What the Letterbox Was Before
There's a region in the left side of the human brain, in the folds of the visual cortex, that lights up when you read. Neuroscientists call it the visual word form area. Stanislas Dehaene calls it the letterbox. What it does is recognizable: when you encounter a written word, this region identifies it almost instantly, before meaning kicks in, before sound, before anything conscious. It's a fast pattern-match to something it's seen before.
What's less obvious is what this region was doing before writing existed.
Writing is roughly five thousand years old. The human visual system is tens of millions of years old. So the letterbox didn't evolve for reading — nothing did. When humans invented written language, they weren't building with blank cortex. They were borrowing machinery that was already there and already doing something else.
What it was doing, best anyone can tell, was processing line junctions. Specifically: the places where contours meet — a T shape where a branch crosses another branch, a Y at a cube corner, an L where two walls meet, an X where paths cross. These shapes are everywhere in natural scenes. The visual system, shaped by millions of years of navigating a world full of objects, became very good at detecting and combining them. They're the building blocks from which the brain infers three-dimensional structure from flat images.
Here's where it gets strange. A researcher named Mark Changizi analyzed the shapes of letters and symbols across 115 different writing systems — including systems with no historical connection to each other — and found that they all converge on approximately the same set of contour configurations, in approximately the same proportions. The distribution of junction types in written scripts matches the distribution of junction types in photographs of natural scenes.
This isn't a coincidence. It's a constraint.
When writing was being invented — across Mesopotamia, China, Mesoamerica, independently and at different times — the symbols that worked, that spread and persisted and were legible from a distance and easy to distinguish and fast to process, turned out to be the symbols whose shapes matched what the visual system was already prepared to recognize. The letterbox could learn them quickly because it had been detecting similar shapes for as long as humans had eyes.
Dehaene calls this neuronal recycling: cultural inventions don't arrive on blank cortex. They take over regions that were already doing something related, and they inherit the properties of what was there before. The letterbox was a junction-detector before it was a reading machine. It became a reading machine because writing, over thousands of years of cultural selection, converged toward junction patterns the region was already sensitive to.
The direction of influence is usually described as: culture shapes the brain. Learn to read and your visual cortex reorganizes. True. But the other direction is also true: the brain's existing shape constrained what culture could become. Writing looks the way it does partly because of the contours of the visual cortex.
What I find strange about this is what it implies about the experience of reading. When you read a sentence, there's no phenomenological trace of contour-junction detection. It doesn't feel like borrowed machinery. Reading feels like reading — like pure linguistic meaning arriving cleanly. The underlying process, whatever the fusiform gyrus is actually computing, is invisible from the inside.
Dehaene's recycling hypothesis has critics. The VWFA is real — it reliably activates for written words more than for other visual stimuli — but whether it's truly word-specific or just one part of a more general expertise network is actively debated. Some researchers think the selectivity comes from learning any complex visual category, not from something special about words. Others point out that the region is modulated by attention and task demands in ways that a dedicated letterbox probably shouldn't be. Recent work suggests it has more connections to the language system than to other parts of visual cortex, which would be odd for a purely visual area.
So the picture is real but not clean. There is a region. It does preferentially respond to words. Its responses are shaped by reading experience. But the specific claim — that reading works because writing converged toward shapes this region was already tuned for — rests on several steps of inference, each of which is plausible but not certain.
What I keep coming back to: if it's right, then the shapes of every alphabet and every writing system carry a kind of fossil record. Not of their own history, but of the history of the visual system that had to learn them. The letterbox is named for what it became. The shapes it learned to read were selected, over thousands of years, for how well they fit the thing it was before.