The dark neuron problem, or mind reading at 90% accuracy

The dark neuron problem, or mind reading at 90% accuracy

By Mark Humphries

Scroll to Article Content

I’m going to read your mind. Right now. Ready? Don’t get freaked out. Deep breath. Here we go…

You’re not thinking of anything.

Wasn’t that cool? No? Was it a hamster? Damn, should have guessed hamster. Actually I was right: most of your cortex was doing nothing. Most of your cortical neurons were silent. Most of you was not thinking of anything.

Stick my brain in a fMRI scanner, what do you see? Well, once you’ve successfully explained to the police why you felt it necessary to first chop my brain out rather than just ask me to climb into the scanner, you nutter, you’ll see a picture of my brain light up. There’ll be lots of colours. In some places there will be very intense colours. Imaging tells us that there is activity all over cortex.

But no, it’s not true. (Let’s leave aside the small but crucial technical issue that fMRI scanners record blood flow and not anything resembling neural activity.) Each dot of colour on the screen showing my brain is a cube, and each cube contains millions of neurons. Tens of millions. And most are silent.

Historically, neuroscientists did not grasp this. In classical neuroscience, each neuron has a role. Every neuron that was recorded responded to something. To the extent that in the 1970s showing the mere existence of a handful of “silent” neurons in the brainstem was enough to get a Science paper.

This was a false picture of the brain. Up until the 1990s, neuroscientists could only record individual neurons in an animal blindly. They inserted their tiny, sharp electrode into some bit of cortex, and only knew they had found a neuron by the noise it made on the speakers in their lab: tick tick tick tick…. That is, they could only find active neurons, because their only way of finding neurons was by their activity.

But then neuron imaging came along. We point a digital video camera at a bit of brain, and in that bit of brain each neuron contains a chemical that lights up when the neuron is active. By filming the bit of brain in very sharp focus, we can see all the neurons—and see which ones light up. Turns out we were only recording the tip of the iceberg. Most neurons we can see in these videos were not active. And responded to no stimulus.

(Early neuron imaging papers showed silent neurons, but still left open many problems: was there some technical problem with the chemicals we used? Did they not get taken up by all neurons? Did they get damaged by the chemicals? But other teams used patch-clamping—“Susan”—to prove they existed. Here they find a neuron within an animal’s brain by attempting to physically attach to it—“patching” to it. Because they are finding neurons by physical contact, the experimenter is not relying on activity. Once attached, they can play the animal some sounds, or get it to touch something, and see if their attached neuron becomes active. Largely, it did not. Most neurons were silent most of the time.)

Whatever the recording technique, the numbers agree. About 10% of cortical neurons produce about 50%—at least half—of all spikes recorded. Most neurons produce less than one spike per minute. The brain communicates using spikes, so these neurons are not communicating anything, to anyone. They are the dark neurons. (1)

Dark neurons are a real pain. Our theories for how bits of the brain work are based on the patterns of neuron activity in them. But the dominance of dark neurons means that our theories of the brain are only about a tiny minority of actual neurons.

So what, you may ask? Ignore them! Embrace the light, turn away from the dark side. Luke… Luke… Sorry, where was I:

These dark neurons must do something. Neurons are expensive to build and expensive to maintain. If we didn’t need them, it is highly unlikely evolution would have brought them into existence.

What, then, could they be for?

In the lab we can only probe a tiny fraction of the real world input these neurons receive. So maybe our experiments are just not rich enough, and we need recordings of neurons in animals behaving naturally over very long times—days, weeks, months—to find out what they are for. Then, if we sample enough of the animal’s life, we can find out what the dark neurons respond to. Technically, this is just about within our grasp. Practically, less so: some poor graduate student has to actually do these recordings, locked in a lab for weeks on end, and lose their entire social life, their romantic partners, and their self-esteem in the process.

Or maybe the dark neurons are a resource to be plundered by the brain. Perhaps whenever we lay down a new memory, some of these neurons are recruited to the cause, and will become active to recall that memory.

Or maybe the few spikes of the dark neurons are crucial after all. Each fires rarely, but they vastly outnumber the active neurons. So collectively they produce a lot of spikes. In this idea, it only matters what the whole gang of neurons is doing, not the individual.

Whatever they do, it’s clear the dark neurons make up most of cortex, and do nothing most of the time. Claims we are close to understanding the brain are truly mad. Sure, grand claims are great publicity for their authors and utterers, but they reveal deep ignorance. Because of the dark neurons, this is like claiming we understand the rules of football because we watched one player running up and down the pitch for one minute. Or understand the narrative of Hamlet because we know five lines—all Hamlet’s—from the play. Or that we know how an electric guitar works because we were once forced to listen to Ynigwie Malmsteen. (Don’t click that link. There are things you can’t unhear. It will alter your relative perception of the musical universe to the extent that Taylor Swift seems like the second coming of The Beatles, instead of that one song with an interesting enharmonic change in the chorus.)

Truly, dark neurons mean we know nothing about how the brain works. We don’t know what 90% of cortex is doing, and we don’t know why we don’t know. Happily, that means we can still have a lot of fun finding out.

This essay was originally published on Medium.

Mark Humphries is Chair in Computational Neuroscience at the University of Nottingham. He is the founding editor of The Spike, a Medium online publication. He lives in Sheffield, England. Twitter @markdhumphries


[1] The existence of so many silent neurons also makes sense when you think about how much energy they consume: the brain’s energy budget per neuron suggests that large brains must have mostly quiet neurons, because there’s only enough energy for the minority to be active at any one time.