6 BULLETIN 1004, U. S. DEPARTMENT OF AGRICULTURE. 
There is, of course, a considerable number of experimental errors 
in the determinations. Even in the most uniform soils, duplicate 
determinations of soil moisture are nearly certain to vary at least 
one-half of 1 per cent, and in less uniform soils variations of much 
greater magnitude occur. The probable error of the total of any 
individual sampling or of the difference between the totals of any 
successive samplings may be relatively large or small, depending 
upon the dispersion of the errors of the units making up the totals 
and the extent to which they compensate each other. An error in 
the same direction of one-half of 1 per cent in each foot in a deter- 
mination to the depth of 6 feet on a soil weighing 75 pounds per 
cubic foot would be equivalent to 0.42 inch of water. In soils lacking 
uniformity the experimental error may be much larger. 
In the present study of the use of water the following practices 
have been followed for the purpose of reducing the experimental 
error as much as possible : 
(1) The depth to which the samples have been included in each deter- 
mination has been limited to the zone of change of water content. For 
example, even if samplings have been made to the depth of 6 feet or more 
but all changes in water ccmtent (either reduction by use or addition 
by rainfall) have been confined to the first 3 feet of soil, the calculations 
to determine the quantity of water used have been limited to 3 feet. 
Inclusion of additional depths would only increase the experimental error 
without contributing anything of value to the results. 
(2) Periods of time as long as practicable have been used to deter- 
mine the r^te of use in each instance. 
To obtain the rate of use per day, the quantity of water used in 
any period is divided by the number of days in that period. Any 
error that may exist in the total quantity of water used is conse- 
quently reduced in proportion to the number of days in the period. 
In a general way, experimental errors of successive samplings tend 
to compensate each other, and the experimental error for the quan- 
tity of water used in a season should be small and is limited to the 
errors in the first and last samplings. 
DAILY RATE OF THE USE OF WATER WHILE THE CROP IS 
GROWING RAPIDLY. 
Under this heading are included the results of attempts made to 
determine the daily rate of the use of water by the wheat crop under 
full and uninterrupted growth. This rate was calculated in all years 
when two soil-moisture determinations at least two weeks apart 
were made during the period of rapid growth of the crop and while 
it was not suffering for lack of water. As long a period as possible 
was selected for each year, in order to make the error in the rate 
per day due to the experimental error of the determinations as small 
as possible. In years when no shortage of water occurred the end 
of the period is determined by harv^est. No years were included (1) 
when the crop had suffered from drought before the period of rapid 
growth was reached; (2) when the second sampling during the rapid- 
growth period was delayed until after the crop commenced to suffer 
from drought; (3) when some condition, such as hail, rust, or insects, 
destroyed the crop. Years when some factor other than a lack oi 
water was responsible for crop injury were excluded, to avoid the 
