The whole premise of a remember-the-color game is a gap. A color appears, you look at it, it disappears, and then, after some delay, you have to put it back. The interesting question is what happens inside that gap. Most people assume the answer is obvious: the longer the wait, the fuzzier the memory, the worse the guess. A color you reproduce after one second should be close, a color you reproduce after ten should be a smear. That picture feels right. It is also, according to the people who have measured it most carefully, mostly wrong.
The intuition: a slow fade
Ask anyone to describe how a remembered color decays and you get the same shape: a smooth slide. The color is vivid at first, then it loses saturation, drifts, washes out, and eventually becomes a vague impression of “something blue-ish.” It is the visual version of a sound fading out. Every score you have ever lost to a long delay seems to confirm it. Wait longer, land further from the target. A clean, gradual relationship between time and error.
The trouble is that an average can hide its own shape. If you take a hundred guesses made after a long delay and average their error, you do get a worse number than the short-delay average. But the average does not tell you how those guesses went wrong. It could be that every guess drifted a little. It could also be that most guesses stayed almost perfect and a few collapsed completely. Those two stories produce the same mean and could not be more different. For decades the field read that mean as gradual fading because the tools in use could not pull the two stories apart.
What actually happens: sudden death
Weiwei Zhang and Steven Luck built an experiment that could. Instead of asking whether a remembered color was right or wrong, they had people report the color directly by clicking on a color wheel, which separates two things the old methods blurred together: the probability that a memory is still there at all, and the precision of that memory when it is. People held three colored squares, waited out a delay of up to several seconds, and then reported one of them.
The result reversed the intuition. The memories that survived the delay were just as precise after several seconds as they were after a fraction of one. The colors that were still there were still sharp. What changed over time was not the resolution of the memory but the odds of it existing. Some representations simply stopped, in full, partway through the delay. Zhang and Luck called it sudden death: a color memory holds at high resolution and then terminates all at once, rather than gently blurring its way out.
It is worth sitting with how counterintuitive that is. Your memory of a color is not a photograph slowly yellowing in the sun. It is closer to a light that is either on or off. While it is on, the color is essentially as good as the moment you looked away. When it goes, it goes whole, and you are left guessing from nothing, which is why the wrong answers in these tasks tend to be not slightly off but wildly off, scattered as if the person were picking at random. Because, in those cases, they were.
Why this changes how a color score behaves
If color memory died gradually, a delay-based game would have smooth, forgiving physics: a little more time costs you a little more accuracy, and you could trade one for the other. Sudden death makes the game bimodal instead. On any given round you are likely in one of two states. Either the memory is alive, in which case a long delay barely hurts you and your guess can be excellent, or it has died, in which case no amount of care helps and your guess is a coin flip dressed up as a decision.
This matches something most players feel but misread. A run of strong rounds followed by one that is bafflingly bad does not mean you suddenly got worse at color. It usually means the light went out on that one round. The skill being tested is partly precision, but it is also survival: keeping the representation alive long enough to use it. That is a different thing from seeing color well, and it is why a steady score in a timed mode like Speed, with its one-second flashes, is harder to earn than people expect. You are not fighting blur. You are racing a memory that might simply switch off.
The second clock: drift toward the category
Sudden death is only half the story, because the memories that stay alive are not perfectly frozen either. They lean. Bae, Olkkonen, Allred, and Flombaum had people reproduce hundreds of colors and found that the reports were systematically pulled toward the centers of color categories, and away from the boundaries between them. A color sitting on the line between blue and green does not stay on the line in memory. It picks a side and slides toward the prototype of that side. Show someone a teal that is slightly more green than blue, and they remember a greener green than they saw.
The reason is that your visual system stores a color as two things at once: the actual shade, and a coarse verbal label for it. When you go to recall the color, those two sources get combined, and the label, being cruder, drags the answer toward the middle of its category. The further a color is from a clean prototype, the more the label distorts it. This is the same machinery we describe in memory colors, where a stored idea of an object’s color bleeds into what you actually see. Here it is the category, not the object, doing the bending.
Put the two effects together and you get a fuller picture of the gap. For as long as a color memory lives, it does not soften, but it does drift, and it drifts in a predictable direction: toward whatever single word you would use to name it. Then, at some unpredictable moment, it may die outright. Gradual bias on one axis, sudden loss on the other.
How to play against both clocks
Knowing the failure modes suggests a way to fight them, and it is not “concentrate harder,” which does nothing against sudden death. Two habits actually help.
Commit early. Because a live memory is already at full resolution, waiting does not improve your guess, it only gives the memory more time to die. The first confident reading you have is very close to the best one you will ever have. Players who deliberate, hunting for an accuracy that is not coming, are mostly just exposing a good memory to more risk. Lock the answer while the light is still on.
Fight your own label. The category drift is driven by the single word your brain tags the color with. If that word is “blue,” you will reconstruct a more prototypical blue than you saw. The counter is to refuse the easy word and store a more specific one in the brief moment you can see the color: “greyed navy, a touch toward purple” pulls toward a different, closer prototype than “blue” does. You cannot stop the labeling. You can choose a better label. We make the same argument at length in training your eye for color, and it is the single habit that separates people who plateau from people who keep improving.
None of this turns memory into a recording. The point of measuring a color guess with a perceptual metric like CIEDE2000 is that it prices a near-miss fairly, and near-misses are what a healthy color memory produces: a shade that is close, slightly category-shifted, but plainly in the right neighborhood. The catastrophic scores are not bad perception. They are the rounds where the light went out, and the only defense against those is to answer before they happen.
So, how long?
The honest answer is that there is no single number, because the question has the wrong shape. You are not asking how long until a color fades, because it does not fade. You are asking how long until a particular memory rolls the dice and loses, and that can be one second or it can be many. What you can count on is this: while the memory is alive it is good, better than you give it credit for, and the moment to use it is now. The worst thing you can do with a color you remember clearly is wait to see whether you remember it more clearly in a moment. You will not. Put it down. Try a few rounds with that in mind, or test the sharp-then-gone feeling directly in Blind Sliders, where there is no live feedback to lean on and the only thing in play is the color you are still holding.