22 



SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. IO7 



above cited, the average cliange of the solar constant used was said 

 to be 0.7 percent. I have remarked above that only the largest solar- 

 constant changes could be discovered individually at that time, and 

 that favorable combinations of them with accidental errors tended to 



L949 



I.94B 



1.947 



/.946 



1.945 



1.948 



fS47 



/.946 



Fig. II. — Average variations of the solar constant, September observations. 

 In the upper curve no phase shifting is admitted. In the lower curve phase 

 shiftings were made corresponding to phase shiftings at identical times in 

 computing mean values of temperature departures at Washington. Evidently 

 no phase shifting is admissible in solar observations. 



give too high an estimate of some of their amplitudes. Having now 

 fixed the phases of the 6.6456-day period, it is possible to pick out 

 from the solar-constant record considerable numbers of cases of 

 small and of large observed change, which were surely real. Such 

 small and large cases could then be averaged separately, in order to 

 find out if the amplitude of the solar changes associated with the 

 6.64S6-day period is really variable. We have already seen that it 

 has no appreciable variability correlated with the temperature effect, 

 if averages of numerous cases are considered. 



