Thursday, June 25, 2009

How are memories formed?

The first thing to clarify here is the vital difference between short-term memory (or working memory) and long-term memory (what we normally think of as memories).

You can further break down memory into different types such as declarative versus procedural (can you say it out loud, or is it knowhow you can use to do something?) and semantic versus episodic (is it a fact you know, or something you experienced?), but I don't think that's important for the basic question of how they are formed.


If you think of the brain as being like a computer, short-term memory is like RAM and long-term memory is like the hard drive. Information in short-term memory is readily accessible and can be manipulated and modified rapidly, but it disappears once we stop using it. Information in long-term memory is held indefinitely, but it requires a complex retrieval process that is actually astonishingly unreliable. (People can very easily forget things or even remember things that didn't happen.)

Short-term memory we actually understand fairly well. We have three sectors of short-term memory, the visual-spatial sketchpad, the phonological loop, and the episodic buffer.

The phonological loop is almost literally an audio buffer. It stores data in a format remarkably similar to a raw audio format such as WAV, where the encoding is done in terms of frequencies. It stores a limited amount of time worth of audio (usually about two seconds), so you can actually remember phrases and sequences of numbers better if you say them faster. If you ever find yourself saying a phone number over and over to keep it in your head, that's your phonological loop at work.

The visual-spatial sketchpad is essentially a 3D modeling system. We're not sure exactly how it is encoded, but we know that it retains spatial information in three dimensions---in some people such as brilliant mathematicians possibly even four dimensions---in such a way that we can visualize how things look, imagine rotating them in our heads, and combine them as if they were physical objects. Some people are better at this than others, and actually most of what we measure when we measure "IQ" is really the capacity of your visual-spatial sketchpad. (Most of the rest is just your overall health!)

The episodic buffer stores events, and here we have basically no idea how it is stored. My favorite theory is narrative; I think we store events as stories we tell ourselves, often quite literally encoded into words. This explains why we love stories so much, why we can remember things much better if we write or talk about them, and even some of the ways we make systematic errors in memory and reasoning that seem to make us into "protagonists" of an "adventure" against "enemies". This would be a form of what's called semantic encoding, where what we store is the meaning of things. That's all very speculative though; we honestly don't know.

Short-term memory goes away when we stop thinking about it, so we need long-term memory to actually retain information for long periods. The process of encoding short-term memory into long-term memory is called memory consolidation. Actually I think a better term might be compression, because I think what's happening is we are quite literally compressing the data so it can fit into a much smaller storage space. Most of this seems to happen in the hippocampus, but the associative areas and rhinal and entorhinal cortexes are also important. (Frankly I'm not a big fan of this sort of neuroimaging-based stuff about where things happen; they are basically just meaningless facts at this point because we have no idea how any of these brain structures work. Again with the computer analogy, it's like saying that most of your visual processing seems to occur in the video card which is six centimeters to the left of the CPU. Okay, sure, but that tells me nothing about how images are actually processed.)

Most of this memory consolidation happens during REM sleep, which we think has something to do with what dreams are for (again, honestly we really don't know), which is why you'll remember things better if you work through them just before you go to bed so that your short-term memory doesn't have time to empty before it starts consolidating to long-term memory in your sleep. Again the actual mechanism of memory consolidation is still a mystery, but my money is again on semantic or narrative encoding; I think we remember things primarily by making them into stories we tell ourselves---or perhaps stories we dream ourselves. This would allow an enormous amount of data compression---somewhere around 10,000 or 100,000 to 1 in terms of gigabytes required---and again fits with a lot of the types of errors we tend to make when we misremember things.

No comments:

Post a Comment

Thomas Jefferson's election in 1800 is sometimes called the Revolution of 1800. Why could it be described in this way?

Thomas Jefferson’s election in 1800 can be called the “Revolution of 1800” because it was the first time in America’s short history that pow...