8 BULLETIN lOSO, U. S. DEPAETMEXT OF AGRICULTURE. 
tained from the last two. All of these were conducted at chronic 
nonhatch points. 
Those clusters placed in the wire cages were entirely removed from 
any influence the trees might have upon them and were fully ex- 
posed to all actions of the elements. Very little information was 
obtained from this series beyond the fact that clusters reacted the 
same way no matter how they were j^laced. 
During the cold years all of these clusters were killed, as were those 
naturally on the trees. After the mild winter they all hatched com- 
pletely. 
These clusters in the inverted jars received a considerable amount 
of protection, as they were in no way affected by storms. Low tem- 
perature could act upon them freely and to a certain extent the at- 
m^ospheric moisture could do so, for the jars were open at the bottom, 
allowing air to ascend into them when it became warm. Xo rain or 
snow could reach them, however, and as a result they remained per- 
fectly dry during the entire winter, as was proved by numerous ob- 
servations. They were therefore not frozen or covered with ice, as 
were many of the clusters in the open. 
These clusters were only exposed to temperature and atmospheric 
moisture. It has been shown 'already that humidity can play very 
little, if any, part in killing the eggs; therefore it may be considered 
that if these eggs were killed temperature must have been responsible. 
To corroborate this conclusion all eggs exposed in this manner failed 
to hatch after the cold winters, but hatched perfectly after the mild 
one. 
Experiments in natural protection were suggested by notes taken 
in connection with the " observation point " investigations, as has 
already been noted. It was found that those placed close to the 
ground at nonhatch points, whether in cavities, under roots, or in 
the open, hatched completely, while those high up were killed. At 
first sight snow appeared to be the protecting factor: and this sup- 
position was borne out later by actual observation, though it was 
sonie time before all variations could be reconciled with this theory. 
To afford protection the snow had to cover the clusters during every 
severe cokl spell, which it did not do because of the countless varia- 
tions in its depth, and it was only after a long series of observations 
in the woods that a true appreciation of the variableness of this 
factor became apparent. Depth even immediately after the cessa- 
tion of a storm, varied enormously, particularly if the snow was light, 
for every breath of air induced drifting. Many times also the eddy- 
ing of the wind around the tree trunks left the snow i:>iled against one 
side and blown away from the other. The resulting depression 
served to expose some clusters while others remained covered, a fact 
wliich explains why clusters close to the ground were killed and 
those much higher hatched. 
A close following of the snow history of a section during the wii^ter 
showed great fluctuations in depth, when measured on the trunk of 
a tree small enough to be unaffected by the wind eddies mentioned 
above. At the end of a storm the snow would be piled to a certain 
height on the tree; then gradually settling would take place, until 
after the lapse of a few days there would be several inches difference 
