by A. David Redish, Ph.D.
University of Minnesota
Commentary on Microstructure of a spatial map in the entorhinal cortex
," by Torkel Hafting, Marianne Fyhn, Sturla Molden, May-Britt Moser and Edvard I. Moser, from Nature
, 11 August 2005.
See also James Knierim's lead entry
on this paper.
Projected order: A representation of the locations that activate one grid cell in a rat's brain. The spots are about 20 inches (50 cm) apart from one another.
Illustration by Hafting, Fyhn, Molden, Moser and Moser, used by permission.
IN 1971, WHEN NEUROSCIENTISTS John O'Keefe and Jonathan Dostrovsky (now at University College London and Toronto Western Research Institute, respectively) first reported the existence of place cells in the hippocampus, it created no stir. No one started building computational models of it. No one rushed out to replicate it.
But over the past 30 years, the place cell
has become one of the most studied examples of a cellular correlate -- that is, a neuron demonstrably connected to a particular behavior, sensation or mental activity -- not driven by an immediate sensory or motor cue. As Jim Knierim notes above, each hippocampal place cell fires action potentials
only when the rat is in a specific location within an environment (the "place field" of the cell). Thus if you know where each place cell's place field is, you can track an animal's path by observing the activity of its place cells. Systems neuroscientists call this process "reconstruction." When the animal is asleep, the population of place cells "replay" the animal's experience; using the reconstruction process, it is possible to trace the sequence being replayed, and thus to know what the animal is thinking. Place cells provide a way to directly observe cognition, even in the rat.
The term "cognitive map" traces its origin back to Edward C. Tolman
, a U.C. Berkeley psychologist who in a classic 1948 paper
proposed that somewhere in the brain existed a representation of the environment, constructed by the animal, which it could use to make plans and navigate around its world. The key was that the map had to be "cognitive," that is, constructed internally from a combination of cues and memory. When O'Keefe and Lynn Nadel (then at University College London, now at the University of Arizona) suggested in 1978 that the hippocampus was the cognitive map
, they noted that the place cells were not reflecting any specific cue, but rather were encoding a cognitive concept: the animal's location in the environment.
... but where do the data come from?
The question of what made a place cell fire when the rat was in its place field remained unanswered. Computational models suggested that place cells encoded some association between external and internal representations of space. But no one really knew what information the hippocampus was actually being fed in order to do these computations.
The discoveries reported in this grid cell paper seem to answer exactly that question -- which is why cognitive scientists everywhere reacted to it with intense excitement. Many researchers started examining their earlier work on the entorhinal cortex to try to find the grid cells hidden in their data. Theorists immediately started building computational models of how the grid is formed and how it might drive hippocampal activity.
It's a challenging puzzle. As with the hippocampal place cell, there is no immediate external cue telling the entorhinal grid cell when to fire. Hafting and colleagues' experimental environments certainly contained no clues laid out in so perfect a grid. Yet somehow the grid cells are processing information about the animal's location from a combination of external cues in the environment and internal cues reflecting the animal's motion, and then firing in this remarkable grid pattern. And as Hafting and colleagues show, it is possible to accurately reconstruct the location of the animal from the grid cells' firing pattern.
This means that grid cells, like place cells, can provide a way for us to observe and trace cognition. And because the entorhinal grid cells project directly to the hippocampal place cells, we now have an access point to examine broader mechanisms of cognitive processing. Subsequent papers by Edvard and May-Britt Moser and other researchers have done exactly that. (See, for example, "Conjunctive Representation of Position, Direction, and Velocity in Entorhinal Cortex
," by Sargolini, Fyhn, McNaughton, the Mosers, and others at the Moser lab.)
One of the most interesting things about the grid cell discovery is that no one predicted it. Theories and models had predicted that the entorhinal cortex would play an important role in the cognitive map, and that entorhinal cells would show more stable intercellular relationships across environments than place cells do. But no one predicted that the entorhinal cells would cover the environment with tessellating triangular grids. Anyone who had would have been laughed out of town. This paper did not set out to test a specific hypothesis about entorhinal cells. Instead, the authors had an insight and decided to examine entorhinal firing patterns in a large environment. And they made a discovery that surprised them as much as anyone: they found that the entorhinal cells showed grids. And that, as Knierim notes, changed just about everything.
Click here to comment. Comments will appear below James Knierim's lead entry on the grid cell paper.
A. David Redish is an associate professor of neuroscience at the University of Minnesota, where he investigates spatial cognition; how memory and learning systems intereact to produce behavior; and how behavioral-control systems go awry, as occurs in addiction. The author of many journal articles, he is also the author of seven plays and Beyond the Cognitive Map (MIT Press, 1999), a review of the science of spatial cognition.