162 PROCEEDINGS OF THE ACADEMY OF [1895. 



layer of retinal pigment. One, therefore, only obtains six to eight 

 sections in sectioning an entire eye; that is, three or four just as the 

 surface of the retina is reached by the knife, and again just as it 

 leaves off cutting through the retina at the opposite side of the eye- 

 ball. In other words, very superficial tangential sections only, or 

 such as are cut in a plane very nearly coincident with the bacillary 

 or Jacob's layer will show the regular repetition of structure that 

 impels me to make a comparison between the compound eyes of 

 Arthropods and the retinal structure of the larval salmon. 



The first thing that strikes the observer in examining such sections 

 is the extraordinary regularity with which the bacillary portions of 

 the retinal elements are arranged. One notices that the rods and 

 cones form together a pattern of the most astonishing regularity. 

 That the rods are in groups of four or five, but that the cones are 

 much larger across and always in pairs, or of the nature of Han- 

 nover's double cones and united together closely at their apices. 

 These double cones are arranged in rows in two ways at right 

 angles to each other, like the squares on a chess-board, Fig. 1. If 



Fig. 1. 



one now imagines the rows of double cones to represent the 

 black squares of a chess-board, the groups of rods, which are 

 much smaller in cross-section, may be regarded as occupying the 

 intervening red squares. This is the appearance of a section taken 

 at the outermost level. Pigment from the pigmented layer extends 

 in between the rods and cones, and the distinctness of the regular 

 retinal pattern is thus greatly heightened. In the next section, 

 only 2. 5 to 5 mikrons below the level at which the extreme regu- 

 larity of pattern just described appears, there is another pattern 

 formed by the rods and cones, which shows that the bodies of the 

 pairs have slightly separated from one another. The continuations 

 of the double cones of Hannover are now also seen to be less 



