Single neurons in the human medial temporal lobe fire selectively for a specific concept (a particular person or place) regardless of how it is presented: a photograph, the written name, or the spoken name all trigger the same cell. This shows the neuron responds to the abstract identity rather than to visual details like edges, colors, or viewing angle.
During epilepsy surgery recordings, researchers found a single neuron that fired for photos of Jennifer Aniston but not for dozens of other images, with other cells tuned to other individuals like Halle Berry or landmarks like the Sydney Opera House. The finding, made in patients with implanted electrodes, became the empirical anchor for the idea that individual neurons represent individual concepts.
The brain does not store experience in rich detail; it discards most specifics and retains only abstracted concepts and gist, which is a feature rather than a flaw. People who recall everything (as in rare hypermnesia or the fictional Funes) are overwhelmed and cannot generalize or think, because meaning depends on throwing information away.
The brain does not passively capture a faithful image of the world; it builds perception from sparse sensory samples combined with prior expectations, so we see a stable scene despite taking in very little detail at any instant. Phenomena like change blindness and the eye's blind spot demonstrate that much of what we experience is filled in by inference.
Each concept is represented by a small set of neurons, and overlapping activation of these sparse groups is proposed as the substrate for linking concepts together into episodic memories and associations. When you remember two things together, the neurons for each concept become jointly active, binding them.
Rather than viewing a concept as spread diffusely across millions of neurons, this lens treats a small, identifiable handful of cells as sufficient to represent it, since experiments could reliably find a responsive neuron by sampling only a limited number. This reframes representation as sparse and partly explicit rather than wholly holographic.