THE MEASUREMENT OF RESULTS 17 



several thousand plants, as in the case of agricultural plots, the 

 area required would be preposterously large. Nor is it probable 

 that, in this country, additional accuracy would be obtained at 

 all commensurate with any great increase in the size of the plots ; 

 for the land in England is so lacking in uniformity, that it is 

 difficult to secure any one field where fruit trees do equally well 

 in all parts of it. More satisfactory results seem to be obtainable 

 by planting small plots in situations where the conditions 

 appear to be strictly uniform, and duplicating the experiments by 

 repetitions in other situations where a similar uniformit}' exists. 

 The first question which arises is, the minimum number of trees 

 or bushes which each plot should contain. Judging by the results 

 obtained at Woburn, this number may be placed at 6 to 12 for 

 fruit trees, 20 to 50 for fruit bushes, and 100 to 200 for straw- 

 berries. But such an estimate requires many qualifications. It 

 applies only to the case where one variety of the plant is under 

 investigation, and to a case where all the trees or bushes have 

 been very carefully selected, so as to be as uniform as possible 

 in the first instance. This may be effected by making a rough 

 preliminary selection, then weighing each plant, and arranging the 

 plants in groups, such that the average weight of each group is 

 the same, and that the difference in the weight of the individuals 

 is but small. Trees thus selected behave in a surprisingly uniform 

 manner during the first few years of their life, i. e. so long as they 

 are occupied in growing, rather than in bearing : as they get older, 

 the peculiarities of the individuals assert themselves, and the 

 behaviour becomes less regular ; hence this estimate of the number 

 of trees required must depend on the nature of the experiment : if 

 the results are to be measured by the crops borne when the trees 

 have matured, the minimum number given above for trees should 

 be increased two- or three-fold. With bush fruits and smaller 

 plants, the results would nearly always be measured by the crops. 

 The nature of the investigation must also in other ways modify 

 the estimate of the number of plants required. For an isolated 

 experiment, where a plot under some special treatment is com- 

 pared with a plot under normal treatment, six trees to each 

 plot would not be enough ; but generally there will be many plots 

 bearing on the same question, and in that case, a minimum number 

 of trees may be used for each. Thus, in manurial experiments, 

 there would probably be a dozen different plots, each receiving 

 different doses of manure, and the regularity of the results in these 

 would usually serve as a check on too readily accepting abnormal 

 behaviour in any one plot. 



