we assumed, correctly as it turned out, that the overall traffic of impulses in the two optic nerves must have been normal.
How, then, could the strabismus have produced such a radical change in cortical function? To answer this we need to consider how the two eyes normally act together. What the strabismus had changed was the relationship between the stimuli to the two eyes. When we look at a scene, the images in the two retinas from any point in the scene normally fall on locations that are the same distance and in the same direction from the two foveas—they fall on corresponding points. If a binocular cell in the cortex happens to be activated when an image falls on the left retina—if the cell's receptive field is crossed by a dark-light contour whose orientation is exactly right for the cell—then that cell will also be excited by the image on the right retina, for three reasons: (1)
the images fall on the same parts of the two retinas, (2) a binocular cell (unless it is specialized for depth) has its receptive fields in exactly the same parts of the two retinas, and (3) the orientation preferences of binocular cells are always the same in the two eyes. If the eyes are not parallel, reason 1 obviously no longer applies: with the images no longer in concordance, if at a given moment a cell happens to be told to fire by one eye, whether the other eye will also be telling the cell to fire is a matter of chance. This, as far as a single cell is concerned, would seem to be the only thing that changes in strabismus. Somehow, in a young kitten, the perpetuation over weeks or months of this state of affairs, in which the signals from the two eyes are no longer concordant, causes the weaker of the two sets of connections to the cell to weaken even further and often for practical purposes to disappear. Thus we have an example of ill effects coming not as a result of removing or withholding a stimulus, but merely as a result of disrupting the normal time relationships between two sets of stimuli—a subtle insult indeed, considering the gravity of the consequences.
In these experiments, monkeys gave the same results as kittens; it therefore seems likely that strabismus leads to the same consequences in humans. Clinically, in someone with a long-standing alternating strabismus, even if the strabismus is repaired, the person does not usually regain the ability to see depth. The surgeon can bring the two eyes into alignment only to the nearest few degrees. Perhaps the failure to recover is due to the loss of the person's ability to make up the residual deficit, to fuse the two images perfectly by bringing the eyes into alignment to the nearest few minutes of arc. Surgically repairing the strabismus aligns the eyes well enough so that in a normal person the neural mechanisms would be sufficient to take care of the remaining few degrees of fine adjustment, but in a strabismic person these are the very mechanisms, including binocular cells in the cortex, that have been disrupted. To get recovery would presumably require protracted reestablishment of perfect alignment in the two eyes, something that requires normal muscle alignment plus an alignment depending on binocular vision.
This model for explaining a cell's shift in ocular dominance is strongly reminiscent of a synaptic-level model for explaining associative learning. Known as the Hebb synapse model, after psychologist Donald Hebb of McGill University, its essential idea is that a synapse between two neurons, A and C, will become more effective the more often an incoming signal in nerve A is followed by an impulse in nerve C, regardless of exactly why nerve C fires (see the illustration on this page). Thus for the synapse to improve, nerve C need not fire because A fired. Suppose, for example, that a second nerve, B, makes a synapse with C, and the A-to-C synapse is weak and the B-to-C synapse is strong; suppose further that A and B fire at about the same time or that B fires just slightly ahead of A and that C then fires not because of the effects of A but because of the strong effects ofB. In a Hebb synapse, the mere fact that C fires immediately after A makes the A-to-C synapse stronger. We also suppose that if impulses coming in via path A are not followed by impulses in C, the A-to-C synapse becomes weaker.
To apply this model to binocular convergence in the normal animal, we let cell C be binocular, nerve A be from the nondominant eye, and nerve B be from the dominant eye. The nondominant eye is less likely than the dominant eye to fire the cell. The Hebb hypothesis says that the synapse between nerves A and C will be maintained or strengthened as long as an impulse in A is followed by an impulse in C, an event that is more likely to occur if help consistently comes from the other eye, nerve B, at the right time. And that, in turn, will happen if the eyes are aligned. If activity in A is not followed by activity in C, in the long run the synapse between A and C will be weakened.
It may not be easy to get direct proof that the Hebb synapse model applies to strabismus, at least not in the near future, but the idea seems attractive.
THE ANATOMICAL CONSEQUENCES OF DEPRIVATION
Our failure to find any marked physiological defects in geniculate cells, where little or no opportunity exists for eye competition, seemed to uphold the idea that the effects of monocular eye closure reflected competition rather than disuse. To be sure, the geniculate cells were histologically atrophic, but—so we rationalized—one could not expect everything to fit. If competition was indeed the important thing, it seemed that cortical layer 4C might provide a good place to test the idea, for here, too, the cells were monocular and competition was therefore unlikely, so that the alternating left-eye, righteye stripes should be undisturbed. Thus by recording in long microelectrode tracks through layer 4C, we set out to learn whether the patches still existed after monocular closure and were of normal size. It soon became obvious that 4C was still subdivided into left-eye and right-eye regions, as it is in normal animals, and that the cells in the stripes connected to the eye that had been closed were roughly normal. But the sequences of cells dominated by the closed eye were very brief, as if the stripes were abnormally narrow, around 0.2 millimeter instead of 0.4 or 0.5 millimeter. The stripes belonging to the open eye seemed correspondingly wider.
As soon as it became available, we used the anatomical technique of eye injection and transneuronal transport to obtain direct and vivid confirmation of this result. Following a few months' deprivation in a cat or monkey, we injected the good eye or the bad eye with radioactive amino acid. The autoradiographs showed a marked shrinkage of deprived-eye stripes and a corresponding expansion of the stripes belonging to the good eye. The lefthand photograph on this page shows the result of injecting the good eye with radioactive amino acid. The picture, taken, as usual, with dark-field illumination, shows a section cut parallel to the surface and passing through layer 4C. The narrow, pinched-off black stripes correspond to the eye that was closed: the wider light (labeled) stripes, to the open (injected) eye. The converse picture, in which the eye that had been closed was injected, is shown in the photograph on the next page. This section happens to be cut transverse to layer 4C, so we see the patches end-on.
These results in layer 4C tended to reinforce our doubts about the competition model, doubts that lingered on because of the geniculate-cell shrinkage:
Cell C receives inputs from A, a left-eye cell, and B, a right-eye cell. The Hebb synapse model says that if cell C fires after cell A fires, the sequence of events will tend to strengthen the A-to-C synapse.
Below: We obtained these sections from a macaque monkey that had an eye sutured closed from birth for eighteen months. The left (open) eye was then injected with radioactive amino acid, and after a week the brain was sectioned parallel to the surface of the visual cortex. (The cortex is dome shaped, so that cuts parallel to the surface are initially tangential, but then produce rings like onion rings, of progressively larger diameter. In the picture on the right, these have been cut from photographs and pasted together. We have since learned to flatten the cortex before freezing it, avoiding the cutting and pasting of serial sections.) In an ordinary photograph of a microscopic section the silver grains are black on a white background. Here we used dark-field microscopy, in which the silver grains scatter light and show as bright regions. The bright stripes, representing label in layer 4C from the open, injected eye, are widened, the dark ones (closed eye), are greatly narrowed.