About

Leah Somerville is the Grafstein Family Professor of Psychology at Harvard University and a Harvard College Professor (Endowed 2021-2026). She is also faculty in the Center for Brain Science. Her research focuses on characterizing adolescent brain development and understanding the consequences of brain development on psychological functioning and well-being. Her work integrates behavioral, computational, and neuroimaging approaches. Notably, she conducts the Human Connectome Project in Development, a large NIH-funded study on brain connectivity development.

Research topics

• Psychology
• Cognitive psychology
• Psychiatry
• Clinical psychology
• Computer Science
• Psychotherapist
• Medicine
• Neuroscience
• Developmental psychology
• Applied psychology
• Linguistics

Selected publications

• Associations Between Pubertal Timing, Loneliness, and Internalizing Symptoms in Adolescence: Integrating Brain and Behavior

  Open MIND · 2026-01-01

  otherOpen accessSenior author

  This project examines how pubertal timing relates to loneliness and internalizing symptoms during adolescence, and whether these associations are shaped by brain network organization. Using cross-sectional data from the Human Connectome Project in Development (HCP-D), we first establish behavioral associations showing that earlier pubertal timing is linked to greater loneliness and internalizing symptoms, with loneliness significantly mediating this relationship. Building on these findings, we test whether default mode network (DMN) connectivity with the salience (SAL) and dorsal attention (DAN) networks is associated with pubertal timing and moderates its association with loneliness. Pubertal timing is indexed relative to same-age peers and standardized such that higher values reflect earlier maturation. We hypothesize that earlier pubertal timing will be associated with greater DMN–SAL and DMN–DAN connectivity (i.e., increased cross-network coupling), and that greater connectivity will amplify the association between pubertal timing and loneliness. This work integrates behavioral and neurodevelopmental approaches to better understand pathways linking early maturation to emotional risk in adolescence.

  PublisherDOI

• Time-Based Cognition and Working Memory Explain Changes in Delay Discounting Through Adolescence

  PsyArXiv (OSF Preprints) · 2026-02-10

  preprintOpen access1st authorCorresponding

  Individuals of all ages face choices between immediate but smaller and delayed but larger rewards. In a decision process known as delay discounting, people tend to devalue options that are realized farther into the future. Several candidate cognitive mechanisms have been proposed to underpin such choices: inhibition (overriding a prepotent response), prospection (orienting to the future), working memory (manipulating information held in mind), reward motivation (activating behavior for reward), and time perception (subjectively judging time horizons). These mechanisms change through adolescence, in tandem with delay discounting decisions becoming more patient from childhood to adulthood. However, the cognitive mechanisms principally responsible for developmental increases in patience remain contested. By measuring delay discounting alongside indices of inhibition, prospection, working memory, reward motivation, and time perception in the same participants (N = 159, 10-20 years, collected in the United States), the present study examined whether these mechanisms – alone or in combination – explained age variability in reward- and time-based choice in different stakes contexts. Mixed- effects modeling revealed that patience indeed increased with age, especially for higher- than lower-magnitude decisions. Dependency modeling identified dissociable and overlapping drivers of age-related changes in delay discounting: prospection mediated patience in lower-magnitude decisions whereas working memory mediated patience in higher-magnitude decisions; time perception mediated patience in both. Although compressed time perception supported older participants’ more patient choices overall, distinct cognitive mechanisms may become available during adolescence to underscore willingness to wait for larger later rewards, depending on decision stakes. These findings advance understanding of how multiple constituent processes scaffold the development of value-based choice.

  Publisher

• Post-training makes large language models less human-like

  ArXiv.org · 2026-05-08

  articleOpen access

  Large language models (LLMs) are increasingly used as surrogates for human participants, but it remains unclear which models best capture human behavior and why. To address this, we introduce Psych-201, a novel dataset that enables us to measure behavioral alignment at scale. We find that post-training -- the stage that turns base models into useful assistants -- consistently reduces alignment with human behavior across model families, sizes, and objectives. Moreover, this misalignment widens in newer model generations even as base models continue to improve. Finally, we find that persona-induction -- a popular technique for eliciting human-like behavior by conditioning models on participant-specific information -- does not improve predictions at the level of individuals. Taken together, our results suggest that the very processes that are currently employed to turn LLMs into useful assistants also make them less accurate models of human behavior.

  PublisherOA PDF

• Post-training makes large language models less human-like

  arXiv (Cornell University) · 2026-05-08

  preprintOpen access

  Large language models (LLMs) are increasingly used as surrogates for human participants, but it remains unclear which models best capture human behavior and why. To address this, we introduce Psych-201, a novel dataset that enables us to measure behavioral alignment at scale. We find that post-training -- the stage that turns base models into useful assistants -- consistently reduces alignment with human behavior across model families, sizes, and objectives. Moreover, this misalignment widens in newer model generations even as base models continue to improve. Finally, we find that persona-induction -- a popular technique for eliciting human-like behavior by conditioning models on participant-specific information -- does not improve predictions at the level of individuals. Taken together, our results suggest that the very processes that are currently employed to turn LLMs into useful assistants also make them less accurate models of human behavior.

  PublisherDOI

• Advances on Design Considerations in Developmental Cognitive Neuroscience

  UNC Libraries · 2026-03-25

  articleOpen access
  PublisherDOI

• Disentangling the Unique Associations of Age, Pubertal Stage, and Pubertal Hormones with White Matter Microstructure in Childhood and Adolescence

  Imaging Neuroscience · 2026-01-01

  articleOpen access

  Abstract Puberty is associated with hormone level changes that influence white matter development. It remains unclear how pubertal stage and pubertal hormones uniquely relate to white matter microstructure development. Further, it is unclear if white matter tracts develop along a hierarchical sensorimotor-association (S-A) axis, similar to other aspects of brain development. We used the Human Connectome Project in Development cross-sectional sample of 1,105 youth (aged 5-21 years) to investigate unique associations of sex, age, pubertal stage, DHEA, testosterone, estradiol, and progesterone with fractional anisotropy (FA) within canonical white matter tracts. The average S-A axis rank was calculated for tract cortical endpoints. Age was the best-fit model for 17 tracts with prefrontal, parietal, and temporal connections. The Pubertal Timing model (age-adjusted pubertal stage) was the best fit for the superior longitudinal fasciculus 1. The DHEA model was the best fit for the splenium, genu, and prefrontal body of the corpus callosum. The Estradiol model was the best fit for the ventral cingulum bundle, extreme capsule, inferior longitudinal fasciculus and uncinate fasciculus. The Full model was the best fit for the rostrum. Radial diffusivity explained more FA variability than axial diffusivity in all tracts. Pubertal Timing and DHEA were more related to FA in association-related tracts than sensorimotor-related tracts. Thus, during childhood and adolescence, Pubertal timing and DHEA may be uniquely related to white matter tract microstructure in association-related tracts.

  PublisherDOI

• Registration-free analysis of diffusion MRI tractography data across subjects through the human lifespan

  UNC Libraries · 2026-03-27

  articleOpen access
  PublisherDOI

• Monetary Rewards Modulate Working Memory Performance During Adolescence

  PsyArXiv (OSF Preprints) · 2026-02-09

  preprintOpen access

  Rewards can be powerful motivators for higher-order cognition, and adults typically improve performance on demanding tasks when high incentives are at stake. However, less is known about how rewards influence executive function during adolescence, and few studies have examined the developmental emergence of when rewards modulate adolescents’ working memory (WM) performance. Because WM continues to mature from childhood to adulthood and differentiates from other executive processes in early adolescence, we examined how rewards modulate adolescents’ WM performance and whether this modulation changes with age. Participants aged 10-20 (Nusable=187) completed a neutral N-back working memory task (no payout) and a rewarded N-back task with low-magnitude and high-magnitude rewards. To assess subjective perceptions of cognitive effort and reward value, participants also completed a cognitive effort discounting task and provided subjective ratings of task demands and monetary payouts. During the rewarded N-back task, participants of all ages exhibited increased accuracy for high- relative to low-magnitude rewards. However, when comparing performance in rewarded versus neutral N-back tasks, younger adolescents’ WM benefited the most from reward, and this benefit decreased with age. Effort discounting and subjective ratings were age-invariant, indicating that age-related differences in reward-modulated WM were not explained by subjective differences in perceived effort demands or reward values. These findings reveal that rewards enhance WM throughout adolescence, particularly at younger ages. This work also reveals that reward-related improvements in cognitive function emerge earlier in development for working memory than other executive function domains.

  Publisher

• Two axes of white matter development

  Nature Communications · 2026-01-23 · 2 citations

  articleOpen access

  Despite decades of neuroimaging research, how white matter develops along the length of major tracts in humans remains unknown. Here, we identify fundamental patterns of white matter maturation by examining developmental variation along major, long-range cortico-cortical tracts in youth ages 5-23 years using diffusion MRI from three large-scale, cross-sectional datasets (total N = 2716). Across datasets, we delineate two replicable axes of human white matter development. First, we find a deep-to-superficial axis, in which superficial tract regions near the cortical surface exhibit greater age-related change than deep tract regions. Second, we demonstrate that the development of superficial tract regions aligns with the cortical hierarchy defined by the sensorimotor-association axis, with tract ends adjacent to sensorimotor cortices maturing earlier than those adjacent to association cortices. These results reveal developmental variation along tracts that conventional tract-average analyses have previously obscured, challenging the implicit assumption that white matter tracts mature uniformly along their length. Such developmental variation along tracts may have functional implications, including mitigating ephaptic coupling in densely packed deep tract regions and tuning neural synchrony through hierarchical development in superficial tract regions - ultimately refining neural transmission in youth.

  PublisherOA PDFDOI

• Low Emotional Awareness as a Transdiagnostic Mechanism Underlying Psychopathology in Adolescence

  UNC Libraries · 2026-04-03

  articleOpen access

  The ability to identify and label one's emotions is associated with effective emotion regulation, rendering emotional awareness important for mental health. We evaluated how emotional awareness was related to psychopathology and whether low emotional awareness was a transdiagnostic mechanism explaining the increase in psychopathology during the transition to adolescence and as a function of childhood trauma-specifically violence exposure. In Study 1, children and adolescents (N=120, aged 7-19 years) reported on emotional awareness and psychopathology. Emotional awareness was negatively associated with psychopathology (p-factor) and worsened across age in females but not males. In Study 2 (N=262, aged 8-16 years), we replicated these findings and demonstrated longitudinally that low emotional awareness mediated increases in p-factor as a function of age in females and violence exposure. These findings indicate that low emotional awareness may be a transdiagnostic mechanism linking adolescent development, sex, and trauma with the emergence of psychopathology.

  PublisherDOI

Recent grants

• NIH Grant U01MH109589

  NIH · $17.1M · 2022

• NIH Grant K99MH087813

  NIH · $169k · 2012

• CAREER: Psychological and Neurodevelopmental Mechanisms of Social Influence on Adolescent Decision-Making

  NSF · $750k · 2015–2022

• NIH Grant R00MH087813

  NIH · $719k · 2015

Frequent coauthors

• Paul J. Whalen

  31 shared

• Juliet Y. Davidow

  Monash University

  26 shared

• Tom Johnstone

  Swinburne University of Technology

  23 shared

• Patrick Mair

  23 shared

• Hackjin Kim

  Korea University

  22 shared

• Margaret A. Sheridan

  University of North Carolina at Chapel Hill

  22 shared

• Koene R. A. Van Dijk

  Digital Science (United States)

  20 shared

• John C. Flournoy

  Harvard University

  20 shared

Labs

• Affective Neuroscience and Development LabPI

  Focuses on understanding why people harm others and act in destructive ways, investigating the developmental processes involved in emotion, psychopathology, and antisociality in children and youth.

Education

• Ph.D., Psychology

  Harvard University

  2002

• B.A., Psychology

  University of California, Los Angeles

  1997

Awards & honors

• FJ McGuigan Early Career Research Prize for Understanding th…
• APS Janet Taylor Spence Award for Transformative Early Caree…
• Everett Mendelsohn Excellence in Mentoring Award
• National Science Foundation CAREER Award