Do neuroscientists need to switch gears to understand how brains make choices?

Last month, the announcement of International Brain Laboratory (IBL) made headlines because of its unusual approach to a fundamental mystery of neuroscience: what happens in the brain when it makes a decision? Associate Professor Anne Churchland, who co-founded the IBL along with Professor Tony Zador, explains how it could help solve a problem in neuroscience.

By Anne Churchland

brain geersCredit: Piyushgiri Revagar/Flickr

Decisions span a vast range of complexity, from really simple ones, like whether I want an apple or a piece of cake with my lunch, going up to truly complex decisions like which car to buy or career to choose. Neuroscientists have found some of the individual brains areas, the individual players, that are relevant to decision-making. But understanding how these players work together as a team is still a challenge, not only in understanding decision-making, but for the whole field of neuroscience.

Part of the reason is that until now, neuroscience has operated in a traditional model, in which individual labs work on their own and usually focus on one or a few brain areas. That makes it very difficult for us to interpret data, even collected by another lab, because we all have slight differences in how we run experiments. So, when we get a different result, we are tempted to say, “Well, is it because we were observing a different behavior, or it was a methodological difference?” We really can’t interpret it at all.

I think the field is starting to move in a more collaborative direction, and I think the BRAIN Initiative was one of the key initiatives that started to encourage that kind of collaboration. I just think it hasn’t gone far enough.

How can neuroscientists go further as a team?

I co-founded a project called the International Brain Laboratory—which is a virtual mega-laboratory, a joint effort of many labs at different institutions—to show that the proverb “alone we go fast, together we go far” holds true for neuroscience.

All of us want to understand how the brain supports decision-making, and especially decisions that combine sensory information, like what you see and hear, with internal signals, like what you “know” is the best decision to get a good reward. And we want to achieve that by bringing together a team of 20 people who will work very closely with one another. We want to jointly collect a large dataset and analyze it to understand the brain-wide interactions that support complex decision-making. The team will record activity from 5,000 to 10,000 neurons throughout the brain in animals performing exactly the same task.

With so many possible decisions, where do you start?

What we we’re hoping to do is to recreate a mouse’s natural foraging environment. In real foraging decisions, an animal is kind of navigating in the world. There are many different paths it can take. It wants to find food, because food is rewarding, so it has to use incoming sensory cues, like, “oh, I see a cricket over there!” Those would be sensory signals. And they might combine that with an internal representation of reward, like, “I know there happens to be a lot of food in this area, I remember that from yesterday, so I’ll go over there.” Or, “I know over here there was a cat last time, so I’d better avoid that area.”

mouse foragingImagining the world from a mouse’s perspective is essential for International Brain Laboratory scientists. This whimsical rendering by Elena Nikanorovna shows a mouse making a foraging decision in a natural environment.

In these real-world foraging decisions, you’re combining lots of different pieces of information—your sensory signals, your internal knowledge about what’s rewarding, what’s risky—but implementing that in a laboratory context is pretty hard. So, in an attempt to recreate that, we have animals that are making two-choice decisions: they either lick a spout on the left or on the right side of our test apparatus, and their decisions are guided both by sensory information—by a visual stimulus that they see—and also by their internal knowledge about which of those two options is more rewarding, and whether they’re becoming more or less rewarding as time passes.

What’s innovative about everyone studying the same simple decision?

We really had to think about the tradeoff between having a behavior that was complex enough to give us insight into interesting neural calculations, and one that was simple enough that it could be implemented in the same way in about 12 different experimental laboratories. The balance that we came up with was a decision-making task that starts simple and becomes more and more complex as an individual animal achieves different stages of training.

Even in the simplest, very earliest stage we’re looking at, where the animals are just making voluntary movements, they’re making a decision of when to make a movement to harvest a reward. I’m sure we can go much further, but even if that’s as far as we get, having electrical recordings of brain activity during a simple behavior like that will be very interesting. We don’t know how it happens in the brain that you decide when to take a particular action and how to execute that action. Having neural measurements from all over the brain of what happened just before the animal spontaneously decided to go and get reward would be huge.