12 



BULLETIN 99, U. S. DEPARTMENT OF AGRICULTURE. 



found that short rows can be sown by hand fairly rapidly, very uni- 

 formly, and much more accurately than by machine. 



In the beginning of the work only one row of each strain was 

 grown, but it was found that the same strain should be replicated in 

 as many rows in different parts of the field as is feasible. Beginning 

 in 1909, the different strains were replicated several times and have 

 been each year since then. After trying different systems it has 

 been decided to repeat each variety or strain 10 times, as any less 

 number is considered too small. Since the average of all the rows is 

 easily obtainable with 10 as a divisor, it makes it all the more desir- 



FlG. 3. — Oat nursery at the Cornell University experiment station, 1912. 



able to use this number. Figure 3 shows the oat nursery at the 

 Cornell experiment station in 1912. 



In regard to the use of short rows for the testing of strains, it 

 may be said that of all the systems tried this method seems the 

 most accurate. This is especially true on soil which is very non- 

 uniform. It is not safe to draw conclusions from one plat of a 

 variety, as the plat may be very favorably or unfavorably located 

 and the results influenced thereby. On tins point attention may be 

 called to the work of Dr. T. L. Lyon. 1 After comparing the errors 

 obtained from rod rows and tenth-acre plats, he concludes: 



1 Lyon, T. L. A comparison of the error in yields of wheat from plats and from single rows in multiple 

 series. Proceedings, American Society of Agronomy, v. 2, 1910, p. 38-39, 1911. 



