CHAPTER V 



FLUCTUATION 



When we desire to compare different sets of data, we 

 njust either admit of no observational error,, or ascertain 

 separately what error exists in each set, or else employ 

 standard methods of known probable error. The last is 

 the most convenient method, since the yeiarly programme 

 of work can be standardised, and comparisons are easily 

 effected. The particular method employed for any one 

 characteristic, such as the form of the leaf, is a compromise 

 between the probable error and the facilities for 

 observation. 



The slide rule and nomographs, measuring dividers, 

 glass plates graduated in angles and lengths, the plani- 

 meter and pantograph, adding machine and Joly balance, 

 with a string of simpler appliances stretching down to the 

 oldest of all — a small-tooth comb — have been successfully 

 pressed into service with the object of economising time 

 in these researches. 



The pure strains on which these observations have beeij 

 .made were derived from single plants of various kinds. 

 A few of the offspring in every year during periods of 

 three to six years have been covered with mosquito netting^/ 

 to prevent natural crossing ; the resulting strains are not 

 visibly impure, so far as our present knowledge can 

 ascertain. 



