Repository logo
 

97. NEURAL AND BEHAVIORAL EVIDENCE FROM REINFORCEMENT LEARNING CONVERGE IN SUPPORT OF RELAXED BELIEFS UNDER LSD


Change log

Abstract

Abstract

              Background
              The REBUS (RElaxed Beliefs Under pSychedelics) and the anarchic brain model of psychedelic drug action proposes that confidence in beliefs – and thus expectations (priors) – are relaxed under these drugs. The brain makes inferences to minimize surprise, or discrepant expectations and outcomes (prediction error [PE]), and better model the world. We recently showed that the non-specific 5-HT2A receptor agonist LSD heightened sensitivity to PEs, reflected by speeded updating of value representations following better and worse than expected outcomes. Indeed, events that are surprising are inherently less expected. There is evidence that LSD produces ego dissolution that is positively correlated with disintegration of the default mode network (DMN), assessed using resting state functional connectivity (RSFC), and is correlated inversely with alpha oscillatory power. The DMN has been proposed to be a seat at the ego’s table and decreased alpha power has been posited to reflect relaxation of the cognitive hierarchy, mediated by 5-HT2A receptors on deep layer V pyramidal neurons in cortex.
           
           
              Aims & Objectives
              Here we aim to demonstrate neural correlates of an objective, behaviorally derived marker of REBUS. Greater speed at which choice value increased following a reward PE – higher reward learning rates (RLRs) – was operationalized as enhanced sensitivity to surprise. We tested whether medial prefrontal cortex (mPFC) RSFC within the DMN – a region often associated with value – and diminished alpha power correlated significantly with the RLR.
           
           
              Method
              Alpha power was derived from magnetoencephalography (MEG). The functional magnetic resonance imaging (fMRI) resting state functional connectivity (RSFC) data of interest were extracted from mPFC regions of the DMN. To assess anatomical specificity, we also predicted that static RSFC of primary visual cortex (V1) would not be correlated with the RLR. We additionally extended our analysis to the brain’s temporal organization, parsing the data into dynamic states where network integration or segregation predominate, again focusing on mPFC regions of the DMN.
           
           
              Results
              The RLR was inversely correlated with MEG alpha power (r(13) = -.618, p = .024), and positively correlated with the extent of disintegration of the DMN as assessed using static RSFC (r(15) = .641, p = .010). V1 static RSFC did not correlate with the RLR, as hypothesized (r(15) = .204, p = .465). The RLR was positively correlated with dynamic RSFC in a mPFC a priori region of interest during a predominantly integrated sub-state of brain dynamics (r(15) = .624, p = .013).
           
           
              Discussion & Conclusions
              A behaviorally derived computational marker of surprise was correlated with neural signatures of relaxed beliefs under psychedelics (REBUS).

Description

Journal Title

International Journal of Neuropsychopharmacology

Conference Name

Journal ISSN

1461-1457
1469-5111

Volume Title

28

Publisher

Oxford University Press (OUP)

Rights and licensing