It has often been noted that with repeated exposure to random-dot ster
eograms the time required to perceived depth decreases. Further, with
extensive practice, stereoacuity thresholds have been shown to decreas
e. For both types of learning some researchers have reported specifici
ty of the improvements to retinal location, and have thus suggested th
at the learning may be localised at early levels of visual processing,
such as in primary visual cortex. However, these studies have not ade
quately ruled out the possibility that the specificity shown may be du
e to the operation of selective-spatial-attention mechanisms. In the p
resent study this possibility was examined by training observers to ju
dge the relative depth of a pair of stereograms presented equally ofte
n in two spatial locations, but stimuli were only presented with one d
irection of disparity (ie crossed or uncrossed) in any one location. R
esults indicated that, as expected, observers' judgments improved with
practice. However, this improvement transferred completely to stimuli
presented with the other direction of disparity in each location. Thu
s, it is argued that previous findings of retinal-location-specific im
provements in stereoacuity may well be due to selective-spatial-attent
ion mechanisms, rather than to learning localised at an early level of
visual processing.