This morning my Facebook news feed was having some kind of end-times meltdown over this tumblr photograph of a dress. Last night a photographer friend posted ‘What colour is this dress? Mind blown.’ It was clearly white and gold on my iPhone, so I was entirely baffled by the number of people claiming it was blue and black.
As the comments rolled in split between two camps I became pretty sure it was some sort of social media flashmob mindgame, so I resolved to get some sleep and view it in a colour managed environment the next morning. It’s blue and black. If like me you saw white and gold, your brain just messed with you. I’ve dealt with colour critical work for years and know the effects that lighting, hues in the surrounding viewing environment, colourspaces, screens and even browser choice have on colour reproduction and perception, but this is probably the weirdest case of white balance gone wrong I’ve ever seen.
The human eye extrapolates colour by context. If you take a blue shirt into a fitting room with strong yellow lighting, you still see a blue shirt. You could hand it to a friend in those same lighting conditions and they’d know it was a blue shirt. Your eye perceives the yellow wash but your brain makes a decision based on context. A camera lens doesn’t have a brain behind it to cancel out the yellow (which is why getting a good white balance is vital), so if you take a photograph of the shirt, a look at the RGB values might tell you you’re looking at a fairly neutral coloured fabric, the yellow cancelling the blue, but with clues from the environment, human skin tones, a wall you understand to be white, the mind can still make the decision that the shirt is blue, even when the actual colour values say otherwise. This is why the colour picker methods I’ve seen used on social media don’t quite solve the issue.
Edward H. Adelson’s famous checkerboard illusion below demonstrates how the mind makes these context decisions. Square A is the same colour as square B;
You can use a colour picker picker tool to verify this, but it’s easy to spot with the aid of a connecting line;
The B tile is a part of the white pattern, you recognise that it’s in shadow, but you know that it’s lighter. The mind understands the context and makes a technically false decision.
The problem with the photograph of the dress is a lack of context clues for the mind to extrapolate its own white balance, and the crop is a large part of that. If we had a better indicator of the global lighting conditions there’d be no debate. For example if there was something as innately recognisable as a human face in the shot, the mind would have less problems instantly putting the puzzle together. It’s a lack of contextual markers that’s thrown everyone’s brain into meltdown, including my own.
In short, the yellow cast from the lighting and overexposure have pushed yellow ‘gold’ tones into the shadows and neutralised the blues in the midtones, giving the appearance of ‘white’, leaving it in a bizarre middle ground where few people can decide with any immediate certainty which of the two options is correct.
Here’s the dress (minus the bolero) by the way, shot under ideal lighting conditions;
Let us know which colour you saw in the comments below, and please share!