WE experience the visual world as being three-dimensional. A major sou
rce of depth information derives from the slightly different views of
each eye, leading to small variations in the retinal images ('disparit
ies'). Neurons sensitive to visual disparities are thought to form the
neural basis of stereo vision1-10. Barn owls2,3 as well as several ma
mmalian species1,4-10 have neurons that are sensitive to visual dispar
ities. But how visual disparities are represented in the brain has bee
n a matter of discussion ever since the first disparity-sensitive neur
ons were found some 25 years ago. Here we adopt a new approach to this
problem and study the neural computation of visual disparities with a
paradigm borrowed from auditory research. The measurement of interaur
al time difference (ITD) has many similarities with the measurement of
visual disparity on the formal, algorithmic level. We speculate that
the similarities might extend to the level of neural computation. The
neural representation of ITD is well understood11-18, and we have stud
ied the representation of disparities with visual stimuli analogous to
those successfully used in acoustic experiments. For example, ITD is
converted in the brain to a pathlength on an axon that, owing to the f
inite conduction velocity in neurons, exactly matches the external ITD
. This pathlength is called 'characteristic delay,12. Our results sugg
est that there is an analogue of the characteristic delay in stereo vi
sion which we propose to call 'characteristic disparity'.