Comment by kazinator

You can estimate that if you can determine at which point the color becomes too ambiguous to call blue on one side, or green on the other. Different people will have a different range. If you want to identify a threshold, you can take the midpoint of the range.

Either of these approaches may be bad. The third paragraph of this page explains why:

https://en.wikipedia.org/wiki/Two-alternative_forced_choice

My suggested approach might not be much better though; it still relies on presenting a single stimulus.

It's not clear how the two-alternative forced choice can be used to find someone's blue-green threshold.

I think a better experiment would be to show the user gradients and ask them to move a bar to where they think is the midpoint in the blue-green transition. Subsequent gradients center on the user's previously identified midpoint, but zoom in more.

There is also this question: by which path do we interpolate from blue to green?

Let's imagine the CIELAB color space. Say that our pure green lies on the red-green axis, all the way on the green end. Blue lies on the extreme of blue-yellow. Do we interpolate through these linearly or what? And using what luminance value?

I suspect that for every given, fixed luminance value, the blue-green boundary is a contour. There are many paths we can take between blue and green, and along each path there is a boundary point. If we join those points we get this contour. Then if we do that for different luminance values, the contour becomes a 3D surface in the color space.