The Photograph That Proved You Don't See the World

In the last week of February 2015, a mother on the Scottish island of Colonsay took a photograph she would never live down. She was preparing for her daughter’s wedding and wanted to show off a dress she planned to wear. The lighting in the shop was poor, the phone was ordinary, and the picture came out washed and ambiguous. She sent it to the bride, and the two women promptly disagreed about what colors were in front of them. One saw blue and black. One saw white and gold. Neither could talk the other out of it.

Unable to settle the argument privately, the family did what families now do. They posted it. Within days the image had leapt from a wedding-planning quarrel to a Tumblr post to the front page of BuzzFeed, and from there into a genuine global event. The photograph was viewed tens of millions of times inside a single week. Celebrities weighed in. Couples argued over dinner. People held phones up to strangers on trains and demanded a verdict. What made it uncanny was not that people disagreed about a color. People disagree about colors all the time. It was that they were looking at the exact same pixels on the exact same screens and still could not converge.

The episode was treated at first as a piece of internet silliness, a viral trifle that would evaporate by the weekend. It did not evaporate. Instead it became something rarer: a spontaneous perceptual experiment on a scale no laboratory could ever fund or ethically arrange. Millions of people, in real time, discovered that the private theater of their own vision did not match the person sitting next to them. And a small group of vision scientists, watching the chaos unfold, realized they had been handed a gift.

The eye is not a camera

The intuition almost everyone carries is that seeing is a kind of passive recording. Light enters the eye, lands on the retina, and a faithful copy of the outside world is delivered to the brain, like a photograph dropping into an inbox. It is a comfortable picture of perception. It is also wrong.

The signals the eye actually sends inward are fragmentary and strange. The retina does not transmit a finished image; it transmits patterns of contrast, edges, sudden changes in brightness, and coarse wavelength information, all of it noisy and incomplete. There is a hole in the middle of each visual field where the optic nerve exits, a genuine blind spot you never notice. Color-sensitive cones cluster near the center and thin out toward the edges, so most of your peripheral vision is nearly colorblind. The smooth, continuous, richly colored world you seem to inhabit is not arriving through your eyes fully formed. It is being constructed, moment by moment, by the brain, which fills every gap with an inference about what is most likely out there.

Seeing, in other words, is closer to guessing than to recording. The philosopher and neuroscientist would say the brain runs a constant prediction about the causes of the light hitting the retina, then corrects that prediction against incoming evidence. Most of the time the guess is so accurate, and so instantaneous, that we mistake it for direct perception. We believe we are looking at reality. We are looking at our brain’s best model of it.

Usually the machinery is invisible because it agrees with itself. Your guess and my guess about the color of a stop sign land in the same place, so we never notice that guessing was involved at all. The dress was extraordinary precisely because it drove two guesses apart and then held them there, stubbornly, so that ordinary people could feel the seam.

The problem of subtracting the light

To understand why the guesses diverged, you have to understand a quiet miracle your visual system performs thousands of times a day without asking permission. It is called color constancy, and it is one of the most useful tricks the brain has.

Consider a plain white shirt. At noon under a clear sky, the light falling on that shirt is heavy with blue. Near a candle or an incandescent bulb, the light is warm and yellow-orange. The wavelengths physically bouncing off the fabric are dramatically different in each case. If your visual system reported raw wavelengths, the shirt would appear to change color every time you walked from a window to a lamp. Instead, it looks white in both. Your brain has quietly measured the color of the ambient light and subtracted it, discounting the illumination in order to recover the true color of the object underneath.

This is a remarkable feat of computation. A surface’s apparent color is a tangle of two things: the color of the object and the color of the light hitting it. The brain’s job is to separate them, to answer the question “what color is this thing, independent of how it happens to be lit right now.” It does this using every cue it can find, the color of the surrounding scene, the brightness gradients, memories of how similar objects look, and assumptions about what kind of light is probably present.

And here is the trap the dress laid. The photograph was overexposed and badly balanced. The background gave almost no reliable information about the light source. Was the dress standing in cool bluish daylight spilling through a window? Or was it lit by warm artificial light indoors? The image did not say. The one crucial clue the brain needed in order to run its subtraction, the color of the illumination, had been stripped away.

So each brain had to gamble. And different brains placed different bets.

Bevil Conway and the two assumptions

Among the scientists drawn in was Bevil Conway, then at Wellesley College and later at the National Institutes of Health, who studies exactly how the brain converts wavelengths into the felt experience of color. He and his colleagues ran a survey of more than 1,400 people, one of three studies published together in the journal Current Biology in May 2015, barely three months after the image first appeared.¹

The numbers were startling. Roughly 57 percent of respondents reported blue and black. About 30 percent saw white and gold. Around 11 percent saw blue and brown, and a small remainder could switch between interpretations. “I couldn’t believe it,” Conway told reporters at the time. In color science, stimuli almost never split a population cleanly into stable camps like this. Ambiguous images usually flicker; people flip back and forth. The dress locked people in.²

Conway’s explanation turned on the hidden assumption each viewer’s brain made about the light. Some brains, he argued, unconsciously assumed the dress was bathed in cool, bluish daylight. To recover the true color of the fabric, those brains subtracted blue from the image, and what remained looked white and gold. Other brains assumed warm, yellow indoor lighting. They subtracted the yellow, and what remained read as blue and black.

Same photograph. Two different assumptions about the illumination. Two genuinely different perceived objects. Neither group was making an error in the ordinary sense. Each was correctly applying color constancy to an image that had been engineered, by accident, to support two incompatible solutions. The dress was a rare case where the brain’s normally invisible arithmetic produced two right answers.

Larks, owls, and a lifetime of light

If Conway explained the mechanism, another researcher went looking for why particular people fell on particular sides. Pascal Wallisch, a neuroscientist at New York University, suspected the split was not random. He surveyed more than 13,000 people and found a pattern that connected perception to something intimate: a person’s history of exposure to light.³

Wallisch’s hypothesis was that the brain resolves the ambiguity by guessing at the most likely illuminant, and that this guess is shaped by what kind of light a person has spent their life under. People who wake early and spend more of their waking hours in natural daylight, the so-called larks, were significantly more likely to assume the dress was lit by daylight. Assuming daylight means subtracting blue, which yields white and gold. Night owls, who spend disproportionately more time under warm artificial light, were more likely to assume artificial illumination, subtract yellow, and see blue and black.³

In other words, the answer you gave to a viral internet quiz was quietly encoding your chronotype, your sleep schedule, and your accumulated lifetime of lighting environments. The dress was functioning as a mirror, reflecting back a private history most people had never articulated. Someone who had spent decades rising with the sun and someone who had spent decades working late under bulbs were not simply choosing different answers. They were, in a real sense, living in different perceptual defaults, built up over years and only exposed by one ambiguous photograph.

Other factors mattered too. Age and eyesight shifted the balance; older viewers, and those whose lenses had yellowed with age, leaned more often toward white and gold. Gender showed a modest effect in some samples. The point that emerged from all of it was uncomfortable for anyone who believes perception is objective. As Wallisch put it, individual differences in perception are the rule, not the exception. We normally do not notice them because the world rarely poses a question ambiguous enough to pull them apart. The dress did.

Nobody was wrong

Here is the part of the story that almost everyone missed while arguing. The argument itself was based on a false premise. People assumed that one camp had to be right and the other deluded, that somewhere inside the pixels lurked a single true color, and that the disagreement was a contest to be won. There was no such hidden truth to win.

The photograph, as a photograph, was genuinely ambiguous. It supported multiple valid perceptual solutions. Both the blue-and-black viewers and the white-and-gold viewers were seeing the image accurately, in the sense that each was running the same normally reliable machinery on data that happened to permit more than one answer. It is tempting to say one group was fooled, but that framing misunderstands what vision is. Nobody was being fooled. Everyone was doing exactly what the visual system always does, which is to make an intelligent guess and then experience that guess as fact.

This is the deeper reason the dress unsettled people. It was not merely surprising that friends disagreed. It was that the disagreement could not be resolved by looking harder, or by anyone being more honest, or by better eyesight. Two people staring at identical information arrived at genuinely different realities, and both were correct. That is not how we like to think perception works. We prefer to believe that if we all just look carefully at the same thing, we will see the same thing. The dress quietly demolished that assumption in front of tens of millions of witnesses.

What the pixels were hiding

There is, admittedly, a fact about the physical garment. The real dress, a bodycon design sold by the British retailer Roman Originals, was blue and black. You can buy it. Under neutral lighting it is unambiguously blue with black lace. In that narrow sense, the blue-and-black camp happened to match the object in the shop.

But it would be a mistake to treat this as vindication, as if the white-and-gold viewers had simply lost the round. The question the internet was actually asking was never “what color is the physical fabric in a warehouse in England.” It was “what color do you see in this photograph,” and to that question there was no single answer. The physical dress and the photographic image are two different objects. Confusing them is exactly the error that made the whole thing feel like a paradox in the first place.

What the pixels were hiding, then, was not a color. It was the ordinarily invisible act of interpretation that stands between the light and the seeing. Every waking moment, your brain is discounting illuminants, filling blind spots, smoothing over the noise, and delivering to your awareness a confident, seamless, colored world that feels like raw reality. It almost never shows its work. The dress, for a few strange weeks in 2015, forced it to.

The humility of certainty

The lasting lesson of the dress has little to do with fashion and everything to do with how much of what we call reality is authored inside the skull. The colors you see are not a transcript of the world; they are your brain’s best inference about the world, shaped by your biology, your age, the state of your eyes, and the accumulated lighting of your entire life. Most of the time your inference and mine align closely enough that we never suspect a difference exists. Beneath that shared agreement, though, we are each running a slightly private version of everything.

The dress did not create these differences. It revealed them, dragging a piece of ordinarily hidden machinery into the daylight where a crowd could watch it operate. And it left behind a small, useful humility. The next time you are utterly certain about what you are seeing, it is worth remembering that certainty is a feeling the brain manufactures, not a guarantee it can honor. Someone standing right beside you, looking at exactly what you are looking at, may be living, quietly and irreconcilably, in a slightly different color.

Sources

1. Lafer-Sousa, R., Hermann, K. L., Conway, B. R., Striking individual differences in color perception uncovered by ‘the dress’ photograph, Current Biology, 2015. — https://www.cell.com/current-biology/fulltext/S0960-9822(15)00420-0
2. Corum, J., A Simple Way to Understand Why We See The Dress Differently, The New York Times, 2015. — https://www.nytimes.com/interactive/2015/02/28/science/the-dress-blue-black-white-gold.html
3. Wallisch, P., Illumination assumptions account for individual differences in the perceptual interpretation of a profoundly ambiguous stimulus in the color domain: ‘The dress’, Journal of Vision, 2017. — https://jov.arvojournals.org/article.aspx?articleid=2617976
4. Gegenfurtner, K. R., Bloj, M., Toscani, M., The many colours of ‘the dress’, Current Biology, 2015. — https://www.cell.com/current-biology/fulltext/S0960-9822(15)00426-1
5. Winkler, A. D., Spillmann, L., Werner, J. S., Webster, M. A., Asymmetries in blue-yellow color perception and in the color of ‘the dress’, Current Biology, 2015. — https://www.cell.com/current-biology/fulltext/S0960-9822(15)00464-9
6. Rogers, A., The Science of Why No One Agrees on the Color of This Dress, WIRED, 2015. — https://www.wired.com/2015/02/science-one-agrees-color-dress/
7. Mahowald, M., et al. (overview), Color constancy, Encyclopedia of Perception / Wikipedia biographical anchor for color constancy. — https://en.wikipedia.org/wiki/Color_constancy

Related reading

• The Man Behind the Door
• The Vanishing Cinema Behind Your Eyelids

More from the Mind edition →