A hungry rabbit sits in a cage with two pedals and two feeders. Step on the left pedal and one feeder turns, with carrots in it. Step on the right and the other turns. While the rabbit learns this, electrodes listen to its brain cells.
Some cells fire only for the left pedal. Some only for the right. And some fire for either one. They answer to an abstract category, "pedal", that in Konstantin Anokhin's words never existed in a rabbit's nature. Anokhin is a neurobiologist, an academician and director of the Institute for Advanced Brain Studies at Moscow State University. He told the story on Think Like a Scientist as an experiment from the lab of his teacher, Vyacheslav Shurkov. The category appeared in front of the experimenters' eyes, during training.
Anokhin's bet is that your brain does the same thing all the time, with everything. Chairs, tables and trees are not in the physical world as things. From the point of view of physics they are not invariants, because nothing in the physics stays the same from one chair to the next. The sameness is made inside the nervous system.
How? A single neuron gets drawn into one activity of the body after another. In childhood it works for play and learning, in youth for something else, in adulthood for something else again. Anokhin's minimum is a handful of such systems over a life, and in reality it is thousands. Each time, a small burst of long-lasting change stores how the cell was wired into that task. After enough rounds the cell has left behind whatever was particular about the jobs, and what remains is the thing they shared. It becomes a cell that sees red. A fire engine, a rose, a hat, a watermelon: it doesn't care which one carries the colour.
We begin to tell apart objects, things, chairs, tables, trees, which do not exist in the physical world. From the point of view of physics they are not invariants.
Мы начинаем различать объекты, предметы, стулья, столы, деревья, которые не существуют в физическом мире. С точки зрения физики они не инварианты.
— Konstantin Anokhin, Think Like a Scientist, 74:16
The deeper in the nervous system, the more abstract the cells' job. Shallow ones track a level of light, a stripe, a line, a colour. Deep ones answer to configurations of those pieces. At the far end sits a particular person. It doesn't matter what they are wearing or where they stand. It doesn't even matter whether you see their face or hear their voice on the phone. The inputs are completely different, and the same deep cells light up.
This also sets a limit on how far these cells reach. The cells that form such elements are local. They are groups with similar inputs and outputs, sitting in one anatomical region: association areas of the cortex, or the limbic system, tied to particular emotions.
Recordings from human brains show something close to this picture. A June 2024 bioRxiv preprint from D. E. Dijksterhuis reports that cells selective for a particular noun are reactivated later by pronouns that refer to that noun. The word "it" wakes the cell that was tuned to the thing. A October 2025 preprint by Katharina Karkowski adds a wrinkle: human concept cells show graded responses that follow semantic tuning curves. They are not simple on-off switches.
Notice what this does to the rabbit. In its million years of inherited experience there was no pedal. After a few sessions in a cage, there is a cell for it, a thing the world did not contain until the rabbit's own history made it.
And you, reading this, are sitting on something your neurons assembled out of a lifetime of thousands of small tasks. Physics never put a chair there.
