In addition to Gupta, the researchers include Carnegie Mellon Computer Science Professor David S. Touretzky and A. David Redish, associate professor of neuroscience, and Matthijs van der Meer, a post-doctoral researcher, from the University of Minnesota.
“Our work provides clues into how animals construct a complete, fully-navigable representation of their environment, even if they’ve only partially explored that environment,” said Gupta, who also is a medical student at the University of Pittsburgh School of Medicine. “The cognitive maps created in this way may allow animals to plan novel routes or shortcuts. As we learn more about the neural mechanism that enable animals to flexibly navigate through the world, we hope to apply those lessons to research in robotics that could improve autonomous navigation systems.”
The team used electrode “hats” to record brain activity of rats as they navigated a maze. In particular, they monitored certain neurons, called place cells, which fire in response to physical locations. That enabled the researchers to identify where an event that was being replayed was located based on which place cells were firing. During an experiment, a rat might be in one portion of the maze, while the firing of place cells in the hippocampus indicated that the rat was replaying information about a different location.
On a task with two behavioral sequences, A and B, the researchers found that the animals would replay sequence B more often though they spent most of their time running sequence A. In other words, the researchers found that the rats were most likely to replay the path they had experienced less often. This suggests that replay is not just a function of helping an animal remember what it has experienced most frequently or most recently, but an important function in helping it map its whole environment.
During the replay process, the research team also was able to observe the animal making connections between paths that it had never physically traveled before. For example, if the animal had physically traveled from point A to point B, and also from point A to point C, but never from point B to point C, they observed the single sequence B to A to C during the replay process, implying that the rat’s brain was able to make the connection between points B and C on its internal map. This further indicates that replay plays a role in helping an animal learn and maintain the entire map of its environment and make connections within it. The rats were not just reviewing recent experience to move it to long-term memory.
“Based on these observations, we have to rethink what is the role of replay for memory,” wrote neuroscientists Dori Derdikman and May-Britt Moser of the Norwegian University of Science and Technology in a commentary also published in the March 11 issue of Neuron. They suggested that replay in the hippocampus may prove to have a dual role — both for memory consolidation and for making cognitive maps of the environment.
In a report published March 11 in the journal Neuron, Anoopum Gupta, a Ph.D. student at Carnegie Mellon’s Robotics Institute and the Center for the Neural Basis of Cognition, and his colleagues say their findings suggest replays in the hippocampus are not merely passive echoes of past events, but part of a complex, active process of decision making. In a study of rats navigating a maze, the researchers found that replays occurring in the hippocampus were not necessarily recent or frequent paths through the maze, as would be expected if the event was being added to memory. Rather, the replays often were paths that the rats had rarely taken or, in some cases, had never taken, as if the rats were trying to build maps to help them make better navigation decisions.