Anxiety- and stress-related disorders can be efficiently treated, but relapse is common - often following stressful events. Relapse can be experimentally modeled using differential fear conditioning, where the unconditioned stimulus is unexpectedly re-presented after extinction training (i.e., reinstatement, RI). Prior cross-sectional research suggests that exposure to recent adversity is associated with reduced distinction between threat and safety cues. Yet, longitudinal evidence is lacking, limiting mechanistic interpretations.

In this longitudinal study, 120 healthy participants completed a fear acquisition training phase, followed by 24h-delayed extinction training and a reinstatement-test (T0). Six months later, 95 participants returned to undergo the same procedure (T1). At both time points, we assessed skin conductance responses (SCRs), BOLD fMRI, fear ratings, and self-reported recent (i.e., during the past six months) adversity.

During acquisition training, reinstatement-test and fear recall (i.e., first trials of the 24h-delayed extinction training), participants who experienced adversity between T0 and T1 discriminated less between the threat and danger cues in SCRs than unexposed participants - consistent with cross-sectional work mainly driven by blunted threat signal (i.e., CS+) responding in exposed individuals. These group differences were accompanied by distinct activation patterns in key fear-related brain regions, including the vmPFC, insula, and amygdala, while fear ratings did not differ between groups.

Given that impaired threat-safety discrimination is associated with stress- and anxiety-related mental health challenges, these results suggest that recent adversity may constitute an important risk factor for relapse. Understanding individual differences in relapse vulnerability could ultimately inform personalized interventions to promote long-term remission.

Background: This study examined learning and memory processes in individuals with clinical anhedonic depression within a randomized trial comparing a 12-week Metta-based online group psychotherapy with an active control (Stangier et al., 2021). Participants completed a probabilistic reversal learning task in which they learned to associate object categories with better outcomes (larger rewards or smaller punishments). Recognition memory was assessed for images presented during feedback, allowing examination of baseline symptom associations and intervention-related changes.

Methods: Participants completed the task before (N = 58) and after the intervention (N = 38, data collection ongoing). Computational parameters - learning rate and choice sensitivity - were estimated using hierarchical Bayesian modeling. Mixed-effects models tested associations between baseline symptoms, symptom change, and learning and memory parameters.

Preliminary Results: At baseline, greater anhedonia, but not depression, predicted lower learning rate and choice sensitivity across conditions, primarily in the control group. Higher anhedonia, but not depressive symptoms, was associated with reduced recognition memory, particularly for reward-related images in the intervention group. Following the intervention, greater reductions in anhedonia, but not depression, were associated with increased positivity bias, driven by decreased learning from negative prediction errors, particularly in the intervention group. Reward-related improvements in memory in the intervention group were predicted by greater reductions in depression, but not anhedonia.

Conclusions: Metta-based therapy was associated with symptom-linked changes in reward-related learning and memory, suggesting that reductions in anhedonia and depression differentially modulate reinforcement learning biases and memory for rewarding experiences.

Effective learning from rewards and punishments is pivotal for the maintenance of mental health, and dysregulated learning mechanisms may contribute to various psychopathologies. Although we often think of reward learning as a trait-like construct, typical reinforcement learning parameters show a surprisingly low test-retest reliability, suggesting considerable variability over time. In this talk, I will argue that this large behavioral state component should be seen as both a psychometric challenge and an opportunity to improve our understanding of aberrant learning mechanisms across mental disorders.

To address the psychometric challenge, I will present data from several studies with repeated runs of a gamified reinforcement learning task, including patients with eating disorders, depression, or both. By using hierarchical models, we can effectively partition trait- and state-like variability to gain novel insights into often neglected processes, such as behavioral flexibility or sensitivity to changes in mood or metabolic states.

To highlight the opportunities, I will present associations of psychopathology with reinforcement learning dynamics, that is, changes and variability over runs. Moreover, I will outline how such dense sampling designs can be used to track the mid-term effects of interventions, taking non-invasive vagus nerve stimulation as an example. These designs provide much-needed tools to monitor individual treatment effects over time.

By developing and refining such assessments for reinforcement learning, we can pave the way for personalized and dynamic adjustments of treatment regimes that use behavioral trajectories to track progress. Such trajectories could also help optimize the timing of adjunct treatment modules.

While learning mechanisms have long been recognized as central to psychological models of primary chronic pain, their causal involvement remains elusive. This presentation will focus specifically on the role of neural mechanisms of learning in the development and maintenance of chronic pain, with particular emphasis on the role of frontostriatal circuits.

The first part of the presentation will focus on findings from a longitudinal fMRI study investigating the predictive values of neural markers of reward learning for the development of primary chronic pain. Patients with subacute musculoskeletal pain underwent fMRI at baseline while completing a reward-learning paradigm assessing functional activation and connectivity within frontostriatal circuits. Pain outcomes were reassessed six months later. Higher frontostriatal activity and stronger frontostriatal connectivity during reward feedback predicted the transition to chronic back pain, while patients with established chronic pain showed reduced frontostriatal activity during reward feedback. This suggests altered neural processing of reward learning across different stages of pain chronification.

The second part of the presentation will focus on a study examining whether these neural learning mechanisms also predict treatment response. Patients with chronic musculoskeletal pain underwent fMRI during reward learning before and after three different behavioural treatments. Preliminary findings suggest that frontostriatal activity predicts treatment success, with the strongest predictive effects in treatments that are grounded in learning mechanisms.

Together, these findings highlight the relevance of neural mechanisms of learning for both the development and treatment of primary chronic pain and highlight the need for more individualized therapeutic approaches targeting maladaptive learning processes.