Watching a paradigm shift unfold: how the brain makes decisions
Three brutal rounds of scientific boxing
Do you want to read this? No? Then here’s the ending: three new papers, all from the same team, have just torn up and stomped up and down on the textbook account of how the brain makes decisions.
Changed your mind? We now have no idea which bit of your brain made that decision. Last year we did; this year, we don’t. The signs are that a true paradigm shift is unfolding right in front of us. So let’s take this once-in-a-lifetime opportunity to observe the rare, majestic “paradigm-shift” creature in its natural habitat. And watch it crap all over everything.
At the start of 2014, the theory of how brains make decisions looked something like this. (Specifically, “perceptual” decisions — deciding what we are seeing, and deciding what to do about it). One or more bits of cortex fire to represent what’s going on in the outside world: like how fast things are moving, what direction they are moving, and what colour they are. (The big, red lion is running away from me. Phew. The bigger, greyer rhino is lumbering towards me with intent. Bugger.) Another bit of cortex adds up all that information to see if it favours one decision (“I see a lion, moving away: stay put”) over other possible decisions (“I see a badger, attached to a dirigible: ring my psychiatrist to up my dosage”). When enough evidence has been gathered for one particular decision, that decision is “made”.
Hopefully, correctly.
(“Hey, Dr Golding, it’s me. Yes, the badgers again. Dirigibles this time. Well, now you mention it, it is better than the gyrocopter. Hang on, something’s moving — oh crap, it was a lio….”)
We’re not allowed to put study participants in a locked room with a lion anymore, so how does one measure decisions and neurons at the same time in an experiment? Take some animals. Make them watch some dots moving on a screen. Most jiggle about, randomly. Only a handful of these dots move in a consistent direction. The animal’s job is to watch these dots for long enough to work out what that direction is — and then look in that direction. That looking direction signals what the animal thought it was seeing.
In this dot-world, we had it all worked out. Neurons in area MT of cortex were the ones that represented what’s going on in the world — the moving dots on the screen. Neurons in area LIP of cortex were the ones that added up all this information — about how many dots were moving in each, different direction. Our evidence was that neurons in LIP increase their firing rate for the direction in which the dots consistently moved. And the eye movement in that direction didn’t happen until those area LIP neurons reached a high firing rate. So, that was how the brain makes decisions: area LIP neurons add up sensory evidence and make a decision.
Simple. Now, torn asunder, piece-by-piece, by the three papers from the combined teams of Alex Huk and Jonathon Pillow:
Paper number 1 (Nature Neuroscience, 2014; Memming Park et al): LIP neurons increase their firing? Nope. Many, perhaps most, LIP neurons do not increase their firing rates when they should be adding up evidence. Many showed weird patterns of increases and decreases. So the simple adding-up evidence story is looking wobbly.
Paper number 2 (Science, 2015; Latimer et al): Still, some neurons in LIP increase their firing as they add up evidence, right? Nope. For those LIP neurons that do increase their firing rates, the activity in a single decision is better explained as a sudden jump from low to high rates than a continual increase. So individual neurons do not seem to add up evidence at all. (For the neuro-hardcore: yes, the population of neurons could still add up evidence. If each jump is one piece of evidence, and different neurons jump at different times, then the population of neurons is adding up evidence).
Paper number 3 (Nature, 2016; Katz & Yates et al): OK, so LIP still, somehow, adds up evidence, by some method we haven’t quite figured out yet? Nope. If we switch off area LIP, in the bit where we see neurons increasing activity, nothing happens to the decision. Nothing. The animals go right on making the same choices. Just as fast and just as accurately as they did when LIP was still working. Area LIP is not needed for decision-making.
(For the neuro-hardcore, two things were particularly fantastic about this paper: first, it got out of its own way — simple text, simple figures, story told cleanly and crisply. If nothing else, read it as a model example of how to tell a scientific story. Second: to really take a kicking to the “decision-making” theory, the authors showed that turning off LIP did make the animals worse at remembering where a light was. So they could break seemingly simple tasks, but not decision-making).
To the neuroscience community this was like watching three brutal punches in a KO round of boxing. Left hook — THUMP — neurons don’t increase their rate — the theory’s knees are wobbling. Roundhouse — SMACK — those that increase actually jump, not add-up— slamming into the ropes, the theory’s chin is exposed. Upper Cut — BLAMMO — that whole brain area is not involved anyway. And the textbook theory is lying prostate on the canvas, wondering why some guy in a stripy shirt is holding up 10 fingers.
Of course, all these studies have issues; science is hard, no individual study is perfect. Paper number 2 in particular is causing a huge, ongoing row. Which, in the end, is pointless if indeed LIP does nothing in a decision. A paradigm shift unfolds.
This is in no way to disparage the researchers that built the original theory. No, it is to praise them. This is how science is supposed to work. Without their astonishing hard work, there wouldn’t be a theory to test. There wouldn’t be a paradigm to shift. So much of neuroscience has no theories or paradigms at all that it is a testament to Mike Shadlen, Bill Newsome and their colleagues that this area has testable, mathematical theories. We should all aspire to have our theories tested. Or, first, to have testable theories.
That said, when the third paper came out you could practically hear the wailing and gnashing of teeth reverberating from certain labs.
If the paradigm is shifting, where is it shifting to? It’s no good just tearing down a theory — that’s paradigm crushing, not shifting. Something has to replace the prevailing theory. The new paradigm for decision-making is emerging: it occurs throughout the brain.
Recent studies have reported activity linked to evidence, and its summing up, in all sorts of bits of cortex. Even in bits of brain that aren’t cortex at all. (Those crazy guys, thinking there’s something cortex doesn’t do all by itself — will they ever learn?) A brain-wide network of activity, not a single, small bit of cortex, makes decisions.
This will not yet be the end of the textbook theory. Like all good fighters, it will demand a rematch. It will train harder, run further, push itself to the limits. There’ll be a montage sequence in a snow-bound log cabin. And then it will step back into the ring. If it somehow emerges victorious, then, pushed to the limits by a younger challenger, it will be a better theory — tauter, leaner, tougher. Science wins, either way.
Want more? Follow us at The Spike
Twitter: @markdhumphries