In amygdala we trust: different contributions of the basolateral and central amygdala in learning whom to trust

Human societies are built on cooperation and mutual trust, but not everybody is trustworthy. Research on rodents suggests an essential role of the basolateral amygdala (BLA) in learning from social experiences (Hernandez-Lallement J et al., 2016), which was also confirmed in human subjects with selective bilateral BLA damage as they failed to adapt their trust behavior towards trustworthy vs. untrustworthy interaction partners (Rosenberger LA et al., 2019). However, neuroimaging in neurotypical populations did not consistently report involvement of the amygdala in trust behavior. This might be explained by the difficulty of differentiating between amygdalas structurally and functionally different subnuclei, i.e., the BLA and central amygdala (CeA), which have even antagonistic features particularly in trust behavior (van Honk J et al., 2013). Here, we used fMRI of the amygdala subnuclei of neurotypical adults (n=31f/31m) engaging in the repeated trust game. Our data show that both the BLA and the CeA play a role and indeed differentially: While the BLA was most active when obtaining feedback on whether invested trust had been reciprocated or not, the CeA was most active when subjects were preparing their next trust decision. In the latter phase, improved learning was associated with higher activation differences in response to untrustworthy vs. trustworthy trustees, in both BLA and CeA. Our data not only translate to rodent models and support our earlier findings in BLA-damaged subjects, but also show the specific contributions of other brain structures in the amygdala-centered network in learning whom to trust, and better not to trust. SIGNIFICANCE STATEMENTIn this fMRI study, the central amygdala was found active during trust behavior planning, while the basolateral amygdala was active during outcome evaluation. When planning trust behavior, central and basolateral amygdala activation differences between the players was related to whether participants learned to differentiate the players trustworthiness. Nucleus accumbens tracked whether trust was reciprocated but was not related to learning. This suggests learning whom to trust is not related to reward processing in the nucleus accumbens but rather to engagement of the basolateral amygdala. This study overcomes major empirical gaps between animal models and human neuroimaging and shows how different amygdala subnuclei and connected areas orchestrate learning to form different subjective trustworthiness beliefs about others and guide trust choice behavior.