MOISTURE CONTENT AND SHRINKAGE OF FORAGE. 5 
Column 6 carries a statement of the percentage of moisture in each 
of the original samples, as determined by the difference between the 
original and the oven-dry weights. It is recognized that this loss 
may not necessarily be exclusively water. Slight losses may have 
taken place through volatilization of substances other than water or 
through fermentation due to enzyms or bacteria, but such losses are 
undoubtedly small when the hay has been quickly cured. The percent- 
ages as given are determined by using the original weights of the 
samples as the base. It is recognized that this practice is open to 
criticism, owing to the fact that the original weights vary in relative 
importance, due to the differing percentages of moisture which the 
samples contain. This criticism is of little importance in the present 
case, however, since the percentage of moisture is very nearly the 
same in each group where a comparison is made. The use of the 
absolute dry matter as a base from which to figure the percentages 
was tried, but this method seemed impracticable, because it makes 
the percentages so at variance with the moisture percentages as 
usually given. Column 6 also gives the means of groups of three and 
sroups of five or six samples, with their probable errors. In tables 
where tliere are only five samples in each class the second group of 
three represented by the second mean includes the remaining two 
samples and the one next above, which has already been considered 
in the first group. For example, in section A the first ‘‘mean of 3” 
is based on samples 1, 2, and 3, and the second ‘‘mean of 3” on 
samples 3, 4, and 5. These means are set in black-faced type, so 
that they will be apparent at a glance. The consistency in the per- 
centages of moisture in each set of samples is remarkable. In only 
one case has the probable error for the group of six samples exceeded 
1 per cent, and the probable error for a single sample averaged con- 
siderably less than 1 per cent, although in exceptional cases it ap- 
proached 2.5 per cent. The probable error was chosen as the most 
efficient measure of the comparative reliability of the different sizes 
of samples and methods of sampling. Since the moisture is here 
stated in percentages, means of practically the same size are dealt with, 
and the need for a term like the coefficient of variability is lacking. 
Column 7, as shown by the heading, is a record of the percentage 
of moisture in the air-dry material, the weight of which is shown in 
column 4. 
Column 8 gives the percentage of moisture which was lost in reduc- 
ing the material from its original state to an air-dry condition. The 
base on which this percentage was calculated is the weight of the 
original material given in column 2. The actual losses of weight in 
100 pounds of field-cured and 500 pounds of green material under 
the same conditions as those surrounding the samples are given fol- 
lowing the tabulation of sample weights. 
