61 
because the scales killed by one insecticide may remain longer at- 
tached to the bark than those killed by another, and there is abundant 
evidence that this is often the case. It is furthermore practically 
impossible to make sure that a tree selected for a scale-count is a 
fair average of its plot with respect to the degree of its infestation, 
and scarcely less difficult to make sure that the parts of its infested 
surface selected for counting are a fair average for the tree. For 
these reasons the counting method was abandoned, in the main, and 
I have relied instead upon a careful estimate of degrees of infesta- 
tion for each tree of both check and experimental plots, these esti- 
mates being recorded in figures ranging from i to 6. From these 
the average degrees of infestation for the trees of each plot were fig- 
ured, and these averages were compared with each other in a way 
to show numerically the effect of the treatment in a reduction of the 
average amount of infestation. 
Conditions Necessary to Exact Experiment. 
In order to an estimate of the precise value of the results here 
reported, it will be helpful to review tlie conditions necessary to an 
entirely satisfactory experiment of this description. In the first 
place one should have at his command an orchard of considerable 
size, either all of one variety or with the separate varieties planted 
in complete and continuous rows all running in one direction. This 
should be divided across its whole width, and across the variety 
rows, into strips of five rows each if moderately infested by the San 
Jose scale, and into seven to ten rows if heavily infested, each strip 
receiving a different treatment except that occasionally strips should 
be left untreated as checks. To avoid the mutual influence of check 
and experimental plots upon each other, which may extend from one 
to three rows in each direction across the boundary, only the interior 
rows of each strip should be used for comparison. Experimental 
trees near the borders of the check strip are likely to receive young 
scales from the latter in larger number than they lose, if the experi- 
mental treatment has had any marked effect, and the reverse is of 
course true of the marginal rows of the check. This is not a theoret- 
ical supposition merely, but is shown by our previous experience, 
and by some of the averages to be reported fa^thei^ on. As the check 
plots will act as centers of general dispersal of the scale insects 
during the whole of the following season, there should be no more 
of them than are clearly necessary. One at each end of the orchard 
will usually be enough. 
The trees of the interior rows of each strip should be graded very 
