"This approach was also able to identify subgroups of children with different levels of cognitive control and performance monitoring, or the ability to modify one’s strategy after making an error."
This should surprise no one. You took a large population and found subpopulations within it. If you want to look at a population average, then use the population data. If you want to look at kids with specific attention needs (guessing ADHD since medical related) then design a study to select for children fitting that criteria, including subtypes.
This seems like the type of thing that should have had a study about study design done long ago that they could have followed to help them structure their own population selection.
The specific counterintuitive result is mentioned toward the end of the article, and I'm having some trouble understanding it:
> when analyzing average trends in groups of children, slower reaction times to the “Go” signal were linked to increased activity in many brain regions, including the default mode network
> However, when an individual had a slower reaction time to the “Go” signal, activity decreased in the default mode network — the opposite of the group-level pattern.
This is not a case of Simpson's Parodox, at least the analogy about accuracy vs speed from the article isn't.
You're not comparing global vs subgroup correlations. On the one hand you're measuring how speed and accuracy are correlated across the population when you ask subjects to solve a problem. On the other hand, (rather than measuring subgroup correlations) you're measuring how accuracy is affected when you ask an individual to speed up or slow down.
Yup, to no one’s surprise (least of all the investigators), doing neuroscience by correlating cortex regions with cognitive activities is extremely clunky at best. Very robust finding confirming this tho, thanks for sharing!
Now that we’re finally moving to the next stage of neuroscience due inscrutable latent systems (aka LLMs), I can’t help but feel some nostalgia. It’s all fun and games until someone makes a lie detection helmet that actually works…
While interesting, I get a few questions from this:
- As another commenter said, this is a known disadvantage of averages. I'm curious if it's possible to get a median result from per-individual averages. I'm not familiar enough with how this research is done to get a result.
- Was any effort made to re-test individuals in a second/third/etc session, showing consistent patterns to the brain activity? I know it was consistent within a session, but I'm curious if it might change week over week.
Mistry, P.K., Branigan, N.K., Gao, Z. et al. Nonergodicity and Simpson’s paradox in neurocognitive dynamics of cognitive control. Nat Commun 17, 3494 (2026). https://doi.org/10.1038/s41467-026-71404-0
19 comments
[ 2.7 ms ] story [ 36.4 ms ] threadThis should surprise no one. You took a large population and found subpopulations within it. If you want to look at a population average, then use the population data. If you want to look at kids with specific attention needs (guessing ADHD since medical related) then design a study to select for children fitting that criteria, including subtypes.
This seems like the type of thing that should have had a study about study design done long ago that they could have followed to help them structure their own population selection.
> when analyzing average trends in groups of children, slower reaction times to the “Go” signal were linked to increased activity in many brain regions, including the default mode network
> However, when an individual had a slower reaction time to the “Go” signal, activity decreased in the default mode network — the opposite of the group-level pattern.
Now that we’re finally moving to the next stage of neuroscience due inscrutable latent systems (aka LLMs), I can’t help but feel some nostalgia. It’s all fun and games until someone makes a lie detection helmet that actually works…
- As another commenter said, this is a known disadvantage of averages. I'm curious if it's possible to get a median result from per-individual averages. I'm not familiar enough with how this research is done to get a result.
- Was any effort made to re-test individuals in a second/third/etc session, showing consistent patterns to the brain activity? I know it was consistent within a session, but I'm curious if it might change week over week.
https://www.nature.com/articles/s41467-026-71404-0
Mistry, P.K., Branigan, N.K., Gao, Z. et al. Nonergodicity and Simpson’s paradox in neurocognitive dynamics of cognitive control. Nat Commun 17, 3494 (2026). https://doi.org/10.1038/s41467-026-71404-0