Tarana Nigam

My absolutely fantastic colleague @auksz.bsky.social just got awarded an ERC consolidator grant !!!!!!!!! Congrats Ryszard ! So well deserved and I can't wait to see the amazing things you will do and discover :)
For PhDs/postdocs interested, I HIGHLY recommend ryszard as a supervisor!

Congrats @nedash.bsky.social

Lastly, our thanks to the funders
Research Alliance Ruhr, @sfb1528.bsky.social
@dfg.de
@FACES_epilepsy
@maxplanck.de

Feel free to reach out if you have any feedback, questions or related ideas!

And a bigger thanks to my fantastic supervisor, Caspar Schwiedrzik who supported me through every part of this project, brought exciting new ideas and perspectives and for bringing this project to this final stage.

My big thanks to all my co-authors : Andrea Campos,
@vanmedge.bsky.social ,Juan Vidal, Marcela Perrone-Bertolotti, Philippe Kahane, Thomas Thesen, Orrin Devinsky. My special thanks to the fantastic & inspiring Lucia Melloni @predictivebrainlab.bsky.social who conceived this project.

@ruhr-uni-bochum.de
@yourUMG
@ae.mpg.de
@uni-goettingen.de
@primatenzentrum.bsky.social
@northwellhealth.bsky.social
@ucly.bsky.social
@nyulangone.bsky.social
@GeiselMed

So my huge thanks to the patients who contributed to this study, & our multi-site collaboration with neurologists, neurosurgeons & the entire surgical teams across continents!

The interesting journey of this project has left me with many new questions related to the incredible & rapid flexibility we humans possess: how fast we learn & generalize. These avenues couldn't be answered without the unique possibility to record directly from the human brain.

⏳The cost? Interference. Task-specialization only emerges through repeated exposure, gradually refining representations and improving performance over time. (10/10)

🧩 Wonder why neural representations don’t instantly become task-tailored? A computational trade-off may explain it: reusing shared representations allows quick, “good-enough” performance across tasks. ⚡(9/10)

💥We find that mental flexibility arises by reconfiguring multipurpose, shared neural representations in the sensory areas to rapidly perform multiple tasks. This happens much earlier in the cortical processing hierarchy than previously thought. (8/10)

👁️Interestingly, we find gradual transformation for all the 3 tasks only in the ventral temporal cortex, which is traditionally thought to be a sensory area. This challenges the notion that sensory neural codes are stable, simply representing the external world. (7/10)

🌀 What we found was surprising! Participants performed the task one-shot, but neural representations don’t become task-tailored instantaneously in any of the brain areas. Instead, the representations become optimal through gradual refinement (6/10)

🧠Humans who do 3 different tasks, flexibly forming categories or abstract concepts on each trial - and we recorded neural activity intracranially from multiple brain areas. (5/10)

❓Do neural representations reconfigure instantly for new tasks, or do neural patterns adapt gradually, needing repeated experience, as one would expect from distributed representations that in neocortex? (4/10)

🎯Humans generalize instantly-kids infer what a green apple looks like, never having seen it. This suggests instant task-tailored transformation.

But hold on, let’s not break old knowledge based on 1 weird experience: this suggests neural representations change gradually (3/10)

⚙️We achieve this by flexibly reuse prior knowledge (compositionality) &reshaping neural geometry. By expanding task-relevant dimensions and compressing irrelevant ones, the brain configures task-optimal geometries, allowing multiple tasks on the same inputs. (2/10)

🔥Humans can perform multiple, including entirely new tasks- just through verbal instructions.
This rapid flexibility is a hallmark of human intelligence and lets us adapt to new situations and changing environments.
How does the human brain do it? (1/10)

Multiple task-demands flexibly optimize neural geometry in human ventral temporal cortex A hallmark of human intelligence is the ability to perform multiple tasks immediately upon instruction. Yet, the neural processes that implement such flexibility remain unclear. Using intracranial ele...

🎉 Excited to share one of my projects that I worked on with Caspar Schwiedrzik! Check out my latest preprint: Multiple task-demands flexibly optimize neural geometry in human ventral temporal cortex biorxiv.org/content/10.6... (0/10)

PhD student in computational modeling of highdimensional visual category learning (m/f/x)

jobs.ruhr-uni-bochum.de/jobposting/3... This position involves digital-twin modeling. It is supervised by Caspar Schwiedrzik & will be in collaboration with the experimentalists in the lab doing non-human primate & human work & with @sinzlab.bsky.social Fabian Sinz who is a deep learning expert.

PhD student in computational modeling of highdimensional visual category learning (m/f/x)

Exciting computational cognitive neuroscience PhD position in our lab on Flexible Dimensionality of Representational Spaces in Category Learning!
Caspar Schwiedrzik, my supervisor is an absolutely fantastic scientist & an incredible supervisor. I strongly recommend this position & working with him!

My incredible friend & colleague @nedash.bsky.social has an interesting open position to use machine learning for predicting decision making of primates in naturalistic scenarios! She's a fantastic scientist, very supportive supervisor & has many ambitious & exciting projects going on in her lab!

Now officially out in JML:

"Shades of Zero: Distinguishing Impossibility from Inconceivability"

(cool kidz summary below)

www.sciencedirect.com/science/arti...

My incredible ex-labmate , @auksz.bsky.social is advertising an interesting PhD position that he will be co-supervising ! Highly recommend Ryszard - He's a wonderful scientist and an absolutely lovely person!!

In case you missed my talk at Cosyne yesterday, drop by to hear me talk about my work on top-down flexibility through predictive context in the macaque face processing system.
Where: Primate Neurobiology meeting @primatenzentrum.bsky.social
When: 4th April at 9am
Looking forward to more feedback!

I'm looking forward to an exciting discussion in this workshop and input on my work! Join us on the 1st April !
toliaslab.org/workshop/cos...

I'm giving a talk on dynamic code-switching and top-down flexibility in the neural codes for face processing in macaque monkeys at #Cosyne2025 @cosynemeeting.bsky.social in the exciting workshop on Object-centric representations! (organized by @katrinfranke.bsky.social & @mdiamantaki.bsky.social )