160 THE DEVELOPMENT OF RAW COTTON 



recent discoveries about natural crossing. There is only 

 one way in which certainty can be attained, and that way 

 is a laborious one, though not at all impracticable ; by 

 dividing the experimental area into small plots (PI. XI.), 

 putting five under each kind of treatment, and scattering 

 these five over different parts of the area, the precision 

 of the results with cotton may be increased to 9 : 11 as 

 the maximum possible dissimilarity due to accident, 

 when the five-plot averages are compared. It is fre- 

 quently objected that the trouble and labour of handling 

 small plots makes them impracticable; there is, how- 

 ever, no escape from the fact that only in this way can 

 a reasonably correct answer be obtained; whether it is 

 "practical " to obtain an answer which has practically 

 no meaning, and " unpractical " to expend a little more 

 on labour to obtain one which has a definable signifi- 

 cance, must be left to the future to decide. It should be 

 noted that the same amount of land and ordinary culti- 

 vation is required in both cases. The additional trouble 

 The Handling * s * n ^ e handling of the small plots, and 

 of Small especially in laying them out and sowing 

 Plots. them. One working suggestion in the 

 former respect may be useful, namely, that no attempt 

 should be made to differentiate the -plots in the field- 

 work; each plot should have a serial number, be observed 

 and treated under that number, and the final grouping 

 effected only when working up the results of the obfeerva- 

 tions. Working in this way it is not impracticable to 

 combine two, or even three, experiments into one ; thus 

 sowing-time and spacing could be handled in one series 

 of plots in the following way : 



