« An Inconvenient Truth | Main | An ontological question about complex systems »

June 29, 2006

A "reverse" color test

In the past few days, I've received a number of emails from people who took the reverse color-blind test on this site. Normally, I've gotten a sparse, but steady stream of emails about it over the years, but the recent deluge appears to be due to Clive Thompson's blog entry on the topic. (Thanks Clive!) For those of you who are curious, here's the story behind the test.

About six years ago over lunch, my friend Nick Yee and I were chatting about color vision and the differences between what a "color sighted" person can see and what a "color blind" person can see. I represented the former, while Nick represented the latter. Now, there are several different kinds of color-blindness, and Nick happens to have the kind that makes him fairly insensitive to variations in red hue, so-called red-green color blindness. One of the things he complained about was the color blind tests that hide images in the hue of an image so that color sighted people can see the image, while color blind people cannot. We mused about whether you could make a reverse test where we hid an image in a way that color blind people could see, but color sighted people could not. But, if color blindness is purely a deficit, how can this be done?

In most color-sighted people, the parts of the retina that are sensitive to the hue of red seem to be a little more sensitive to intense red hues than the corresponding parts for blue and green. So, our idea was that we could overwhelm (saturate) the red channel in color-sighted people and thus hide information in an independent channel (luminosity) that red-green color-blind individuals would be able to detect. That evening, we each went to work in photoshop and produced some test images to try out on each other. My images and writeup are on this site (here), while Nick maintains his own page here.

What I've learned over the years, from emails by people finding and taking the test, is that a person's sensitivity to red hue varies tremendously from individual to individual. Most people can see the corner of the secret image in the first picture, but few can see the remainder. Some individuals have a more balanced sensitivity to hues, and their eyes aren't fooled by the intensity of the red - they can see the secret image just fine, even though they can also see the full color spectrum (these people seem to be quite rare). Some people have trouble with the second picture - their eyes are confused by the heterogeneous structure of hue - while others have no trouble at all. Even a color-sighted person can see the secret image in the second picture if they know what to look for, that is, if they have a clear expectation about what's hidden and where to look for it, then they can pick out the subtle variations in luminosity.

I'm very curious to hear from someone who knows more about how the retina works, where this variation comes from, and what other possibilities exist for making a better "reverse" color test. For those of you coming from Collision Detection and taking the test, feel free to leave a comment about your experience with the test.

Update, 21 July 2006: The test was also featured recently on Veronica Magazine's Netnews. Welcome folks from the Netherlands!

Update, 13 March 2007: Turns out that the test was dugg sometime around last July, which explains some of the spike in traffic the test received around that time. The Digg comments are kind of amusing - many people discovered that their laptop monitors let them see the "hidden" images even though they aren't colorblind.

posted June 29, 2006 11:53 PM in Self Referential | permalink


Just done the colortest. I am colorblind. Funny thing is: I normally do not see any figures in the color blind tests, now I was wondering what I was missing except the "6" and the circle I was looking at....Answer was: I wasn't missing anything: These figures were almost as clear to me as they were on your "answer page". For me a new phenomenon and I cannot understand that others cannot see the figures in the original picture. Finally a great way to explain to not-colorblind people what colorblindess actually is, as they most often do not understand...

If you want, contact me for more testing: I'd be a happy subject for this!

Posted by: Michael at June 30, 2006 05:23 AM

I thought this was so funny! Both my Dad and Sister are colorblind, but I thought I wasn't. As it turns out, I could see the 6 and circle, but when I looked at your friend's page, I could also see the stuff I was and wasn't supposed to see. I guess I'm both! I had no ides I was mildly colorblind! :) Good work, let me know if I can help out in any way!

Posted by: Leslie at June 30, 2006 11:44 PM

A very interesting test indeed. By no means am I an expert in the areas of color perception or video display technology, but I do have a small amount of experience in those areas as my undergraduate education primarily focused on signal processing, particularly image and video processing.

I took the test after seeing a post about it on the "collision detection" blog. I did not initially see the embedded image in your "red gradient" test, which is displayed much smaller at collision detection. However, as I shifted the angle of my laptop's LCD display the image began to reveal itself, to the extent I could fully identify it even at the reduced size.

When I took the "speckle" test on your blog page I did not need to move the LCD at all. I could perceive very faintly the embedded image at all angles when I did move the LCD.

I believe it may be possible for different display technologies to enhance or diminish the degree to which the embedded images are apparent to the viewer. I have not tried the test using a CRT yet, but I'll try some different displays when I'm next at work. I do know that some LCD displays produce light with a noticeably non-uniform distribution of polarizations, but I don't know if this is responsible for my experiences with the two test images.

Fascinating stuff!

Posted by: Jonathan at July 1, 2006 03:37 AM

I've had a couple of comments from people via email about the effect different display technologies have on the tests - Nick and I originally designed the tests using CRT monitors, but LCD seems pretty ubiquitous today (I've been exclusively on a laptop for about 4 years now). Given my tiny understanding of how these technologies work, it makes a lot of sense to me that different techs and different parameters for those techs would vary the efficacy of the test, mostly by either cranking up the luminosity contrast or by attenuating the hues in the red spectrum.

Posted by: Aaron at July 1, 2006 02:34 PM

I could see the 6 but not the circle. Very interesting!

Posted by: Susan M at July 6, 2006 02:55 PM

Interesting test. It's funny to find out that you are colorblind at 45! In fact, it is such a fascinating area of study, it could become a wonderful topic for many research papers. I would have my wife pass this test too.: )

Posted by: BEST at July 10, 2006 07:47 AM

I found this great Doc in Maryland who was able to help me with my color vision, and now I can pass the ishihara color test! :) www.colormax.org

Posted by: Kirby at July 11, 2006 03:36 PM