400 Mr. 0. H. Latter on the secretion of 
the shield. My friend^ Mr. Archer Vassall, informed 
me, shortly after the publication of that paper, that he 
had observed the imago remove the shield by means of 
its front pair of legs. I have now myself repeatedly 
seen that this is the normal process. Almost as soon 
as the imago is free of the cocoon it halts for an 
instant, and raises its front pair of legs over its head, 
crossing the two tarsi dorsally to the shield; it then 
makes a rapid forward and downward stroke with the 
two legs simultaneously, and pulls the shield away, 
leaving it on the ground. 
(iii.) Mode of attacking the cocoon. 
In collecting the secretion for quantitative analysis, 
I was enabled to observe more fully the exact behaviour 
of the emerging imago. The pupae were enclosed in 
glass tubes (to be more particularly described below) 
and were thus easily watched. The first violent 
struggles of the imago liberate its anterior regions from 
the pupa case, this latter is then thrust back a little 
until its posterior extremity rests against some solid 
resisting surface, which would in nature be formed by 
the posterior end of the cocoon. A series of peristaltic 
contractions then sweep over the abdomen of the imago 
from behind forwards, urging the head forward against 
the anterior end of the chamber until the pressure 
becomes considerable. Between each peristaltic move- 
ment the body slips backward a little by the partial 
telescoping of the abdominal segments. These move- 
ments constitute the strokes made by the labral prongs 
agamst the cocoon wall, and, at the same time, the 
muscular contractions compress the contents of the 
body and expel drops of potassium hydroxide from the 
mouth. If, after the imago has emerged, the action 
be imitated by gentle pressure between the fingers, 
more of the fluid can generally be driven out. 
(iv.) Quantitative analysis of the potassium hydroxide 
solution. 
The investigation of the strength of the solution 
proved difficult, and was attended by many failures. The 
following is the method which I found most satisfactory. 
