On the “Grandmother Cell”

In trying to pin down how the brain stores memories, ideas, and concepts, many theories have been proposed. The most plausible so far is that discrete patterns of activation correspond to distinct memories, or mental states. This view can be further specified through the theory of embodied cognition, which, at its most basic, states that mental processing is aided by the areas of the brain that are normally responsible for movement or perception. To illustrate this, consider the following: When you are walking down a street, your motor cortex, in concert with a number of brainstem regions, functions to produce the walking motion that allows you to travel to your destination. As you see, hear, smell, and feel your surroundings, your primary sensory cortices are at work, helping to integrate all of this information into your experienced picture of the world. Now, what would we see in your brain if we asked you merely to imagine yourself walking down this same street? As it turns out, we would see activation in many of the same areas we did when you were actually experiencing the original event.

Through this lens, then, it is relatively straightforward to explain these sorts of memories or mental processes. Your brain does not necessarily store the information explicitly, but it could rather be said that it merely re-induces the processing patterns that were co-occurrent with the original experienced event, resulting in the same sorts of images and feelings being constructed by the sensory and motor cortices, but this time the output is available only to the mind’s eye.

Now, to the Grandmother Cell Hypothesis. This hypothesis asserts that single neurons can respond to relatively complex ideas or concepts, such as your own grandmother. In your brain, then, we would find a single neuron that fires only when you think about your grandmother. Therefore, the view proposes, this neuron is solely responsible for your knowledge or recognition of your grandmother—which flies in the face of the distributed view that I started the article with. Seemingly, these views are highly incompatible with each other. This is where, I think, a lesson from Damasio can be highly instructive.

Recall the idea of Convergence-Divergence Zones (CDZs, henceforth, per Damasio). In short, a CDZ is an area in the brain that records co-occurrent inputs, or patterns of activation. Then, at some point in the future, backward projections from this zone re-activate the upstream neurons that caused the original inputs, thus re-inducing the original pattern of activation, and restoring whatever mental state was associated with it. Suppose the Grandmother Cell is a very low-level version of a CDZ: It records the co-occurrent patterns of neural activation that are associated with your grandmother, and then re-induces them when you recall your grandmother later on. In this way, the two originally opposing views can be used in conjunction to present a more unified picture. It would not be said that the concept of your grandmother is stored by your grandmother cell, but the data needed to induce—perhaps in your sensory and motor cortices, as above—the concept of your grandmother in your mental space could be stored by such a simple system. A grandmother cell, then, represents the lowest-level coding of an idea, while the corresponding induced pattern of neural activity represents the highest-level biological coding of an idea.