Activity associated with choice showed up in cortical areas, in line with previous findings, but it also occurred in subcortical areas such as the hindbrain and cerebellum, challenging the notion that only a few select areas encode information about decision-making and supporting the idea that it is widespread.
“It is interesting how much choice selectivity is everywhere,” says Long Ding, research associate professor of neuroscience at the University of Pennsylvania, who was not involved in the work.
Movement- and feedback-related signals also pervaded across the brain: 81 percent of recorded brain regions contained information that could predict the animal’s wheel speed, and activity from nearly all recorded brain regions—including those beyond the associated reward areas—accurately predicted whether the mouse had received a reward or not, with stronger activity in the thalamus, the midbrain and the hindbrain.
If the mice saw the circle more often on one side of the screen than the other, they eventually integrated that prior information into their next decision. This information was represented broadly across 20 to 30 percent of the brain, including in sensory processing areas, such as the dorsal lateral geniculate, that are located early in the visual pathway, the team reported in the second study.
The findings contradict the long-standing idea that prior information is integrated into the process only in higher-order cortical or decision-making regions “at the very last step,” Churchland says. Instead, priors shape decisions all along, the new findings suggest.
A
ltogether, the studies suggest that the current model of decision-making and the brain regions that control it might be limited in scope and that other, unexplored brain areas might also be important, Churchland says.
And although the analyses show that a distributed network of brain regions contains information about decision-making even at early stages of sensory processing, the results do not show causality, so future studies need to determine how that information is used, Ding says. “Yes, [the information is] reflected everywhere, but where is it actually used for the next decision, for learning?”
The comprehensive map sets the stage for those next experiments and could even act as a “library” to help neuroscientists double-check results in their own labs, de Lange says, and ultimately, these studies underscore the importance of large-scale, multilab efforts, particularly for studying brain activity.
The global consortium has since expanded to include 21 experimental and theoretical neuroscience labs and has established a new group called IBL 2.0 that plans to share the tools and expertise it has amassed with new partners, Churchland says. “I hope that our work makes clear that when larger groups of folks team up, they can accomplish things that are beyond the scale of a single laboratory and that really generate critical insights for the field.”