Hacker News Reader

Comment by jacobolus | Hacker News Reader

This is bogus. Basically the popular combinations are "liked" because (a) they are colorful, and (b) they have value (lightness) contrast, whereas the others are both less colorful, and also have low value contrast. Completely removing the color information from these suggested schemes reveals that value contrast alone accounts for nearly all of the suggested effect.

It's easier to make combinations that satisfy conditions a and b using light yellow and dark red or blue, etc., ... but all the part about picking colors that "shift in accordance with the natural brightness" is nonsense. (I can pick colors that shift "opposite the natural brightness" which look just as good.)

Anyone wanting to really understand color should hit the books, or—if confined to the internet—should take a look at http://www.handprint.com/HP/WCL/wcolor.html and in particular this page: http://www.handprint.com/HP/WCL/color11.html

View on HN

For those who don't believe me, here's an explicit graphical rebuttal to the article:

http://www.hcs.harvard.edu/~jrus/ycnews/nonsense.png

I actually find the original palettes to be more pleasing than your reconstructed ones due to the hue shift. I can't help but go with what my eyes are telling me.

Sure, I only swapped the values and not also the chroma information. But the point was to refute the original article, not to set up my own controlled experiment.

As another example... if this hue shift was the main thing that mattered... well, I took his "ugly" combination and flipped the color information vertically, so now the hue shifts in the opposite direction from before. Does the new combination on the right really look better? http://www.hcs.harvard.edu/~jrus/ycnews/nonsense2.png

The original idea is that the ones on the left are uglier than the ones on the right because of their brightness. The rebuttal switches the brightness levels, which results in the same progression, from ugly to less ugly.

Nice observation (and diagram) about the greater value contrast on the right side.

Note also there is greater saturation contrast on the left side. (I'm not sure how this effects aesthetics.) You've inspired me to create a corresponding diagram:)

http://peoplesign.com/content/colorSaturationHN.png

You should use a measure of chroma (CIELAB C*, the C from CIECAM's JCh, Munsell's "chroma", etc.), rather than HSB/HSV "saturation", as the latter is not really meaningful from a human visual perception standpoint. CIELAB isn't ideal for this, but it's easy to compute:

http://www.hcs.harvard.edu/~jrus/ycnews/nonsense3.png

Respectfully, you're too quick to dismiss saturation. Unless you have a color very close to either white or black, decreasing saturation makes the color more gray, and increasing saturation makes the hue more vibrant/noticeable.

In our example, balancing the saturation of the left colors makes a meaningful perceptual difference (again, I'm not saying anything about aesthetics).

Here is a side-by-side comparision with saturation values shown for each color. http://peoplesign.com/content/colorSaturationHN2.png

Yeah, but "saturation" as you're using it is arbitrary, based on HSL or HSV, which are extremely simple transformations of RGB space, which means that it's based on the (arbitrary) choice of R, G, and B primaries for your screen. HSL/HSV were developed because computer hardware of the 1970s couldn't do large lookup tables or complex math with any kind of reasonable performance. There is really no excuse for them to persist to 2009, and that they are still being baked into specs is something of a travesty. They're essentially pre-20th century color theory pseudo-science, resurrected because anything better was too computationally expensive 35 years ago.

What you're looking for (to be perceptually meaningful, based on real psychometric data) is something like the "chroma" of the Munsell Color System. See http://en.wikipedia.org/wiki/Munsell_color_system

While I agree that simple contrast accounts for much of the effect, I'm not sure this is totally bogus. I've heard similar advice in the pixel art community that suggests cooler shading and warmer highlights. I think it's the same concept at work.

If the goal is to show that, then the experiment should be set up carefully to equalize the value & chroma, and then test the hypothesis that such hue shifts make a difference. Any effect they have pales in comparison to the value/chroma effect. The explanation as given in this article is unscientific, and shouldn't be taken as serious advice, especially not by "color newbies" looking for wisdom from experts.

(The chromas have arbitrary relation to one another, the hue shifts are non-uniform and essentially arbitrary, and as I showed the values are exactly selected so to "prove" the author's claims. Setting up this experiment with attention to human color perception wouldn’t be that hard.)

I concur, there's no science here. Just a pretty good system for picking complimentary colors with nice contrast. "How to pick colors everyone likes" is overplaying the hand a bit.

Be careful. "Complementary" has a technical meaning in terms of color science.

Reply on news.ycombinator.com