contra the modular view of the brain:

there's this idea that specific brain regions 'do' things.

'the amygdala generates fear' or 'prefrontal cortex is responsible for executive function'

how do we know this? put people in huge metal detectors and see which parts of their brains get more oxygen during certain tasks you can try to do the same thing with a microprocessor; each transistor is kinda like a neuron, and 'display donkey kong' is kind of like the behavior of a brain.

You can then get the average activity ('local field potential') in a specific spatial region, which is similar to fMRI.

if you do this, you get confusing conclusions like "this cluster of transistors is responsible for the donkey kong image"

if you apply a bunch of other common neuroscience techniques like connectomics, lesion studies, tuning curves, you get a bunch of obviously wrong results like "there are distinct transistor cell-types with different functional roles" when there is only one type of transistor or "this transistor encodes visual brightness" — when really it just happens to correlate in a highly nonlinear way

we don't have good models for the brain, but an even deeper challenge is that our tools aren't even adequate for distinguishing between a good and a bad model of the brain