MOISTURE CONTENT AND SHRINKAGE OF FORAGE. 17 
_ made with only one sample. Table II also shows that the samples 
of field-cured material are less consistent than samples of green © 
material when compared with bulk lots of the same forage dried 
under similar conditions. 
RELATIVE VALUE OF SAMPLES OF DIFFERENT SIZES. 
The figures on the relative value of samples of different sizes as given 
in Table III are not conclusive. There is a general, though not 
consistent, decrease in the probable error as the size of the sample 
is increased, but what would otherwise have been an expressive array 
of averages has been spoiled by the excessive probable error in the 
8-pound field-cured sample of alfalfa at Chico, Cal. The average 
probable error for the 8-pound samples, including the Chico results, is 
0.515 per cent; if we eliminate the Chico results it would be 0.281 per 
cent, which perhaps is nearer what might ordinarily be expected. It 
will be noted that as the green samples of alfalfa and of tall oat-grass 
increase in size, the greater was the percentage of loss in curing, as 
indicated by the column headed ‘‘Moisture in original samples’ in 
Table III. This fact makes it seem probable that there was a loss, 
by fermentation, of matter other than water, but such a loss would 
not mean an increase of error in the use of samples when the samples 
are of a uniform size. 
On account of the difficulty of curing samples of green forage they 
must necessarily be comparatively small, and when used in correcting 
actual field weights the samples, whether green or field cured, must be 
small enough to admit of easy handling. From the data presented 
in the table, it seems that the 4-pound field-cured and the 8-pound 
green samples are nearly as accurate as the larger ones. Considering 
accuracy, the facility of handling, the ease of figuring percentages, 
etc., 5-pound samples of field-cured and 10-pound samples of green 
material are recommended as the most desirabie for practical work. 
EFFECT OF REPLICATING TRE SAMPLES. 
The data on the effect of replicating the samples are found in Table 
IV, where the probable error has been expressed for single samples, 
replicates of three, and replicates of five and six. 
TaBLE I1V.—Average of the probable errors of one, three, and five or six samples. 
Field-cured material. Green material. 
Number of replica- grend 
aE ela G2 1. 8 Atenas 8 | foes | 162) lEAvers| ager 
pound.|pound.|pound./pound.| age. |pound./pcund./pound./pound.| ace. 
ae ct.| Per ct.| Per ct.| Per ct.| Per ct. Per ct.| Per ct.| Per ct.| Per ct.| Per ct.| Per ct. 
CAD SE Te A ee ee ee 1.006 | 1.243 | 0.818 | 0.994 | 1.015 | 0.686 | 0.586 | 0.407 | 0.486 | 0.527 | 0.777 
athireetes o.8 2S ST 458 | .668] .397) .473 | .499) .331 | .277|] .156} .256] .255 377 
Pivean@sixs ioc 25.22. 445 | .548 353 | .515 | 2465) 9.278 | .257 | 2187 |) 2212 )) .234 - 349; 
21216°—Bull. 353—16——3 
