Is There Such a Thing as Color Harmony?

Some colors seem to go together; others do not. The same goes, even more strongly, for notes. Some form pleasant harmonies, and others are dissonant. In the case of sound, it's more mathematically obvious what's different: harmonious combinations of sound frequencies (like a major third) are constructively interferent, and dissonant combinations are destructively interferent. Light waves have frequencies as well, though we usually think of them in terms of wavelengths.

Another interesting difference in our perception of sound is that our nervous system automatically does a Fourier transform for us, before the sound reaches conscious awareness. This is why you can listen to a chord, and hear the individual notes, instead of a mess of superimposed frequencies. That doesn't happen when we're seeing multiple colors - others, there would be no such thing as non-spectral colors like pink, and your screen couldn't fool you into thinking you're seeing more than the three frequencies it's actually sending to your eyes. There are also biological reasons that certain colors are more important to us just by themselves, without context (like red, which could be fruit or blood), as opposed to certain isolated frequencies of sound (what cause would our ancestors have had to be really attuned to a sound at 640 Hz?) Still, I've always wondered if colors "going together", or pleasant mixtures of pigments used in famous paintings, are actually making some kind of harmony, and we're attracted to those harmonies without realizing.



Above: what comes into your ear. Below: what your brain receives, post-transform. This is a major chord.



I've considered doing a proper color harmony experiment. As I just said, computer monitors don't allow this - I can't just make images online and ask people which they find more pleasing, because computer monitors just emit red, green and blue, and we're not actually looking at those pure light frequencies. I would have to buy specific LEDs and have people do the test in person. The expense of this hardware exceeds my interest in the result; and, if you think somehow subconciously we can tell the difference between the fake RGB hue, and the perceived hue, you're already agreeing that we are actually doing a Fourier transform with light perception.

So what I did, using the fake color imitations coming out of our screens, was to look at a simple artwork with just a few colors, and alter them so that they are making harmonious or dissonant chords. Below is Piet Mondrian's Abstact Cubes. Since he uses three colors, that means we can make a chord.



My approach was as follows.

• I found a website that would convert a wavelength into a perceived color (here.)
• I took the red spaces as the root note (e.g., the C in a C chord.) I chose the red spaces as the root of the chord because red does tend to be dominant, analogous to the way the root note establishes the basis of the chord, and because red is the lowest frequency. (Why didn't I choose tones for the white and black? While we're stretching the analogy, let's say that's black and white are the spatial equivalent or rhythm, an atonal drum beat.)
• I started with "C" (arbitrarily) at 780nm, which means the high C octave will be at 390nm. (For musicians this is counterintuitive. These are wavelengths, and as we go up the scale, the frequency gets higher but the wavelength gets smaller.) Also notice that humans can just barely see one visual octave.
• I used equal temperament, calculating the frequency as frequency = root frequency * 2^((number of half steps up)/12.)
• My hypothesis, which I expect will be falsified: individuals will display the a consistent preference for each color harmony (i.e., will consistently like major, minor, or discordant best.)
• As a side note: to highlight the likely inaccuracy of the colors (remember, we're looking at RGB monitors - you're not really looking at real shade) - I use two monitors when I work and even between the monitors, the colors are noticeably different. I hoped that the ratios of the wavelengths would remain the same, but on one of my monitors there was a difference between colors that looked the same on the other. Further complicating things, our color vision is not equally sensitive across the spectrum - it's much easier to discrimate a 1nm difference in the middle than at the edge of the frequency band we can see.

Above is the "scale" that results from this. Below are chord compositions based on the arbitrary C at 780nm. Major and minor are obvious enough; the discordant one is the root, a diminished second, and a diminished fifth. Do any of these compositions look more or less pleasing to you? And is this in accord with the harmony (that is, is major more pleasing than minor, minor more pleasing than the discordant one?) For reference, below each composition is each respective chord, made from the Szynalski tone generator. On the sounds, click to pop out if you want to hear them.


To avoid any special effect from these particular colors, I did another set of three (major, minor, and discordant) starting on the arbitrary E. If there's a harmony effect, the order of preference should be the same for both sets.

(Here I kind of like the discordant one, but I'm also a metalhead who likes diminished fifths.)

Maybe Mondrian knew what he was doing when he chose the root note for his composition. What would happen if I started with that one as the root, that is, normalized his red to the C below A440, and built chords from that? Here is the Mondrian color scale:



Here are the three harmonies, with the root note as his red, and the three same chords. Which one do you like most? Also, converting colors back into sounds - here instead of a piano I used an online tone generator to compare the major, minor, and discordant chords to the chord made by the actual colors in Mondrian's painting. The reason I used a tone generator instead of a piano is that, going in this direction (from light to sound) the color frequencies are likely to land between the notes,, and I wasn't about to de-tune my piano.
Interestingly, of these, the major chord is the one that most resembles the original work, though the original chord (the harmony from the colors of the original work, normalizing the red to the C below A440) is not a major one:



I hope you've escaped this without developing color-sound synesthesia. For next steps I may put up these harmonies (blinded) for a vote to see if people choose them consistently. If you want more, at some point I'm going to automate the pixel-counting of some famous paintings and assign wavelength values to each pixel (the Bridge at Giverny, the Scream, Starry Night, School of Athens) and see what kinds of chords those make.