<--back--      home      --next-->

So we conducted a user study to find out. We used the minimally distinct border method to compare the face-based method against, because it is easy to implement on a standard CRT.

This image is the minimally distinct border stimulus. Its formed by taking the double face image, flipping it upside down, and rearranging the connected components. This is more complicated that the usual minimally distinct border stimulus, but we felt it allowed a more controlled comparison with the face-based method:

• the length of the boundary is the same,
• the number and shape of pieces is the same,
• so even the amplitude spectrum is the same.

(Running demo)

And here's how it looks in action. You can probaby see the boundary distinctness increase as the color is either more or less luminant than the gray, and minimally so somewhere in the middle. But just looking at this, you can imagine the face-based method might be more precise, because there's this powerful directional cue, to point you towards the middle, based on which side you see the face on. That directional cue is lacking in the minimally distinct border task.