Monday, March 16, 2015

No simple genetics for thedress - but does color vision affect compensation for lighting?

In my previous post, I asked whether the way that one sees the now famous dress might have a genetic influence and invited people to send me family data. I got data from 28 families (thank you very much!), and have some conclusions.

First, this cannot be strictly genetic. 
There are examples of monozygotic twins that see the dress differently, and there is a significant minority of people who see it differently from one time to another. These observations are inconsistent with a purely genetic basis. In my own data, I have four families where both parents see blue and black; four of the 12 children in these families see white and gold. I also have nine families where both parents see white and gold; here seven of 20 children see white and gold. Thus, neither trait breeds true. 

However, there are some hints. 
I noticed that 64% of sibling pairs, and 8 of 9 sister pairs, see the dress the same way.  I also noticed that in families where the parents differ, 10/11 daughters see the dress as their father does, which is suggestive of an X-linked partially dominant factor. This led me to ask whether daughters preferentially show the paternal phenotype in families where the parents have the same phenotype. Of cases where I knew the gender of the child, daughters saw the dress as their father did 4/6 times (evenly divided between the two phentoypes). So, overall 14/17 daughters agree with their father. This is significantly different from expected (a simple two-tailed chi-square test with one degree of freedom yields a p value of 0.008).

Since the X chromosome carries the highly polymorphic cone opsin genes that are known to affect color vision, I’m wondering if how one perceives the lighting on the dress (dark vs. light) is affected by these genes. Mechanisms by which women always use color to correct for lighting as their fathers do because of X-linked opsin genes are nearly ruled out by the observation that mutations affecting these genes (red/green color blindness) are recessive. If women used only their father's cone opsin genes, then they would inherit color blindness from their fathers. However, I say "nearly ruled out" because I can think of (admittedly unlikely) scenarios whereby a specific subset of cone cells that plays a role in compensation for lighting also preferentially inactivates the maternal X.  Of course, the limited data here are also consistent with partial dominance, with some other X-linked gene, or, indeed, with no genetic influence at all.
Does normal variation in color vision affect compensation for lighting?
Since the dress illusion is understood to involve compensation for lighting, I am drawn to the question of whether or not variation in color vision affects this compensation. To address this, I’ve come up with a second form, which involves two things:

1) reporting how you see the same image without color

2) taking a vision test (the PANTONE® Online Color Challenge).

I'm asking people to say
-- How they see the dress without color
-- How they see the dress with color and
-- How they score on the online color test (both the numerical score, and, if possible, a screenshot of the results showing the pattern of color discrimination across the spectrum).

The form is available at

This second survey is about the limit of what I can or should do informally through social media. I’m pleased to hear that 23andMe is asking people about how the see the dress. If you have an account with them, then you can contribute at

Note (March 23): 23andMe has posted results from 25,000 responses. They find no strong genetic associations, but an effect of age and an association with whether one lived as a child in a rural (blue and black) or urban (white and gold) setting. They did not look at transmission within families.

Data Summary  March 16
I collected data from 28 families.
The total frequency was approximately half and half. 44 saw blue and black while 54 saw white and gold (six were some sort of intermediate or other; three went back and forth).
In 4 families both parents saw dark colors (blue and black): 6/10 of their children saw colors; 4/10 saw light colors
In 10 families both parents saw light colors: 13/21 children saw dark colors; 7/19 children saw light colors and one child saw the dress differently over time.
In 3 families the mother saw dark colors while the father saw light colors. In this case, 6/6 children saw light colors. All were daughters.
In 6 families the mother saw light while the father saw dark. In this case, 8/11 children (four daughters and four sons) saw dark colors and 3/11 (two sons and one daughter) saw light colors. 3 of 4 daughters saw as their father did.

Mom Dad Families Dark Light Other
Dark Dark 4 8 4 0
Light Light 9 13 7 0
Dark Light 3 0 6 0
Light Dark 6 8 3 0


Unknown said...

What if somebody sees blue and gold ? There have been a few individuals who have claim to have seen these colours.

Steve On Genetics said...

I classify them with white and gold. In fact, I see a light blue and gold myself. In the survey, I have five categories, but I group them into "light" and "dark," intermediate (very rare) and alternating (less rare).