320 THE AMERICAN MONTHLY [Dec 
been recently studied with the 4 mm. and a 5-6ths solid 
axial cone. The most difficult structural features have 
not been seen with a lesser cone, but I do not assert that 
they may not possibly be so resolved, although the re- 
sults of my observations have strongly inclined me to 
the belief that, with axial illumination, structure just 
within the capacity of the lens employed ean only be 
seen witha very large cone. It has appeared to me that 
closing down the cone, while greatly strengthening the 
contrast of the coarser, causes the finer detail to disappear 
altogether, and material)y reduces the separating power 
of the objective. With reference to this matter the fol- 
lowing experiment may prove interesting :—Arrange a 
Cherryfield Navicula rhomboides, mounted in a mixture 
of monobromide of naphthaline and balsam, under a good 
-semi-apochromatic }” of N. A. -77, and 27 ocular, so that 
the valve shall lie longitudinally along and on the sharply 
focussed edge of the lamp flame. With slightly under # 
cone the longitudinal striae will appear conspicuous 
throughout the entire length of the valve, while the closer 
transverse striae, although they may be seen to a certain 
extent, are far less satisfactorily defined, no thoroughly 
clear separation being apparent. Now replace the smaller 
by a 5-6ths cone. The coarse strongly-defined longi- 
tudinal striae disappear, and at the first glance all struc- 
ture may seem to have disappeared with them, but a little 
careful scrutiny will reveal the presence of a faint dotted 
resolution, the transverse divisions of which are as fully 
and cleanly shown as the longitudinal. 
I am aware that the results dealt with in this paper 
can not meet with general acceptance until they receive 
- confirmation at abler hands than mine, nor indeed would 
it be desirable that they should be so accepted, involving 
as they do important theoretical considerations, until in- 
dependent practical experience shall have placed their 
truth beyond doubt, 
