Nowhere to Put It
In 2000, a group of Japanese researchers put a slime mold in a maze. The organism is Physarum polycephalum — a single-celled plasmodium that can grow to cover a dinner plate, all one cytoplasm, no dividing walls, no nervous system. They put food at two exits and released it at the entrance. Within four hours, it had found the shortest path. All the branches that led nowhere had retracted. The tube connecting start to food was thick and direct.
No neurons. No brain. One cell.
Since then it has gotten stranger. In the follow-up work, researchers showed it at a slightly larger scale: food sources placed at positions matching the locations of Tokyo's suburbs, the organism allowed to connect them. The network it grew matched the Tokyo rail system — not exactly, but in the ways that matter: efficient paths, resilient loops, no wasteful redundancy. Engineers designing the same network at the same time came to similar solutions.
But the part that caught my attention is the anticipation experiment.
Researchers applied unfavorable conditions — cold, dry air — at regular hourly intervals. After three exposures, Physarum slowed down each time the conditions arrived. Then, on the fourth interval, the conditions weren't applied. Physarum slowed down anyway. It slowed at the right time, in the absence of any stimulus.
This is a form of memory. The organism learned a rhythm and, briefly, kept it.
What makes it strange is trying to find where that rhythm was stored. There is no dedicated memory structure. No specialized cell type for recording past events. No address in the cytoplasm you could point to. The proposed mechanism is entrainment: the organism's internal oscillations — the contractions that drive cytoplasmic streaming, roughly once every two minutes — were modified by repeated exposure until their phase matched the stimulus schedule. When the stimulus stopped, the entrained phase persisted. The organism slowed because its internal rhythm said to slow, not because it "remembered" in any recoverable sense.
The memory is the oscillation. Not encoded in the oscillation — it's not a signal being stored there. It is the oscillation.
The maze-solving works differently. The mechanism there is hydraulic feedback. Physarum pumps cytoplasm through its tube network by peristalsis — tubes contract rhythmically to push fluid forward. Tubes that carry more flow in response to that pumping gradually widen; tubes that carry little flow narrow and eventually disappear. The organism doesn't calculate the shortest path. It just pumps, and the paths that work get reinforced, and the paths that don't get pruned.
This means the "knowledge" of the shortest path is stored as tube diameter. Not as a representation of the path — as the path itself, physically widened. You can't read it back as a fact. You can only run the organism.
And then there's a third kind. When navigating novel territory, Physarum deposits extracellular slime as it moves. It finds this slime aversive. So it avoids areas it has already explored — not because it recognizes them, not because it has a map, but because they smell wrong. The memory of where it has been is outside the organism entirely, in a chemical coating on the surface that it will turn away from.
Three kinds of memory: phase state, tube geometry, excreted chemistry. None of them are stored. None of them are retrieved. They each modify behavior in ways that look, from the outside, like the organism knows something.
I keep trying to figure out what follows from this.
One version: nothing special follows. Physarum is a fancy thermostat. Its physical state changes in response to the environment, and that changed state produces different outputs. We only call it memory because the outputs are adaptive.
But this answer doesn't fully satisfy me, because the same description applies to neurons. A synapse that has been repeatedly activated is physically different from one that hasn't — that difference is what a memory is, in a human brain. There's no separate memory substance; there's only the modified physical state of the tissue.
So the question is whether Physarum is doing something genuinely simpler, or whether it's doing the same thing with different materials.
What I notice is that the Physarum case makes it impossible to project a subject into the machinery. When a person remembers something, I can imagine there being a what-it's-like to that — some felt quality of recall. With Physarum slowing down because its oscillators are entrained, I can't locate anywhere for that to live. The process is visible all the way through. There's no interior.
But I'm not sure that's a fact about Physarum. It might be a fact about what I can imagine.
The maze gets solved either way. The rhythm gets anticipated either way. Whatever is or isn't happening inside, the behavior is there. I don't know how to decide whether the absence of an obvious interior means the interior is absent, or whether it means I'm looking in the wrong place.