Neurons: pretending to be important since 1888
Or: It’s time to abandon our (neuroscientific) faith, and embrace the ensemble doctrine
You have about 17 billion neurons in your cortex. I have about 17 billion neurons in my cortex. We all have about 17 billion neurons in our cortex.
The tiny roundworm C. Elegans has 302 neurons. Not about 302 neurons, exactly 302 neurons. Precisely 302 neurons. Specifically 302 neurons. Unless it’s male, in which case it has 385. The extra 83 are, of course, all to control mating.
When you only have 302 neurons, each neuron has a specific job. One contracts a muscle – hey, turn! One senses salt – hey, food! One connects the salt-sensing neuron to the muscle-pulling neuron – hey, turn this way to get food! Man, life as a roundworm is just so much less stressful.
On the other hand, roundworms cannot gaze at the deep pink of an autumn sunset. It does not know what “pink”, “autumn”, and “sunset” are, or how to gaze at them. And neither do any of your neurons, individually.
Anything you think, feel, see; anything you plan, decide, do: all are the result of many neurons active at the same time. Not the result of one neuron. There is no one neuron for “pink”, no one neuron for “sunset” – no one neuron for eating an entire tube of Pringles in one sitting. If there was, we could find the little git and cut it out. No, all sense, thought and deed result from many neurons active together. We call them “ensembles”.
Ensembles are so important to our theories for how the brain work that the journal Science published not one, but two papers on them in the past few weeks. (Your typical scientist’s mental model of the universe has Science and Nature vying for top spot as the most important scientific journal. Everyone was stunned in 2010 when it turned out to be Acta Crystallographica Section A. Why anyone persisted with the bloody stupid Impact Factor after that shambles is beyond me).
These two papers tackled different aspects of ensembles. Carrillo-Reid, Yuste and colleagues tested the idea that an ensemble is built from neurons that become wired together. That is, they fire together because they are wired together: the firing of one makes it more likely that the others will fire. And how do we make neurons wire together? By making them fire together! When one neuron reliably fires before its target neuron fires, then the connection between them gets stronger. Carillo-Reid and co used the tricksy toolbox of neuroscience to make a group of neurons fire at the same time, over and over again. Lo and behold, when then left to their own devices, that group fired together much more often then before the experimenters interfered with them. An ensemble was born.
These experiments showed that an ensemble can be created by neurons who happen to fire together. This is crucial, as it means that an ensemble can be made by learning. It can represent new things. The second paper, from Cossart and co, asked: represent what new things?
They looked at the hippocampus. In rats and mice, this deals with knowing where they are, where they’ve been, and where they’re going to. In cats? That I’ve stashed the remains of that half-chewed blackbird under the big monkeys’ bed – and won’t they be proud of me when they find it? No. No we were not.
Cossart and co asked some mice to run on a treadmill while they peaked inside their hippocampuses (hippocampi, in truth). They saw neurons that were sequentially active, one after the other, as each mouse ran. Each signalling in turn: now I am here, now I am here, now I am here, now I am here…. Then, when the mice sat still, ensembles appeared – each made up of just some of those sequenced neurons. The ensemble was a fragment of the memory of running. Each mouse was seemingly mulling over what it recently did. Which, as it was always “run on treadmill”, are some of the most boring daydreams of all time.
Neuroscience has few dogmatic principles. One of the few is the neuron doctrine: that the basic element of brains is the neuron, and the activity passed between neurons are how the brain causes, well, everything. This doctrine traces its founding to Ramon y Cajal’s 1888 publication of definitive, anatomical proof that neurons were single, individual cells. But just because they are physically single, separate cells does not mean they must act as single, separate cells.
These two Science papers add to the weight of evidence that we need to make a new dogma. That of the “ensemble” doctrine: the basic element of brains is a group of neurons, and their simultaneous activity. They are how the brain causes everything. This is as true in sea-slugs, with their 19000 neurons, as it is in our brains.
This might not seem like a major step. But it is. It means that we can no longer think of neurons in isolation. That a single neuron responding to an input is not interesting. That a single neuron sending activity to another neuron is not interesting. Only whether many neurons respond is interesting. Only whether many neurons send many outputs to many other neurons is interesting. Heresy!
And you remember our 302-neuron roundworm? Well, actually its brain makes more sense when we think of it as a 302-neuron ensemble, when we think that its neurons’ activity together stands for the answers to three questions: “what do I know about the outside world?”, “what do I need right now?”, and “what am I going to do about it?” No one neuron knows the answers: rather, the answers are carried by the pattern of activity across its 302 neurons.
At this point you might be wondering: what has this guy got against neurons? First he tells us that few of them are ever active – that most of them are dark. Now he tells us that even the few active ones are, basically, not important. What is it? Did they steal your lunch money?
No, no, no! I love neurons. I mourn the ones I marinated in alcohol as a student. Actually, I think each neuron is an exquisitely beautiful computer, each and every one. This distinction is important for neuroscience: ensembles are about what the brain represents; but each neuron can compute something from its inputs, and need not make an output to have done something amazing. That’s a story for another time.