348 



Amounts of tenter evnporatrd from soil and plant dnrintj thf firowing season. 



Estimated normal amounts — 



Per ])lant 



Per kilogram of air-dry plant 



Estimated amount per hectare — 



Season of ]x«5 



Season of 1XH6 



Season of 1887 



Barley. Wheat. Oats. Peas. [Fallow 



Liter t. 

 2.0 

 350 

 Cuhi/; 



vieterg. 

 2. 46.3 

 3.475 

 2.900 



Liter*. Liter*. 

 1. 6 2. 7 



400 3.'»0 



(\ibie Cubic 

 meter*, mrter*. meter*. 

 2.702 2.666 ; 2.245 

 3.625 i 3.625 

 3.050 



Liter*. 



2.975 



Liter*. 



5.6 



300 _^. 



Cubic C^iliie 



meter*. 



2. 075 



2. 800 



2. 775 



3. 575 

 2.000 



The cxperimenfnl mrfhod. — D('si>ite tlic niis1ia]>s in thes»' ]»inti('nlar 

 e.xiH'iinuMits, Woliltinaiin ic;:ai<l.^ tin- nu'tliod as having innvt'd itself 

 fitted for its purpose. 



(1) Tlie eliief ditticulty was cansed l>y leaks. The sides of th»» boxe,s 

 next to the passagt-ways beiii}; suhjeet to outward pressure from tlie 

 soil within ami not ]»roteeted by eorres])ond in j; lateral outside i)ressure, 

 bulged out and pressed against the ui)right glass tubes. The joints of 

 the latter were thus loosened and in some »ases the tubes were broken. 

 This ami defective stopjx'rs were the causes of the leaks. A simi»le 

 arrangement is proposed for reiiu'dying these dilliciilties and providing 

 at the same time for drawing off the drainage water at the bottom 

 instead of at a jioint 2(1 cm. ab(»ve the bottom, as was tlone in these 

 experiments. This would al.>^o allow better regulation of the height of 

 the watci tabic. 



(2) One essential lest o| the value of the nu'thiMl is the agreement of 

 duplicate trials as shown by the amount of juoduce. In each ot" the 

 three seasons' experiments four kinds of plants — barley, wheat, oats, 

 and i>eas (the trials with lieaus and lupines were left out <»f account in 

 the comi>aiison because of failure «tf a ninuber of plants, as alwtM- 

 f;tate<l) — were grown, each in ."» duplicate series with ditferent fertilizers 

 or unmanure(l. Tliis made fbr each year 40 single «»r 20duplicat«' trials. 

 In thes«' l>ll cases there were ".. all with ])eas the secoml season, in which 

 the ])lants of one of the duplicat<'s wt'ie .so (lete<ti\f as to nuike the 

 results unsatisfactory. The agreement of the duplicates in the other ."»7 

 cases was computed by taking the larger yield of the 1 dupli<'ates in 

 each case at 10(1 and the smaller at the corresiKmding nniuber. and 

 calling the ditference between these two numbers the difference of 

 duplicates in ])er cent. This itereentage dittereme ot duplicate trials 

 averaged in ISS."), ;i.4; in 1XS(», the moist season, 5.7; in 1SS7, .'i.H; and 

 in all the ~>7 cases of the three seasons together, \.'^ per cent. It 

 exceeded 5 per cent in 1(> eases, of whieli 8 were in ISSfJ. and exceedetl 

 10 per cent in 7 cases, of which .■> were in ISSH. The differences of over 

 5 per cent <lid not occur in the same ])airs of boxes in 2 diffei-ent years 

 in more than 2 cases, and in ouIn 1 of these did the diffeiences a]»itear 

 in 2 successive years. 



It was thus clear that these difterences in yield of duplicates were not 

 due to diflfereuees iu soil or any other constant factor. The inference 



