Eye, Brain, and Vision
Home
Book
Illusions
Biography
Publications
David Hubel's
Home
Book
Illusions
Biography
Publications
HMS
                                                                                    
       
   
 
The various parts of the two retinas project onto their own areas of the right lateral geniculate body of the cat (seen in cross section). The upper geniculate layer, which receives input from the opposite (left) eye, overhangs the next layer. The overhanging part receives its input from the temporal crescent, the part of the contralateral nasal retina subserving the outer (temporal) part of the visual field, which has no counterpart in the other eye. (The temporal part of the visual field extends out farther because the nasal retina extends in farther). In monocular closure (here, for example, a closure of the left eye), the overhanging part doesn't atrophy, presumably because it has no competition from the right eye.




either the competition hypothesis was wrong or something was faulty somewhere in our reasoning. It turned out that the reasoning was at fault for both the geniculate and the cortex. In the cortex, our mistake was in assuming that when we closed the eyes in newborn animals, the ocular-dominance columns were already well developed.



                             NORMAL DEVELOPMENT                           OF EYE-DOMINANCE COLUMNS

The obvious way to learn about ocular-dominance columns in the newborn was to check the distribution of fibers entering layer 4C by injecting an eye on the first or second day of life. The result was surprising. Instead of clear, crisp stripes, layer 4C showed a continuous smear of label. The lefthand autoradiograph on the next page shows 4C cut transversely, and we see no trace of columns. Only when we sliced the cortex parallel to its surface was it possible to see a faint ripple at half-millimeter intervals, as shown in the righthand autoradiograph. Evidently, fibers from the geniculate that grow into the cortex do not immediately go to and branch in separate left-eye and right-eye regions. They first send branches everywhere over a radius of a few millimeters, and only later, around the time of birth, do they retract and adopt their final distributions. The faint ripples in the newborn make it clear that the retraction has already begun before birth; in fact, by injecting the eyes of fetal monkeys (a difficult feat) Pasko Rakic has shown that it begins a few weeks before birth. By injecting one eye of monkeys at various ages after birth we could easily show that in the first two or three weeks a steady retraction of fiber terminals takes place in layer 4, so that by the fourth week the formation of the stripes is complete.

















We easily confirmed the idea of postnatal retraction of terminals by making records from layer 4C in monkeys soon after birth. As the electrode traveled along the layer parallel to the surface, we could evoke activity from the two eyes at all points along the electrode track, instead of the crisp eye-alternation seen in adults. Caria Shatz has shown that an analogous process of development occurs in the cat geniculate: in fetal cats, many geniculate cells temporarily receive input from both eyes, but they lose one of the inputs as the layering becomes established.
The final pattern of left-eye, right-eye alternation in cortical layer 4C develops normally even if both eyes are sewn shut, indicating that the appropriate wiring can come about in the absence of experience. We suppose that during development, the incoming fibers from the two eyes compete in layer 4C in such a way that if one eye has the upper hand at any one place, the eye's advantage, in terms of numbers of nerve terminals, tends to increase, and the losing eye's terminals correspondingly recede. Any slight initial imbalance thus tends to increase progressively until, at age one month, the final punchedout stripes result, with complete domination everywhere in layer 4. In the case of eye closure, the balance is changed, and at the borders of the stripes, where normally the outcome would be a close battle, the open eye is favored and wins out, as shown in the diagram on this page.
We don't know what causes the initial imbalance during normal development, but in this unstable equilibrium presumably even the slightest difference would set things off. Why the pattern that develops should be one of parallel stripes, each a half-millimeter wide, is a matter of speculation. An idea several people espouse is that axons from the same eye attract each other over a short range but that left-eye and right-eye axons repel each other with a force that at short distances is weaker than the attracting forces, so that attraction wins.
With increasing distance, the attracting force falls off more rapidly than the repelling force, so that farther away repulsion wins. The ranges of these competing tendencies determine the size of the columns. It seems from the mathematics that to get parallel stripes as opposed to a checkerboard or to islands of left-eye axons in a right-eye matrix, we need only specify that the boundaries between columns should be as short as possible.
One thus has a way of explaining the shrinkage and expansion of columns, by showing that at the time the eye was closed, early in life, competition was, after all, possible.
Ray Guillery, then at the University of Wisconsin, had meanwhile produced a plausible explanation for the atrophy of the geniculate cells. On examining our figures showing cell shrinkage in monocularly deprived cats, he noticed that in the part of the geniculate farthest out from the midline the shrinkage was much less; indeed, the cells there, in the temporal-crescent region, appeared to be normal. This region represents part of the visual field so far out to the side that only the eye on that side can see it, as shown in the diagram on

 

 

 

 

 





this page. We were distressed, to say the least; we had been so busy legitimizing our findings by measuring cell diameters that we had simply forgotten to look at our own pictures. This failure to atrophy of the cells in the geniculate receiving temporal-crescent projections suggested that the atrophy elsewhere in the geniculate might indeed be the result of competition and that out in the temporal crescent, where competition was absent, the deprived cells did not shrink.

   
 
This competition model explains the segregation of fourth-layer fibers into eyedominance columns. At birth the columns have already begun to form. Normally at any given point if one eye dominates even slightly, it ends up with a complete monopoly. If an eye is closed at birth, the fibers from the open eye still surviving at any given point in layer 4 take over completely. The only regions with persisting fibers from the closed eye are those where that eye had no competition when it was closed.






Previous Page
Next Page


Here, in a different monkey, the closed eye was injected. The section is transverse rather than tangential. The stripes in layer 4C, seen end on and appearing bright in this dark-field picture, are much shrunken.

 
 
 
 
 


Next Page
Previos Page
Left: This section shows layer 4C cut transversely, from a newborn macaque monkey with an eye injected. The picture is dark field, so the radioactive label is bright. Its continuity shows that the terminals from each eye are not aggregated into stripes but are intermingled throughout the layer. (The white stripe between the exposed and buried 4C layers is white matter, full of fibers loaded with label on their way up from the lateral geniculates.) Right: Here the other hemisphere is cut so that the knife grazes the buried part of the striate cortex. We tan now see hints of stripes in the upper part of 4C. (These stripes are in a subdivision related to the magnocellular geniculate layers. The deeper part, {3, forms a continuous ring around a and so presumably is later in segregating.)