“The brain works the way it does because it’s made of meat, and meat is not deterministic.” — Anthony Movshon
At first glance, the brain is a mess. More like a tangled ball of yarn than a finely woven tapestry, every combination of neuron-to-neuron is in there, somewhere. Yet look a little closer and this complex structure devolves into very clear regularity. I could take you on a tour of the waves of Purkinje cells, straight-backed like military men, reaching their arms out to passing fibers shooting up from a distant province. I could show you the shapes of the hippocampus where memories are created, messages washing down step by step. I could show you the round columns of barrel cortex, clear to your eye, that precisely mirrors the pattern of whiskers that eventually stimulate them. There is so much visible structure in here that we’re still attempting to unlock.
Until recently, studying the brain was like studying individual water molecules by throwing a rock into a pond. We didn’t really understand that the brain caused behavior until the 1600s when Thomas Willis coined the term neurology. We didn’t know neurons were individual units that talked to each other until a century ago. We didn’t know how neurons spiked until half a century ago, nor what they were doing.
We’re only starting to glimpse what the brain is capable of, how it sends messages without sending noise, how it takes the overwhelming set of sensations and pairs it down into manageable feelings and desires. We’re still grasping at straws here, so many, many straws, and then throwing those straws at physicists and theoreticians and telling them to explain that. And they’ll come up with this explanation or that explanation but really we shrug our shoulders and don’t know what to do with what they say.
Now people are telling us that we’ll never have a theory of the brain like physicists have. That biology isn’t elegant like physics is. You’re going to look at this picture, this hippocampus, precisely structured to create memories, the same in all mammals, and tell me, “it’s too complicated! no one could understand this!”? It’s not elegant and beautiful and clearly structured?
It’s not like there are not clear explanations already. In fact, we know exactly how neurons spike to signal to their partners that something important happened. In 1952, Alan Hodgkin and Andrew Huxley developed a series of equations that not only describe how the membrane voltage of a neuron changes to generate action potentials, but the equations also predicted the existence and kinetics of ion channels that cause those action potentials. By combining these equations with an equation used to describe how electrons move through cables, neuroscientists can simulate an entire neuron. This is exactly how the Blue Brain Project is attempting to simulate large networks of neurons in concordance with physical reality.
So we know exactly how neurons behave! At the end of the day, though, it can only tell us so much. It cannot tell us how networks of neurons move in their intricate dance to generate the right pulses and combinations that cause you to cough, or to turn your head, or to smile and say “hello”. It is like knowing the behavior of a particle but asking why the planets turn the way that they do or why when I push coffee off my table it splatters on the floor.
We have some other equations that represent what the brain does (see above). Some are physical realities, others are algorithms the brain is appears to be approximating. I have gone into what I think of them elsewhere, but two of them are especially important.
First is the general principle that sensory neurons — that first layer of neurons that learns about the world — maximize their information about the world. If you look at Shannon’s information theory, the equations that he deduced to describe signals transmitted down a noisy cable, it turns out that they also have an impressive explanatory power for sensory neurons. This is what these neurons do! They want to grab as much information about the world as possible and bring it into the nervous system to be used by the rest of the brain. But how they can maximize their information is contingent on the environment that the animal is in. The statistics of the natural world vary between cave dwellers, and bats, and people hiking through the bright desert sun, and the nervous system has to take that into account.
Second is the idea that we learn the value of an event or an object by continually predicting how much we will like it and then responding not to the actual value — but to how different that ends up being from our expectations. This is a theory that was proposed from observations of behavior and ended up having a very clear neural analog! There are neurons that release dopamine, but only when something unexpected happens. And this dopamine is a signal that tells other parts of the nervous system: the world is different than we thought; we need to change how we are behaving. And again, I can write down these equations for you — equations that are similar to those that programmers use to build modern, machine learning algorithms — and they will tell you how one part of the nervous system learns.
These are very different explanations of the same brain, and one is not going to be helpful in understanding the other.
We could ask for a ‘theory of the brain’, but it’s not even clear what a ‘theory of the brain’ would mean. There’s a theory of evolution; but that does tell you everything you want to know about evolving creatures? Evolutionary theorists have built the replicator equation, the Price equation, Fisher’s Fundamental Theorem, use game theory, use a bewildering variety of theories to explain — evolution. There’s a theory of electromagnetism: but looking at the equations won’t be useful if you want to build computer without referring to anything else.
Now imagine that there was a theory of the brain. And let’s say that I thought the brain was an information-processing machine that was trying to generate heaps of information about the future in order to minimize harm and maximize reward. And, if you thought through the consequences of this theory long and hard enough, BAM, you could explain everything about the brain.
And say we have these equations, these equations of space or the brain or evolution. Will you look at them and instantly say, “oh, of course we act in such-and-such a way. Of course we are greedy and angry and quixotic in our quest for happiness.” No! You need a large body of background knowledge — the permissivity of space, the characteristic impedance of a vacuum, the likelihood of finding edges in your visual environment, or social environment you grew up in. These are the constants and boundary conditions implicit in using any equation or theory to understand something about the world. Understanding the theory of the movement of a magnetic field through space is useless unless you know a whole host of other things to constrain and describe your world.
So what would be a satisfactory theory of the brain? What do we want it to tell us? What will it look like? Can we just shrug and say, sure, language is just the logical end-result of maximizing information transmission given certain boundary conditions? Would that be… satisfying? Useful? Or would you want something else, more specific?
When it comes to the physical world, we don’t look at one fundamental set of equations to describe everything. It is pointless and impractical. Different levels of explanation need different levels of theory. Sure, if we know every particle and really want to crank through some computations we can take quantum mechanics and particle physics and throw them together and figure out the pressure inside of a box. But why do that when you can use the much simpler equations of thermodynamics? Similarly: to understand the brain will require multiple scales of explanation. A “theory of the brain” might sound interesting but would ultimately be useless, if not pointless.
We need to find a set of loosely interconnected theories of the brain. In order to get at these will require a number of things. First and foremost, we cannot just study one species. What we can explain to ourselves as a general principle when we study a single species often turns out to be contingent on some aspect of that animal’s brain. Take the organization of neurons within visual cortex that respond to the direction of edges. These cells come in beautiful pinwheels, wheeling about so cells that respond in similar ways are near each other. When they were discovered in the visual world of the monkey, we theorized that of course this was how visual cortex was laid out. When we moved to the mouse it turned out that this isn’t a necessary organization: these murinae have a salt-and-pepper organization, cells scattered about almost willy-nilly. Why? The constraints on the nervous system, on the size of the animal and the distance between cells!
It is clear that we need neuroethology: there is nothing about studying a cricket or cockroach or a microscopic worm that won’t shed light on who we are and how our brains work. We need to understand the contrast before we can understand ourselves. To do otherwise would be to say we can understand everything about planets from studying the Earth; who needs Jupiter or Mars?
We also need to understand the boundary conditions that we operate under. What is the distribution of light, and how it speckles across the waves? What is the booming laughter that we hear and the collection of friends that we commune with?
So when you hear someone say that the brain is a mess, that it is just a product of evolution and will never give us ‘beautiful’ laws or truths: tell them that there are regularities, there are consistencies, there are things that it should be doing. Physics is not just a set of equations but measured constants and facts that are contingent on where we are. The nervous system is no different, even if it seems so right now. Tell them that looking for one ‘theory of the brain’ is pointless: ask what they want to know, and then we can try to explain that.