G. K. Gilbert—Special Processes of Research. 461 
of geologic formation, or of railroads and telegraph lines. 
There is in fact no limit whatever to the number of possible 
classifications of this nature, and as very few. of them can yield 
other than negative results, it is evident that the investigator 
cannot afford to take them at random. Some criterion of dis- 
crimination is necessary, and this he finds in the employment of 
hypotheses. The only bases of classification affording him 
results of positive value are the phenomena which are con- 
nected in a necessary or causative way with the phenomena 
under investigation, and the only bases he can afford to test are 
those which hypothetically have such connection. If he enter- 
tains the hypothesis for example that arboreal vegetation indu-. 
ces precipitation, he will classify the rainfall with reference to 
the distribution of forests; and if he finds a general corres- 
pondence but suspects that it may be deceptively produced 
through the influence of altitude on both rainfall and forest 
preservation, then he will also classify both rainfall and the dis- 
tribution of forests with respect to altitude. If he entertains 
the theory that the moon is concerned in the generation of 
storms, he will group his phenomena according to the phases 
of the moon. 
We find from the record that Prof. Davis employs succes- 
sively two external bases of classification. On the hypothesis 
that solar heat is directly concerned in the production of the 
storms, he groups them with reference to the hours of the day, 
finding a well marked correspondence. On the hypothesis that 
the conditions of their occurrence are produced by the more 
general movements of the atmosphere, he groups them with 
reference to cyclonic storms traversing the same district, and 
here again he finds a correspondence. In each case the corres- 
pondence is of such nature as to suggest other hypotheses and 
lead to other classifications. 
In platting the data concerning rain-front on the map of the 
district, and in the preparation of the composite portrait of the 
storm Professor Davis employed what are called graphic pro- 
cesses ; and the general graphic method to which these pro- 
cesses belong is of such importance in research that no apology 
is needed for devoting the remainder of my time to its consid- 
eration. 
It is a familiar principle of analytic geometry that every 
equation involving two variables may, through a process of con- 
struction, be represented by a line on a plane surface. The 
values of the two variables are referred each to an axis, or 
origin of distances, and each pair of corresponding values de- 
termines the position of a point. The line drawn through or 
constituted by all the points is then a graphic expression of the 
