FIELD CROPS. 39 



tions in nitrogen content were attributed therefore to variations in environ- 

 ment beyond the ordinary range of observation. The centgeuers, rows, and 

 small plats were found to vary almost as much as the individual plants, and 

 the most practical way of overcoming this variation seemed to be by replicating 

 the plats or rows a sufficient number of times to reduce the errors to less than 

 one-half the real variation. To bring the experimental error within proper 

 bounds most suceessfuDy, single plants should be replicated 40 or more times, 

 16 ft. rows from 5 to 10 times, and blocks 5.5 ft. square from S to 16 times. 



In order to eliminate the undesirable strains, it was found that the experi- 

 mental error should be less than one-half the real or expected variation. The 

 easiest and most practical method found of growing strains to compare for the 

 nitrogen content was to plant in rows from 12 to 16 ft. in length and repeat 

 10 times in different parts of the field. Several check plats should also be 

 inserted. 



Practically no relation was observed between the yield of grain and the 

 nitrogen content, although there was a slight irregular tendency for the yield 

 to increase as the nitrogen decreased. Tabulated data of nitrogen content and 

 yields, and diagrams showing the results of different methods employed to re- 

 duce experimental errors, are given. 



II. Experimental error in the nursery and variation in yield (pp. 33-61). — 

 This paper deals with work performed chiefly in 1909 and 1910, the results of 

 which are discussed with the idea of pointing out sources of error in connec- 

 tion with the centgener, row, and small plat method employed in cereal breed- 

 ing and comparison of varieties, which seem to be due to unexpected variations 

 in soil and climate. Suggestions of methods of correction of these errors have 

 been deduced from studies of many hundreds of plants of wheat and oats 

 covering the effect of repetition in reducing errors. Tabulated data show com- 

 parison of yields, variations, and coefficients of variations. 



It was found that when 14 or 16-ft. row plats were used as checks or sown 

 in duplicate, there was great variation in yield, owing to natural unequal 

 effects of the environment. When the row plats were repeated 5 or 6 times, 

 the extreme error was still large, owing to the chance combination of high or 

 low variants. Systematic repetition constantly reduced error as the number of 

 repetitions increased, but with 16-ft. row plats from 10 to 20 repetitions should 

 be made, depending upon the degree of accuracy desired. 



Small blocks, 5.5 ft. square, gave results similar to those of the row plats. 

 except that the reduction of experimental error was somewhat greater as the 

 result of repetition. Blocks repeated from 8 to 10 times gave results apparently 

 about as accurate as rows repeated from 15 to 20 times. Increasing th<' size 

 of the block, up to a certain limit, rapidly decreased variability ; but error 

 could not be indefinitely decreased by continuing to increase the size of the plat, 

 as it could be by repetition. Variability was not constant from year to year 

 on the same plats. 



To alternate with check rows gave a high degree of accuracy, with a few 

 extreme variations, when as high as 10 or more checks were used. The total 

 number of plats required for the same degree of accuracy, however, was greater 

 by this method than by systematic repetition. By increasing the length of the 

 row or the size of the block, the number of repetitions necessary was decreased, 

 but the total area required to secure the same accuracy was increased. 



The rate of planting within certain wide limits had little influence on the 

 yield. There was some competition between adjacent rows, especially when 

 varieties very different in habit of growth were planted side by side. The use 

 of blocks did away with this source of error. 

 95099°— No. 1—13 4 



