Will the study of the most fine-grained neural workings of the brain, thanks to ever more powerful techniques, enable researchers to claim they understand the brain and the behaviors it produces? There are serious reasons to doubt it, argues a group of neuroscientists.
A few months ago, five neuroscientists published, in the journal Neuron, a paper under the title “Neuroscience Needs Behavior: Correcting a Reductionist Bias”. Nowadays, they argue, countless neuroscience labs around the world are studying the neural basis of behavior in living animals thanks to new and spectacular techniques that allow them to turn neurons on an off, or make them glow when they are activated – and to record the activity of hundreds of them simultaneously – in order to correlate all this data with behavior.
This is giving neuroscientists a grip on the working brain that a few years ago would have seemed impossible to achieve. But, as they marvel at the technological breakthroughs, add the authors, they are ignoring a crucial component of the study of behavior: the careful “dissection” of the behavior itself.
Unless such studies of behavior come first, they write, and are then used to formulate hypotheses that can guide experiments at the neural level in animal models, the current approach – based on the assumption that you can infer what goes on in the brain “from the bottom up” – will not pay off.
An example: when dopamine is depleted in the brain, patients will experience slow movement, rigidity and tremors, all of them hallmarks of Parkinson’s Disease. Knowing this has led to the therapeutical use of dopamine to alleviate the symptoms. But does it mean scientists understand why depleted dopamine leads to those specific behaviors? No, says American neuroscientist John Krakauer, one of the co-authors of the Neuron paper.
Krakauer, who for several years has been spending part of the summer as visiting scientist at the Champalimaud Centre for the Unknown, in Lisbon, talks about this state of things, which he considers very worrisome for the future of neuroscience.
Do you think neuroscience is being limited by its tools?
I don’t think it’s limited by its tools, I think it’s limited by its obsession with tools. Tools are very important, there’s just this view that if you just have the right tools and the right analytical methods, paradoxically you don’t have to actually think as much, the answers will just pop out. It’s like striking gold, you just bang away at the mountain and suddenly you’ll find diamonds and gold.
There’s a wonderful book – the biography of Alexander von Humboldt, by Andrea Wulf. And at some point in the book, Humboldt guesses where diamonds will be found, based on some geological pairing of diamonds with some other substance. So he says something like “if you go where there’s substance A, you will find diamonds too”. And they did! Far better than by just looking and looking and looking and looking. I think it’s good to have an idea of what you are looking for before you go looking.
The analogy we give in the Neuron paper is that it would be like saying that if you analyse chess pieces to death, you would work out the rules of chess. You wouldn’t. And as the British neuroscientist David Marr famously wrote in the 1980’s, however much you study feathers, you’re not going to discover flight. You may think you are, but you’re not.
So, in your view, you should first look at a behavior at the level of the whole organism, and that would then guide studies at more basic levels. Can you give us some examples?
Take feathers and flight. There are birds that have feathers but don’t fly. Dodos had feathers, but their feathers had an insulation role. So if you were to pick up a dodo feather in Mauritius, when they existed, and study it, you’d never come up with the idea of flight. Whereas, if you understand flight and some of the physical principles of flight, you can begin to see – when you now go back and look at a feather – how instead of being only an insulator it could actually be useful for flight.
What we argue is that you should first come up with notions of flight and flight behavior, and then go looking at feathers. You can then find out interesting things about feathers and go back and say something more refined about flight in general.
Ant colonies are another example: you would never predict the intelligence of an ant colony and the structures they are able to build by just studying one ant. The intelligence that arises from their aggregate behavior is qualitatively distinct from what any one ant ever does.
But if you know that ants, when they are aggregated, do these amazing things, and now, armed with that knowledge, go back and look at a single ant, you begin to notice that it has things which you wouldn’t have looked for – like the pheromones it releases, that may be used to signal to other ants. And that became possible because you knew what questions to ask, what to look for.
Is the same true with the brain?
The same is true with neurons. The interesting work on neurons is going to be done once you know what to look for, and then that work will give feedback to refine the theories. I’m just talking about the directionality of the work – not that one shouldn’t do it.
I think we need pluralism, we need to accept that there are many ways of doing science. And unfortunately, we’re in the grip of a totalizing belief in data and techniques that’s crowding out other ways of doing science. That’s my objection.
Do you think that, however powerful the technologies become, we will not be able to go from studying neurons, or circuits, to any meaningful insight on behavior?
It depends what you’re asking. If what you want is to cause parkinsonian symptoms in a mouse, it is both necessary and sufficient to deplete dopamine from the substantia nigra to the striatum. That is going to cause the symptoms. And I think that from a standpoint of intervention and manipulation and therapeutics, we use that already. But does that mean we understand those phenomena? That’s what I’m questioning.
If you want to know why depleting dopamine in the substantia nigra leads to rigidity and tremor, we have no answer. I agree that it caused them, but why did those particular symptoms arise? Dopamine depletion is the starting condition, but I don’t know why it leads to tremor, rigidity or bradykinesia [slow movement].
Causality and understanding are not synonyms. But now, because the tools are so powerful, causality has become the reigning idea in understanding.
In the case of Parkinson’s, how can dissecting the behaviors themselves serve as a guide?
In our lab, we came up with an idea. It was basically that, for example, when someone lifts this pen up, they pretty much do it with the same tempo – but a Parkinson’s patient may just do it more slowly. You can then ask: why did the Parkinson patient do it slowly?
But we could also ask a more general question: why do we do things at a certain pace? Why do we usually walk at a certain speed – but when we need to catch a bus before it leaves, we speed up? Might it be because we’re motivated to speed up, whereas Parkinson patients slow down because they’re unmotivated?
That’s exactly what we found. And then Josh Dudman, at Janelia [Howard Hughes Medical Institute’s research center, in the US], inspired by our paper on humans, did lovely work in the mouse, reproducing our results and beginning to look at the neural substrate for it. So he was inspired by that framework to go looking.
It’s just about being thoughtful. But we’re not really promoting thoughtful attitudes nowadays. I always joke that if Darwin were to be interviewed for a job, and say that he wanted to journey for a couple of years on a boat and go looking in forests and take notes, nobody would hire him.
What we are saying is that neural data is better gathered and interpreted if it’s under the framework of ideas and well-characterized behaviors.
But neuroscientists are not just looking at neurons, they’re looking at circuits of neurons. Doesn’t this afford a broader view of the workings of the brain?
Some people will now say circuits instead of neurons, but otherwise the sentence has the same structure. Somehow, by inserting the word circuit you think you conceptually moved further – but you haven’t, in my view. It poses exactly the same problem. You think you’re closer, but conceptually the gap is as large.
Of course it comes down to what you’re trying to do. If you’re trying to intervene causally, ultimately with some therapeutical goal in mind, then fine. All I’m saying is that when you utter the sentence “I want to understand how the brain works”, what I think that means is “I want to know how the brain leads to behavior”. And I say there’s more to that question than just correlating a basic level with a higher level.
In my opinion, there’s just not enough history, sociology and philosophy of science being taught in graduate programs, so the words “understanding” and “explain” are just not taught. And scientists get impatient, because they’re being so ingenious at the level of engineering and mathematics, and techniques… “Why are you making me have to think about it?”, they’ll say. “I don’t want to think about it, I want to be clever doing it!” It’s bizarre! And problematic!
Is this going to change?
I think it is, I think people are moving towards more ecological tasks, beginning to be able to use the very same tools that were being used to intervene to actually be able to have better behavioral characterization.
People are going to use technology to allow more elaborate free behaviors and better quantification of behavior at the same fine-grained level of the neural work. I don’t know whether that’s going to substitute for the conceptual frameworks that are going to be needed for the behavioral work. But I would say that really good scientists, for example here at the Champalimaud, are actually doing thoughtful work at multiple levels, and it would be unfair to characterize them as just data collectors. They are actually aware of this.
So not everybody is obsessed with technology…
I think a subset of the “circuit crackers” are in fact beginning to be psychophysically more thoughtful. And it’s interesting that they’re doing a lot more work with people like me, who do human work and develop concepts. So the pipeline is also flowing in the opposite direction. Of course, therapeutic science is being generalized from the animals to the humans, but some of the conceptual work is going from the humans to the animals. Which is exactly what I think should happen!
I don’t think it’s going to work to do lots of sophisticated neural work with overly simplistic behavior that has been under-conceptualized. It’s an argument that a lot of people haven’t yet quite understood. But I’m very much looking forward to thoughtful rejoinders to our Neuron paper, from philosophers and from scientists.
Nonetheless, there’s something of a religious ilk when it comes to neuroscience. We want it to explain our life. And it promises to explain our life: “Why am I sad?”, “Why am I impulsive?” Why… And this is just a false promise in my view. It’s not going to do that anymore than religion did. In fact, it will be inferior to religion. So it’s that totemic explanatory promise that people imbue neuroscience with.
But will we manage to better understand the brain – or at least some parts of it?
We’ll be able to do more intervention medically. But I don’t think it’s going to make our art any better, I don’t think it going to make us know why we think things are beautiful. The criticism in our paper is that if you just said “I want to do causal things”, “I want to do therapeutic things”, “I want to stop the tremor”, that would be fine. But neuroscientists want to be able to say “I’ve understood the brain”, “I’ve understood art”!
Why all that hype on their part?
They rely on that mystique to get their grants. In the New York Times science section, there’s always something about the neuroscience of morality, the neuroscience of crime, the neuroscience of love. It’s annoying! I think we should deliver on what neuroscience can actually do.
Ana Gerschenfeld works as a Science Writer at the Science Communication Office at the Champalimaud Neuroscience Programme
Edited by: Catarina Ramos (Science Communication Office). Photos: Gabriela Martins (CCU), João Camilo.