Quote:
Yeah, that's automatic constancy, which I actually think is an issue with fast-switching, time-aligned ABX tests. My assertion is - even if it isn't - those variables increase the risk either way, yet of course this is just a "theory" and I haven't "proved it", so I'd get banned from hydrogenaudio for questioning their templar of servitude.
Another way to think about it, continuing my color analogy:
Green is the label we give to a particular range of nanometer-length lightwaves.
If a given surface or cast light is in that range, it's green.
However, we don't experience green -- we experience
greenness -- something either looks green or it doesn't. We do not experience greenness in isolation, because the greenness is in a context of other colors (or absence of colors). And so we do not evaluate greenness on its own, but only in context. And we are experiencing the color-ness of the other colors in the same way we are experiencing the greenness of the thing among them.
If everything was exactly green, green would have no meaning: Our perception of green comes in part from the contrasts and complements of green.
As a consequence, it is easy to fool our eyes. Something can look green --
emulate greenness -- when the context around that non-green thing can successfully exploit our assumptions about the non-green thing that appears green.
Objectively we can analyze that green-seeming thing, sniff and say "it's not green", and totter off to measure some reds. Or subjectively we can relish its peculiar greenness.
