They were present in all the frog species studied, except

Hyla moovei, and were distributed broadly all over the retina, ^ conforming to the overall density distribution of the ganglion cells.

Thus there were more in the visual streak area than elsewhere. At equivalent retinal locations, these cells had a larger

dendritic tree, and in most cases a larger soma area, than the a^b cells described below. The soma size and dendritic tree size usually

increased away from the visual streak, and within the streak from centre * to periphery. The soma shape varied mostly according to the orientation pattern of the primary dendrites.

Each cell had from two to four thick primary dendrites. Their calibre, and the dendritic calibre in general, was greater than in the (%ab cells. Occasionally, there were one or two additional thin and short primary dendrites. In some cases, an elongated extension of the soma gave rise to more than one primary dendrite.

After arising from the soma, the primary dendrites ran a steep course through the IPL and branched somewhat away from the soma to arborise in the outer part of the IPL. The tapering of the primary and other dendrites was gradual rather than abrupt. Webbing was frequently present at branch points. The branching pattern was sparse as compared with the Qab cells.

The shape of the dendritic tree varied from circular to more or less triangular, and was often very irregular. This shape reflected the bias and overall orientation of the tree. Thus, both symmetrical and asymmetrical cells were found among the cells. The cells at the retinal margin looked almost bipolar, with very elongated trees oriented parallel to the margin.

The Qa cells arborised m os tly in the outer/deeper part of the IPL (sublamina a of Famiglietti and Kolb, 1976). The depth variation between the dendrites was not usually gre at.

The axon originated either from the soma or from a primary dendrite. In some cases it arose from an elongated extension of the soma along with one or more primary dendrites. It was thick beyond the proximal segment.

3:3.2 The 0^,^ cells in general

These were the large ganglion cells that arborised in both sublaminae a and b (see Introduction, Section 1:2.4). These large cells were present in all five of the frog species studied. They were distributed across the whole retina, being denser in the visual streak area and sparser peripherally. The a^b cells were the most numerous of the three a-cell types in each of the four species in which all three types were studied.

The Q^b cells had a smaller dendritic tree and, in most cases, a smaller soma than the cells at equivalent retinal locations. Both

these sizes increased with distance from the visual streak, and within the streak, from centre to periphery. The soma shape was less irregular than that of the cells, still varying according to the orientation of the primary dendrites.

Usually, two to four thick primary dendrites emerged from each cell. The dendritic calibre in general was less than in the cells. However, in cases where a cell had one primary dendrite giving rise to a very large arbor with the others having small arbors, the former usually had a larger calibre than in the cells. Occasional thinner and

shorter primary dendrites also occurred.

The primary dendrites ran a less steep course in the IPL than the cell primary dendrites.- They usually branched closer to the cell soma than in the case of the or Qq cells. The primaries as well as the other dendrites tapered more rapidly than in the cells. The dendritic tree was more profusely branched than the or tree. At the branch points, the dendrites frequently showed webbing.

By definition, the dendrites arborised in both sublaminae a and b. In Rana esculenta and Rana pipiens, two strictly segregated sets of dendrites, the outer and inner subtrees, were found in the two

sublaminae. In the only Bufo mavinus retina available, the existence of two subtrees was also apparent. However, dendritic terminations being not very well-labelled, no certain comment can be made about any strict segregation. In Xenopus laevis and Eyla moorei, however, the sublaminar dendritic distribution of the cells was comparatively diffuse (see Sections 3*5*2; 3*9*2).

The participation of the primary dendrites in forming the two subtrees varied. The arbors of some primary dendrites ramified almost exclusively in one sublamina. In the others, either substantial

contribution was made to each sublamina or arborisation was made in a diffuse fashion.

Considerable differences were found between the outer and inner subtrees. In the species where these subtrees were well segregated

{Rana esculenta, Rana pipiens and Bufo marinus) the outer was usually larger and more elaborate than the inner. The shapes of the two subtrees frequently differed from each other, and the outer dendrites

appeared thicker than the inner ones. However, in many cases, poorer labelling of the inner subtree could have accounted for these

differences.

Both symmetrical and asymmetrical types of o^b cells were found, the two subtrees did not always match in their orientation or bias, if any. In some cases, the bias in one subtree was directly opposite to that in the other (see Fig. 55)*

The axon arose either from the soma or from a primary dendrite. Sometimes an elongated extension of the soma gave rise to the axon and one or more primary dendrites. The axon was thick beyond the proximal segment.

The two different types of cell (al^b and found in Eyla moorei are discussed in Section 3*9*2.

3:3.3 The a,, cells in general

These cells were defined as the large ganglion cells arborising mainly in the vitreal third (sublamina c) of the IPL (see Introduction, Section 1:2.4). They were found in all five of the species studied.

The best-labelled Qq cells were found to have a very large

dendritic tree. They had a large soma, sometimes around the same size as, or even larger than, the cell somata. Each cell had three or four thick primary dendrites that tapered gradually and arborised very sparsely. As the primary dendrites spread in a layer close to the GCL, these cells had the most 'spread-out' soma shape among the three large cell types. The primary dendrites were usually thicker and branched at a greater distance from the soma than the a^b cell primaries.

The dendritic tree was usually more irregular than that of the and Qab cells. However, the cells in the Bufo retina differed from the same type in Xenopus and Rana retinae in terms of primary dendritic thickness and branching and in tree shape (see Section 3:8*2).

From highly asymmetrical to moderately symmetrical trees were found in the Qq cells, with variations in orientation pattern as well. Most of the dendritic tree remained in the vitreal third of the IPL, very close to the ganglion cell layer, giving the cell a look like a

’c r eeper’ growing up a wall. Some parts of it were somewhat deeper, co- stratifying with the inner dendrites of the a^b cells.

The axons of the cells arose from the soma or a primary dendrite and became thick beyond the proximal segment.

Although some well-labelled cells were found in more than one species, most of the cells were poorly labelled. Many more of them must have been left unidentified for type because of inadequate

labelling. In Xenopus, this inadequacy created a particular problem as discussed in Section 3:5.2.

In terms of labelling, the cells were in general the most variable, and often the poorest, of the three large cell types.

Probably limited penetration of the developer during retinal processing has something to do with this phenomenon. On the vitreal side of the U retina, the developer had to pass through the nerve fibre layer and, often, through remnants of the vitreous membrane as well. The scleral side, on the other hand, might have been more easily penetrable because of the comparatively loose arrangement of cells and neuropil within the various laminae. Several lines of evidence are in favour of this

suggestion. The outer dendrites were almost consistently better

delineated than the inner ones in other ganglion cells; and many cells with quite well-labelled dendrites showed very poor somal labelling in the glucose oxidase-DAB-nickel-processed retinae, where penetration of the large glucose oxidase molecule was potentially a limiting factor. A well-stained receptor layer and the virtual absence of well-labelled ganglion cells in central retina, where the nerve fibre layer is thickest, also reinforce the idea.

3:3.4 Displaced large ganglion cells in general

Some large cells had their somata displaced into the INL or close to it into the IPL. They were present in every species studied. In most cases the dendrites arborised in the outer third of the IPL only, like the Qa cells (for example. Fig. 13B,C). In the other few cases they were like the a^b cells (Fig. 40C) in arborising in both sublaminae a and b. No displaced cells were found to arborise like the cells.

The displaced cells fitted well into the gaps in the mosaic of the equivalent types of ortho topic cell (Fig. l6) and did not tend to form pairs or clusters with them. This conformity in morphological and population criteria with the two established types of orthotopic cell led to the assignment of the displaced cells in the present study to either the or the a^b population rather than to separate types (see Discussion, Section 4:9.0).

3:3.5 An interesting type of cell

One exceptional type of ganglion cell was found in the present study, most conspicuously in Bufo mavinus, Rana esculenta and Rana pipiens.

Although it had a large dendritic tree in sublamina a, it could not be designated as an cell for reasons mentioned in the Discussion

(Section 4:5-2). Figure 56B shows some dendrites in sublamina a from such a cell.

3:4.0 SOME OTHER 'POINTS TO PONDER'