In conventional stereo photography, the depth clues we keep 
or lose are:

Keep
1) parallactic discrepancy
2) convergence (in a properly set up pair)
3) atmospheric haze
4) perspective/texture density gradient
5) stadiametric information  (familiar size of objects)
6) overlap (obscuration)
7) clues from shadows of familiar objects toward or away 
   from the viewer

Lose
1) motion parallax
2) accommodation (focus)

Ques:

What's texture density gradient and how does it help?

Jim C answers:

It's a generalization of linear perspective--if you have a 
surface covered with similar things of roughly similar size 
(gravel, say), their images get smaller with distance. And 
if they have a preferred orientation (I'm having trouble 
thinking of a good real-life example) then changes in the 
amount of foreshortening tell you about changes in the slant 
of the surface they're on.  To take a silly example, if you 
had a perfectly flat ground plane covered with regularly-
spaced pizza pies, then the images of the farther pizzas 
would be smaller & more elliptical.  The images of the 
farther pizzas would also be angularly closer together 
(denser).