Experimental Error in Crop Tests 
9 
effectiveness of check plats for reducing test plats to com- 
parable yields; (2) the reduction of error by the replication 
of plats; (3) the relative reliability of plats of various sizes 
and shapes; and (4) the significance of the ''probable error" 
as a measure of confidence which may be placed in mean 
results. 
When the odd and even numbered plats of these 207 are 
regarded as check plats and test plats respectively and the 
grain yield of each test plat is corrected by the mean of the 
two adjacent check plats, the coefficient of variability for the 
actual yields of these test plats is reduced from 7.85 per cent 
for the actual yields to 7.01 per cent for the corrected yields. 
Assuming every third plat to be a check, and correcting the 
intervening plats by the one adjacent check plat, the coef- 
ficient of variability was reduced from 7.79 per cent to 7.35 
per cent. 
With every third plat regarded as a check plat, and the 
intervening plats corrected progressively by the two near- 
est checks, the coefficient of variability is reduced from 7.87 
to 6.57 per cent. Thus it is seen that none of the three 
methods of check plat correction have been very effective. 
The yield of systematically distributed check plats can- 
not be regarded as a reliable measure for correcting and es- 
tablishing correct theoretical or normal yields for the inter- 
vening plats. 
21. Systematic replication of plats is the most effective 
and satisfactory means for reducing error caused by soil or 
other environmental variations. When 200 thirtieth-acre 
plats were planted to a uniform crop of Kherson oats, the 
coefficients of variability for the grain yields of single plats 
and for the mean yields of two, four, and eight plats were 
6.30, 4.59, 2.91, and 2.13 per cent respectively. The extreme 
variation between yields was also reduced from 20.7 bushels 
for single plats to 7.5 bushels for the means of eight plats. 
Reduction of error by averaging adjacent plats (which 
is equivalent to increasing the size of the plat) was far less 
effective than systematic replication. The coefficients of vari- 
ability for single plats and for the mean yields of two, four 
and eight adjacent plats were 6.30, 5.46, 5.28, and 4.78 per 
cent. 
Variation between long, narrow plats was less marked 
than for short, wide plats of the same area. The coefficient 
of variability for tenth-acre oats plats 48 rods by 5.50 feet was 
