Tim Behrens

Interested in cognitive maps? Wondering how you should best train your participants? Luckily for you, we have investigated just that! Very excited to share a preprint from my PhD in which we show how initial training curriculum affects the cognitive maps we form

osf.io/preprints/ps...

[1/6]

Thrilled to finally share this work! 🧠🔊

Using a new reinforcement-free task we show mice (like humans) extract abstract structure from sound (unsupervised) & dCA1 is causally required by building factorised, orthogonal subspaces of abstract rules.

Led by Dammy Onih!
www.biorxiv.org/content/10.6...

SWC is the best place to do neuroscience in the world

We will consider anyone but you will need to convince us you can lead a program that will link the world class systems neuro at SWC with human ephys work in London and more globally,

We're hiring! This is a unique opportunity to translate our understanding of neural computation - from circuit-level mechanisms to computational principles - into the human brain, through the establishment of cutting-edge human neural recording capabilities with collaborators in London and abroad.

We’re hiring a Group Leader!

Join us to lead a transformative initiative in human systems neuroscience.

Find out more and apply ⤵️

www.sainsburywellcome.org/content/curr...

Awesome new opportunity to join SWC if you are into human ephys.

If you work at the intersection of computational neuroscience and machine learning, consider applying for this postdoc position (January 2027 start date):
academicpositions.harvard.edu/postings/15868
An opportunity to work with a great group of people across Harvard, MIT, and UC Berkeley.

At eLife, we have always prioritised quality, constructive reviews.

Today, peer review is part of the paper. Reviews are published as soon as they're in, rejection doesn’t follow review, and the focus stays on strengthening the science in a way that works for authors.
@oxfordpsychiatry.bsky.social

Why don’t neural networks learn all at once, but instead progress from simple to complex solutions? And what does “simple” even mean across different neural network architectures?

Sharing our new paper @iclr_conf led by Yedi Zhang with Peter Latham

arxiv.org/abs/2512.20607

Our work with @georgkeller.bsky.social on testing predictive processing (PP) models in cortex is out on biorvix now! www.biorxiv.org/content/10.6... A short thread on our findings and thoughts on where we should move on from PP below.

This is super cool!

We think cortex might function like a JEPA. It looks like prediction errors in layer 2/3 are not computed against input (as is the idea in predictive processing), but against a representation in latent space (i.e. like in a JEPA arxiv.org/abs/2301.08243 or RPL doi.org/10.1101/2025...).

Let me believe it’s magic for at least a year before you show me how it works please!

The supplementary videos for this preprint are fantastic. Some wild examples of decoding the animal's attentional focus and/or intent
www.biorxiv.org/content/10.6...

This is totally wild. Remember the object they are attending to is presented egocentrically, but the allocentric theta sweeps follow it. The whole system is wired up to provide something like an "integrated attention reflex".

Localizing the brain systems involved in geometric shape perception.

A geometric shape regularity effect in the human brain.

🔗 buff.ly/4UsILev

Authors can now include video explainers in their papers with eLife!

Take a look at our first example from @mathiassablemeyer.bsky.social in ‘A geometric shape regularity effect in the human brain’
buff.ly/LzvvVb9

Unpicking the second Minneapolis shooting frame by frame BBC Verify has analysed footage of the shooting from multiple angles, piecing together a detailed picture of what happened.

www.bbc.co.uk/news/videos/...

(BBC verify is known for clear fact-based analysis without any political angle).

*Multi-region computations in the brain*
When two regions are better than one...
doi.org/10.1016/j.ne...
#neuroskyence

Dick you need to read the paper! This paper goes way beyond the monkey data in understanding the interregional interactions. It fits a multi-regional network model which makes a host of very precise predictions and then it tests them in fresh optogenetic experiments!

At @elife.bsky.social you can now include explainer videos with every figure. Like going to a seminar while you engage with the paper. First example here elifesciences.org/articles/106...

Click the arrows next to each figure to get a video of @mathiassablemeyer.bsky.social explaining it for you!

On research careers in academia and industry The epilogue to a series on Cognitive Science and AI

Should you go to academia or industry for research in AI or cognitive science? It's the most common question I get asked by PhD students, and I've written up some of my thoughts on the answer, as an epilogue to my research-focused series on these fields: infinitefaculty.substack.com/p/on-researc...

Just to add though, that in general the @elife.bsky.social model does not allow reviewers to gatekeep what authors say, only to comment on it. So if we did peer review the videos, it would be same for these.

yes it totally changes the vibe when everyone is thinking of nice things to say!

We have decided not to peer review the videos at the moment, and to explicitly declare under the videos that they are not peer reviewed for scientific content (see below). They are vetted by journal staff for taste etc. But we are open to discussion about this. We can see advantages both ways.

@elife.bsky.social ftw!!

These videos are much much easier to produce and maintain. We're hoping people use them. We have a paper under review right now, and we have made them :)

:) Executable articles are exciting and still a priority, but the way we implemented them last time was too difficult for the authors and maintainers. We have discontinued them in the short term. There are actually new platforms that look much better though, so we may be back in the game soon.

Not sure if you're being serious :). It seems to suggest you need 200-700 plots per paper .....

I like journal clubs where you are only allowed to say positive things about a paper. They are so much more satisfying.