Professor Donald Macleod
- I try to understand the process of human vision in
physiological or mechanistic terms, using the tools of psychophysics in
conjunction with electrophysiological and anatomical data from animals.
This involves tracing the sequence of operations that occurs as
information flows from retina to brain.
- One representative project asks:
Why isn't vision perfect? Bypassing optical losses by using
interference fringe patterns directly generated on the retina as stimuli,
we have shown that considerable information about the finest details may
survive in the retinal image but be lost in neural processing, and that
all of this neural loss occurs later than the primary
sensitivity-regulating processes of cone vision (which must therefore be
strictly local--either internal to the cones or fed by single cones).
Most recently and most surprisingly, we find that unresolvably
fine patterns can activate primary visual cortex and there produce
pattern-specific aftereffects (such as tilt aftereffects or
orientation-selective losses of visual contrast sensitivity), even though
the subject can not discriminate their orientation. It follows that
activation of single orientation-selective neurons in visual cortex is
not a sufficient condition for perception of orientation, and that our
stimuli are penetrating the visual system as far as primary visual cortex
(the region of cortex thought to be most
critical for the perception of detail), yet fail to penetrate to
conscious experience. We hope that this can be confirmed by MRI
experiments using these laser stimuli.
- In another line of work, the neural coding of
color and luminance is being investigated, both absolutely and in its
dependence on context, with attention to known physiological
nonlinearities. We are trying to characterize quantitatively the
nonlinearities in the neural representation of color, and relate them on
the one hand to mathematically optimal solutions to the problem of
representing colors with the sort of distribution that is environmentally
typical, and on the other hand to color
difference data through a neurally constrained
form of multi-dimensional scaling.
Back to Donald Macleod Home