Re: weighting depth clues

[T3D Jim Crowell <crowell@xxxxxxxxxxxxxxx>]

john bercovitz wrote:

>The visual cliff experiments I saw used a piece of acrylic plastic
>(Plexiglas, Lucite, Perspex, whatever) joined to the edge of a table.
>The infant stopped where the acrylic met the table because he could
>see that the floor was a few feet below the acrylic.  Interesting that
>some pretty advanced stuff is hardwired.

Right, I think Eleanor Gibson did those.

I think most of these things sit on a continuum between purely learned and
purely hard-wired.  For example, in most infants stereo starts to kick in
pretty regularly around 5 months, but proper binocular experience is
necessary; babies with a strabismus (crooked eye) around this time may not
develop stereo.  Motion parallax is more primitive evolutionarily speaking
& much more robust, it may be that just about any visual experience is
sufficient to establish it.  I don't remember if Gibson tried to figure out
exactly which cues the babies were using (did she make the texture on the
ground bigger so that it would project to the same size on the retina?),
but I'd guess motion parallax was important.

>Comparing cars and faces, I had both in the lecture.  I shot with
>lenses that were correct for the audience's seating distance, 1/3
>of correct, and 3 times correct length.  The faces shot with too
>long or too short looked fairly normal in spite of the geometric
>reconstruction being _way_ out of whack.  The cars looked much less
>normal.  I think this goes along with the difficulty of seeing a
>pseudo face.  Perhaps the face is a hardwired model?

Most likely.  It appears there's an area of the brain specifically
dedicated to detecting faces...

-Jim C.


------------------------------