< Symposia

Symposium Session 10 - Neural Computations of Motivated Behavior in Youth

Chair: Dr Jeremy Hogeveen¹; ¹The University of New Mexico
Presenters: Alexandra Cohen, Juliet Davidow, Catherine Hartley, Jeremy Hogeveen

Adolescence is a critical developmental period for calibrating motivated behavior—from learning to seek rewards to avoiding prospective threats. This behavioral shift is paralleled by the significant maturation of neural circuits underpinning motivation and goal-directed behavior. This symposium integrates four complementary research programs that merge behavioral modeling and functional neuroimaging to characterize the developing neurocognitive mechanisms of motivated behavior in adolescence. Dr. Cohen will first detail how rewards enhance memory via shifting contributions from “offline” subcortical consolidation and “online” cortical encoding mechanisms across development. Dr. Davidow will build from this work, providing evidence that adolescence can be a period of superior motivated learning for both initiating actions and inhibiting responses, supported by the coordinated recruitment of multiple neural systems. Next, Dr. Hartley will discuss the critical role of valence, detailing how positive versus negative outcomes differentially shift reinforcement learning during development and shape the prioritization of information in long-term memory. Lastly, Dr. Hogeveen will integrate these themes to examine the 'explore-exploit tradeoff' when adolescents learn from rewards and punishments. This talk will present evidence that the directed exploration of novel choice options expands in a valence-dependent manner across adolescence, driven by the maturation of prefrontal circuits that underpin the hierarchical control of goal-directed behavior. Together, these findings highlight adolescence as a distinct period characterized by unique policies driving motivated behavior relative to childhood and adulthood—providing critical insights into the adaptive strengths and vulnerabilities that characterize this pivotal life period.

Presentations

Reward memory mechanisms vary across development

Alexandra Cohen¹, Susan Benear², Camille Phaneuf-Hadd³, Lila Davachi⁴, Catherine Hartley⁵; ¹Emory University, ²New York University, ³Harvard University, ⁴Columbia University, ⁵NYU Center for Neural Science

Rewards influence behavioral and neural memory mechanisms. In adults, memory for high-reward memoranda is related to increased activation and functional connectivity of mesolimbic dopamine systems, centered around the ventral tegmental area (VTA), the anterior hippocampus, and cortical areas both during and after encoding. Additionally, prior work conducted in adults has shown that rewards alter multivariate activation patterns in the hippocampus during encoding and that cortical pattern similarity between encoding and retrieval is associated with better memory for both neutral and emotional stimuli. Few studies have examined how rewards influence these memory mechanisms across development. To address these knowledge gaps, we had 89 participants ages 8 to 25 years-old complete a reward-motivated encoding and retrieval fMRI paradigm with baseline and post-encoding active rest periods. Participants then returned 24-hours later for a behavioral memory retrieval test. We found that reward enhances associative memory across all ages and that there were age-related differences in less detailed memory. Neuroimaging analyses suggested that reward enhanced memory through differential activation and functional connectivity of mesocorticolimbic systems and differential multivariate pattern reinstatement in anterior hippocampus across development. We found evidence for greater early reliance on “offline” subcortical post-encoding consolidation mechanisms and increasing contributions of “online” cortical encoding mechanisms supporting reward-motivated memory with increasing age. We also found evidence that distinct representational schemes in anterior hippocampus support reward-motivated memory over development. Taken together, our findings demonstrate that reward motivation enhances memory across age through overlapping cognitive and neural routes with varying contributions across development.

Neurocognitive adaptivity in adolescent goal-directed learning

Juliet Davidow¹, Haley Hegefeld¹; ¹Northeastern University

The brain is equipped with multiple learning systems attuned to motivational influences. These systems undergo protracted development across adolescence, with little understood about how they interact. Adolescence is a crucial life stage for understanding motivated behaviors including learning in the service of goal identification and attainment. Adolescent neurocognitive development of multiple subcortical systems, including the striatum and the hippocampus, constrains strengths and challenges in learning, revealed by associations between actions, reinforcement learning model estimates, and fMRI. We examine motivated learning behavior in four independent samples (N = 72, N = 94, N = 124, and N = 174. Total N = 462) finding that adolescent learning can exceed performance of younger children and adults (age range across samples 8 to 30 years old). Surprisingly, age-related benefits in learning were observed both for approach responses and inhibition of motor responses across paradigms, and across motivational contexts of reinforcement from rewards, losses, and threats. Better learning is supported in part by age-related differences in cognitive processes that contribute to learning, suggested by reinforcement learning parameter estimates, and by age-related differences in the coordinated recruitment of multiple neural systems for learning in a subset of participants. Taken together, these findings highlight the potential adaptivity of adolescent goal-directed learning. Collectively, this work has implications for tailoring environments to leverage adolescent strengths in motivated learning and for translation to age-effective clinical interventions.

Developmental shifts in valence biases in reinforcement learning and memory

Catherine Hartley¹; ¹NYU Center for Neural Science

As individuals learn through trial and error, some are more influenced by good outcomes, while others weigh bad outcomes more heavily. Reinforcement learning (RL) models mathematically quantify such valence biases in learning as the relative influence of positive or negative prediction errors (i.e., better or worse than expected outcomes) on subsequent value estimates. In this talk, I will present findings from several recent and ongoing developmental studies in which we leverage the RL framework to characterize developmental shifts in valence biases, their influence on episodic memory, and how they are influenced by early-life experience. In this work, we show that valence biases in RL shift across development from childhood to adulthood, with adolescents exhibiting greater sensitivity to worse-than-expected outcomes. Moreover, these biases in learning are reflected in what is retained in long-term memory, such that those who prioritize worse- or better-than-­expected outcomes during learning are also more likely to remember images paired with those outcomes. In ongoing analyses, we are characterizing the patterns of neural activation and connectivity during learning and post-learning rest that shape these biases in learning and memory. Finally, in a large developmental cohort who recalled specific dimensions of their early-life experiences, we find that, in line with theoretical predictions, self-reported prevalence and controllability of rewards during childhood predict individual differences in valence bias. Together, these findings provide mechanistic insight into the developmental mechanisms that shape valence biases in RL, and the neural mechanisms through which these biases shape the information prioritized in memory.

Neurocomputational Maturation of Directed Exploration in Adolescence

Jeremy Hogeveen¹; ¹The University of New Mexico

Goal-directed learning requires a critical tradeoff between exploring novel options with uncertain future value that may be beneficial in the long-term, versus exploiting familiar rewards with a high immediate expected value. In adolescence, a heightened drive to explore new opportunities is adaptive for self-discovery, but noisy exploratory behavior may also confer vulnerability to real-world risks. We merged behavior, computational modeling, and task fMRI to probe the neurocomputational maturation of explore-exploit decision-making across adolescence (N=135 13-21 year-old participants). We also probed for distinct mechanisms underpinning exploration to maximize gains versus exploration to avoid losses. Our data reveal a valence-dependent developmental shift in exploration: Younger adolescents engaged in more random exploration in loss avoidance contexts, whereas older adolescents deployed more consistent, goal-directed exploration across valence. This shift was paralleled by maturation of functional response in the frontopolar cortex (FPC) and several co-activated regions at the time of choice—these regions became increasingly engaged to encode the relative future value of exploration as a function of age. Notably, these mechanisms held clear clinical relevance. Diminished goal-directed exploration and increased random exploration were associated with more severe recent alcohol use determined via 30-day timeline follow-back. Therefore, the maturation of FPC systems for motivating adaptive, goal-directed exploration of uncertain choice options shapes vulnerability for hazardous substance use in youth.