Monday, July 16, 2018

Slime molds and plants

A while ago I mentioned work done suggesting that Gallistel's conjecture that cognitive computation does not require neural nets is correct. The work discussed "learning" in single cell slime molds and plants. At any rate, this stuff i going mainstream in that Quanta brings this research together in this review (republished in Wired).

The piece focuses on the controversy of whether this can actually be "primitive cognition" noting that for many cognition is only something that brains can do (by brains kogneuro types mean ensembles of neurons). The fear is that this kind of research amounts to "'devaluing' of the specialness of the brain" (12). Others comfort these kogneuro fears by claiming that the "debate is arguably not a war about science, but about words" (13).

Both claims are wrongheaded. These studies are direct challenges to the standard cogneuro paradigm that brain computation is fundamentally inter-neuronal. This is what the Gallistel-King conjecture challenges. The work on slime molds and plants indicates that what fits the behavioral definitions of learning exist in organisms without the requisite neural nets. The conclusion is that neural nets are not necessary for learning. This surely points to the possibility that the standard picture in cog-neuro concerning the centrality of neural nets to cognition needs a fundamental rethink. In fact, it would be biologically amazing if intra-neuronal/cellular cognitive computation was possible and extant in lower organisms but higher organisms didn't use this computational power at all.

Read the review. The content is not news to FoLers. But the reactions to the work and the weird attempts to either discredit, downplay or reinterpret it is fun to look at. The significant thing, IMO, is that this stuff is becoming more and more mainstream. I think we might be on the edge of a big change of mind.


  1. ...This seems really clear to me. Habituation works like this: a stimulus triggers a chemical cascade, and part of that cascade is a negative feedback loop, i.e. a process that produces an inhibitor of some earlier stage in the cascade unless certain things happen (e.g., the stimulus disappears and then reappears later, or consequences like actual damage after exposure to a noxious stimulus happen). Such an inhibitor can obviously diffuse through a plasmodium, indeed it'd be hard to imagine a way to stop it. Keep in mind that plants are actually plasmodia, even though that's less obvious than in plasmodial slime molds: the individual-looking cells are all connected through simple pores in the cell walls.

    This is how habituation works in our own neurons. What I don't understand is the assumption that all memory and all learning is habituation.

    1. Nobody said that it was. But as you can see by the reaction to the work by the less accepting, until this was demonstrated, it was assumed that habituation was an inter neural process. We now know it is not. The next question is how much more "cognitive" things these brainless systems can accomplish. There is at least some evidence that in plants it is quite a lot; including anticipating where the sun is going to be. This is fancier than habituation. So, quite right, not great yet. But then until very recently even this was questioned.