One of the keys to modern thinking about choices and values is a small part of the brain called the nucleus accumbens, which is part of the ventral striatum, itself part of the basal ganglia. It’s sometimes called the reward center of the brain, and neuroeconomists generally believe the nucleus accumbens is responsible for recognizing and processing the rewards and punishments that follow from our actions.
Like much of what you read about neuroscience these days, that’s only partly right. Nucleus accumbens isn’t just a reward processor, according to a recent study. It’s more like a coincidence processor.
When a reward follows an action, that action gets reinforced, and we’re more likely to take that action in the future.
Interest in nucleus accumbens, or NAc, grew in the 1990s, when monkey studies suggested that getting a sip of juice as a reward for a correct response to a problem caused dopamine neurons in and around monkeys’ NAc to fire. FMRI studies showed something similar happening in our brains, too, leading theorists to suggest that NAc was doing some kind of reinforcement learning: When a reward follows an action, that action gets reinforced, and we’re more likely to take that action in the future. But humans and other animals can learn all kinds of associations—things drop when we let go, October is crunch-time in baseball, and the better your Skee Ball score, the more prize tickets you get. Even dogs can learn food’s coming when a bell rings. Curiously, there are few studies that look at whether NAc might be recording all of these associations, not just the action-reward ones.
To press the question, Dominic Cerri, Michael Saddoris, and Regina Carelli conducted a standard experiment. First, they taught 20 rats a variety of second-order stimulus associations. For example, a rat might first learn that white noise followed right after a light flashed, and in a second session, they’d learn that a food pellet would be available following white noise. Finally, the team tested the rats—if they went looking for food after a flashing light, but not other signals, they’d learned the light-noise-food pattern. All the while, the researchers monitored NAc activity using electric probes implanted in the rats’ brains.*
The team found that NAc neurons in the rats fired not only in response to the food pellets in the second session, but also during the first training session when there were no rewards of any kind. Next, the team divided the animals into groups of good and poor learners based on how well they’d performed during the test phase, a process by which they found that good learners’ NAc neurons fired more during learning than either poor learners’ brains or those of a control group. In other words, NAc does more than just encode rewards—it tracks other sorts of connections in the world, too. Though questions remain, that insight might throw a small but intriguing wrench into our understanding of how rat—and maybe human—choices work.*
*UPDATE — October 14, 2014: We originally wrote that the experiment was conducted with mice. That language, in both the body of the post and subheadline, has been corrected to rats.
