I am an undergraduate studying pure mathematics taking a class on computational neuroscience. My default lens for looking at the brain is in terms of universal computation (in the Turing machine sense) and algorithms for this being implemented in a neural substrate (the cortex). In this class, and in the literature in general, we seem to assume that neurons are "enough" for the brain to compute everything that it does. That is to say, in focusing on studying neural networks as a way to understand brain function, we are implicitly saying that no other neural structures are central to neural computation (ie. memory, association). For instance, we are presuming that eg. astrocytes, or even the chemical composition of extra-neuronal brain tissue, do not play a role in neural computation. Why is this justified?

I understand that we know neural networks can compute any function, and so in theory they are ``enough", but how do we know that there is not some physiological/biochemical mechanism involved in other types of cells that is critical to actual neural computation in the brain? How confident are we that neurons are the only important cells to study in terms of understanding human cognition, for instance?

  • $\begingroup$ Not an easy question. I point you to this engaging seminar from Josh Tenenbaum titled "What kind of computation is cognition": youtube.com/watch?v=NsID1iM8gRw Hopefully this helps. $\endgroup$
    – JFR
    May 4, 2022 at 10:00

1 Answer 1


For what it's worth, I don't think the biologists are making this assumption; the biology clearly shows it is not correct to exclude, say, astrocytes from "computation" in the broadest sense. Here are just some non-comprehensive examples when I searched "astrocyte computation":

Volman, V., Ben-Jacob, E., & Levine, H. (2007). The astrocyte as a gatekeeper of synaptic information transfer. Neural computation, 19(2), 303-326.

De Pittà, M., Volman, V., Berry, H., & Ben-Jacob, E. (2011). A tale of two stories: astrocyte regulation of synaptic depression and facilitation. PLoS computational biology, 7(12), e1002293.

Patrushev, I., Gavrilov, N., Turlapov, V., & Semyanov, A. (2013). Subcellular location of astrocytic calcium stores favors extrasynaptic neuron–astrocyte communication. Cell calcium, 54(5), 343-349.

Mederos, S., & Perea, G. (2019). GABAergic‐astrocyte signaling: a refinement of inhibitory brain networks. Glia, 67(10), 1842-1851.

For a reduced model, though? Consider George Box:

“All models are wrong, but some are useful”

It may be reasonable to exclude glia from a model of neural function where their role is abstracted out. Glia tend to be slow relative to neurons, so if you're talking about the specific moments leading up to a specific action potential, they can be reasonably excluded. If you're abstracting away all the other biology that happens at a synapse: metabolism of neurotransmitters, stochasticity of different parts of release machinery, spatial distributions of contacts between cells, etc, excluding glia seems to be pretty far down the chain of things to worry about.

Be wary a bit of: https://xkcd.com/793/ - I think there is sometimes an assumption among people approaching biology from a more quantitative/algorithmic direction to think that the simplifications they encounter are representative of the whole world, and they forget that biologists are focused just as much on the details.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.