Also, Ila Fiete gave a talk today on this exact study. I asked the very question we are discussing and she said they are planning perturbation experiments to causally verify whether these signals originated in distributed dynamics or if they are just readouts. Very cool!

That there are functionally specialized modules, but those modules interact dynamically to produce a given behavior. So it's like a compromise between functional specialization and distributed dynamics. Something to appease everyone :)

This way signals from other regions are used to refine the original function of visual cortex that is perception. There was also another view from the workshop: cognition is both modular and distributed, where modules communicate across regions in a distributed manner to produce behavior.

I was just in a neurips workshop where Rajesh Rao was talking about efference copies of impending motor actions communicated to visual systems to correct for visual expectations that are distorted by self-generated motions. He says perception and action are tightly coupled.

Yes. makes sense that signal B should be there as an indicator of function B. It was a point @earlkmiller.bsky.social was making earlier that we do not know all possible functions that are being executed, and we should listen to the brain more closely? Perhaps, we'll find signal B if we look for it?

Ok, how about this? Say signal A originated from function A in region A. But it is now read out in all regions. So region B uses signal A for another function B. So region A is still specialized for function A, and region B is specialized for function B.

I'm inclined to agree as well. The signals in a region (even if it is a readout from a signal originated in another region) are likely to serve functional roles in different downstream functions. Like subroutines for various functions. But it's not an argument against functional specificity.

I might be wrong, but I felt that the evidence presented wasn't sufficient to question the assumption. All it said is that signals related to cognition are found everywhere, but that could just be downstream effects of signal transmission from the point of origin where the signal originated.

We can look for whether specific computations are happening 'somewhere' in the brain for that behavior. Lesioning a 'function' rather than region is an interesting thought (not sure how that can happen). That said, there is strong evidence of region-specific anatomy in facilitating 'functions'.

In 'functional' specialization in producing a specific behavior - if a region is replaceable,either 1. the 'function' performed by the region is integral but is now being done by other region. 2. The function is not crucial. Thinking in specific computations rather than regions might help?

Fully second this!

A simple reason for motor signals to be everywhere is that they need to be cancelled out to model meaningful behavior. As body is the frame of reference from which every cognition is executed and those reference distortions need to be accounted for everywhere.

How about we just let everyone write them the way they want. If Blake and I want to write in bullets we should be allowed to, and if others want to write art that should be allowed too. At the moment we all have to write this generic and boring pseudo objective style.