Dark adaptation

Why does seeing in the dark take twenty minutes to acquire and one second of light to lose?

The usual explanation is the pupil: it opens in the dark, closes in the light, and there is your answer. Test it against the numbers. A human pupil runs from roughly two millimetres across to roughly eight — a sixteen-fold change in area — and it makes that change in a second or two. Full dark adaptation improves sensitivity by something on the order of a hundred thousand-fold, over twenty to thirty minutes. The pupil accounts for a rounding error of the effect and none at all of the timing. So something else is doing the work, and whatever it is has a violently lopsided clock.

Go down to where light is actually detected. A photoreceptor is packed with molecules of visual pigment, each a protein holding a small bent molecule derived from vitamin A. A photon arrives and straightens the bend. That single geometric change is the detection event: it deforms the protein around it, setting off the cascade that eventually says light.

Now notice what has happened to the detector. The straightened molecule no longer sits in the protein properly and no longer responds to anything. That pigment molecule is spent. To detect light is to consume the means of detecting it, and a receptor's sensitivity is, roughly, its stock of unspent pigment.

So ask what sets each rate. Spending is driven by the number of photons arriving, and there is no ceiling on that whatsoever — a bright light can bleach a large fraction of the stock in a moment, and a brighter light does it faster. Restocking is a pipeline of chemistry: the straightened molecule must be released from the protein, chemically reduced, carried out of the receptor into the layer of pigment epithelium behind it, bent back into shape by an enzyme, carried in again, and recombined with a waiting protein. Every one of those steps runs at whatever speed the enzymes and the transport allow, and that speed has a ceiling.

That, and not any brute fact about fast and slow, is the asymmetry. One process scales without limit with its input; the other cannot exceed its maximum rate no matter how dark it gets. Any stock built like that empties in the time it takes to open the tap and refills only at the pump's one speed. Nothing about the darkness is being learned, and nothing is being remembered; a supply is simply being rebuilt.

Two refinements make the twenty minutes recognisable. First, it is not one curve but two. Cone pigment recovers quickly, most of the way in the first several minutes, but cones bottom out at a fairly high threshold. Rods recover far more slowly and go far deeper. So the measured curve has a kink in it, the rod-cone break, usually around seven to ten minutes in — before it you are watching cones finish, after it rods continue, and it is that second limb that runs on and delivers most of the sensitivity.

Second, a bleached receptor is not merely empty. The protein left holding nothing weakly activates the same signalling cascade on its own, so it behaves as though a faint light were falling on it. You are not only less sensitive after a bright light; you have added a haze of internal noise that has to drain away as the pigment is restored.

And there is a corollary that tests the model neatly. Rods are most sensitive around 500 nanometres and fall away steeply toward the long wavelengths, so dim red light barely touches the rod stock while cones can still use it. That is why darkrooms and ships' bridges are lit red — exactly what you would expect if the mechanism were a wavelength-specific chemical inventory, and not at all what you would expect if it were the pupil.

The correction worth making is that "the eyes adjusting" sounds like something the mind does, and none of it is. It is an inventory, plus a gain setting, plus a pupil — in that order of importance.

The ordering matters because the fast parts are real. The pupil moves in a second, and neural gain adjustments in the retina and beyond are quicker than the chemistry. After a small bleach those fast mechanisms carry most of the recovery and you are back to normal in moments; the twenty-minute figure is the price of a large bleach, where the slow tail is governed by how fast pigment can be resupplied. Exactly how the pigment and neural contributions divide the curve at each moment has been argued over in considerable technical detail, and this simplifies it — but the long tail after a bright light belongs to the resupply.

Which dissolves the original framing. There is no separate skill of night vision that is acquired and then lost. There is a stock that can be spent at any rate the world chooses and rebuilt at only one.

• What happens when the enzyme that re-bends the pigment is faulty, and why some inherited night-blindness traces to exactly that step.

Every term the collection defines is gathered in the glossary.