460 G. K. Gilbert—Special Processes of Research. 
of the storm itself. In the case of cyclonic storms the area of: 
lowest pressure has been thus used, its terrestrial meridian 
serving as an axisof orientation. In the case of thunderstorms 
no such area of lowest pressure had been determined, and it 
was necessary to select some other feature or features to serve 
the same purpose. The second classification was therefore 
made as a means of overcoming a practical obstacle to the 
accomplishment of the desired classification. 
The preliminary classification, or more exactly group of clas- 
sifications, having been decided upon, its particular nature was 
determined by the antecedent knowledge that thunderstorms 
ordinarily move across the country, and by the hypotheses that 
the first rain, or the heaviest rain, or the loudest thunder, or 
some other definite event might be found to be so constant in 
its relation to the storm as to constitute a suitable basis for the 
third classification. 
The change of basis in classification, whether it belonged to °* 
the original plan of the investigation or was devised under the 
stimulus of a discovered obstacle, finds its analogues not only 
in other researches but in other departments of human expe- 
rience. The quarryman who wishes to overturn a heavy block 
inserts his bar beneath it wherever it finds a crevice, even 
though the point of insertion be not the most advantageous for 
his general purpose. Having done so, he lifts until there is 
room for the lever at the point of vantage, and then shifts his 
position. The algebraist by premeditation eliminates from his 
equations all but one of the quantities whose values he seeks, 
and having determined that one, undertakes with its aid the 
solution of one of the original or intermediate equations. 
If then our interpretations are correct, the bases of clasifica- 
tion, like the methods of observation, sprang from analogy with 
other researches, from previous knowledge of the subject in 
hand, and from hypotheses in regard to the results, and were 
conditioned by obstacles. 
As already remarked, the basis of classification need not 
itself belong to the subject under investigation. The phenom- 
ena of the New England thunderstorms may be compared with 
respect to distribution in time with any other phenomena or 
events occurring in the same time, or they may be compared 
with respect to distribution in place with any feature of their 
district. Thus itis entirely possible to group any of the thun- 
derstorm phenomena according to the days of the week, or 
with reference to the daily records of marriages in London, or 
the fluctuations of a mining stock, or the success of the Phila- 
delphia base ball nine, and it is equally possible to compare 
their distribution with the topographic relief of the country, 
with the density of population, with the distribution of forests, 
