842 THE AMERICAN NATURALIST. : [Vou XXXII. 
While it is clearly an advantage to have stationary eyes 
directed towards the three planes of space, it is not so clear why 
the fibrils should be oriented to these planes. As chiten is said 
to be doubly refractive, it seemed probable that there was some 
relation between the presence in arthropod ocelli of chitenous 
polarizing lenses and parallel retinidial fibrils. But in the cases 
that I examined there was apparently no polarization in the 
principal axis of the lens, although I am not satisfied that 
renewed experiments may not give different results. In mak- 
ing the experiments I was strongly inclined to believe that in 
these ocelli the light was polarized in such a way that it would 
vibrate at right angles to the retinidial fibrils. 
(5) The Retinidial Fibrils are Arranged in accurately 
Graded Series according to their Length. —The visual rods, 
in all the cases studied, could be resolved into a series of simple 
or compound wedges, hence the retinidial fibrils ought to vary 
in length according to their position in these wedges. The dif- 
ference in length between the adjacent fibrils, although exces- 
sively small, should vary with the angle, and the number of 
fibrils in the series should vary with the altitude of the wedges. 
FART If. 
There is no reason to doubt that the structure of the visual 
apparatus in the higher vertebrates is essentially like that in 
the invertebrates, for we have succeeded in demonstrating that 
the retinophorz in fishes and amphibia are twin cells like those 
of molluscs and arthropods (Fig. 1-*). We can also demon- 
strate that the twin cone cells of amphibia are nothing more 
than extreme types of the same kind of cells (Fig. 1—), 
Furthermore, the presence of retinidial fibrils, like those of 
invertebrates, is clearly indicated by the well-known transverse 
cleavage of these rods and cones, but the fibrils themselves are 
probably too minute and too numerous to be clearly seen under 
any conditions we are as yet able to command. They are pre- 
sumably arranged in distinct layers, alternating with the clear 
matrix of the rods and cones. Without further discussion of 
this point here, let us assume that the retinophorz of verte- 
