I04 BOTANICAL GAZETTE [February 



Many of the Illinois experiment fields include three check plots 

 in each series. These check plots are all untreated and are only a 

 short distance apart, yet some of them differ widely in crop yields. 

 It is reasonable to assume that neighboring plots receiving the 

 same fertilizer treatment would differ as widely. These differences 

 due to factors not under the control of the investigators make the 

 probable error large, and when only one plot of each treatment is 

 used, the differences between plots with different treatments must 

 be great before one can assume that the treatment has been effec- 

 tive. Where the differences are as great as in the work of Reimer 

 and Tartar (58) and of Olson (54), there is no doubt that the 

 treatment has been effective, but in many of the field experiments 

 in different parts of the country the differences are too small to 

 justify the conclusions drawn from them, as the probable error is 

 so great. Where a number of plots of each treatment are used, the 

 uncontrollable factors tend to neutralize each other and the prob- 

 able error is reduced. As the number of plots of each treatment 

 increases, smaller average differences are necessary to be signifi- 

 cant. It seems probable that three plots of each treatment are 

 necessary if satisfactory results are to be obtained. In the past 

 investigators have had a tendency to scatter field experiments over 

 a number of widely separated fields on the same soil type. It seems 

 probable that more satisfactory results would be obtained if the 

 work were confined to one field on each soil type, and each field 

 had from three to five plots of each treatment. 



Summary 



1. Composite soil samples from Indiana, Kentucky, Michigan, 

 Ohio, and Wisconsin were analyzed for total sulphur, total phos- 

 phorus, and volatile matter (loss on ignition), and cooperative 

 fertilizer experiments with gypsum were conducted in fields in 

 Indiana and Kentucky. 



2. The analytical data show a general relation between the sul- 

 phur content and loss on ignition in soil samples from the same soil 

 type or closely related soil types, but the relation is not apparent 

 when different soil types are compared. 



