Nature Human Behaviour

While creativity is widely recognized as multifaceted, genome-wide association studies have largely focused on individual measures. Here, we used genomic structural equation modeling of six creativity-related GWASs and identified two latent genetic factors: a general factor and an artistic factor, with 22 and 13 genome-wide significant loci, respectively, nine of which are novel. The general factor was almost perfectly genetically correlated with educational attainment (rg = 0.96), suggesting that it primarily captured genetic influences shared with cognitive and educational processes rather than creativity-specific variation. In contrast, the artistic factor showed only a weak genetic correlation with educational attainment (rg = 0.11) and consistent positive genetic correlations with psychiatric disorders, whereas associations for the general factor were largely null or negative. Biological annotation analyses further identified distinct enrichment patterns unique to the artistic factor, providing initial evidence that different dimensions of creativity may be characterized by partially distinct neurobiological pathways.

Language is vastly ambiguous, yet in context our brains resolve ambiguity effortlessly. This is particularly evident at the lexical level: Although most words can convey a wide range of related meanings, humans rapidly and automatically select the one appropriate for the context. We investigated the neural mechanisms underlying this process by focusing on a case where, intuitively, one meaning appears more basic than another: nouns describing animate entities, such as tiger, which can also refer to representations of those entities when combined with a material modifier, as in plastic tiger. Using magnetoencephalography, we decoded animacy from neural signals and found that even in such cases, our approach revealed no evidence that the tiger in plastic tiger is represented as a sentient animal in any stage of processing. Instead, phrase-level animacy--plastic tiger as inanimate--was decodable as early as 210 ms post-stimulus onset and persisted for several hundred milliseconds. These findings challenge theories assuming a shift from a basic sense of a word to a more derived one, highlighting instead the role of compositional context in shaping word meanings.

How do psychedelic drugs alter human brain function? Answering this question relies on our ability to understand how information is stored, transferred, and transformed across the cortex. Here, we leverage a precise mathematical framework for multivariate information decomposition to construct the first comprehensive picture of these processes across five compounds: LSD, psilocybin, ketamine, DMT and 5-MeO-DMT (N=80). We provide a formal and empirical explanation of major prior results, revealing sets of competing underlying phenomena masked by extant approaches. We extend this analysis to novel higher-order phenomena, and we demonstrate how all of these observations can emerge from a minimal generative model in which neural populations operate with reduced self-coupling. In sum, our results suggest that the psychedelic brain is one in which cortical populations become alike, but not necessarily together. This has important implications for models of how psychedelics reorganize brain dynamics to alter subjective experience and alleviate mental health conditions.

Intelligence is often considered to be fundamentally different from other personality traits. Twin studies show that intelligence and the Big Five personality traits are partly influenced by the same genetic factors, but it remains unclear to what degree shared genes, SNPs, and biological pathways influence them. To address these questions, we analyzed the genetic associations of all six traits. First, we calculated their genetic correlations using LD score regression. We then quantified polygenic overlap using a bivariate causal mixture model, finding that approximately 90% of the genetic variants influencing the Big Five personality traits also impact intelligence. Finally, we performed gene prioritization and gene set enrichment analysis on each trait, once again finding substantial overlap. At each level, intelligence was as similar to the Big Five traits as they were to one another. Genetically and biologically, intelligence fits in with the rest of personality.

In language, not only do sentence meanings depend on how words are structured together, but word meanings also depend on their roles in the structure. While we have some understanding of how the brain grasps complex meanings built from simpler parts, we know far less about how linguistic structure reshapes the meanings of the words themselves. Here we used magnetoencephalography (29 human participants: 16 women, 9 men, 4 nonbinary) to track the neural dynamics of animacy, a fundamental semantic feature that guides conceptual organization. Regardless of how animacy was operationalized (as discrete semantic categories or more abstractly), the same pattern emerged: When word-level animacy conflicted with the structural context, it disappeared from the word's neural representation. Our findings show how compositional processes rapidly reconfigure neural traces of word meanings, allowing phrase-level meanings to even override word-level features.

Language can be experienced in highly serial form, as in speech, or in fully parallel form, as in text, yet how our brains adapt to differences in seriality is not understood. Here, we held modality constant (visual) and examined whether processing mechanisms underlying grammaticality detection are similar under Rapid Serial Visual Presentation (RSVP) and Rapid Parallel Visual Presentation (RPVP). To identify shared neural representations of grammaticality, we trained a classifier on magnetoencephalography (MEG) responses to grammatical and ungrammatical three-word phrases in one presentation mode and tested its generalization to the other. Behaviorally, accuracy was higher and reaction times faster for grammatical vs. ungrammatical phrases across presentation modes. Neurally, cross-presentation decoding revealed shared representations of grammaticality at word 2 across train-test directions--albeit with different latencies and durations--and an early shared representation at word 3 when trained on RPVP. We take the bidirectional representations at word 2 to capture distinct but presentation-invariant combinatorial processes and the asymmetry at word 3 as reflecting stronger global structure encoding in RPVP.

Free recall of naturalistic stimuli, such as films and stories, reveals what people remember and how they organize it. Scoring such recall requires matching each recalled utterance to the encoded stimulus, and doing this manually often makes large studies impractical. Existing automated methods mostly return similarity or aggregate scores rather than explicit matches between recalled and stimulus content. Here we introduce an open-source, model-agnostic pipeline that uses a large language model (LLM) to divide the stimulus and the recall into information units and to match units between the two texts. The output is a table of unit-to-unit matches with accuracy labels, from which standard analysis scripts compute recall measures. We validated the pipeline with four LLMs (Claude Sonnet 4.5, GPT 5.4, Gemini 2.5 Pro, and Gemini 2.5 Flash) on three datasets. On typed recall of four short stories, agreement between each model and two trained raters ({phi} = .73-.79) was close to the agreement between the raters themselves ({phi} = .77). A second run reproduced each model's matches ({phi} = .91-.97) more closely than the raters agreed with each other. On spoken recall of 10 films, every model attributed recall to the correct film with high agreement. On uncorrected speech-to-text recall of a television episode, agreement varied more across models. The median cost was under $1 per participant for every model. These results suggest that the pipeline can serve as a reliable, low-cost alternative to manual scoring of free recall.

Cerebral cortical maturation supports emotional, social, and cognitive development during adolescence1. Neurodevelopmental theories of psychiatric illnesses, which peak in onset around age 182, implicate subtle alterations in adolescent cortical development3. However, mechanistic links between cortical maturation and emergent psychopathology have remained elusive. This gap likely reflects a need for more nuanced understanding of cortical development4, larger, longitudinal samples powered to detect small effects,5 and more uniform image quality control.6 Using 29,482 structural MRI scans from 11,720 adolescents in the Adolescent Brain Cognitive Development (ABCD) study, we demonstrate reproducible differences in the contour and tempo of maturation across the cortical surface. Regional trajectories are robust to image quality variance and align with increasing functional specialization over time. Thicker cortex in early-maturing sensory regions predicts psychiatric symptoms as early as four years prior to their emergence, a pattern that subsequently spreads to association cortex. Overlap with spatially and temporally organized gene co-expression networks implicates prenatal programming of this sequence of maturation and its links to psychopathology. These findings present a novel, versatile neurodevelopmental framework for cortical maturation during adolescence that sequentially implicates sensory and association cortex in risk for emergent psychiatric symptoms.

Effective verbal communication requires individuals to align their communicative intentions through the precise use of utterances, yet how this is achieved in dynamic, open-ended communication remains unclear. Combining predictive deep language models (DLMs) with cognitive computational modeling, we tested a computational division of labor between context-sensitive global constraint on the utterance space and partner-sensitive local adjustment of selection priorities. We formalized local adjustment using an active inference inspired updating rule within a statistical-pragmatic hybrid model. Thirty dyads performed a dynamic, open-ended collaborative word-generation task during functional near infrared spectroscopy (fNIRS) hyperscanning. Interlocutors converged in their lexical choices, and lexical alignment was best explained by a hybrid model integrating contextual selection priors with interaction-specific local adjustment among the models tested. This hybrid process formed a dynamic hypothesis landscape comprising a global semantic search region and local selection priorities. Critically, interaction did not simply reduce uncertainty: within individuals, the hypothesis space of global semantic search progressively narrowed while local uncertainty among plausible alternatives increased; between partners, both global hypothesis space and local selection became increasingly aligned, with alignment of local selection priorities most strongly predicting subsequent lexical alignment. The contribution of local adjustment was stronger during precise verbal communication than in the control tasks. Neural analyses further identified correlates of the local adjustment process: temporoparietal and anterior temporal regions supported partner-related inference and semantic prediction error, respectively, and their activity during thinking predicted subsequent inferior frontal activity associated with self-updating during speaking. The participant's own hypothesis landscape was also represented in the left inferior parietal-supramarginal cortex. The main findings generalized across language models and an independent Czech dataset. Together, these findings reveal that alignment in open-ended communication emerges from complementary statistical and pragmatic computations, providing a framework for understanding human and human-AI communication.

The human brain transforms a brief glance into a neural representation flexible enough to serve many behavioural goals. This flexibility may reflect multidimensional visual representations read out selectively across tasks, rather than goal-specific instantiation. Combining a large-scale EEG dataset with a behavioural battery of 12 tasks spanning 200 objects, including perceptual, categorical and social decisions, we tested how behaviours are supported by stimulus-evoked representations. Information relevant to 11 of 12 tasks was accessible in the neural signal, emerging early and overlapping in time. Orthogonal behavioural dimensions showed distinct temporal profiles, indicating a genuinely multidimensional representation. The neural code converged on a human-relatedness axis, while feature-level information remained accessible. Behavioural accessibility was thus broad but structured. Object familiarity alone was reliable in behaviour yet absent from the shared neural signal, revealing a limit on what a glance makes accessible. We propose that sustained, recurrent processing serves this broad behavioural accessibility.

Extracting stable individual-level representations from noisy and variable trial-wise EEG remains a major challenge in computational cognitive neuroscience. We developed a spatiotemporal graph attention network to analyze Go/NoGo EEG from children and adolescents with and without ADHD. The framework combines ERP-informed temporal windows, phase-synchrony graph priors, and trial-to-participant transfer learning, making both information routing and representation geometry explicitly testable. The model learned reproducible attention trajectories, yet ADHD and control participants did not differ reliably in spatial attention allocation. Instead, group-related structure emerged in compact participant-level embeddings that distilled trial-level variability and retained discriminative information beyond convolutional representations. Performance exceeded a convolutional neural network comparison, but the central advance is methodological: neurophysiologically constrained graph learning can reveal where diagnostically relevant information emerges within the analytical pipeline. This provides a framework for developing auditable cognitive-neurophysiological markers of neurodevelopmental variation.

Childhood adversity has enduring consequences for health and development across the lifespan, yet the organizing structure linking diverse environmental experiences to the developing brain remains poorly understood. We jointly modeled 104 measures of children's environments, spanning interpersonal, household, and neighborhood contexts, with 7 neuroimaging modalities of brain structure and function in 5,976 youth. Despite this breadth, only a single adversity-brain dimension emerged and replicated in held-out data. This dimension was defined primarily by socioeconomic and neighborhood deprivation and selectively expressed in lower cortical surface area and distributed functional connectivity concentrated in sensorimotor systems. Total cortical surface area reproduced 96% of the regional association, indicating a global rather than regionally specific structural signature. This dimension predicted poorer cognition and cardiometabolic health six years later. These findings reveal a strikingly selective, low-dimensional relation between children's environments and neurodevelopment, with socioeconomic deprivation emerging as the principal environmental axis associated with the developing brain.

Emotion regulation via cognitive reappraisal is essential for socioemotional development and mental health, yet the organizational logic of its supporting brain systems across youth remains unclear. Using a large multisite cohort, we demonstrate that maturation of reappraisal capacity reflects reorganization of a distributed architecture rather than progressive strengthening of prefrontal control. Reappraisal robustly reduced negative affect, with non-linear gains concentrated between 6 and 10.5 years. Bayesian systems identification resolved this architecture into four functionally distinct systems-reappraisal-specific, common-appraisal, modifiable-emotion, and non-modifiable-emotion components-spanning prefrontal-control, salience, default-mode and dorsal-attention networks. Age-related variation in anterior temporal cortex, emerged as essential hub, statistically accounted for developmental gains in reappraisal success. Multivariate modelling together with intersubject analyses further identified a cross-site-generalizable whole-brain reappraisal signature and revealed an age-dependent shift in brain-behaviour coupling, with neurobehavioural convergence increasing with age. These findings establish distributed network reorganization as a core mechanism of emotion regulation development and provide mechanistic targets for stage-matched interventions.

Motivational incentives are known to improve working memory (WM) when items differ in value, but whether they enhance performance when all items are equally valued, and if so, by what mechanism, remains unclear. Improvements could reflect enhanced memory-based representations or speed-accuracy trade-offs, whereby incentives promote slower, more cautious responding without improving memory itself. Here, we tested these alternatives across two experiments using an auditory WM task in which incentives varied across trials, but all items on a trial were equally incentivised. In Experiment 1,we allowed participants to resubmit their responses if desired. Reward-related improvements in WM performance were selectively observed when memory load was high, and only at serial positions associated with greater interference. These improvements were also dissociated from systematic response slowing and resubmission behaviour. This indicated that the benefits of incentives were unlikely to be attributable to response caution. Critically, in Experiment 2, motivational benefits persisted under strict response deadlines, where speed cannot be traded for accuracy. Computational modelling showed that reward selectively improved the precision of memory-based responses without reducing guessing. Together, these findings indicate that motivational incentives improve WM by enhancing the quality of memory-based representations, rather than by strategic slowing or altering response strategies.

The auditory oddball protocol is widely used to study neural responses to unexpected events. In conventional designs, however, stimulus statistics are stationary: deviants do not indicate environmental change and therefore should not drive learning. Here, we introduce change-point oddball, a passive, no-report protocol in which the standard-oddball mapping reverses at unpredictable change points, requiring inference over latent environmental states. We measured pupil-linked arousal in three human cohorts. Pupil responses were larger for state-specific oddballs than standards, decayed across newly established standards after state changes, and scaled with model-derived change-point probability. A many-standards control experiment implicated a prediction-based rather than repetition-based mechanism. Finally, within participants, stronger pupil encoding of change-point probability during the passive task was associated with greater change-point-sensitive belief updating in a separate active navigation task. Thus, passive change-point oddball preserves core signatures of adaptive belief updating, enabling cross-species and clinical investigations.

While much of the neural and cognitive machinery for visual perception is in place within weeks from birth, we identified and quantified a fundamental difference between infant and adult vision: the temporal rate at which meaningful visual information can be reliably extracted. Using frequency-tagging electroencephalography with rapid serial visual presentation, we found that object categorization at 4 and 9 months is respectively seven times and two to three times slower than in adults. Using quantitative information from infants experiments, we created and presented accelerated visual streams to adults, to match infants perceptual experience. We used deep convolutional neural networks to identify object representations that match those elicited in adults when accelerating the information flow. Results show that acceleration compromises finer-grained object distinction while sparing broad, general object representations. We propose that slow vision is a design feature of early development for preserving and highlighting the most stable and general features of the world.

Conformal prediction is increasingly proposed for clinical decision support because it provides distribution-free coverage guarantees that hold for any model and any data distribution. The guarantee routinely reported, however, is marginal, and we show that in multi-class diagnostic classification it can be satisfied exactly while individual diagnoses are covered very unevenly. On a four-way mood and psychosis classification task over 1,520 subjects from three studies and 14 acquisition sites, marginal split-conformal calibration attained 0.9000 empirical coverage against a nominal 0.90, while healthy controls were covered at 0.941 and schizoaffective disorder at 0.818. The surplus and the deficit cancel, so the aggregate figure is uninformative about either. Class-conditional (Mondrian) calibration reduced this 12.3-point disparity to 0.4 points at a cost of 0.07 labels in mean set size, under 3%. Holding coverage fixed across diagnoses has a further consequence: the residual variation in set size can no longer be attributed to class prior or per-class accuracy, and becomes interpretable as a property of the subject. We show that the resulting prediction sets separate subjects into confident, boundary, ambiguous and unresolved strata, and that the proportion of subjects independently flagged as label-ambiguous by a structural-MRI model trained separately on the same cohort rises monotonically across these strata (34.1%, 57.3%, 68.1%, 81.8%; p = 8.8e-18). No schizoaffective subject and 0.9% of bipolar subjects reach the confident stratum, against 18.0% of controls and 13.4% of schizophrenia subjects. Set size at matched coverage also provides a comparison between representations that accuracy cannot. Applied to structural MRI, functional MRI and their fusion, it reveals a sign reversal that accuracy conceals: fusion reduces set size for bipolar (-0.25) and schizoaffective (-0.23) subjects and increases it for controls (+0.16), and the aggregate fusion benefit itself changes sign below alpha = 0.10, while top-1 accuracy rises monotonically from 0.495 to 0.578 to 0.611 across the three models. Reporting marginal coverage alone is insufficient whenever diagnostic classes are unbalanced.

Humans adapt to changing environments by understanding the latent causal structure of events, and modern large language models (LLMs) increasingly exhibit analogous abilities. However, elucidating how causal knowledge is represented and integrated in the human brain and in LLMs poses a substantial challenge for cognitive science. To address this, we combine topological analysis of representational manifolds with neuroscientific analyses of reactivation. We find that event-level causal relations modulate temporally structured neural reactivation in the human brain, particularly in the angular gyrus, precuneus and prefrontal cortex. By contrast, LLMs express these relations through positionally structured representational reactivation. Despite these differences, the resulting reactivation patterns exhibit similar intrinsic manifold structure across the two systems. Together, these findings suggest that the human brain and LLM share a common principle for organizing causal knowledge.

While polygenic gene-by-environment (GxE) signals are increasingly detected in biobank data, it is often unclear whether they reflect the phenotype's measurement scale or one of two biologically distinct mechanisms: heterogeneous genetic effects across environmental strata (effect heterogeneity) or systematically larger but correlated effects in one stratum (amplification). We introduce REGENA (Robust Estimator of Gene-ENvironment Architectures), a variance-component method that disentangles these contributions at biobank scale by learning a phenotype transformation that minimizes scale-induced GxE, thereby assessing scale-dependence while jointly estimating the cross-environment genetic correlation and environment-specific genetic variances. In simulations, REGENA is well-calibrated and powerful while recovering the transformation underlying scale-dependent GxE. Applied to 43 quantitative traits and seven environmental exposures in 291,273 UK Biobank individuals, REGENA finds that roughly half of trait-environment pairs with significant GxE are scale-dependent. Among the remaining scale-independent trait-environment pairs, GxE heritability accounted for an average of 15% of additive heritability and was driven primarily by effect heterogeneity (accounting for 71% of GxE heritability on average) although there was substantial variation in its relative contribution across trait-environment pairs. Environment-specific effects tend to make a larger contribution at lower allele frequencies, accounting for 6% of genetic variance in low-frequency variants versus 3.9% at common variants (signed test p = 1.7e-24). Modeling these distinctions improves environment-specific polygenic prediction (14% on average), with the largest gains for pairs whose genetic effects are least correlated across environments (p = 4.0e-3; 3.8-fold for testosterone stratified by sex and 2.32-fold for waist-hip ratio stratified by alcohol consumption). Further, environment-stratified GWAS detects 174 loci missed by pooled GWAS with 67 having opposite-signed effects across strata.

Human perceptual decision confidence has been associated with brain-wide signalling, albeit with inconsistencies across studies. It remains unclear whether more robust neural latent representations exist, particularly for implicit confidence. We address this by applying complementary machine learning approaches to comprehensive, brain-wide intracranial electroencephalography (iEEG) data acquired without explicit confidence reports. Using explainable feature selection methods, we elucidated key spatiotemporal iEEG components, especially in the frontal cortex, that readily distinguished confidence characteristics related to decision correctness, difficulty, and speed, and that were transferable across motor modalities. Despite inconsistencies in the features identified across methods and conditions, dimensionality reductions revealed robust low-dimensional neural latent representations. Furthermore, their neural trajectories within a shared subspace enhanced the distinguishability of confidence characteristics. Overall, our results uncover low-dimensional spatial and temporal neural latent representations underlying human implicit perceptual decision confidence that are robust to analytical methods and decision conditions.