That Dress

In: What do you consider the most interesting recent (scientific) news? What makes it important?

Ed. John Brockman, New York, Harper, 2016

Also online at EDGE The annual question (2016) – my response


Could the colour of a cheap dress create a meaningful scientific controversy? In 2015 a striped, body-hugging, £50 dress did just that. In February, Scottish mother Cecilia Bleasdale sent her family a poor quality photo of a dress she bought for her daughter’s wedding. Looking at the same image, some people saw the stripes as blue and black; others as white and gold. Quickly posted on Tumblr ‘that dress’ was soon mentioned half a million times. On Buzzfeed it gained over 1.5 million mentions within 24 hours and by March 1st had over 37 million hits. On Twitter it had ten million tweets in the first week and over eleven million altogether. This simple photo had everything a meme needs to thrive: it was easy to pass on, accessible to all, and sharply divided opinions. #thedress was indeed called the meme of the year and even a ‘viral singularity’. Yet it did not die out as fast as it had risen. Unlike most viral memes, this one prompted deeper and more interesting questions.

Scientists quickly picked up on the dispute and garnered some facts. Seen in daylight the actual dress is indisputably blue and black. It is only in the slightly bleached out photograph that white and gold is seen. In a study of 1,400 respondents who’d never seen the photo before, 57% saw blue and black, 30% saw white and gold, and about 10% saw blue and brown. Women and older people more often saw white and gold.

This difference is not like disputes over whether the wallpaper is green or blue. Nor is it like ambiguous figures such as the famous Necker cube which can be seen tilted towards or away from the viewer, or the duck/rabbit or wife/mother-in-law drawings. People can typically see these bistable images either way, and flip their perception between views, getting quicker with practice. Not so with ‘that dress’. Only about 10% of people could switch colours. Most saw the colours resolutely one way and remained convinced that they were ‘right’. What was going on became a genuinely interesting question for the science of colour vision.

Vision science has long shown that colour is not the property of an object, even though we go on speaking as though it is. In fact colour emerges from a combination of the wavelengths of light emitted or reflected from an object and the kind of visual system looking at it. A normal human visual system, with three cone types in the retina, concludes ‘yellow’ when any one of an indefinite number of different wavelength combinations affects its colour-opponent system in a certain way. This means that a species with more cone types, such as the mantis shrimp which has about sixteen types, would see many different colours when humans would see only the same shade of yellow.

When people are red-green ‘colour blind’, with only two cone types instead of three, we may be tempted to think they fail to see its real colour. Yet there is no such thing. There are even rare people (mostly women) who have four cone types. Presumably they can see colours that the rest of us cannot even imagine. This may help us accept that the dress is not intrinsically one colour or the other, but still provides no clue as to why people see it so differently.

Could the background in the photo be relevant? In the 1970s, Edwin Land (inventor of the Polaroid camera) showed that the same coloured square appears to be a different colour depending on the other squares surrounding it. This relates to an important problem that evolution has had to solve. If colour information is to be useful, an object must look the same colour on a bright sunny day as on an overcast one, yet the incident light is yellower at midday and bluer from a gloomy or evening sky. So our visual systems use a broad view of the scene to assess the incident light and then discount that when making colour decisions, just like the automatic white balance (AWB) in modern cameras.

This, it turns out, may solve the great dress puzzle. It seems that some people take the incident light as yellowish, discounting the yellow to see blue and black, while others assume a bluer incident light and see the dress as white and gold. Do the age and sex differences provide any clues as to why? Are genes or people’s lifetime experiences relevant? This controversy is still stimulating more questions, and in 2016 the journal Current Biology is to publish a special edition on ‘that dress’.

Was it a step too far when some articles suggested that #thedress could prompt a “world-wide existential crisis” over the nature of reality? Not at all, for colour perception really is strange. When philosophers ponder the mysteries of consciousness they may refer to qualia – private, subjective qualities of experience. An enduring example is ‘the redness of red’ because the experience of seeing colour provokes all those questions that make the study of consciousness so difficult. Is someone else’s red like mine? How could I find out? And why, when all this extraordinary neural machinery is doing its job, is there subjective experience at all? I would guess that ‘that dress’ has yet more fun to provide.