20 THE PHYLOGENETIC METHOD IN TAXONOMY. 



accurate judgment is capable of, together with a clear indication of the relation to 

 adjacent units as revealed by the bases of the curve. Of equal importance is the evidence 

 furnished by the modes as to the direction of evolutionary movement. This is true not 

 only of the species or variad, but also of the criteria themselves, and one of the most 

 significant services of the statistical method is the opportunity it affords of scrutinizing 

 criteria as they occur in nature. This constitutes an invaluable service at this time, when 

 experiment is still too rare to provide the results desired. The testing of criteria should 

 be the first task of quantitative studies, were it not for the fact that this can be done at 

 the same time that the modes and limits of species and variads are determined. 



The methods of statistical analysis can be applied equally well to species and their 

 variants as they exist in nature, to results of such outstanding natural experiments as 

 ecads, and to the forms obtained by experiments under control. However, the chief 

 value at present Ues in the application to the study of variation as it occurs in nature, 

 owing to the enormous mass of material demanding analysis. In this it is of the greatest 

 value in determining whether a segregate represents an actual fact of evolution and in 

 confirming its suggested relationship to the species as well as the nearest variads. While 

 the statistical method applies only to characters that can be measured or counted, or to 

 sequences, these are found to include practically all characters of importance in species 

 and variads. It demands a large amount of material to insure the most accurate and 

 comprehensive results, and the selection of this is of the first consequence. Random 

 selection is without value for careful analysis, and gives usable results only where it is 

 wished to obtain merely the total range of variation in a particular area or locality. 

 A preliminary examination is indispensable to permit the recognition of variants and 

 ecads, and to insure that the analysis runs parallel with these instead of cutting across 

 them. While analysis on the spot is always desirable, owing to the abundance of material 

 or the opportunity of checking, both as to forms and habitat factors, this is possible only 

 at field laboratories and substations. In the case of regions remote from these, material 

 must be collected with the maximum discrimination, and with especial reference to the 

 object sought. This may be the range of variation of any character or any organ with 

 respect to the individual, community, species, or variad. The use of fresh material is 

 always preferable, but often impossible, and the question of the use of dried or preserved 

 material depends somewhat upon the plant and part concerned, as well as upon facilities. 

 The chief precaution necessary is to measure while living a few of the parts to be studied, 

 in order to check out any alteration of size due to the method of preservation. As 

 statistical methods come more into use in evolutionary taxonomy, the results will be 

 regularly expressed in biometrical form, but at present tabulations and simple graphs 

 furnish all the facts required. 



Experiment. — The r61e of experiment in evolutionary taxonomy is even more funda- 

 mental than that of statistics, though in the beginning it rather supplements the latter, 

 owing to the time involved in complete experiments. Its unique importance hes partly 

 in the control possible throughout the entire process, but it arises chiefly from the fact 

 that it alone yields conclusive evidence as to descent. The exact use of statistical 

 methods affords practical certainty as to the origin and relationship of variads, for 

 example, but the actual origin has already occurred, and its nature and causes can only 

 be inferred. To bridge this gap, experiment is indispensable, and hence it must be the 

 final arbiter in all questions of origin and descent. In actual practice, experiment is 

 begun with the statistical studies or as soon after as possible, and the two are carried 

 forward hand in hand, so that each may profit from the other. 



As a matter of convenience, experiments may be designated as natural and artificial, 

 though the line between them is slight. Natural experiments are those in which various 

 migration agents have carried migrules into adjacent habitats and the individuals have 



