Unlike ordinary optical illusions, the McCollough effect alters your color perception for days or even weeks after you've finished adapting to it. The mechanism is still not understood.
By training your eye on a vivid color, then withdrawing the stimulus, and presenting an opposite or complementary color, you can experience physically impossible colors, like blue-yellow or red-green.
Oh no, I'm not looking at that again. It took a several days to get normal vision, and I was a bit worried it would never go away. I's astonishing how similar to a software glitch this is.
What a fantastic page. When I was teaching (music technology) in schools, I used to use optical illusions like these (and the illustration of the blind spot, which many of the kids hadn't seen before!) to illustrate that our senses are not calibrated, scientific instruments, but that we experience the world in a somewhat different manner to reality.
After seeing things like this, the pupils were a lot more open to the idea that their hearing may not be completely accurate, and that the harmonic series in a single note (which we hear all as one tone) could be perceived by another intelligent species as a series of discrete sounds. They also got how lossy audio encoding could throw away masked sounds because our hearing can also be fooled. It was always a great moment seeing them realise that a door had opened to a new understanding of how they navigated the world, and that their assumptions were incorrect (for some of them, of course!)
The harmonic series is never perfectly in phase in acoustic instruments. In many instruments it isn't even perfectly in tune, never mind perfectly in phase.
In fact it isn't always in phase in electronic instruments. As soon as you add a filter sweep the harmonics are going to have small time-varying phase shifts because the filter phase offsets change with frequency.
None of this alters that this is a perceptual issue. Timbral fusion - which is a form of perceptual modelling - means we hear closely related harmonic patterns as single sound objects coming from a single physical source. We use our models to identify sound-generating objects/animals/events around us.
That happens in brain firmware, possibly because it's evolutionarily optimal. But it's possible - at a stretch - to imagine evolutionary contexts in which timbral fusion wouldn't be a useful answer, and it would be more effective to hear individual overtones.
Or possibly both. Some trained musicians, piano technicians, organ makers, and so on can pick out individual overtones.
Thanks for the detailed response. Let me do my best to address some of your points and explain my reasoning.
> The harmonic series is never perfectly in phase in acoustic instruments.
This is a misunderstanding. Let's take a stringed instrument as our example. When a string vibrates at a certain frequency this is called the fundamental, which is the pitch we perceive. However, there are additional frequencies that we also hear called overtones. These come from perfect subdivisions oscillating within the same string. In other words, they are in phase, because they're literally in phase. If you pluck a string at first it has a lot of energy, then it slowly decays, the string wiggles and vibrates in and out of tune, and all the while every subdivision within that string is doing the exact same thing, in phase.
> In many instruments it isn't even perfectly in tune, never mind perfectly in phase.
I'm not 100% sure what you mean by this. The Harmonic series will be derived from a single string regardless of whether it's in tune with concert A or C, that won't matter, it works at any fundamental. I think what you mean is regarding multiple strings played very closely in phase (e.g. note on a piano). This will not work unless they're relatively in tune, otherwise you'll have harmonic dissonance.
Getting back to phasing and what we perceive as a single note vs what we perceive as multiple notes. Let's consider the reciprocal arrangement and imagine a chord that consists of the same notes as the harmonics that occur in a single string. (Let's also pretend for the sake of argument that the notes are strummed with the exact same strength that their harmonic counterparts are present and that they're perfectly in tune.) Since it's a chord, we would hear, by definition, multiple pitches, not a single fundamental. The reason for this is that each note is oscillating independently, I.e. out of phase.
> In fact it isn't always in phase in electronic instruments.
So, try this. Use audacity or whatever and generate pitches according to the harmonic series and tell me what you hear. I promise you, 100%, you will hear a single note. Since a computer is generating a sine wave, and the harmonics are by definition ratios of the fundamental, they will be perfectly in phase, and you will therefore hear a single note, try it.
> None of this alters that this is a perceptual issue.
No disagreements here, and I don't think anything I'm saying is mutually exclusive with this. If you take the above example and start shifting the phasing, at a certain point you will no longer perceive the other tones as harmonics but as pitches within a chord. Our brain however will stitch them together when they're close, that's the perception aspect. Another way to think about this is with pianos. If you take my chord example from above, and make a special piano with a single hammer that strikes multiple strings tuned according to the harmonic series independently, then we will perceive a single pitch, despite them not being perfectly in phase. They're not perfectly in phase because they're actually separate strings, but our perception will stitch them together, like you said.
> Some trained musicians, piano technicians, organ makers, and so on can pick out individual overtones.
I'm not really sure what the point you're making here, if you care to clarify. We all hear individual overtones, that's called timbre. I guess people who are trained can identify them (octave, 5th, etc...), but we all have a sense of what's a bright vs dark instrument.
Edited to fix formatting issues that came from a little ascii art thing I tried to draw showing how overtones are perfect subdivisions of an oscillating string.
It's missing one of my favorite mind blowing color illusions - the "checker shadow illusion": https://en.wikipedia.org/wiki/Checker_shadow_illusion
The square A and B are exactly the same color, even though you would be 100% sure they are not. I used to print it out and bring it to school and everyone would argue with me that they are different color. I would also have a cut-out mask to cover the rest and prove they are not :-).
My senior research thesis in undergrad was based on this illusion, using it to render basic heightmaps in a constant top-down orientation. This made the technique usable on PDAs and Web games at a time when even 2D isometric rendering was too slow/battery intensive.
Not the case. The base site is in simple HTML, the illusions are made with the Cappuccino framework and Michael Bach (the creator) is a frequent participant on the Cappuccino mailing list as he creates these illusions.
https://michaelbach.de/ot/col-McCollough/index.html
Unlike ordinary optical illusions, the McCollough effect alters your color perception for days or even weeks after you've finished adapting to it. The mechanism is still not understood.