COTTON-GROWING 159 



In the last five years we have seen a revolution in our 

 knowledge of field experiments, the methods of con- 

 ducting them, and the errors inherent in 

 S ^ aU f lot them. The two chief features of this new 

 knowledge are as follows: The increased 

 precision obtained by increased size of plot is practically 

 negligible beyond about one-tenth of an acre; the 

 errors in comparison of plots which are supposed to be 

 identical is such that with English wheat the plots may 

 differ as 84: 116 by pure accident, and only half of them 

 will be as closely similar as 95 : 105. 



A great part of this error is due to soil variations 

 which cannot be eliminated. When we deal with a deep- 

 Errors of r 00 ^ 6 *! plant like cotton, which sends its 

 Plot- roots through more than two metres depth 

 Comparison. o f so j] m a se ason, the errors are much 

 greater from this cause. Identical plots of cotton may 

 differ by nearly 75 : 125 through normal accidents alone, 

 and half the plots are bound to differ more than 93 : 107, 

 this degree of difference being shown on total yield, and 

 being proportionately more on separate pickings (Fig. 19). 

 Recognition of the existence of this .very high degree 

 of uncertainty in comparison will account for the un- 

 Need for certainty which attends on our present 

 Accurate knowledge of the cotton crop. Practically 

 Results. t jj e W h i e of the work of the past fifty years 

 on experimental crops of cotton will have to be repeated 

 in this new light, just as Mr. Leake has pointed out that 

 nearly all the failures in introduction of new cottons into 

 India have become devoid of significance in the light of 



