The science of ‘dressgate’

White and gold or blue and black? That’s the question behind a record-breaking twitter storm – so which is it, and what can this tell us about the visual system? Here, Dr Neil Parry explains the science behind the colour enigma.

The first week in March was a strange one for vision scientists all over the world. By and large we sit in darkened laboratories conducting obscure experiments on visual perception on small handfuls of willing volunteers. I recall in my early days pinning down an interesting effect we had seen on one subject, by conducting what we called, with no irony whatsoever, the ‘mass data project’. This required all of 10 subjects. But between the Thursday and Friday evenings of the previous week, an accidental experiment had been conducted with about 30 million subjects!

For the two readers who have been living in a cave, this all revolves around a simple photo, which became known as ‘the dress that broke the internet’. Just google ‘what colour is this dress’. Last week just ‘dress’ would have got you there. The public went crazy arguing about the colour of a dress, about half the world seeing a white dress with gold stripes, and the other half a blue dress with black stripes. The photographer saw one colour, her mother the other, so she sent it to some friends via social media, and it went record-breakingly viral, thanks to Buzzfeed.  By Saturday afternoon, 38 million people had viewed the original Buzzfeed posting.

What, then, has all the fuss been about? What is so special about this dress that it has excited so many people? Why are people so divided about its colour? Why is this debate so important?

There is in fact nothing particularly significant about the dress, which is uncontroversially royal blue with black lace trimmings. The photograph is what caused the debate.  It is horribly overexposed but contains ambiguous clues about the environment it was in. These combined to produce what turns out to be a very ambiguous image. There are some subtle cues as to the lighting but these are insufficient to be able to deduce whether the photo was taken of a blue dress in direct artificial light (indoors) or a white dress outside in shaded sunlight (thus reflecting the blue sky).  Have a look at the cartoon to see this in action using two different artificial lights (with thanks to Bryant Arnold and Cartoon-a-Day).

Discounting the illuminant is is something the brain is normally extremely good at, without any help from us, thanks to a mechanism called Colour Constancy.  Google it.  There are some fantastic demonstrations, particularly Beau Lotto’s ‘Rubik’ cube (www.lottolab.org/illusiondemos/Demo%2012.html).

The mechanism is not perfect but, with sufficient clues about the environment, we are able to discount the colour of the illuminant and extract the ‘real’ colour of the object.  That’s why a piece of white paper still looks white under yellowish tungsten (artificial) light or under blueish daylight.   Take the object out of context, though, and colour constancy fails. If you are in the ‘white dress’ camp, take a small piece of dark card and roll it up so you can view a small piece of the dress on its own.  It should appear to be its true colour, rather a pale blue. My friend Michael Bach has a good demo of this (www.michaelbach.de/ot/col-dress). Colour constancy is a powerful mechanism because there is wide variation in the spectral content of daylight during the day, from reddish to blueish, depending on the height of the sun. Without colour constancy, the bus you get to work in the morning would appear to be a different colour to the one you take home at night. In evolutionary terms, imagine the effect this would have on the apparent ripeness of fruit.

There is little disagreement in the vision science community that the failure of colour constancy is at the bottom of this phenomenon. But what we don’t really understand is, firstly, why people are so divided and, secondly, why the percept is so ‘sticky’. Perception can be said to work by allowing the brain to make and test models of the world based on experience, i.e. prior knowledge. We know that a ripe banana is supposed to be yellow, so even if the illumination is extremely distorted, we still perceive it this way. But no-one viewing that dress had any prior model of its colour, so it may be that, whichever model ‘won’, that was the initial percept. This may also be swayed by interpretations of the way the light is reflected off the lace.

But why so sticky? Most so-called bistable illusions allow one to flip-flop freely between one percept and the other, for example the famous duck-rabbit figure (right) or the face-vase illusion. We don’t really know why this doesn’t happen with the dress image. Most people cannot force themselves to go from the blue camp to the white camp. It may tell us something about the perceptual power of colour. There are at least four major international conferences this year covering colour vision where we can expect the debate to continue. The Vision Sciences Society meeting in Florida in May will have a section devoted to this phenomenon (and have offered a prize to anyone turning up wearing the dress). The International Colour Vision Society has its biennial meeting in Sendai, Japan in July, the European Conference on Visual Perception meets in Liverpool in August, and the OSA Fall Vision Meeting (in San Jose in October) always has a strong colour theme. And I can guarantee you that, in three year’s time, we will start to see some PhD work come out of this.

Why the viral interest in a relatively obscure phenomenon? That’s not for me to say, but it must be key that many slebs, such as Kim’n’Kanye and Taylor Swift weighed in with their view.  And I think it is rather sweet that the whole world joined in this singularity where the internet was being used very publicly for its original ‘private’ purpose – the interchange of academic ideas and data. I’m sure that will also generate a lot of research into the power of social media. And speaking as a colour vision scientist it was great to see this level of excitement. I know that there are bigger things to worry about but there has to be room to debate stuff that is simply interesting and tells us something about perception. It turns out the cameras do lie. It also turns out that scientific truth is not absolute and that, like beauty, colour is in the eye of the beholder.

The author: Dr Neil Parry, SRCS is a Consultant Clinical Scientist at the Vision Science Centre of Manchester Royal Eye Hospital