386 A STUDY Ob' CORRELATION'S AMON(; TERRESTRIAL TEMPERATURES. 



have resulted in either a rise or a fall of the temperature of the earth in general. 

 Hence even if we accept as unquestionable the correctness of the bolometric measures, 

 it does not follow that there would be any corresponding change in the terrestrial 

 temperature. 



But Langley has brought forth wliat seems to be very strong evidence of a corre- 

 lation between the temperatures in widely separated regions of the globe, using a 

 method identical in principle with that of the present work, but including only the 

 year 1903. His material was derived from the Dekadenberichie of the Deutsche Seewarte 

 which gives ten-day temperatures in a great number of regions in various parts of the 

 globe. The latter was divided by Langley into seven great regions and the mean 

 departure found in each, on the same general plan that has been followed in the pres- 

 ent work. The fluctuations in the seven regions were expressed in the usual way by 

 curves, fi'om a study of which the conclusion that there was a marked synchronism 

 between the curves of temperature inter se seems quite plausible. The bolometric 

 measures suffered so many interruptions that the curve representing them is frequently 

 doubtful but, so far as it can be compared, there seems to be some correspondence 

 between it and the temperature curves. Yet, the method of eye estimates through 

 curves is one in which there is too much room for bias, and which does not admit of 

 sufficient precision of determination. The correlation thus exhibited is quite at vari- 

 ance with the general conclusions of the present work, though these would not preclude 

 the possibilit}' of a marked chance correlation through any one year. But even for 

 the special year 1903, it will be seen that the criterion of correlation is only 



A = 78 - 69 = 9 



which does not rise above the expected result of chance accumulation of accidental 

 deviations. 



In view of the fact that, in the present work, the year 1903 does not show any 

 well-marked correlation among ten-day temperatures, it will be of interest to trace out 

 tlie cause of the seeming divergence. It would be better that this should be done by 

 another ; but some comparisons by the present writer may serve at least as suggestions 

 on the subject. 



We remark at the outset that there is no inherent necessity that the fluctuations 

 in the seven regions selected by Langley should show any close relation with those of 

 the three regions chosen in the present work. Such a relation can only be regarded 

 as more or less probable according to circumstances. 



The question now presents itself how far the seeming divergence arises from 

 accidental fluctuations in the special data made use of, and how far to differences 



