Examine the dress top left: it has stripes small and wide, let us concentrate on the wide stripes at the top. What colours do you see in these stripes:
Gold on top, white below? Or:
Black / dark brown on top, blue below?
You will have a clear opinion on that, so where is the illusion? This is a SOCIAL illusion – it only becomes interesting in company. Seek out spouse, family, friends, colleague, whatever and ask them. [Make sure that all look straight at the screen (many LCDs distort colours when watched from an angle.]
If you find their colour impression to markedly differ from yours, don’t start a row :). I hear couples have separated just because they disagreed on this figure! It is perfectly normal to differ in opinion here – lengthy explanation below.
What to do
Drag the mask with the two holes right of the photo. Find a position to reveal the top stripe in the upper hole, the lower wide stripe in the lower hole. Looking at the colour patches in isolation, what do the colours look like now? Do you still strongly disagree? Chances are, now there will be less debate, I assume both factions will now see it more brownish / light bluish.
Or: Move the mouse over the image (or tap it for touch devices). Now the disk at bottom-right will give the local colour thus sampled. Another option to see the local colour in another context.
Dress colours (w/o background) in the CIE chromaticity diagram (→full R code )
Comments
This has nothing to do with colour deficiency / colour blindness. One could call it the “Necker Cube of colour vision” (Moukheiber). To understand this, two main steps are needed: (1) how do we judge colours anyway, especially textiles under different illumination, and (2) why do we have such strong disagreements for this specific photo, although in all other cases we usually can roughly agree on colour appearance (excepting colour deficiency).
The colour of any non-luminous object, the dress’s textile here, rests on two factors:
One is the spectral distribution of the incident light (is it white, reddish or bluish; or specific parts of the spectrum are very strong or missing {neon lighting}). The spectral distribution –reddish in the morning/evening, bluish around noon– varies more than we perceive.
Second are the light-absorption properties of the textile. Does it reflect incident red or not? And so on for all wavelengths. The light that reaches our eye is a spectrum, where per wavelength the incident light strength is multiplied with the reflectivity. Thus the light emitted by a textile in reddish light can be very different from the light emitted by the same textile in bluish light.
Our visual system “knows” about these two factors! It cleverly tries to detect the colour of the incident light and then “subtract” it. This process is called “colour constancy”, and usually is quite successful. A case in point: we usually don’t perceive the colour change across the day, but we are surprised when a photo, taken on a glacier at noon, looks markedly bluish when viewed at home.
In a nutshell: Colours are judged rather correctly, even under coloured room illumination, because our visual system guesses on incident light direction and colour and “subtracts” it. This is quite similar for all of us.
Why then does it differ so much between people in this case? Now I enter the field of hypothesis, having not had the chance to perform careful experiments on this (the photo only appeared yesterday and immediately went viral). I assume 3 factors to play a role here:
Bad photo :). This photo is poorly lit (luckily, otherwise we wouldn’t talk about it…) and incomplete – only the dress itself. The right part is clearly overexposed. Since with overexposure all colours desaturate, and shadows vanish, our visual system has too little information to guess at the direction and colour of the illumination and ambient lighting. So in different persons the guess at the illuminant will differ, and after its “subtraction” the perceived dress colour will differ. Using the mask puts you into another mode: no lighting can be reconstructed anyway.
This specific textile is glossy, has lustre (shimmering). Here experience with this type of textile might differ, and the visual system knows less about this, adding to the interindividual differences of the previous item.
Colour naming is categorical. We say red-orange-yellow-green, although of course in between there are more hues. But we lump together a range of wavelength compositions into colour terms, where these terms in part depend on your cultural background. Now imagine a colour bright & bluish – for some this falls into “white”, for others this falls into “blue”. Thus totally different words are used, although the actual difference in perception might be subtle. Gold is a difficult colour anyway: it is a brown (or dark orange) with some glossy sprinklings.
So why would this be the Necker Cube of colour vision? Because the picture is bistable, two interpretations are about equally likely. The difference to the Necker Cube is that is is difficult for oneself to switch between the two possible interpretations.
Finally: what is the real colour after all??? Answer, well lit with white light: black & blue. See neighbouring figure.
Interested to partake in a survey on your perception of the dress? →Here.
Want to see the code that produced the chromaticity colour sampling diagram above? →Here is the complete R code.