Michael Ragone

What are place cells?

Imagine you enter a room floored with large tiles. On the edge of the room is a screen filled with static—it buzzes, thumps, chirps, and occasionally melodizes. You walk in and notice that whenever you stand on a tile, a green dot shows up on the upper-left corner of the screen and you hear a drum beat. This particular drummer is either very avant garde or very new to his craft—he plays an erratic rhythm, one that’s almost regular, but not quite. You take your foot off the tile. The beat fades and the dot dissolves into the static.

You step on another tile: now another dot brightens the bottom of the display, again accompanied by drummer of uncertain skill. Already noticing a pattern (you clever human, you), you quickly guess that every tile in the room has a corresponding screen dot. You adjust your foot—huh, the drummer speeds up the tempo. You skooch back…he slows down. Ah ha! The drummer plays faster when you stand near the center, and slows to a stop as you approach the edges of the tile.

Place cells are neurons in your brain that represent space.¹ Each place cell has, by definition, a corresponding place field, which is simply a region of space in which the cell tends to fire—in the analogy above, the tiles fill the role of place fields. Since place cells are neurons, they’re inherently electrical, meaning we can listen to their patterns of spiking—their drumming—by recording from electrodes positioned in regions known to have place cells. We often hunt for place cells in the CA1 subregion of the hippocampus, a region known to be involved in spatial navigation and memory.

Recording place cells

So—we have a rat, and we have some electrodes in CA1. These electrodes just pick up any electrical signals around them, so we have to do some sifting to identify cells we’re interested in. Like the screen in the analogy, it’s a noisy process—kind of like listening to someone in an overcrowded room. You have to use some clever listening to hear the right conversations, like sorting by volume (signal strength) and vocal timbre (spike waveform). This process of identifying neurons in a recording is known as spike sorting, and it’s a fascinating and vibrant area of research, with rather huge implications for neural science—it can easily affect the outcome of any recording experiment, and since verifying the success of sorting is highly nontrivial, mistakes in sorting are very hard to detect. It’s massively important, and pretty cool. Just saying.

Back to our problem: we have our rat and some electrodes. How do we find these pesky place cells? It’s actually quite easy once this part is set up: just let the rat walk around, and listen to neurons along the way. Start to record where each neuron is firing (drummer is drumming), and at what rate the each neuron is firing (tempo). Then, after a few minutes of wandering, you’ll have a nice map of where each cell is firing—some will fire selectively to regions of space, others won’t. So great—now that we’ve found place cells, we completely understand rat spatial navigation…right?

A murkier story

Well…there’s a lot more to place cells than that. They’re far more finicky than they seem at first. For instance, place cells are known to occasionally remap, a phenomenon wherein following some significant change in context—e.g. adding/removing an object—place cells’ fields reposition, expand/contract, or even vanish. Weird.

But there’s more. When I say “place cell”, it’s a functional description of a neuron—neurons are called place cells only if they have a corresponding place field. A place cell isn’t a cell type (most place cells are hippocampal pyramidal cells), and indeed there’s evidence that place cells can take on different roles, spatial and nonspatial. For example, these same cells seem to be capable of encoding:

• Odor sequences
• Sound spaces—the fascinating work of Dr. Dmitriy Aronov at Columbia demonstrates that these cells are capable of establishing fields in “sound space”, i.e. cells that fire selectively to ranges of the frequency of a tone.
• Time—an analogue for place cells for the time dimension, such “time cells” would have fields corresponding to periods of time. The coordinates of these fields would be anchored by the start of some experience, like a rat starting a maze.
  • It’s worth noting that the interpretation of the experiments of this paper are hotly debated, with many arguing that these cells aren’t truly encoding time. The theorist in me strongly suspects that the brain has time cells with features like these, for the sake of encoding coordinates in time-space for episodic memory, but I think we need more evidence for the time being. But I digress.

The TL;DR is that even though place cells are clearly can represent spatial information, they’re probably involved in more processing than they let on. Experiments catching “place cells” doing things other than encoding space lead some neuroscientists to suggest that these would be better dubbed “experience” or “sequence” cells.

Why theorists love place cells

While there’s interesting theoretical work on many areas of the brain, place cells (alongside visual cells) seem to garner the attention of a disproportionate number of theorists. With all of this confusion regarding the true nature and purpose of place cells, why do theorists spend so much time on them? While there’s certainly several reasons (wealth of experimental data, well-researched properties), I suspect that it’s because place cells offer something that very few neurons in the cortex offer: readily interpretable information at a cellular level. Often the neurons we study are clearly involved in something, but the significance and information content of individual neurons can be difficult or impossible to tease out, so we are forced to make statements about large groups of neurons and their statistics. Place cells freely display at least a significant subset of the information they are storing—it’s clear that their firing rates are directly connected to locations in space in some way. So even if we don’t have the complete story, place cells still offer a unique look at information processing in individual neurons.

And besides, place cells are cool.

* Coffee: a cortado (espresso with a little steamed milk) and a cold brew soda at the wonderful James Coffee Co. in San Diego, California

^{1. While place cells have been observed in humans, a vast majority of place cell experimentation has been in rats and mice, so we’ll just talk about that. Extrapolation in science always needs justification, and while some properties of place cells likely show up in both cases, research in human subjects is sparse and we can’t just assume that all properties carry over.}