732 EXPERIMENT STATION RECORD. 



FIELD CROPS. 



The experimental error of field trials, W. B. Mekceb and A. D. Hall 

 {Jour. Affi: ScL, .'/ {1911), Xo. 2, pp. 107-l.i2, jiys. 10). — The authors discuss two 

 methods of reducing experimental error, (1) by repetition over a number of 

 years, and (2) by multiplying a number of plats on a given field. Tables state 

 the relative location and yields secured from 20<) plats of mangels and 500 

 plats of wheat. In connection with the mangels, tables state the varying weight 

 of rows of plats, the relation between the weights of roots and leaves, and the 

 standard deviation with different numbers of competent units. In connection 

 with the wheat yields, tables state the relation between grain and straw, vary- 

 ing weight of rows, influence of ix)sition on land, and the standard deviation in 

 one-fifth acre plats composed of different numbers of units. 



In view of the data presented the authors draw the following general con- 

 clusions : 



" In all field trials, however uniform the land and careful the management 

 of the experiment, there will be an exiDerimental error attached to the result, 

 due to so many incalculable factors that it may be described as casual. The 

 error diminishes with the size of the plat, but the reduction is small when the 

 plat grows above one-fortieth of an acre. The error may be best diminished 

 by increasing the number of plats similarly treated and scattering them about 

 the area under experiment, but there is not much to be gained by increasing 

 the number of plats above five. 



" For practical purposes the authors recommend that in any field experiment 

 each unit of comparison (variety, method of manuring, etc., according to the 

 subject of the experiment) should be given 5 plats of one-fortieth of an acre 

 each, systematically distributed within the experimental area." On land at 

 all suitable, this will reduce the experimental error to within 2 per cent of the 

 result, but will not eliminate the unequal effects of different seasons upon the 

 varieties or the action of fertilizers. 



An appendix by "Student" outlines a plan w^hereby "the experimental 

 error may be still further reduced when only a single comparison is desired, as 

 for example between 2 varieties or 2 methods of manuring, by taking advantage 

 of the correlation which exists between adjacent areas." 



Field trials and their interpretation, A. D. Hall and E. J. Eussell {Jour. 

 BJ. Agr. [London'], Sup. 7, 1911, Nov., pp. 5-11,, figs. ,?).— The authors dis- 

 cuss the probability, causes, and magnitude of the experimental error in 

 field trials. They arrange the sources of variation in two groups, (1) those 

 which decrease as the size of the plat decreases, such as soil variation and 

 conditions of growth, and (2) those which Increase as the size of the plat 

 decreases, such as individual variations and the incidence of disease. 



Differences are least in the first grouji when plats are narrow and lie side 

 by side. On the Broadbalk field at Rothamsted such plats similaiiy treated 

 differ by from 5 to 7 per cent as compared with differences of from 7 to 12 

 per cent in the case of plats placed end to end. 



A table states the probable error for simple and scattered unit plats of 

 various sizes. In the case of simple one-fiftieth acre plats, for instance, the 

 probable error is 4.2 per cent as compared with 2.4 per cent when made up of 

 5 scattered units. One-fifth acre plats made up of 5 scattered one-twenty-flfth 

 acre plats have a probable eri-or of 1.3 per cent as compared with 1.1 per cent 

 when made up of 10 one-fiftieth acre plats, 1.7 per cent when made up of 2 

 one-tenth acre plats, and 3.1 per cent as simple plats. 



The interpretation of experimental results, T. B. Wood {Jour. Bd. Agr. 

 {London], Sup. 7, I'Jll, Nov., pp. 15-37, figs. 2). — These pages contain the sub- 



