tjvr

On Brain Structure

06 Dec 2012

Here’s a weird thought. How is the brain structured?

In Alex’s Adventures in Numberland, he describes tests done on monkeys that showed that there was a specific group of neurons in their brains which “lit up” when they saw each number of objects. And surrounding numbers would light up a bit, too: for example, if they saw five shapes, the area representing “five” would light up strongly, and four and six would light up a bit, and so on.

Leaving aside the different kinds of number (eg. the difference between hearing the word “five” or seeing the digit “five”, and seeing five objects) – I was wondering: if each number has its own neuron, or cluster of neurons, how can you think about more than one number at a time? How do you represent addition of two numbers, or more complex operations?

And most crucially: does the brain have any way of copying information?

Let me give an example. First, how do neurons work? My limited understanding of them, based on what I’ve heard of neural networks, is that the brain is a network of neurons. Each neuron is linked to certain others as inputs, and certain others as outputs. When the inputs reach certain values, the neuron “fires”, sending an electrical impulse to those neurons connected to its output.

So, say I want to add the numbers 4 and 8. When I think of those numbers, I almost immediately think of the result: 12. Does this mean there’s a neuron connected to both 4 and 8 that triggers 12 into firing? But then how does my brain “know” the difference between the operands, 4 and 8, and the result, 12? Or maybe there are different “copies” of each number for different circumstances – the neuron for “12 as a result” is different to the 12 used as an operand.

And what about calculations I’ve never done before, like 18 + 26, that I don’t immediately know the answer to? Presumably the brain has some sort of procedure, even an algorithm, to work out the result.

But this leads me to the question I asked earlier: can the brain copy information?

Consider memory: my understanding was that information was stored in short-term memory, and then either copied into long-term memory or forgotten. But are these actual physical locations within the brain, or just different states of neurons? After all, I also thought that connections were strengthened when learning takes place; certain pathways get built up through repetition, for example.

And if “memory” is a separate physical location in the brain, like RAM or hard disk storage in a computer, does information have to be “copied” out of memory into a working area to do operations on it? For example, recalling some number – for example, the gravitational constant, g – and using it to calculate the weight of an object using the equation W = mg, or mass times g. In such a design, g would have to be recalled from memory and transmitted across the brain to the working area – in effect, copying the information. So there would have to be at least some neurons concerned with encoding, transmitting, and decoding information at different points in the brain.

Or am I trying to apply the principles of a computer, designed by humans, to the human brain?

It would certainly seem more “brain-like” (or rather, less like a computer) to have a model where the neurons represent the information directly – for example, clusters of neurons corresponding roughly with concepts – than for them to represent the information indirectly, so to speak. Phenomenons like short-term memory would emerge from all the many neurons, only making sense when one considers the brain as a whole.

But in either case I still have no idea how the brain can perform operations on things.