The Geometry of Hallucination

Here is a fact that should be stranger than it is: when people take very different psychedelics, in very different centuries and cultures, and are asked what they see behind closed eyes near the peak, they describe the same handful of shapes. Not the same visions — the elaborate content is endlessly personal — but the same underlying geometry. Honeycombs and lattices. Cobwebs. Tunnels and funnels. Spirals winding into the distance. That this universal vocabulary exists at all is a clue, and following the clue leads to one of the most elegant explanations in all of neuroscience: these forms are a picture of the visual brain’s own wiring, projected onto the darkness.

Klüver’s four constants

The story begins with a German-American psychologist named Heinrich Klüver, who in the 1920s did what serious scientists of the era occasionally did: he took the drug himself. Chewing peyote and cataloguing what he saw, Klüver noticed that beneath the shifting content, the elementary geometry kept repeating — not just for him, but across the reports of others. He distilled it into four recurring categories he called form constants: (1) gratings, lattices, honeycombs and chequerboards; (2) cobwebs; (3) tunnels, funnels, cones and vessels; and (4) spirals (Klüver, 1966). His radical claim was that every more elaborate hallucination is a modification and transformation of these four primitives. If that is true, the geometry does not come from the drug or the vision. It comes from the perceiver — from the machine that does the seeing.

The map from eye to cortex

To understand why, you have to know one deep fact about the brain’s visual wiring. The image on your retina is re-drawn onto the primary visual cortex (V1) at the back of the head — but not faithfully. The center of your gaze, the fovea, is given a huge share of cortical real estate, while the periphery is crushed into a sliver. In 1977 Eric Schwartz captured this distortion mathematically: the retina-to-cortex map is approximately complex-logarithmic (Schwartz, 1977). The consequence is the hinge of the entire theory, and it is astonishing: under a logarithmic map, concentric circles, radiating rays, and spirals in the visual field all become families of simple, parallel straight stripes on the flattened cortex. A ring becomes a set of vertical bars; a starburst becomes horizontal bars; a spiral becomes diagonal bars. Run it the other way, and one boring thing — parallel stripes of cortical activity — unfolds in perception into tunnels, funnels, and spirals depending only on which way the stripes tilt.

The math that closed the loop

In 1979, Bard Ermentrout and Jack Cowan put the two halves together. They modeled V1 as a sheet of excitatory and inhibitory neurons wired to one another, and asked what happens when the balance between excitation and inhibition is disturbed — say, by a drug that changes the cortex’s gain. The answer is that the flat, quiet sheet becomes unstable and spontaneously breaks into a periodic pattern of activity: stripes, hexagons, rolls. This is a self-organizing instability, the same mathematics Alan Turing described in 1952 for how a featureless embryo spontaneously grows spots and stripes (Turing, 1952). Feed those spontaneous cortical stripes backward through Schwartz’s logarithmic map, and they bloom in the visual field as precisely Klüver’s tunnels and spirals (Ermentrout & Cowan, 1979). Two decades later, Paul Bressloff, Cowan, Martin Golubitsky and colleagues extended the model to include the cortex’s orientation columns — the machinery that detects edges and angles — and out came the remaining form constants too: the contoured lattices, honeycombs and cobwebs (Bressloff et al., 2001). All four of Klüver’s classes, generated from cortical wiring alone.

Turing patterns, in a brain

It is worth pausing on how deep this goes. The engine underneath the whole picture — a uniform system spontaneously breaking into a regular pattern with a characteristic spacing — is one of nature’s most universal tricks. The same reaction-diffusion mathematics paints the spots on a leopard, the stripes on a zebra, the ripples in windblown sand, the spacing of desert shrubs (Kondo & Miura, 2010). In the visual cortex the “ingredients” are not diffusing chemicals but spreading neural activity — short-range excitation and longer-range inhibition — and the intrinsic wavelength of that instability sets the scale of the pattern, which is why the honeycomb has the cell size it has. The hallucinated lattice is a Turing pattern rendered in living tissue. You are watching the same law that decorates animals decorate your own field of view.

Where the drug comes in

So what does the psychedelic actually do? It supplies the push that tips the cortex over the edge. Classic psychedelics are agonists at the serotonin 5-HT2A receptor, which is densely expressed on the large excitatory pyramidal neurons of the deep cortical layers, including in visual areas (Weber & Andrade, 2010). Switching those receptors on increases the excitability and gain of the cortex and shifts the delicate balance between excitation and inhibition — exactly the kind of nudge that, in the Ermentrout-Cowan framework, drives V1 past its stability threshold so its intrinsic patterns emerge as visible geometry. There is even direct human imaging: under LSD, with eyes closed, activity in the early visual cortex becomes more governed by its own internal retinotopic structure — patches of V1 and V3 that represent the same location start communicating as if they were seeing, when there is nothing there to see (Roseman et al., 2016). The visual brain, unmoored from the eyes, falls back on the only thing it has left: its own wiring. (One honest caveat: each link in this chain — receptor, gain, instability, retinotopic map — is well established, but the full causal loop from 5-HT2A to a specific spiral has not yet been closed in a single experiment. It is a superbly motivated hypothesis, not a settled proof.)

The brain, seeing itself

The most satisfying evidence is that the forms are universal across triggers. The same spirals and lattices show up under mescaline, LSD, psilocybin and DMT; in near-death experiences; in the shimmering fortification zigzags of migraine aura, now understood as a wave of altered activity sweeping literally across the visual cortex (Hadjikhani et al., 2001); from stroboscopic flicker; at the edge of sleep; and from simply pressing on your closed eyes. The trigger barely matters, because the trigger only supplies energy. The form is supplied by the architecture — the logarithmic map, the orientation columns, the intrinsic wavelength of the cortical sheet — and any perturbation that destabilizes the excitation-inhibition balance will express the same small family of patterns. That is the quiet revelation at the center of all this. A geometric hallucination is not a picture of anything external. It is the visual cortex rendering its own structure into experience — the wavelength of its circuits, the lattice of its columns, the logarithm of its map. When you close your eyes and watch the spiral turn, you are not seeing the world. You are, for a few luminous minutes, seeing the shape of your own seeing.

OOTW Journal is educational and does not provide medical advice. Psychedelics remain controlled substances in most countries and are not safe for everyone, including people with a personal or family history of psychosis or bipolar disorder, and those taking serotonergic medications. Nothing here is a recommendation to use any psychedelic. If you are considering this work, please consult a qualified professional.