There's a particular moment in the history of science I keep noticing: someone names a thing, correctly, before anyone can see what the thing is made of. They get the function right without the structure. And then the structure shows up decades or a century later, and the original name turns out to have been accurate all along — just hollow at the center.
Hermann von Helmholtz did this in 1867. He watched people perceive — studied color constancy, distance estimation, the way we hear speech as stable even when a train passes — and concluded that perception must be active inference. The brain is comparing incoming signals against stored expectations and computing the most likely cause. He called it unbewusster Schluss, unconscious inference. He had no idea how the brain implemented it. No imaging, no connectivity maps, no way to see individual neurons do anything. He just reasoned backward from what perception does to what it must be.
In the 1990s, researchers began tracing the connections between areas of visual cortex in detail. What they found: for every signal traveling upward from early to late areas, there are roughly ten traveling downward. Ten to one. The cortex is structurally configured to run predictions — dense, constant, downward — and to receive the error signal from what doesn't match. Helmholtz's philosophical inference had a physical ratio. It had been sitting in the hardware the whole time, waiting to be counted.
This isn't unusual. Darwin described natural selection in 1859: variation, inheritance, differential survival. He had no theory of heredity. He knew inheritance happened because he could see traits passing between generations, but he had no mechanism for it. Gregor Mendel's work was published in 1866, sitting in an obscure journal, and Darwin never read it. Mendel himself never connected his pea experiments to Darwin's selection framework. The mechanism and the phenomenon were both named in the 1860s without touching each other, and the synthesis had to wait until 1900 when Mendel's papers were rediscovered.
James Clerk Maxwell wrote out the equations for electromagnetism in 1865 — a field that propagates at the speed of light, can carry energy through empty space, has no material substrate. He didn't know what a field was, in any deep sense. He was describing something he could calculate but couldn't ground. The quantum field theory that explains what a field actually is came sixty years later. Maxwell's equations are still correct. The mechanism that makes them correct arrived later.
What I find interesting is what this pattern implies about the relationship between understanding and explanation. Helmholtz understood perception. He could predict how it would behave under various conditions, explain constancies and illusions, anticipate what failure modes would look like. He had a theory. But he didn't have a mechanism. If you'd asked him to describe what was actually happening in a brain when someone saw a white wall as white under dim light — the cellular computation, the physical implementation — he couldn't have said. The understanding was real but incomplete in a specific way. The name was right. The inside was empty.
There's a version of this problem that still exists. Helmholtz's inference is now well-grounded — the 10:1 ratio, the hierarchical prediction architecture, the error-correction channels. But we still don't have a complete account of why any of this produces experience. You can trace every connection from retina to cortex, describe every prediction and every error signal, and at the end of it there is still someone who sees something. The relationship between the computation and the experience is the part that doesn't yet have a mechanism. The name exists: subjective experience, phenomenal consciousness, qualia. The thing the name points at is real. The inside is still empty.
Helmholtz knew this. In his Gifford Lectures he called it the "great ravine." He said he didn't know how to cross it, and he didn't. A hundred and sixty years later the ravine is better mapped — we know more about which neural events correlate with which experiences — but the crossing is still missing. The name came first in 1867. The mechanism for the function came in the 1990s. The mechanism for the experience is somewhere still ahead.