THE MEASUREMENT OF RESULTS 



21 



that of the length of new wood formed ; for, if the leaf-size were 

 reduced by 100 per cent, it would mean that there were no leaves, 

 i.e. that the trees were dead, whereas the growth may well 

 be reduced to nil while the trees are still very far from being 

 dead. But it has been well established by the observations them- 

 selves that all these measurements yield results of the same 

 character in ordinary cases, and that any one of them may gener- 

 ally be taken as a criterion of the condition of the tree. One 

 example of the numerous instances available may be quoted here : 

 this refers to eight sets of experiments with apple trees grown in 

 pots (only two or three trees in each experiment) under various 

 conditions (XIII, 82). The results are entered below in the 

 order of the vigour of the trees as measured by the length of 

 new wood formed, whilst the values in the other columns are 

 those given by taking measurements of seven other features. All 

 of these eight sets of measurements lead to practically the same 

 order of merit, there being only two out of the eight experiments in 

 which any anomalous results were obtained (enclosed in brackets), 

 and, as a matter of fact, in one, if not in both of these, there was a 

 good reason why the results should be anomalous. 



60 



(76) 



63 



(84) (53) 



The features of growth which were generally found to afford 

 the simplest means of ascertaining the well-being of a tree, were 

 the length of new wood formed, or the leaf -size, selecting the sixth 

 leaves from the shoot ends. 'The former exhibits far greater 

 differences (often several hundreds per cent.) than the latter with 

 trees under different treatment, and it is free from the obvious 

 source of experimental error which is inherent in any method 

 dependent on the selection of particular leaves ; but it cannot well 

 be applied to trees of any considerable size. Taking the results 

 with 200 apple and pear trees of six years of age, the measure- 

 ments of those which had been similarly treated indicated a 

 probable error of i 20-5 per cent, in any one tree due to its 

 individual peculiarities, or an error of db 8-3 per cent, in the 

 values from a plot of six trees. 



