CYCLONIC STORMS IN TEMPERATURE LATITUDES 513 



haps more mobile than the lower portion of the sun^s atmosphere where 

 sun-spots apparently originate. Therefore, when the exciting cause be- 

 gins to act, the earth may respond immediately, while a certain amount 

 of time must elapse before the sun becomes sufficiently disturbed to show 

 much evidence of it. This perhaps explains why in 1890 storminess has 

 become highly pronounced, although the spots of the sun are only begin- 

 ning to become active. The degree of importance to be attached to this 

 explanation is not great, however, for in 1894 and 1908 the maxima of 

 storms come a year after those of spots. The increase in storminess after 

 the spots have passed their maxima is, to be sure, only slight, but it de- 

 serves notice. At present our knowledge is too scanty to enable us to 

 determine the minor causes which induce the sun-spots and storms to act 

 differently in certain cases. 



To resume our analysis of the curves, in 1891 both rise, but the follow- 

 ing year shows disagreement. Thereafter for 20 years a high degree of 

 agreement is observed. It pertains not only to the major fluctuations, 

 but to such little variations as the minor maxima of 1898 and 1900, both 

 of which appear as irregularities in the sun-spot curve. In the last of 

 the three sun-spot cycles the agreement is more striking than anywhere 

 else. Each of the three irregularities in the upper curve is plainly appar- 

 ent in the lower. 



A test of the shifting of storm tracks by means of correlation coefficients 

 with sun-spots. — The relationship shown in the two curves of figure 9 

 forms, so to speak, a concrete summary of the main fact of Kullmer's law 

 of the shifting of storm tracks. It thus illustrates the most important sin- 

 gle piece of evidence on which is based the hypothesis here under discus- 

 sion. Therefore it should be tested most rigorously. The most severe 

 mathematical test which can be applied to two such curves is that of cor- 

 relation coefficients. It will be remembered that in computing these it is 

 first necessary to find the amounts by which each of two phenomena dur- 

 ing each of a series of years departs from the average for the entire series 

 of years. From these values a standard mathematical process makes it 

 possible to determine the exact degree of relationship. If the two phe- 

 nomena are absolutely unconnected the coefficient is zero. If they are 

 absolutely connected so that one never occurs without the other, and so 

 that their relationship can be expressed by an invariable mathematical 

 formula, the coefficient is 1. Such a relationship is like that of the rising 

 of the sun to the rotation of the earth, or of the driving wheels of an 

 engine, which can not turn except in harmony with the stroke of the pis- 

 ton. Other related phenomena, such as the amount of food which a man 

 eats and the amount of work that he does, are expressed by coefficients 



