When you look at a person, what you perceive is not a series of shapes and colors that correspond to what's there, but rather a bunch of hastily constructed symbols that convey the information that the brain thinks is important.

A way to emphasize this vividly (though it may trigger a physical brain fallacy) would be to see what video reconstructions from brain activity look like, especially when they are noisy and somewhat error prone so slippage between different cognitive alternatives is visible in the video. It would be interesting to try calibrated computer reconstruction of imagery on people who have had 0, 10, 100, 1000, and 10000 hours of training in artistic seeing. Maybe even the same people as they rode the learning curve? :-)