visual systems are exquisitely sensitive to information about the shapes
of objects, including depth within an object, that is, which parts are
farther and which closer. Consider a piece of paper lying on a table.
You can easily tell which side is closest to you. How does your brain
know? In other words, what information is available to construct your
percept? This information comes in several forms, which are often called
"visual depth cues." One cue comes from stereoscopic vision. Slight differences
in the two eyes' retinal images are interpreted as depth by the brain.
Another cue comes from an unconscious assumption that the paper is a rectangle.
This cue is a variant of the "linear perspective" cue. (Try this: trim
a thin triangular wedge from one side of a rectangular piece of blank
paper. Put it on a table or surface that doesn't have much texture. Look
at it from a sharp angle [not from overhead] with just one eye. Looks
like a rectangle, doesn't it?) Another cue is height within the visual
field objects that are seen from above (things on tables and floors,
for example) are generally farther from you if they are higher in the
visual field. Now, notice that those last two cues almost always tell
you the correct answer as to what's near and what's far, for real objects
in the real world. They are what vision scientists call "statistically
valid." Because they almost always give correct information, it makes
sense for the brain to pay attention to them and use them during percept
construction. Just as it's statistically valid, but not logically necessary,
to conclude there was a power failure if the alarm clocks, microwave oven
display, and VCR are all blinking when you get home (it's unlikely, but
possible, that a cat jiggled all the plugs instead). Now, what if a clever
artist were to construct, artificially, a peculiar but very real object
in which the depth cues give false testimony? Your brain brain would be
perfectly right to conclude that the most likely object, the one it should
construct in the percept, is not the real object but the one suggested
by the cues.