm spheres around individual interaction peak voxels (see Materials and Methods). Confidence bounds around regression lines reflect 95% confidence intervals around the model fit.

  • Fig. 4 The impact of aversive affect on choice domain–independent neural correlates of decision-making.

    We tested the main effect of aversive affect on the neural correlates of decision-making independent of the choice domain (social and NS). This analysis revealed a domain-general network consisting of the bilateral posterior dlPFC [right peak at xyz = −62, −4, 18 (A); left peak at xyz = 62, −6, 28 (B)] and two clusters in the vmPFC [left peak at xyz = −10, 44, −8; right peak at xyz = 6, 21, −14 (C)] and left vlPFC [peak at xyz = −48, 41, −8 (D)]. These regions show significant threat-related suppression (no threat > threat; regions shown in green) in choice-related activity during both trust and NS control trials. Additional regions that are shown in fig. S3 include the left amygdala (B) (peak at xyz = −24, −15, −23) and posterior paracentral lobule (peak at xyz = 4, −36, 69). Time courses reflect choice domain–independent activity that shows suppressions due to the aversive affect during decisions in both trust and NS control trials. To illustrate the equivalent effect of aversive affect, figs. S3 and S4 show activity for both trust and control trials in separate graphs. Time courses were extracted from 6-mm spheres around peak voxels. The 5.5-s choice period is displayed in yellow and is shifted for 4 s to account for the hemodynamic lag.

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/5/3/eaau3413/DC1

    Section S1. Ordinary least squares regression investigating the influence of experienced electrical stimulation on choice

    Section S2. Behavioral differences between trust and control decisions

    Section S3. Assessing complexity differences across game types using choice latency

    Section S4. Assessing the lateralization of neuroimaging results

    Fig. S1. Manipulation checks.

    Fig. S2. Additional post hoc inspection of the significant interactions reported in the main paper within all voxels of independent TPJ and amygdala masks.

    Fig. S3. Main effect of threat 1: Suppression of game type–independent neural correlates of decision-making.

    Fig. S4. Main effect of threat 2: Enhancement of game type–independent neural correlates of decision-making.

    Table S1. Ordinary least squares regression results reflecting the influence of experienced electrical stimulation on choice behavior.

    Table S2. ROI analyses investigating trust-specific neural correlates in regions associated with social cognition and valuation.

    Table S3. ROI analyses investigating TPJ-amygdala connectivity patterns.

    Table S4. ROI analyses investigating brain-behavior relationships.

    Table S5. Whole-brain analyses investigating brain-behavior relationships.

    Table S6. Whole-brain analyses investigating the main effect of threat.

    Reference (62)

  • Supplementary Materials

    This PDF file includes:

    • Section S1. Ordinary least squares regression investigating the influence of experienced electrical stimulation on choice
    • Section S2. Behavioral differences between trust and control decisions
    • Section S3. Assessing complexity differences across game types using choice latency
    • Section S4. Assessing the lateralization of neuroimaging results
    • Fig. S1. Manipulation checks.
    • Fig. S2. Additional post hoc inspection of the significant interactions reported in the main paper within all voxels of independent TPJ and amygdala masks.
    • Fig. S3. Main effect of threat 1: Suppression of game type–independent neural correlates of decision-making.
    • Fig. S4. Main effect of threat 2: Enhancement of game type–independent neural correlates of decision-making.
    • Table S1. Ordinary least squares regression results reflecting the influence of experienced electrical stimulation on choice behavior.
    • Table S2. ROI analyses investigating trust-specific neural correlates in regions associated with social cognition and valuation.
    • Table S3. ROI analyses investigating TPJ-amygdala connectivity patterns.
    • Table S4. ROI analyses investigating brain-behavior relationships.
    • Table S5. Whole-brain analyses investigating brain-behavior relationships.
    • Table S6. Whole-brain analyses investigating the main effect of threat.
    • Reference (62)

    Download PDF

    Files in this Data Supplement:

Navigate This Article