David Baxter PhD
Late Founder
Three years of research into #thedress, digested – a lesson in humility for perceptual science
by Christian Jarrett, BPS Research Digest
May 16, 2018
Three years ago, in a time before Trump or Brexit or This Is America, someone posted an overexposed photograph of a black and blue striped dress on Tumblr. Soon millions of people had seen it and started arguing about it. The reason? It quickly became apparent that about half of us – more often women and older people – perceive the dress, not as black and blue, but white and gold.
In a neat example of real life echoing a classic psychology experiment (I’m referring to Asch), #thedress was enough to make you think your friends were gas lighting you – how could it be that you and they were looking at the exact same picture and yet seeing entirely different things?
The squares marked A and B are actually the same shade of grey, via Wikipedia Of course there are many optical illusions, including others that involve colour (see, for example, the “checker shadow” illusion, pictured right). What was special about #thedress was that it triggered a bimodal split in perceptual experience among the population. Also, many illusions trigger a fluctuating percept, but once someone perceives the dress one way, they usually keep seeing it that way.
Viral hits happen overnight. Science is slow, but it’s catching up. With the passing of the years, numerous studies into #thedress have now been published – 23 according to a new review. Here we present you with a fascinating digest of what’s been discovered so far about the famous frock – researchers have made progress, certainly, yet much remains mysterious, making this a humbling experience for perceptual science.
A big part of it how you see the dress has to do with a process called colour constancy
Before any studies had been published, psychologists and vision experts were quick to explain that the #thedress illusion is related to a process called “colour constancy” whereby your brain takes environmental lighting effects (which are ambiguous in #thedress photo), and your own past experiences, into account when interpreting the precise wavelengths it believes are being reflected off a surface. The same automatic adjustment process allows us to recognise grass as green whatever the weather or time of day. The process sometimes goes wrong, though, like when the blue top you bought at the clothes store turns out to be black when you get home.
How you see the dress depends partly on what inferences you make about the lighting
If our inferences about the background lighting are relevant, as the initial expert reaction suggested, then manipulating the background illumination in an unambiguous way ought to have a direct effect on viewers’ perceptions of the dress. A team led by Rosa Lafer-Sousa found this to be the case. When they presented the dress against an obviously cool, blue-ish background (see image, left), most people saw it as white and gold, but when they presented it against a warm, yellowish background, most saw it as black and blue.
The researchers Andrey Chetverikov and Ivan Ivanchei later developed this idea by showing that people’s perception of the dress are related to their assumptions about the light source in the original photograph. In their survey, people who saw the dress as black and blue were twice as likely as the white and gold group to believe that the dress is illuminated from the front, as if with a flash. Conversely, white and gold perceivers were more likely to believe there was a window behind the dress (implying under exposure).
In the survey, time of day seemed to make a modest difference to these assumptions, but still, the fundamental reasons why we come to make different inferences about the lighting remains largely unknown. In their new review of research into #thedress, published in Archivos de la Sociedad Española de Oftalmología, ophthalmologist Julio González Martín-Moro and his colleagues state that the “dichotomic behaviour of this optical illusion is surprising and leads us to consider two different versions of the software in charge of maintaining chromatic [colour] constancy.”
Your initial impression of the dress are likely to stick
I already mentioned how whichever camp we are in (white/gold or blue/black), most of us seem to stay seeing the dress that way. This stubbornness of the percept could be because our brains continue to make the same assumptions about the scene, but alternatively it might be what Leila Drissi Daoudi and her colleagues describe as an example of “one shot learning”.
To test this, they manipulated naive participants’ perception of the dress by using occluders to block out the background – when they used black occluders (see above), this induced the perception of the dress as white and gold in nearly everyone; when they used white occluders, the opposite was true. Crucially, when the researchers took the occluders away, very few participants experienced any change in their perception of the dress’ colours. “There seems to be a one-shot learning mechanism in play, which leads to an imprinting of the color perceived”, the researchers wrote, although they were unable to explain what (without the use of occluders) induces the initial and lasting perception – they tested the location of people’s initial eye movements to the dress, but this made no difference.
The brains of gold/white perceivers are engaged in more interpretive processing
Consistent with the idea that our inferences about the background lighting are important to #thedress illusion, the results from a brain imaging study that was published in 2015 suggested that people who see the dress as white and gold may be engaged in more “top down” interpretive processing, albeit that this processing leads to a misleading perception. The researchers scanned the brains of volunteers while they looked at the dress and found that those who perceived it as white and gold exhibited more neural activity in a raft of brain areas, including in frontal, parietal (near the crown of the head) and temporal (near the ears) regions.
What this brain scan study couldn’t answer is whether the extra neural processing is the cause or consequence of the gold/white perceptual experience of the dress. However, a survey of patients with multiple sclerosis suggested the neural activity may be causal. Researchers in Italy found that with greater progression of the disease, people were more likely to perceive the dress as blue and black.
Perhaps, the researchers speculated, this is because greater impairment of colour processing regions in the cortex led to “a less demanding black and blue vision of The Dress in more advanced MS”.
Differences in the eye may also account for the illusion
The way we perceive the dress might not all be about colour constancy and top-down brain processing. A 2016 study found that “front end” factors might also be relevant. Specifically, the researchers found that “macular pigment optical density” was higher in volunteers who saw the dress as white and gold. Macular pigment is found in the retina and a greater density increases the amount of short wavelength light that is absorbed by the eye. “Our results indicate that early-stage optical, retinal and neural factors influence perception of the Dress”, the researchers said. Consistent with this, another study found that white and gold perceivers tended to have smaller pupils potentially limiting the amount of light that falls on the retina, thus suggesting another early stage influence upon the perception of the dress.
Past experience seems to be more important than genetics
Last year, researchers in London and Cambridge published the first twin study into how people perceive the dress. By comparing the similarities in perception between identical and non-identical twins, they estimated that around 34 per cent of the variance in people’s perception is due to genetic differences, while 66 per cent is due to environmental factors (i.e. including learning and past experience). This was a small study, however, and the findings are tentative. Omar Mahroo and his colleagues concluded that the environmental factors “…may include prior lifetime experience to different spectral and luminance environments, shaping retinal and higher neuronal processing, and the development and evolution of [how come to name] object colors in various contexts.”
In other words, there are lots of possibilities for how experience shapes our varied perceptions of colour in ambiguous scenes, all of which remain to be studied. We also no nothing about how common visual deficits, such as colour blindness and cataracts, interact with the perception of #thedress.
A lesson in humility
Of all the senses, vision is the one that’s been studied most intensively by psychologists and neuroscientists. Despite this, there are no tests available that would allow even the best experts in the land to predict whether you will see the dress as black and blue or white and gold. And while we have some good ideas about the processes that are involved in #thedress illusion, so much remains mysterious about what leads each of us to view it one way or the other.
At the end of their review, González Martín-Moro and his colleagues come to a stark conclusion: “…the fact is that researchers have failed to come up with a theory that can satisfactorily explain the dichotomic behaviour of the population when viewing said optical illusion.”
by Christian Jarrett, BPS Research Digest
May 16, 2018
Three years ago, in a time before Trump or Brexit or This Is America, someone posted an overexposed photograph of a black and blue striped dress on Tumblr. Soon millions of people had seen it and started arguing about it. The reason? It quickly became apparent that about half of us – more often women and older people – perceive the dress, not as black and blue, but white and gold.
In a neat example of real life echoing a classic psychology experiment (I’m referring to Asch), #thedress was enough to make you think your friends were gas lighting you – how could it be that you and they were looking at the exact same picture and yet seeing entirely different things?
The squares marked A and B are actually the same shade of grey, via Wikipedia Of course there are many optical illusions, including others that involve colour (see, for example, the “checker shadow” illusion, pictured right). What was special about #thedress was that it triggered a bimodal split in perceptual experience among the population. Also, many illusions trigger a fluctuating percept, but once someone perceives the dress one way, they usually keep seeing it that way.
Viral hits happen overnight. Science is slow, but it’s catching up. With the passing of the years, numerous studies into #thedress have now been published – 23 according to a new review. Here we present you with a fascinating digest of what’s been discovered so far about the famous frock – researchers have made progress, certainly, yet much remains mysterious, making this a humbling experience for perceptual science.
A big part of it how you see the dress has to do with a process called colour constancy
Before any studies had been published, psychologists and vision experts were quick to explain that the #thedress illusion is related to a process called “colour constancy” whereby your brain takes environmental lighting effects (which are ambiguous in #thedress photo), and your own past experiences, into account when interpreting the precise wavelengths it believes are being reflected off a surface. The same automatic adjustment process allows us to recognise grass as green whatever the weather or time of day. The process sometimes goes wrong, though, like when the blue top you bought at the clothes store turns out to be black when you get home.
How you see the dress depends partly on what inferences you make about the lighting
If our inferences about the background lighting are relevant, as the initial expert reaction suggested, then manipulating the background illumination in an unambiguous way ought to have a direct effect on viewers’ perceptions of the dress. A team led by Rosa Lafer-Sousa found this to be the case. When they presented the dress against an obviously cool, blue-ish background (see image, left), most people saw it as white and gold, but when they presented it against a warm, yellowish background, most saw it as black and blue.
The researchers Andrey Chetverikov and Ivan Ivanchei later developed this idea by showing that people’s perception of the dress are related to their assumptions about the light source in the original photograph. In their survey, people who saw the dress as black and blue were twice as likely as the white and gold group to believe that the dress is illuminated from the front, as if with a flash. Conversely, white and gold perceivers were more likely to believe there was a window behind the dress (implying under exposure).
In the survey, time of day seemed to make a modest difference to these assumptions, but still, the fundamental reasons why we come to make different inferences about the lighting remains largely unknown. In their new review of research into #thedress, published in Archivos de la Sociedad Española de Oftalmología, ophthalmologist Julio González Martín-Moro and his colleagues state that the “dichotomic behaviour of this optical illusion is surprising and leads us to consider two different versions of the software in charge of maintaining chromatic [colour] constancy.”
Your initial impression of the dress are likely to stick
I already mentioned how whichever camp we are in (white/gold or blue/black), most of us seem to stay seeing the dress that way. This stubbornness of the percept could be because our brains continue to make the same assumptions about the scene, but alternatively it might be what Leila Drissi Daoudi and her colleagues describe as an example of “one shot learning”.
To test this, they manipulated naive participants’ perception of the dress by using occluders to block out the background – when they used black occluders (see above), this induced the perception of the dress as white and gold in nearly everyone; when they used white occluders, the opposite was true. Crucially, when the researchers took the occluders away, very few participants experienced any change in their perception of the dress’ colours. “There seems to be a one-shot learning mechanism in play, which leads to an imprinting of the color perceived”, the researchers wrote, although they were unable to explain what (without the use of occluders) induces the initial and lasting perception – they tested the location of people’s initial eye movements to the dress, but this made no difference.
The brains of gold/white perceivers are engaged in more interpretive processing
Consistent with the idea that our inferences about the background lighting are important to #thedress illusion, the results from a brain imaging study that was published in 2015 suggested that people who see the dress as white and gold may be engaged in more “top down” interpretive processing, albeit that this processing leads to a misleading perception. The researchers scanned the brains of volunteers while they looked at the dress and found that those who perceived it as white and gold exhibited more neural activity in a raft of brain areas, including in frontal, parietal (near the crown of the head) and temporal (near the ears) regions.
What this brain scan study couldn’t answer is whether the extra neural processing is the cause or consequence of the gold/white perceptual experience of the dress. However, a survey of patients with multiple sclerosis suggested the neural activity may be causal. Researchers in Italy found that with greater progression of the disease, people were more likely to perceive the dress as blue and black.
Perhaps, the researchers speculated, this is because greater impairment of colour processing regions in the cortex led to “a less demanding black and blue vision of The Dress in more advanced MS”.Differences in the eye may also account for the illusion
The way we perceive the dress might not all be about colour constancy and top-down brain processing. A 2016 study found that “front end” factors might also be relevant. Specifically, the researchers found that “macular pigment optical density” was higher in volunteers who saw the dress as white and gold. Macular pigment is found in the retina and a greater density increases the amount of short wavelength light that is absorbed by the eye. “Our results indicate that early-stage optical, retinal and neural factors influence perception of the Dress”, the researchers said. Consistent with this, another study found that white and gold perceivers tended to have smaller pupils potentially limiting the amount of light that falls on the retina, thus suggesting another early stage influence upon the perception of the dress.
Past experience seems to be more important than genetics
Last year, researchers in London and Cambridge published the first twin study into how people perceive the dress. By comparing the similarities in perception between identical and non-identical twins, they estimated that around 34 per cent of the variance in people’s perception is due to genetic differences, while 66 per cent is due to environmental factors (i.e. including learning and past experience). This was a small study, however, and the findings are tentative. Omar Mahroo and his colleagues concluded that the environmental factors “…may include prior lifetime experience to different spectral and luminance environments, shaping retinal and higher neuronal processing, and the development and evolution of [how come to name] object colors in various contexts.”
In other words, there are lots of possibilities for how experience shapes our varied perceptions of colour in ambiguous scenes, all of which remain to be studied. We also no nothing about how common visual deficits, such as colour blindness and cataracts, interact with the perception of #thedress.
A lesson in humility
Of all the senses, vision is the one that’s been studied most intensively by psychologists and neuroscientists. Despite this, there are no tests available that would allow even the best experts in the land to predict whether you will see the dress as black and blue or white and gold. And while we have some good ideas about the processes that are involved in #thedress illusion, so much remains mysterious about what leads each of us to view it one way or the other.
At the end of their review, González Martín-Moro and his colleagues come to a stark conclusion: “…the fact is that researchers have failed to come up with a theory that can satisfactorily explain the dichotomic behaviour of the population when viewing said optical illusion.”
