534 The American Naturalist. [July, 
Although Petit’s system is by no means perfect, it is at least 
a step in the right direction. He bases it upon characters that 
have some physiological significance, while the other systems 
are wholly or in greater part based on merely accidental char- 
acters. A clue to the genetic relationships of Diatoms, as of 
other plants, will be most certainly found in their method of 
reproduction. The shape of the frustules, or their markings, 
will serve for specific, or in some cases for generic characters, 
but they have no significance that will warrant their use in the 
erection of higher groups. Absolute shape and size will not 
serve as definite characters, for a single species between one 
auxospore stage and the next varies greatly in both these 
respects. Owing to the peculiar mode of cell division in which 
each new valve is formed inside the old one, each new frustule 
is smaller than the parent, hence the size gradually decreases 
until an auxospore is formed. Schumann”, out of 470 species 
found ten in which the length of the largest was five times 
that of the smallest; twenty-nine in which the largest were 
from three to four times as long as the smallest, and the rest 
showing less variation. The variation in form is even as great 
as the variation in size. This is probably due to the difference 
in the thickness of the girdle, 7. e. the part of the valves that 
overlaps, in different parts of the frustule. Navicula iridis Ehr. 
is a good example of a variable species. Its different forms 
have been described as species by most writers. In the typical 
form the valves are elliptical with gracefully curved margins. 
The first variation from this type has apices cuneate, and a still 
further deviation shows them acuminate-cuneate ; and from this 
it varies to rostrate or capitate ; and a diminution in size goes 
step by step with this change in form. These forms are repre- 
sented by Navicula iridis Ehr., N. amphigomphus Ehr., N. affinis 
Ehr., N. amphirhynchus Ehr., and N. producta W. Sm. If the 
overlapping portions of the valves are slightly thicker near the 
ends than elsewhere, this variation would be the necessary 
result, for each new valve formed inside an old one would be 
slightly constricted opposite this thickened place, at first chang- 
ing the rounded ends to cuneate, and as the narrowing pro- 
12 Pfitzer, l. c., p. 441. 
