1. With enough practice,
you almost completely stop using Color Working Memory. Your brain simply stores all the colors used in the app as objects in
Long-Term Memory, and when you see a color during the game, you simply remember—let's call it a
"link to object in long-term memory"
2. With even more practice, your brain goes even further: since the list of colors is small, number of double color combinations is also small, therefore
your brain begins to remember combinations (for example,
red-
blue,
blue-
red, or
yellow-
green)
as a single object because you've seen them so many times. And at some point, this scales even further. So instead of train Color Working Memory you just make chunks bigger (this could be very useful if only you could apply this chunking skill to other colors that you've never practiced with, but you can't)
I think n-back apps should have option in settings that enables random colors usage. So problem will disappear.
Another interesting thing that can be done is to combine the color n-back and color-discrimation task (correlation with g ~ 0.2-0.4) in the following way: the minimum distance between colors in RGB 3D space can be gradually reduced, so at a certain point of training you will learn to effectively remember and distinguish very close shades that are very easy to confuse, for example 5 different shades of green