ou gee ay 
Punt 
DIRECTOR'S REPORT.  — 
One of the most important, nay the very dandamenen condition of 
experimental work, is that we shall correctly interpret the results of 
our work, If guesses are to take the place of method, then we might 
as well guess at our work in advance instead of going through the 
formality of measurings and weighings. ‘The plat system of experi- 
menting, from its apparent simplicity, has entered into general practice 
amongst experimenters, and the results of figures gained are, in general, 
received without question. The figures gained under this system of — 
equal plats are, however, more apt to afford incorrect than correct 
‘conclusions, as usually inter preted, and it is well worthy of discussion 
whether there may not be some method which will enable us to detect — 
inaccuracies and decide as to the values of our figures in their relations” 
to greater or less. 
In the station work an attempt in this direction is being made — to 
a certain extent provisional in its nature, as subject to improvement 
as experience is gained. Through harvesting our plats in separate por- 
tions in such a way that adjoining sections admit of comparison of 
yield, it seems probable that some idea of the uniformity or lack of 
‘thus derived be made available. Exactness of figures cannot be 
expected to prevail between duplicate plats, as absolute identity of — 
conditions cannot be forecast; in every such trial there must exist 
some differences between the yield of independent areas, as arising 
from the complex agencies which affect plant growth, in itself 
‘dependent npon.a complexity of hereditary environment, If, then, 40% § 
we cannot expect to obtain identity between two plats under. dupli- 
cated conditions, the query comes as to how closely the results must’ 
figure in order to serve the purpose of an experiment. When our 
query takes this form, we at once seem to recognize that for each class 
of events there are certain limits within which normal variations may 
take place, limits which are to be defined by a careful investigation, as 
deduced by experiment, of the variations which actually may happen ~ 
in a given number of plants upon a given area, under conditions of a 
presumed equivalency. 
The errors which most frequently occur in plat experimelt etm + 
may be grouped as follows: first, those dependent upon the lack of © 
uniformity of character in the seed used ; secondly, those which arise — 
from the number of plants which survive and attain maturity being — 
different in each area; thirdly, from the area of soil not being the 
same for each plant that produces the crop. As a corollary to this © 
statement we may then say that the first endeavor in plat experimen- 
tation must be to secure uniform seed, and to establish for each ae 
the like number of maturing plants occupying similar spaces for root — 
*:\ 
P 
* 
uniformity of soil conditions can be obtained, and the information \ ~ 















Oe oe 
me is 
