VISUAL ADAPTATION AND RETINAL GAIN CONTROLS
not-unreasonable number which must be checked
experimentally in the future.
Notice that, in this model, the dark noises and
the criterion fluxes for the turning on of the gain
control are distinctly different. This is in accord
with actual measurements on ganglion cells (Enroth-
Cugell and Shapley, 1973a; Barlow, 1977) which
show clearly that the dark light is too small to
account for the behavior of the gain vs background
relationship. A different fundamental constant, ten
to one hundred times larger than the dark noise
must be hypothesized.
The point of the model is revealed in Fig. 62
which shows predictions of the model for the
increment threshold curves of two types of cell, one
with a small center and one with a larger center.
The cells' sensitivities are plotted on the same
coordinates as psychophysical data from Aguilar
and Stiles (1954) and Barlow (1957) which exhibit
characteristic steep and shallow slopes, respectively.
Clearly, the model accounts adequately for the
transition from square root to Weber behavior
contingent on target size. It provides a way to
understand why this transition is a function of the
square of the spatial frequency of a sinusoidal target
(van Nes and Bouman, 1967). It also provides a
framework with which to understand why noise
sometimes controls sensitivity and why, under other
circumstances, the gain determines sensitivity.
7. R E T R O S P E C T I V E AND CONCLUSION
We have considered the " w h y " , the " w h a t " , and
the " h o w " of visual adaptation, and even attemp-
ted to say " w h e r e " and " w h e n " it takes place. At
the outset we demonstrated how automatic
adjustment of retinal gain would cause brightness
constancy under different conditions of
illumination, by making retinal responses dependent
on cqntrast. Then we examined what happens to
the visual performance of humans under different
conditions of background or mean level of
illumination. In this section, we pointed out that
" n o i s e " from the stimulus may sometimes limit
performance, and that at other times retinal
adaptation determines the limits of sensitivity. The
visual responses o f retinal ganglion cells,
particularly cat ganglion cells, as a function of mean
337
or background level, were considered next. In these
cells, gain control seems the dominant factor in
limiting their capacity to detect stimuli as the mean
level of illumination varies. The relation of this
work to human vision, where both gain and
" n o i s e " seem significant, is an open question for
future research. In seeking to answer the question,
" w h e r e " the gain control is located, we next
considered what is known about adaptation in
retinal interneurons. Then we reviewed the topic of
adaptation in photoreceptors. Finally, we discussed
some theories of how the gain of retinal neurons
is controlled, and how such a gain control might
contribute to visual performance.
There are several neural gain controls in the
retina. Surprisingly, what evidence we have about
the time course of gain reduction after an increase
in mean illumination, together with the change in
response time course with change in gain, suggests
that all these gain controls share the property of
attenuating the components of response to slow
variation of the stimulus, and sparing the
components of the response to rapid variations of
the stimulus light. However, the different gain
controls do seem to have different spatial
integration areas, and this allows one to distinguish
them.
Future work is required to establish the answer
to " w h e r e " in the retina some of the gain controls
live. However, it will be even more important to
discover or invent models for " h o w " the gain
controls manage to make the retina exceptionally
sensitive to contrast and yet insensitive to large
swings in the mean level.
APPENDIX 1 -- RETINAL NEURONS
The details of retinal circuitry vary from species
to species although many of the general principles,
at least in higher mammals, are the same. Much of
what we have written about the physiological
mechanisms of retinal adaptation stems from
experiments on the cat retina and any fruitful
consideration o f these mechanisms requires some
familiarity with retinal functional connectivity. The
brief comments about retinal circuitry which follow
draw heavily on the detailed studies of the cat retina
undertaken over the past ten years or so (for