532 



SCIENCE 



[N. S. Vol. XXXVIII. No. 981 



solves itself into the measurement of the 

 yield given by a definite area. The actual 

 measurement, therefore, is the same as that 

 made in manurial trials, and is, of course, 

 subject to the same probable error of about 

 five per cent. 



It follows, therefore, that it is subject to 

 the same limitations. Variety trials on 

 single plots, and that is the method com- 

 monly used, will serve to measure varia- 

 tions in yielding capacity of thirty per 

 cent., or more, but are totally inadequate to 

 distinguish between varieties whose yield- 

 ing capacities are within ten per cent, of 

 each other. 



Numbers of such single-plot trials have 

 been carried out, with the result that many 

 varieties with yielding capacities much be- 

 low normal have almost disappeared from 

 cultivation, and those commonly grown do 

 not differ greatly from one another — prob- 

 ably not more than ten per cent. 



Ten per cent, in yielding capacity, how- 

 ever, in cereals means a return of something 

 like 15 shillings to 20 shillings per acre — a 

 sum which may make the difference be- 

 tween profit and loss ; and if progress is to 

 be made in manuring and variety testing 

 some method must be adopted which is ca- 

 pable of measuring accurately differences in 

 yield per unit area of the order of ten per 

 cent. 



The only way of decreasing the probable 

 error is to increase the number of plots, and 

 to arrange them so that plots between which 

 direct comparison is necessary are placed 

 side by side, so as to reduce as much as 

 possible variations due to differences in soil. 

 Thus it has been shown that with ten plots 

 in five pairs the probable error on the aver- 

 age can be reduced to about one per cent., 

 in which case a difference of from five to 

 ten per cent, can be measiired with con- 

 siderable certainty. 



Such a method involves, of course, a 



great deal of trouble; but agricultural 

 science has now reached that stage of devel- 

 opment at which the obvious facts which 

 can be demonstrated without considerable 

 effort have been demonstrated, and further 

 knowledge can only be acquired by the ex- 

 penditure of continually increasing effort. 

 In fact, the law of diminishing return holds 

 here, as elsewhere. 



It appears, then, that for questions in- 

 volving measurements of yield per unit 

 area, such, for instance, as manurial or 

 variety trials, further advance is not likely 

 to be made without the expenditure of much 

 more care than has been given to such work 

 in the past. The question naturally arises: 

 Is it worth while? I think the following 

 instance shows that it is : 



Some years ago an extensive series of 

 variety trials was carried out in Norfolk, in 

 which several of the more popular varieties 

 of barley were grown side by side at several 

 stations for several seasons. In all, the 

 trial was repeated eleven times. As a final 

 result it was found that Archer's stiff -straw 

 barley gave ten per cent, greater yield than 

 any other variety included in the trials, and 

 by repetition of the experiment the prob- 

 able error was reduced to one and a half 

 per cent. The greater yield of ten per cent., 

 being over six times the probable error of 

 the experiment, indicates practical certainty 

 that Archer barley may be relied on to give 

 a larger crop than any of the other varieties 

 with which it was compared. One difficulty 

 still remained. It was almost impossible to 

 obtain anj^thing like a pure strain of 

 Archer barley. Samples of Archer sold for 

 seed commonly contained twenty-five per 

 cent, of other varieties. This difficulty was 

 removed by Mr. Beaven, who selected, again 

 with enormous trouble, a pure high-yielding 

 strain of Archer barley. Since this strain 

 was introduced into the eastern counties 

 the demand for it has always exceeded the 



