FIELD PLOT TECHNIC 



59 



ported which for the particular condition of the experiment show 

 the number of replications desirable. 



Mercer and Hall (1911), of England, recommend the use of 

 five systematically distributed plots of J^Q acre each. Mont- 

 gomery (1913), in his work at Nebraska, found that 16 ft. rows 

 gave best results when repeated from 10 to 20 times. At the 

 Cornell Experiment Station, when a careful yield test is desired 

 each strain is grown in 10 distributed rod rows. 



In the plant-breeding nursery of the Minnesota Experiment 

 Station the practice is followed of growing each strain in a plot 

 consisting of three rod rows. The plots are replicated three 

 times, making four plots in all. The central rows only are 

 harvested. Table XII taken from Hayes and Arny (1917) 

 shows the effect of replication based on the yield of the central 

 rows of the wheat checks grown in 1916. 



TABLE XII. VALUE OF REPLICATION BASED ON 72 CENTRAL Rows OP 



THREE-ROW PLOTS OF TURKEY WINTER WHEAT (MINN. 529) 



GROWN IN THE PLANT BREEDING NURSERY 



While the standard deviations do not decrease according to 

 theoretical expectation they do show a marked decrease up to and 

 including three replications. The table shows that variability 

 is rapidly diminished by replication up to a certain number. 

 In general, beyond this point it is questionable whether the 

 relatively small gain in accuracy warrants the additional work. 

 The results obtained at the Minnesota Station indicate that 

 plots of three rod rows each or J^o-acre plots sown with the or- 

 dinary grain drill, give about as accurate a comparison for yield 

 when replicated three times as when replicated eight times. 



The manner of making replications is another factor to be con- 

 sidered. If the experimental plots are all planted in a single 



