SPERMATOGENESIS OF EUSCHISTUS 743 
of the spermatocytes, nor is a continuous spirem produced, and 
the univalent autosomes preserve their long axes throughout. 
Chromosomes most difficult of interpretation are ring-shaped 
ones, and fortunately Euschistus has none such in the maturation 
mitoses. Gemini of such form are, however, frequent in the pro- 
phases as shown in figs. 69, 70, 74, 75 and 82 to 84, even though 
these later invariably straighten out into rod form. In earlier 
papers I have shown that such a ring is a geminus in which the 
middle part has widened out while its ends have remained closed; 
other gemini differ in having the univalents separated at one end 
while apposed at the other. The inception of a ring is very simple. 
From such a stage as that of figs. 49 to 55 where every two uni- 
valents are wound around each other, is attained the condition 
shown in figs. 56 and 64 where the univalents begin to separate 
along their middle portions. The latter are incipient rings and 
they may remain such until a late prophase, or may change into 
V-shaped gemini by breaking their contacts at one end. Neither 
rings nor Vs are produced by longitudinal splitting. It is quite 
possible that certain gemini may be regularly ring-shaped and 
other regularly V-shaped during the prophases. For, as shown in 
figs. 69, 74, 75, 82, and 83, the rings are usually of conspicuously 
large size, and usually if not invariably in the number of one to a 
nucleus, as if such a geminus always represented a particular 
large pair of autosomes and were not a mere variation. 
It is now necessary to examine at what period the autosomes 
become longitudinally split. In the growth period through the 
pachytene stage there is no longitudinal splitting, for what I had 
previously (’01) interpreted as such, I now find to be the line of 
conjugation. This is proven sufficiently by the fact that the 
twelve autosomes of the spermatogonia are represented in these 
stages by six double rods. Frequently at certain points along a 
geminus the chromatin granules appear accurately paired (figs. 
51, 52). But this does not appear until a rather advanced stage 
of the strepsinema and is by no means regular. The true longi- 
tudinal splitting of the univalent elements takes place at a much 
later stage than I had previously supposed. Occasionally in a 
very early prophase there is an indication of it, as in fig. 64a. 
