Heading in the right direction: the brain’s internal compass
A tale of the invisible hand of theory in neuroscience
Take a walk through the centre of an old city — meander through Le Marais, stumble through the Shambles, saunter through the City. Twisting, turning, looking for that hidden treasure — a church, a museum, a secret garden. A bar. Now, head for home. Off you go, retracing your steps, turning and twisting in the reverse order. For you to get home again, your brain has kept a record of all those twists and turns. And to do that, it needed to know how much you twisted and turned at each one.
We have an internal compass, always keeping track of which direction we are facing, and always keeping track of changes in that direction. The brain can use this compass to guide the way. Just like your satnav keeps track of each turn you make in the car — so that it can sulkily work out the new route from the direction you are now facing. (Unlike satnavs, our brains don’t constantly narrate our wrong turns with barely disguised sarcasm. Well, mine does, but I suspect that’s not normal).
A deep question in neuroscience is how the brain can possibly build an internal compass: how can a bunch of neurons keep track of turning and twisting? The answer, it turns out, is simple: make a ring of the things. And it also perfectly demonstrates the invisible hand of theory in neuroscience.
Theorists have long loved this internal compass problem. Their many solutions all have one thing in common: there’s a bunch of neurons arranged in a ring. And the connections between the neurons are arranged so that a few neurons next to each other on the ring are active, while everything else is quiet. This bump of activity then represents the current direction in which the brain is heading (with luck, with a body attached too):
And when the input to this ring changes, so the bump moves to a new location on the ring — and that new location is the new direction in which the brain is heading. When the input isn’t changing, the bump is stable — it just sits there, constantly saying “hey, we’re heading this way”. Noise could not shift it. This stability of the bump means that this ring is an “attractor”, a type of system that can sustain stable activity. That was the theory. The main disagreement between theorists was exactly how the neurons were connected together to make this happen.
Experimentalists everywhere: ha ha ha, very funny, bunch of neurons in a ring. Next please!
But recently a group of researchers at Janelia Farm were looking at the bits of the fly’s brain that help the fly navigate. They noticed that a group of neurons in the fly’s brain were arranged in a circle. Hmmmm…. they thought… hmmmmm… is that a ring attractor?
Testing this was ingenious. Take a fly, flying. Record all the activity in that ring of neurons while the fly flew. And then see if the activity around the ring of neurons changed when the fly changed direction.
Did it? Hell, yes. The constantly moving activity in that ring exactly matched the constantly changing direction of the fly.
So this fly-brain ring at least looks like a compass. But is it the compass? Or is this just a read-out of input from somewhere else? To test this, they turned to ideas from the theories: if it’s really the compass, it’s a ring attractor. And a ring attractor can only have one bump on it, which makes sense as there can only be one heading direction (unless you’re Zaphod Beeblebrox. Actually, damn good question: how do multi-headed creatures do path integration?).
So they forced a second bump onto the ring of neurons. They stimulated a few close together neurons with light, driving up their activity, making a fake bump. And, lo, the other, naturally occurring, bump disappeared.
(Even better, when they stopped stimulating the neurons with light, the new bump drifted away, just as the theories predict too).
The fly brain has in it a ring attractor that is an internal compass.
But what type of ring attractor? Of the many ways of wiring up the neurons to create this bump, which one is right? Here the experimenters rolled up their sleeves and got stuck into the nitty-gritty of testing the different options. Again, they came up with a stroke of genius. The models predict that unnatural moves of the world, turns so violent that they could not physically happen, occasionally make the bump jump. Instead of rotating around the ring, it simply disappears from the old location, and reappears in the new location.
The different ways of wiring the models make different predictions about the jump. Key here was how easy it should be to make the bump jump. The “local” model says it should be equally easy, no matter how far away the violent turn was from the current bump. That’s because, in this model, the inhibition of neurons that only allows one bump is the same everywhere. The “global” model predicts further jumps should be harder. That’s because, in this model, the inhibition of neurons for the opposite heading direction is strongest.
So, which is it? One final great experiment. The experimenters created a first small bump in the ring of neurons using optogenetics. Then straight after they created a new bump at 90 or 180 degrees — as though the fly had suddenly spun round to the right, or completely rotated to face the other way in the blink of an eye. (History does not record if the flies were motion sick at this point). They found it was just as easy to make the bump at 180 degrees as at 90 degrees, matching the predictions of the local model.
So, to the best of our knowledge, the fly’s internal compass is this ring of neurons, and they are wired locally, not globally. A great study, advancing our understanding of the brain, built on deep theory: science at its best.
An observation on the modern scientific enterprise, and one of its major flaws. This fly brain paper was published in Science, one of the holy trinity of journals for biomedical scientists to aspire to (Nature and Cell being the other two). The entire paper is about testing theoretical ideas, and rests on decades of theory work. It references 7 theory papers. Yet none of these theory papers are in flashy journals. They are in the standard archival journals of science. Indeed, the basic idea for the “winning” model, the local model, was published in 1977 in Biological Cybernetics.
You won’t find that down your local newsagent.
This Science paper perfectly demonstrates that theories in biology can take decades to come to fruition. And that neither how many times they are cited by other papers nor their place of publication has any necessary correlation with their influence (the Biological Cybernetics paper was cited less then 100 times in the first 20 years of its existence). Yet as scientists we are all judged by where our papers are published, or how many citations they have, or both.
So, the next time someone asks what theory has ever done for neuroscience, you have my permission to smartly bop them over the head with this copy of Science, and yell “ring attractors you muppet” at their rapidly departing form.
The ring attractor in the fly has an actual physical ring of neurons. For the record, this is cheating. There is no reason why the ring attractor has to be actually laid out in a ring. It’s not in our brains. We’ve long known that the internal compass in mammals is spread over multiple brain regions. And does not have a literal ring.
Yet the same ideas apply. There is a stable bump of activity representing head direction in the mammalian brain. And thanks to this tour de force study in the fly, there’s even shorter odds now that we too have a ring attractor buried in our noggin.
In your more philosophical moments, feeling adrift in life, you may ask yourself “Where am I heading?” Fortunately for you, one part of your brain always knows exactly where you’re heading, even if you don’t.
Follow The Spike for more dispatches from the frontline of neuroscience