Eye, Brain, and Vision

One obvious way to decide between these alternatives is to address the
question head on and simply record from a newborn cat or monkey. If learn-
ing is necessary for the wiring up to occur, then we should fail to find any of
the rich specificity that we see in adult animals. A lack of specificity would
nevertheless not decide the issue, because we could then ascribe the lack of
connections either to immaturity—still-incomplete genetically programmed
wiring—or to lack of experience. On the other hand, finding such specificity
would argue against a learning mechanism. We did not expect the experiments
with kittens to be easy, and they weren't. Kittens arc visually very immature at
birth and make no use at all of their eyes before about the tenth day, when the
eyes open. At that time, even the media of the eye, the transparent substances
between the cornea and the retina, are far from clear, so that it is impossible to
get a clear image on the retina. The immature visual cortex indeed responded
sluggishly and somewhat unpredictably and was on the whole a far cry from a
normal adult visual cortex; nevertheless we found many clearly orientation-
specific cells. The more days that elapsed between birth and recording, the
more the cells behaved like adult cells: perhaps because the media were clearer
and the animal more robust but perhaps because learning had occurred. Inter-
pretations differed from one set of observers to another.
The most convincing evidence came from newborn monkeys. The day after
it is born, a macaque monkey is visually remarkably mature: unlike a newborn
cat or human, it looks, follows objects, and takes a keen interest in its sur-
roundings. Consistent with this behavior, the cells in the neonate monkey's
primary visual cortex seemed about as sharply orientation-tuned as in the
adult. The cells even showed precise, orderly sequences of orientation shifts
(see the graph on this page). We did see differences between newborn and adult
animals, but the system of receptive-field orientation, the hallmark of striate
cortical function, seemed to be well organized.
Compared with that of the newborn cat or human, the newborn macaque
monkey's visual system may be mature, but it certainly differs anatomically
from the visual system of the adult monkey. A Nissl-stained section of cortex
looks different: the layers are thinner and the cells packed closer. As Simon
LeVay first observed, even the total area of the striate cortex expands by about
30 percent between birth and adulthood. If we stain the cortex by the Golgi
method or examine it under an electron microscope, the differences are even
more obvious: cells typically have a sparser dendritic tree and fewer synapses.
Given these differences, we would be surprised if the cortex at birth behaved
exactly as it does in an adult. On the other hand, dendrites and synapses are
still sparser and fewer a month before birth. The nature-nurture question is
whether postnatal development depends on experience or goes on even after
birth according to a built-in program. We still are not sure of the answer, but
from the relative normality of responses at birth, we can conclude that the
unresponsiveness of cortical cells after deprivation was mainly due to a deteri-
oration of connections that had been present at birth, not to a failure to form
because of lack of experience.
The second major question had to do with the cause of this deterioration. At
first glance, the answer seemed almost obvious. We supposed that the deterio-
ration came about through disuse, just as leg muscles atrophy if the knee or
ankle is immobilized in a cast. The geniculate-cell shrinkage was presumably
closely related to postsynaptic atrophy, the cell shrinkage seen in the lateral
geniculates of adult animals or humans after an eye is removed. It turned out
that these assumptions were wrong. The assumptions had seemed so self-
evident that I'm not sure we ever would have thought of designing an experi-
ment to test them. We were forced to change our minds only because we did
what seemed to us at the time an unnecessary experiment, for reasons that I
forget.

We sutured closed both eyes, first in a newborn cat and later in a newborn
monkey. If the cortical unresponsiveness in the path from one eye arose from
disuse, sewing up both eyes should give double the defect: we should find
virtually no cells that responded to the left or to the right eye. To our great
surprise, the result was anything but unresponsive cells: we found a cortex in
which fully half the cells responded normally, one quarter responded abnor-
mally, and one quarter did not respond at all. We had to conclude that you
cannot predict the fate of a cortical cell when an eye is closed unless you are
told whether the other eye has been closed too. Close one eye, and the cell is
almost certain to lose its connections from that eye; close both, and the chances
are good that the control will be preserved. Evidently we were dealing not
with disuse, but with some kind of eye competition. It was as if a cell began by
having two sets ofsynaptic inputs, one from each eye, and with one pathway
not used, the other took over, preempting the territory of the first pathway, as
shown in the drawing below.

Such reasoning, we thought, could hardly apply to the geniculate shrinkage
because geniculate cells are monocular, with no obvious opportunities for
competition. For the time being we could not explain the cell shrinkage in the
layers corresponding to the closed eye. With binocular closure, the shrinkage
of geniculate cells seemed less conspicuous, but it was hard to be sure because
we had no normal layers to use as a standard of comparison. Our understand-
ing of this whole problem did not move ahead until we began to use some of
the new methods of experimental anatomy.