But suppose, as is perhaps 
more typical, something 
stimulates each eye in 
corresponding places. Let us say 
the left eye received the 
stimulus XXAXXBXX and the 
right eye received the stimulus 
XXAXXXBX. If the observer 
fixates A, there is the 
possibility of fusing the left 
eye’s B with the right eye’s X 
and the left eye’s X with the 
right eye’s B, since these fall 
on corresponding points. But 
that will not happen: The left 
eye’s A and B will be fused with 
the right eye’s A and B, because 
of the similarity between them. 
Apparently, the perceptual 
system scans the two images 
and decides, on the basis of 
similarity, which units in each 
most probably correspond, the 
implication being that those 
that correspond are produced by 
the same contours in the outer 
world. (David Marr and Tomaso 
Poggio developed an algorithm, 
based on certain 
"assumptions" about the outer 
scene, that would enable a 
computer or the brain to solve 
just such a correspondence 
problem.) Once this scanning 
takes place, the perceptual 
system can evaluate the 
disparity in terms of depth. If B 
is relatively near to A in the 
left eye’s image but is farther 
from A in the right eye’s image, 
then it follows that B is an 
object behind the plane of A. A 
process similar to reasoning 
must occur in arriving at the 
depth interpretation. If some 
agency of mind has available to 
it the sensory information 
reaching the two eyes, and if it 
"knows" in which eye each 
retinal stimulus originates, it 
can compute depth. A process of 
this kind would render 
understandable stereoscopic 
depth constancy. For example, 
if the perceptual system 
"knows" about diminished 
disparity as a function of 
distance, it can take distance 
into account in inferring the 
magnitude of depth.