I just picked up a copy of Douglas Hofstadter's most recent book, I Am a Strange Loop. He's the author of the Pulitzer winning Godel, Escher, Bach, which is always fun to flip through and read sections here and there (though I wouldn't try to read it from cover to cover).

Anyway, the reviews are mixed on Loop, but I picked it up anyway, and already found one metaphor in the book that I kind of like. Hofstadter talks about an analogy of the brain to rows of dominoes, set up so that they cause each other to fall, but also have a spring mechanism so that once they fall over, they can set back up to fall over again.

It's a nice analogy for a couple of reasons. One, it highlights the causal nature of neurons. Two, it's a nice, vintage, low-tech analogy, for all those people who are tired of the computer/mind metaphor. Some people have pointed out that the human mind has always been compared to the most complicated technology available at the time: steam engines, electrical communication systems, computers. Well, dominoes date back nearly 1000 years, so you can't really say the analogy is a lazy reflex to the most currently complicated tech.

So while the analogy is interesting in some ways, it of course breaks down pretty fast. The average cortical neuron is connected to about 10,000 other neurons. So in the analogy, a single domino would need to be able to cause thousands of others to fall (and vice versa). Likewise, it is often the firing of many neurons that push another neuron over the threshold to generate a spike, so you would need to envision a system where many dominoes would strike a single domino (and not necessarily knock it over unless they all fell against it). And finally, for learning to be built into the system (which is kind of an important aspect of the brain) there would need to be some way of modifying the likelihood that a particular domino will be able to knock over another. And that's really where the metaphor crumbles, because synapses and their plasticity are highly specific. The mechanisms we have a lot of good evidence for, long-term potentiation and long-term depression, adjust the likelihood that a single neuron will cause another single neuron to fire by modifying the synaptic strength between them based on the ordering of their firing. In the domino analogy, you wouldn't be able to simply adjust the physical weight of dominoes in order to change the way the system works. You would have to have a system in which, when domino #45928 falls against domino #29078 and contributes to it falling over, the likelihood of domino #45928 causing domino #29078 (but not any other dominoes) to fall over in the future is increased.

I don't think Hofstadter meant for the metaphor to be taken that seriously. As with any good metaphor, there are some salient conceptual features that make it useful, but if you stretch it too far, it breaks down.