The study team constructed a generative whole-brain model that ran dynamics on each participant's structural connectome to produce both neural recordings and behavior. Short statement: anatomy fed dynamics. Longer clause: these participant-specific network simulations generated single-trial electroencephalography (EEG) signals and corresponding reaction times, linking structural wiring to moment-to-moment perceptual performance. The authors fitted the model to data from 17 participants who performed a motion-discrimination task. The fitted models reproduced individual event-related potentials and the distributions of reaction times observed across trials, according to the preprint.
The model was not trained on average signals alone. Instead, investigators fit dynamics so that simulated EEG and behavioral outputs matched empirical single-trial data concurrently. This joint fitting permitted an additional test: whether the model could classify held-out trials as relatively fast or slow within the same stimulus condition. Data indicated that joint fitting enabled the models to predict trial-by-trial speed differences, suggesting that anatomy-constrained dynamics carry information relevant to behavioral variability beyond stimulus-driven effects.
After fitting, the study team performed in silico perturbations—silencing individual nodes within each participant's model—and observed the consequences for simulated decision speed. Short result: the silencing experiments predicted where and when decision speed would be altered for each person. More granularly, the model output provided spatiotemporal predictions about control points in each network: which regions, if disabled within the model, would slow or speed perceptual decisions and at what times those effects would unfold.
Across participants the sensorimotor regions produced relatively consistent effects on decision speed. The cohort showed that impairing sensorimotor nodes tended to influence speed in similar directions across most participants. By contrast, a specific parietal region had opposing effects in different individuals. The authors noted that these opposite outcomes reflected two mechanistic roles: in some participants that parietal region exerted influence primarily through shaping distributed network dynamics, while in others it acted mainly via the decision readout pathway. Thus the same anatomical node could change a person's decision speed in opposite ways depending on its functional embedding and the readout mapping in that individual's model.
Temporal analyses of model activity and empirical EEG suggested a timing dissociation. Early trial activity—occurring soon after stimulus onset—appeared to set the eventual decision speed. Short finding: early dynamics bias speed. Longer sentence: later EEG components, in contrast, mostly reflected the unfolding decision process rather than determining how quickly the decision would be reached. The authors interpreted this pattern as indicating that initial network state and fast-evolving dynamics have a causal role in shaping response speed, while later signals index ongoing accumulation or readout of the decision.
This report is a preprint and has not undergone peer review; the authors declared no competing interests. The abstract and accompanying information describe the approach and main results but do not report every methodological detail in the public summary: for example, the preprint abstract does not provide exhaustive parameter settings, model-fitting hyperparameters, or full cross-validation procedures in the abstract text alone. The model was fitted to data from 17 participants; the source document states that number but does not, in the abstract, elaborate on participant demographics, trial counts per condition, or potential generalizability to other tasks or larger cohorts. Future steps that the authors did not document in the abstract include external replication, peer review validation, and testing whether the model's in silico predictions correspond to perturbation outcomes in vivo. Readers should consider those outstanding questions when interpreting the reported links between individual brain dynamics and perceptual decisions.