Summary: Averaging brain-scan results across groups can mislead researchers about how individual brains function. By analyzing functional MRI (fMRI) data from more than 4,000 children at the single-subject level, Stanford researchers found that many children—especially those who struggle with inhibitory control—show brain dynamics that often run opposite to group-average patterns. These findings have major implications for cognitive neuroscience and for personalized approaches to conditions such as ADHD.
This study challenges long-standing assumptions in neuroscience and offers a theoretical basis for more individualized psychiatry and targeted behavioral interventions.
Key Facts
- Speed-Accuracy Paradox: Group-level measures can show that faster performers also tend to be more accurate, while within a single person the trade-off often goes the other way—individuals slow down to improve accuracy. Brain dynamics exhibit a similar group-versus-individual discrepancy.
- Default Mode Network (DMN) Reversal: When data are averaged across children, slower reaction times correlate with increased DMN activity (linked to mind-wandering). But within an individual child, a slower reaction often coincides with decreased DMN activity—the opposite effect.
- Proactive versus Reactive Control: Cognitive control is not a single, unitary ability. The brain engages multiple sub-processes—proactive control (preparing to stop) and reactive control (stopping in the moment)—each supported by different neural circuits.
- Multiple Neural Pathways to Success: Children with weaker cognitive control can compensate by recruiting alternative neural strategies. Inhibitory control thus appears to be a flexible, trainable set of processes rather than a fixed capacity.
Source: Stanford
Overview
Conventional brain-mapping studies typically combine data from many participants and report average effects. This new research shows that such averages can obscure—and sometimes reverse—the true brain-behavior relationships that occur within individuals. By examining the temporal dynamics of brain activity on a per-child basis, researchers uncovered distinct patterns linked to how each child regulates attention and behavior.
The full results are scheduled for publication in Nature Communications.
“Studying dynamics within individual brains reveals important insights about human variability that group averages miss,” said Percy Mistry, PhD, a research scholar in psychiatry and behavioral sciences and a lead author on the paper. Nicholas Branigan, MS, shares lead authorship; the senior author is Vinod Menon, PhD.
The team focused on inhibitory cognitive control—the brain’s ability to suppress distractions and irrelevant impulses so a person can complete a goal-directed task. Poor inhibitory control is a common feature of disorders such as ADHD, bipolar disorder and addiction, so understanding its mechanisms can guide more precise interventions.
Study design and main findings
Researchers analyzed fMRI and behavioral data from over 4,000 children (ages 9–10) collected during the baseline visit of the Adolescent Brain and Cognitive Development (ABCD) study. Children completed a stop-signal task, which required pressing a button quickly for a “Go” cue and withholding the response if an occasional, unpredictable “Stop” signal appeared.
Rather than relying solely on between-subject averages, the team examined trial-by-trial neural dynamics within each child and compared those within-subject patterns to group-level trends. They found multiple brain-behavior associations that differed—and sometimes reversed—when analyzed within individuals versus across the group.
For example, the default mode network (DMN) showed increased activity in slower trials when data were averaged across children. Yet within individual children, slower responses were associated with reduced DMN activity. The authors emphasize that such reversal of effects is a manifestation of nonergodicity and Simpson’s paradox in neurocognitive dynamics.
The researchers also applied a Bayesian computational model to track how children updated their expectations during repeated trials. Children who adapted well showed increasing readiness to stop on trials following an initial Stop signal, producing faster stopping responses. Children with maladaptive regulation displayed the opposite pattern—reduced expectation of a second Stop signal—and their brain activity reflected that difference. In many cases, group-average effects were driven predominantly by one subgroup, masking alternate dynamics present in others.
Multiple control pathways and educational implications
The study shows that proactive and reactive control are dissociable at the neural level and that children with weaker control may rely on different circuits or adopt alternative strategies. “This reframes cognitive control as a set of flexible processes that can be regulated or trained, rather than a single immutable capacity,” Mistry said.
These insights have practical implications for classrooms and clinical settings: instead of labeling a child as simply having “poor attention,” practitioners could identify which component of control is weaker (for example, reactive stopping versus proactive preparation) and design strategies that leverage the child’s strengths.
Vinod Menon adds that neuroscience should pay closer attention to how each person’s brain responds in specific situations. “There is no such thing as an average brain,” he said. “We must understand how an individual responds to changing contexts that demand attention and adaptive regulation—what response to take, when, and why.”
The ABCD data used in this study are stored in the National Institute of Mental Health Data Archive.
Funding
The ABCD study is supported by the National Institutes of Health and additional federal partners under numerous awards (for example, U01DA041048, U01DA050989, U01DA051016 and others). A full list of supporters is maintained by the ABCD study. At Stanford Medicine, this work received support from NIH grants MH121069 and MH124816, NSF grant 2024856, and the Stanford Maternal and Child Health Research Institute. Stanford University and Stanford Research Computing provided computational resources.
Key Questions Answered:
A: Averaging produces a composite brain that does not exist in any one person. Like using an average clothing size to outfit a diverse group, group averages obscure meaningful individual differences—so they can miss how children with ADHD or other conditions actually process information.
A: Rather than treating a child as having a single global deficit, clinicians could identify which specific control pathway is weak. If a child is strong at proactive preparation but weak at reactive stopping, interventions and classroom strategies can be tailored to capitalize on proactive strengths.
A: Earlier work is not necessarily wrong but can be incomplete. Group studies reveal general trends, but they may misrepresent or fail to predict how any one individual will respond in a specific situation. This study acts as a corrective by highlighting within-subject dynamics.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- The journal paper was reviewed in full.
- Additional context was added by staff editors.
About this neuroscience and ADHD research news
Author: Erin Digitale
Source: Stanford
Contact: Erin Digitale – Stanford
Image: The image is credited to Neuroscience News
Original Research: Open access. “Nonergodicity and Simpson’s paradox in neurocognitive dynamics of cognitive control” by Percy K. Mistry, Nicholas K. Branigan, Zhiyao Gao, Weidong Cai & Vinod Menon. Nature Communications. DOI: 10.1038/s41467-026-71404-0
Abstract
Nonergodicity and Simpson’s paradox in neurocognitive dynamics of cognitive control
Nonergodicity and Simpson’s paradox create significant and often overlooked challenges for cognitive neuroscience. Using imaging and behavioral data from over 4,000 individuals alongside a Bayesian computational model, the study examined brain-behavior relationships governing cognitive control at both the between-subject and within-subject levels.
Results revealed striking reversals of brain-behavior associations across levels of analysis, demonstrating pervasive nonergodicity. Within-person analyses exposed distinct neural representations of reactive and proactive control and showed that individuals who adaptively versus maladaptively regulated cognitive control exhibited different brain-behavior relationships.
These findings indicate that between-subject analyses can mischaracterize mechanisms operating within individuals. The work underscores the need to distinguish between- and within-subject inferences in neuroscience, with important consequences for understanding cognitive processes and designing personalized interventions.