iiinova — where your digital self keeps evolving

Scott Marek @smarek0502.bsky.social · 15/06/2026

Thanks! Yes. It will be posted to Youtube. I can share once it is.

Reposted by Scott Marek

Lucinda Sisk @lucindasisk.bsky.social · 14/06/2026

🧠‼️ Socioeconomic variables show the strongest associations with the brain out of 649 measures—surpassing cognition and mental health. Grateful for the chance to write a perspective with @ted-satterthwaite.bsky.social about @smarek0502.bsky.social and team's important findings! tinyurl.com/bdhk6bc8

science.org

Childhood environments shape the brain

Socioeconomic factors top the list of influences on children’s brain structure and function

04018

Scott Marek @smarek0502.bsky.social · 11/06/2026

Of course not. Nor are we saying that.

110

Scott Marek @smarek0502.bsky.social · 11/06/2026

Right - bc you knew that socioeconomics dominated brain-behavior associations to the point that they render brain-IQ associations largely meaningless.

210

Reposted by Scott Marek

Evan Gordon @gordonneuro.bsky.social · 11/06/2026

This is one of the most socially important scientific findings I've ever been a part of. SES has a larger cross-sectional effect on brain function than any other variable. Prediction of IQ from brain data is actually mostly just predicting SES. Mind blowing.

1238

Scott Marek @smarek0502.bsky.social · 11/06/2026

In sum, SES dominates childhood brain organization. Zip code matters most. These findings have been at the forefront of my mind daily, with Darwin’s quote living rent free in my head: “if the misery of the poor not be caused by the laws of nature, but by our institutions, great be our sin.”

2268

Scott Marek @smarek0502.bsky.social · 11/06/2026

This was evidence of shortcut learning, a well-known problem in machine learning & AI. Rather than models learning brain-IQ associations, they are learning links between the brain and SES. So much so that models trained to predict brain-IQ associations were better at predicting SES than IQ.

1132

Scott Marek @smarek0502.bsky.social · 11/06/2026

When the correlation between SES and IQ was r<0.1, brain-IQ prediction performance was poor (out of sample r < 0.1). This was not the case for varying the training sample by the p-factor. Brain-IQ predictions only performed well when IQ & SES were correlated in the training sample.

1112

Scott Marek @smarek0502.bsky.social · 11/06/2026

To dig in, we performed cross-contextual analyses, in which we trained models on varying sample compositions with respect to SES (high SES only to All). Critically, we tracked the SES-IQ correlation in every training sample.

1141

Scott Marek @smarek0502.bsky.social · 11/06/2026

However, when models were trained on HIGHER SES strata only, model performance plunged to near 0 (r=0.01).

1121

Scott Marek @smarek0502.bsky.social · 11/06/2026

Likewise, when models were trained on LOWER SES strata only, model performance was high (r=0.3).

1121

Scott Marek @smarek0502.bsky.social · 11/06/2026

First, we stratified training samples (ridge regression, split-half) by SES, matching the replication (test) sample IQ distribution to the training sample IQ distribution. When models were trained on a broad spectrum of SES, model performance was high (r=0.3).

1121

Scott Marek @smarek0502.bsky.social · 11/06/2026

But, SES & IQ are fairly strongly correlated (r=0.31).

1111

Scott Marek @smarek0502.bsky.social · 11/06/2026

Next we used machine learning to further dig into the potential confounding of brain-IQ associations by SES. The most common multivariate BWAS is predicting IQ scores from brain data. And it turns out it can be done quite well (r ~ .40).

191

Scott Marek @smarek0502.bsky.social · 11/06/2026

We compared the SES-adjusted & unadjusted IQ BWAS map to arousal and reasoning. Before SES adjustment, the IQ BWAs map was similar to arousal (r=.42) and dissimilar to reasoning (r=-.04). Adjusting for SES made it much less similar to arousal (r=-.08) and more similar to reasoning (r=.16).

1121

Scott Marek @smarek0502.bsky.social · 11/06/2026

We first asked what happens if we control for SES in regression models? Tldr; the majority of brain-IQ associations became much weaker, so that 70% were no longer statistically significant. The strongest IQ associations were those most affected by SES.

1132

Scott Marek @smarek0502.bsky.social · 11/06/2026

This indicated that socioeconomics may affect the brain via sleep deprivation & stress. But why did IQ show this motor & sensory pattern and not the fronto-parietal one of higher-order cognition?

static.klipy.com

ALT:

2141

Scott Marek @smarek0502.bsky.social · 11/06/2026

The BWAS maps of SES, exposome, and IQ did not match those of known neurobiology related to higher-order cognition (fronto-parietal). Rather, they overlapped with arousal - norepinephrine receptors, sleep, & stimulant use.

1183

Scott Marek @smarek0502.bsky.social · 11/06/2026

What does it mean that the SES pattern (motor & sensory) is pervasive across most brain-association maps? To infer meaning, we used pattern-based inference & compared BWAS maps to evidence from other modalities (PET, EEG, drug fMRI, task fMRI).

1162

Scott Marek @smarek0502.bsky.social · 11/06/2026

The BWAS strength of a variable is dependent on how similar that variable’s brain map is to the SES brain map. This suggests that BWAS strength is directly proportional to the degree to which it resembles SES.

1152

Scott Marek @smarek0502.bsky.social · 11/06/2026

The principal exposome pattern was identical to SES (r =0.97)!

1152

Scott Marek @smarek0502.bsky.social · 11/06/2026

To dig into the brain association map similarity structure, we performed a PCA across all maps. The first component accounted for a whopping 34% of the variance across ALL brain maps The exposome (first component) also mapped onto motor & sensory regions.

1152

Scott Marek @smarek0502.bsky.social · 11/06/2026

What about the brain association maps for other variables? When we compared all association maps with each other, an interesting profile emerged. Highly similar brain association patterns (yellow) across many domains suggested a pervasive, low dimensional mode of brain variation.

1162

Scott Marek @smarek0502.bsky.social · 11/06/2026

Other variables with strong brain-wide associations (sleep duration, screen use, IQ) showed a similar motor & sensory pattern, but little association with fronto-parietal cortex. That was surprising - especially for the IQ map.

1192

Scott Marek @smarek0502.bsky.social · 11/06/2026

We mapped all 649 variables to cortical regions of interest. The SES brain map contained the strongest associations w/ motor & sensory regions. The weakest were w/ fronto-parietal ones.

1182

Scott Marek @smarek0502.bsky.social · 11/06/2026

The strongest association with function (resting state functional connectivity; RSFC) was with the social & economic domain of the Child Opportunity Index (hereafter referred to as SES). It represents financial, occupational, & community resources available in a neighborhood (zip code).

1253

Scott Marek @smarek0502.bsky.social · 11/06/2026

Using ABCD data (~12K, 9-10 yo), we performed mass uni- & multivariate Brain-Wide Association Studies (BWAS) for 649 exposures + phenotypes & ranked the results. SES variables had the strongest associations. 37/40 top associations reflected children's SES. SES: The elephant in the brain.

1212

Scott Marek @smarek0502.bsky.social · 11/06/2026

What matters most for childhood brain organization? We analyzed 649 variables. The answer: Socioeconomics (SES); with brain patterns pointing at sleep & stress as drivers. Even brain-IQ associations were better explained by SES. In Science today: www.science.org/doi/10.1126/...

6319169

Scott Marek @smarek0502.bsky.social · 19/05/2026

Partially. In the twins analyzed so far, the data are sick. Definitely going to try to get something together sooner rather than later, but I spend most of my time grant writing now 🙂

120

Scott Marek @smarek0502.bsky.social · 19/05/2026

I have 3 hours in each of 10 adolescent twin pairs (5 MZ pairs, 5 DZ pairs)

110

Reposted by Scott Marek

Evan Gordon @gordonneuro.bsky.social · 22/04/2026

I always assumed that brain function had to line up with cytoarchitectonics. It turns out I was wrong. Human cortex, especially PFC, is tiled by chains of functional patches that subdivide and interlink architectonic areas into parallel processing streams. www.biorxiv.org/content/10.6...

14234122

Reposted by Scott Marek

Roselyne Chauvin @roselynechauvin.bsky.social · 13/03/2026

The 6x US memory champion – Nelson Dellis – can memorize a deck of cards in 40 seconds and knows the first 10K digits of pi. To figure out how, he let us peak inside his brain. Here is what we learned in our precision brain mapping study www.biorxiv.org/content/10.6... youtube.com/shorts/MryMq...

youtube.com

how does his brain do it ? #neuroscience #memory #sport Nelson Dellis 6x US memory champion

YouTube video by Roselyne Chauvin

58532

Reposted by Scott Marek

Bratislav Misic @misicbata.bsky.social · 13/02/2026

Inter-individual variability of neurotransmitter receptor and transporter density in the human brain | doi.org/10.1007/s004... How do neuroreceptor distributions vary across people?

13414

Reposted by Scott Marek

cjl2007.bsky.social @cjl2007.bsky.social · 12/02/2026

🚨 New preprint: Objective Quality Assessment for Precision fMRI Precision functional mapping (PFM) enables individual-level brain network studies — but demands more, and better, data. We introduce an objective framework to determine when a dataset truly supports interpretable, replicable PFM.

13517

Reposted by Scott Marek

Evan Gordon @gordonneuro.bsky.social · 05/02/2026

I’m very excited about this new work describing the specific cortical-striatal circuits that are altered in Parkinson’s Disease and that should be targeted for optimal PD treatment. www.nature.com/articles/s41...

14825

Reposted by Scott Marek

weigangcui.bsky.social @weigangcui.bsky.social · 06/01/2026

Check out our new work on the relationship between glioma and system-level functional networks! We identify that nearly all gliomas localize within a common brain functional network, Action-Mode Network (AMN) . biorxiv.org/content/10.64898/2026.01.05.697608v1

8229

Scott Marek @smarek0502.bsky.social · 24/12/2025

Benjamin, why do I suspect that this will not be the last time we see this pattern?…

130

Reposted by Scott Marek

Benjamin Kay @benjaminkay.bsky.social · 24/12/2025

This is your brain on Ritalin. Got your attention? Stimulant medications like Ritalin (methylphenidate) do, but not in the way you might think. They don't act directly on the brain’s attention systems! Find out what's really happening in @cellpress.bsky.social. doi.org/10.1016/j.ce...

1626990

Reposted by Scott Marek

Benjamin Kay @benjaminkay.bsky.social · 30/09/2025

Ever wondered if your interesting brain-behavior correlation was over- or under-estimated due to head motion, but were afraid to ask? We’ve created a motion impact score for detecting spurious brain-behavior associations, now available in Nature Communications! doi.org/10.1038/s414...

15722

Scott Marek @smarek0502.bsky.social · 28/08/2025

I would love to know the real reason(s) why

010

Reposted by Scott Marek

Damien Fair @drdamienfair.bsky.social · 10/08/2025

I still get chills Meet Mike *30+ years severe depression *first hospitalized @ 13y *20 meds *3 rounds of ECT *2 near-fatal suicide attempts Mike felt joy for the first time in decades after we turned on his new brain pacemaker or PACE see videos, read paper, follow thread doi.org/10.31234/osf...

16390136

Scott Marek @smarek0502.bsky.social · 17/07/2025

I’ve yet to receive a PFM grant.. perhaps the era is not yet over @rodbraga.bsky.social 😅

150

Reposted by Scott Marek

Thomas Yeo @bttyeo.bsky.social · 17/07/2025

1/11 Excited to share our @Naturestudy led by @leonooi.bsky.social @csabaorban.bsky.social @shaoshiz.bsky.social AI performance is known to scale with logarithm of sample size (Kaplan 2020), but in many domains, sample size can be # participants or # measurements... doi.org/10.1038/s415...

317182

Reposted by Scott Marek

Damien Fair @drdamienfair.bsky.social · 26/06/2025

Alert!!!! “An Action Networks Model for Pain” We propose a new model for chronic pain — and highlight two functionally connected cortical networks that could revolutionize how we treat it. 👉 thread below 🧵 osf.io/preprints/ps...

29042