Every minute of the day, our brains are bombarded with enormous amounts of information. Lights, sounds, smells and tastes of all sorts are rapidly coming into our brains, often at the same time. Yet somehow we are able to effortlessly process our complex environments and accomplish our goals. How does this happen? To get around this information overload, the brain selectively processes a subset of the incoming information that it deems important. This clever trick is what we call selective attention.
In their 2017 study, Leong et al. explored the ways that attention influences learning and decision-making . To do so, they invented a game for participants to play during a brain scan with fMRI. In this game, three objects are presented at a time, each with three pictures: a famous face, a common tool, and a landmark. In each play of the game, one picture was associated with a high probability of reward (75%) while the other eight pictures were associated with a low probability (25%). The trick was that participants did not know which picture would be the most rewarding in each play, so they had to learn which objects to choose each time by trial-and-error. In this way, the authors could study how attention helps people quickly learn what pictures are rewarding, and how attention influences each choice that a person makes.
How could the researchers know what pictures the participants were attending to? Luckily, previous research has given us a pretty good handle on which areas of the brain care most about faces, tools, and places. The amount of activity in these face, tool, and place areas was used as a proxy for participants' attention to each type of picture. Additionally, eye-tracking measurements revealed which pictures the participants looked at throughout the game.
The authors predicted four possible ways that attention might affect behavior in this task: either attention affected only choices, only learning, neither, or both. The authors compared their data with four mathematical models that described each of these possibilities to see which one best explained the participants' behavior. The final model won out, suggesting that attention biases both choice and learning. Leong et al. (2017) also found two brain areas whose activity correlated with the effects of attention on choice and learning separately.
Knowing that attention is critical for both learning and decision-making is an exciting advance for the field, but it doesn't explain how brains are able to "switch" attention among features in the environment. To learn what's important in real life, our brains need to realize when we're paying attention to the wrong thing and switch focus. For example, suppose you're a participant in the experiment, and you're exploring a game in search of the most rewarding picture. At first you may focus on the picture of Abraham Lincoln, but you may switch your attention to the picture of the wrench when you start to realize that the wrench yields reward more often than Lincoln does. What happens in the brain during this switch of attention?
When the researchers compared brain activity between "switch" and "stay" trials defined by the participants' attention measures, they found a whole set of brain regions that were more active on switch than stay trials. Remarkably, these particular areas are part of an "attention control" network that behaves like a telephone operator switching between calls.
One shortcoming of this paper is that the brain activity seen in the fMRI scans is correlated with behavior but does not provide evidence of a causal relationship. The areas shown to be active during attentional switches, for example, may not be necessary or sufficient for these switches. Nevertheless, the results of this study lay the groundwork for future explorations into the underlying mechanisms of attention's influence on learning and decision-making and the role of the attention control network in shifting focus toward rewarding information.
 Leong, Yuan Chang, et al. "Dynamic interaction between reinforcement learning and attention in multidimensional environments." Neuron 93.2 (2017): 451-463.
Splash image courtesy: