OCTOBEB n, 1913] 



SCIENCE 



531 



listed on this subject, it appears at once 

 that in many cases the increase in live- 

 weight of sheep fed on plots manured with 

 a suitable dressing of phosphate has been 

 twice as great as the increase in weight of 

 similar animals fed on plots to which phos- 

 phate has not been applied. Now about a 

 difference of this magnitude between two 

 plots there can be no mistake. It has been 

 shown by more than one experimenter that 

 two plots treated similarly in every way are 

 as likely as not to differ in production from 

 their mean by five per cent, of their pro- 

 duce, and this may be taken as the prob- 

 able error of a single plot. Where, as in 

 the ease of many of the phosphate experi- 

 ments, a difference of 100 per cent, is re- 

 corded, a difference of twenty times the 

 probable error, the chances amount to a 

 certainty that the difference is not an acci- 

 dental variation, but a real effect of the 

 different treatment of the two plots. The 

 single-plot method of conducting field 

 trials, which is the one most commonly 

 used, is evidently a satisfactory method of 

 measuring the effects of manures which are 

 capable of producing 100 per cent, in- 

 creases. It was good enough to demonstrate 

 with certainty the effects of phosphate 

 manuring on many kinds of grass land, and 

 it is to this fact that we owe one of the most 

 notable achievements of agricultural science 

 in recent years. 



Another notable achievement is the dis- 

 covery that in the case of most of the large- 

 cropping varieties of potatoes the use of 

 seed from certain districts in Scotland or 

 the northern counties of Ireland is profita- 

 ble. This is another instance of an increase 

 large enough to be measured accurately by 

 the single-plot method. Reports on the sub- 

 ject show that seed brought recently from 

 Scotland or Ireland gives increased yields 

 of from thirty to fifty per cent, over the 



yields pi'oduced by seed grown locally for 

 three or more years. 



That the single-plot method fails to give 

 definite results in many cases where it has 

 been used for manurial trials is a matter of 

 common knowledge. Half the reports of 

 such trials consist of explanations of the 

 discrepancies between the results obtained 

 and the results which ought to have been 

 obtained. The moral is obvious. The 

 single-plot method, which suffices to demon- 

 strate results as striking as those given by 

 phosphates on some kinds of pasture land, 

 signally fails when the subject of investi- 

 gation is concerned with differences of ten 

 per cent, or thereabouts. 



Before suggesting a remedy for this state 

 of things it will be well to consider the 

 allied subject of variety testing, which has 

 been brought into great prominence re- 

 cently by the introduction of new varieties 

 of many kinds of farm crops. In testing a 

 new variety it is necessary to measure two 

 properties — its quality and its yielding 

 capacity — for money-return per acre is ob- 

 viously determined by the product of yield- 

 ing capacity and quality as expressed by 

 market price. I propose here to deal only 

 with the determination of yielding capacity. 

 The determination of quality is not allied 

 to manurial trials. 



In attempting to determine yielding 

 capacity there has always been a strong 

 temptation to rely on the measurement of 

 obvious structural characters. For in- 

 stance, in the case of cereals many farmers 

 like large ears, no doubt with the idea that 

 they are an indication of high yielding 

 capacity. Many very elaborate series of 

 selections have been carried out, on the 

 assumption that large grains, or large ears, 

 or many ears per plant implied high yield. 



We may take it as definitely settled that 

 none of these characters is reliable, and that 

 the determination of yielding capacity re- 



