The experimental reason is that damage to most areas of the brain causes consistent and remarkably narrow impairment, ie aphasia.
The reason you should predict localization just from first principles is latency. The bulk of your brain is connective tissue networking the cortex to itself, but even still signals can make many round trips within a small patch of cortex before it can cross the brain. Neurons are large and slow, and the more distant the neighbors they have to interact with the larger they have to be.
Does that mean that the regions of the brain kindly sort themselves into exactly the high-level categories we want them to be in? Absolutely not! Concepts like "attention" or "memory" might not even make sense when looking at how the brain is actually implemented. But regions of the brain probably perform specific tasks.
Probably need to consider I/O too: whatever is doing the front-end processing of vision is probably not too far away from the end of the optic nerves, otherwise you'd have a routing nightmare.
Actually that's sorta not the case, and it is indeed a routing nightmare. Most vision processing is actually done in the far rear, with the signals traveling all the way from the front to the back. However, there are two big reasons why this isn't as bad as you'd think:
- The data from the eyes is already lightly processed. The way the neurons hook into the rods and cones is basically running a few convolutions all at once, and the result is what's sent back
- Most sensory info gets centralized in the LGN before it's routed to the various parts of the brain, so it's about the same hop it would have to go through anyway
We actually still have a lot of questions about what exactly is going on in the middle bit. It seems like structures deep in the brain have duplicate versions of all the sense processing and motor controlling centers. What do they do? Well reflexes probably, and maybe somehow motivation? Reward is tucked in there somewhere too, or so we think.
The reason you should predict localization just from first principles is latency. The bulk of your brain is connective tissue networking the cortex to itself, but even still signals can make many round trips within a small patch of cortex before it can cross the brain. Neurons are large and slow, and the more distant the neighbors they have to interact with the larger they have to be.
Does that mean that the regions of the brain kindly sort themselves into exactly the high-level categories we want them to be in? Absolutely not! Concepts like "attention" or "memory" might not even make sense when looking at how the brain is actually implemented. But regions of the brain probably perform specific tasks.