20 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. I07 



1/6 percent, as given at page 163 of volume 6 of the Annals. But 

 the reader should recall that, on the average, 7/10x310 days of ob- 

 servation are included in each column of the tabulation giving the 

 weighted general mean. Hence the probable error of a single day is 

 reduced to iV2i7=i/i5 of 1/6 percent for each point in figure 9. 

 Therefore its smoothness is no surprise, despite its small amplitude. 

 The form of the curve of figure 9 is interesting. There is the rapid 

 rise and slower fall so characteristic of Cepheid variable stars, and of 

 the sunspot curve also. But if we suggest the same cause of variation 

 for the sun as for the Cepheid stars (i.e., expansion and contraction 

 of the orb) I am informed privately by Dr. Merrill of Mount Wilson 

 Observatory that the period of 6.6456 days is not theoretically possible 

 with such a star as the sun. Furthermore the Cepheids are from 100 

 to 100,000 times as bright as the sun. So, for the moment, we must 

 regard the sun as the type specimen of a new class of regular variable 

 stars. Stellar photometry does not reach, as yet, I think, accuracy 

 sufficient to enlarge this class to include other objects, for the ampli- 

 tude of this variation of the sun is only about 1/200 stellar magnitude. 



12. ANSWERS TO FOUR INQUIRIES PROPOUNDED UNDER (9) 



Answering the four questions proposed above regarding solar 

 variation : 



1. The form and amplitude of the mean curve of solar variation in 

 the 6.6456-day period is as given in figure 9. 



2. The sun's emission of radiation began to increase on December 

 31, 1923, not on January 3, 1924, as supposed. It reached maximum 

 on January 3. 



3. The solar-constant values are too imperfect and scattering to 

 give directly a satisfactory answer to the question whether different 

 recurrences of the period are of different amplitudes of variation. I 

 shall return to this point below. But it is clear that the large differ- 

 ences of amplitude in terrestrial temperature responses are not paral- 

 leled in the solar emission. In proof of this, see figure 10, where are 

 given two curves. One is the mean of the solar-constant changes on 

 80 recurrences of the period, when large temperature variations were 

 observed. The other curve is the mean of solar-constant changes on 

 54 occasions when small temperature variations were observed. There 

 is little difference in the amplitudes of the two curves, and indeed the 

 one corresponding to small differences of temperature happens to be 

 of slightly the greater amplitude. 



4. The solar changes agree in phases precisely with computed times. 



