The last question in this specific context was, whether our assumption regarding the oblique projection of the cells might have been an artifact caused by bad tissue handling. The next picture shows that also in tangential sections all stained bipolar cells run at exactly the same oblique angle through the inner retina.

Tangential section of the inner nuclear layer showing that all labeled bipolar cells are running in exactly the same direction. The input location is in the upper right direction and their output projection is towards the lower left.

However, there is a severe problem with all these assumptions. We have been looking at only one class of bipolar cell, whereas there are a total of 11 distinct classes. If those would behave differently, the entire model would fall to pieces. There is only one way around that which is to stain the other classes as well. Unfortunately, what today is an easy undertaking, was an almost hopeless task 15 years ago. Yet, fortune struck us and we came across a novel method developed by Sarthy and Detwiler (**) which relied on the idea that nerve cells, when stimulated, “get hungry” and try to fill up their reserves with anything floating around in their environment. So what these colleagues of mine had done was to add the non-toxic fluorescent dye “lucifer yellow” (originally used by the fashion industry, we even used it for some strange Disco effects but I won’t go into details here) to the bathing solution (Ringer’s solution) of the retina. The stimulation was provided by removing calcium from the Ringer which selectively stimulates bipolar cells that signal light ON.

We successfully repeated their experiment and were able to stain a whole new number of bipolar cells, all of which conformed to our model. A nice surprise was that, in addition to orthotopic bipolar cells (those that have their cell bodies within the inner nuclear layer), we were able to identify another class of cell which are not really bipolar in appearance but have their cell body among the photoreceptors and only one long process protruding into the inner retina. Suffice it to say that also these cells follow the general scheme.

Section of the turtle retina stained by lucifer yellow uptake upon stimulation of the ON-center bipolar cells. Note the high number of stained "displaced bipolar cells" in the upper tier.

As always, every new piece of data generates new questions. Could it be possible that we were staining the same cells but that the staining method would alter their appearance? This question was answered easily by applying both staining methods simultaneously. As shown below, there is no overlap between cells stained with lucifer yellow (tentatively classified as light ON cells) and those stained by the immunoreaction against serotonin (tentatively classified as light OFF, here stained with the red dye tetramethyl rhodamine isothiocyanate)

Retinal sections double-stained for both serotonin (in red) and lucifer yellow (yellow). In addition to bipolar cells, there are two Müller cells visible in the upper picture which are a specialized form of neuroglia (supportive cells). Those cells run perpendicularly through the retina, whereas the bipolar cells maintain their oblique course.

Finally, what good is all this for? Below is a plot which takes into account the relation between the size of the optical signal that hits the photoreceptors and the size of the neuronal signal that actually reaches the ganglion cells. Within the visual streak, the signal is magnified, simply by the architecture of the connecting bipolar neurons by a factor of approximately 400 %. Away from the streak, there is a 1:1 projection which is, however pushed outward by the over representation of the central retinal area. Thus, the only way of how this can be accomplished is to move the output location away from the center by means of an oblique projection. Since input and output must finally equal 100 percent, the periphery must be subject to a signal compression which is achieved by an inverted fan-like arrangement. We originally called this phenomenon a neuronal lens, since the functional consequences of the retinal architecture are comparable to what would be achieved if oned placed a complicated lens in front of it which magnifies in the center and compresses the image in the periphery.

What are the consequences of this neuronal architecture for vision? Wouldn’t it be far better to have the same resolution throughout the entire visual field? There is certainly something to this argument, however, in order to accommodate the signals that have to be transmitted to the brain, our optic nerves which are about 5 mm in diameter, would have to be approximately 10 cm in diameter. In addition, since about 0.5 % of the total retinal area occupy 30 % of the brain as the situation stands now, we would have to enlarge our brain by several orders of magnitude which have to be carried around.

Another interesting aspect of the way of how the retina is organized relates to motion perception. If one assumes that a spot of light travels at constant velocity across the photoreceptors, one only has to look at the distortion of the signal through the morphologic properties of the retinal neurons to see that the velocity of the output must follow entirely different rules. In other words, in the retinal periphery, the output signal travels much more slowly than the input but catches up in speed when it reaches the midperiphery. Still, the output lags behind. However, upon approaching the center of the visual field, the output signal accelerates and, at the center, actually gets ahead of the input. After passing the center, the neuronal image starts slowing down again which makes it easier to keep track of objects excaping the focus of attention

Psychophysically, it is very easy to prove that such a mechanism, indeed, exists also in humans. One just has to place oneself into a rotating drum painted with vertical stripes and keep the eyes steady. The sensation will be that the stripes enter the visual field at slow speed and then become faster and faster until they pass the center of vision. After that, they start to slow down again until the exit the visual field in the periphery. A simpler but not exactly scientific approach would be, on a train ride to look on a parallel track, keep the eyes steady and watch the rail bars as they approach slowly first, then accelerate until they become too fast for resolution and finally slow down again on the other side.
Why could anyone profit from such a mechanism? There is no animal which is not threatened by other predators (some of the worst being NYC cab drivers, no offense). So, if they intrude the visual field in the periphery, it is of great advantage if one sees them moving slowly first, because that gives one time to adjust and to react, that is, to look straight at them. Once they have the attention they deserve, they also deserve the highest motion sensitivity, even at the risk of losing track,. If that happens, they are on their way out of the visual field center, i.e. the output signal is slowing down which, again, makes it easier to capture them again.
How does the turtle retina compare to the human retina in this regard? In an excellent study, Heinz Wässle and collaborators showed a similar arrangement of bipolar cells in the monkey retina. They further counted all retinal ganglion cells and did a quantitative assessment of the magnification achieved by the neuronal architecture within the retina and came to the conclusion that, indeed, the so called cortical magnification factor is really a retinal magnification factor.