6 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 8/ 



of approximation is shown by the smallness of the residuals in curve B. 

 It is to be noted that larger residuals are found in the earlier years 

 when the solar observing was less perfected than it became later. 



These results on periodicities have been obtained by Mrs. Bond with 

 an instrument which we call the periodometer, shown in Plate 2. It 

 was constructed with the aid of a grant of $1,000 from the Research 

 Corporation of New York. Its purpose is to discover and evaluate 

 periodicities in long series of observations. It does not recognize the 

 reality of any period until tested, and it evaluates its distribution in 

 amplitude without regard to any assumed mathematical expression. 

 It appears to us, for instance, that as the curve of sun-spot frequency is 

 well known not to be of regular sine form, there is no reason to 

 suppose that other solar periodicities should have a sine form. Hence 

 our instrument is designed to evaluate their forms as Nature fixed 

 them, not according to the forms assumed in mathematical series and 

 harmonic analyzing machines. 



Curves C, D, E, F, G, H, of Figure 3 show the periodicities actually 

 discovered in the solar radiation by aid of the periodometer. It will 

 be seen that the 21 -month period betrays also one of 7 months. In the 

 cases of the shorter periods, we have been able to separate the data into 

 several groups and independently evaluate the periodicities at several 

 epochs. These partial determinations are shown in curves Ci, C2, C3, 

 Di, D2, D3. In such cases we have been encouraged to find that the 

 maxima and minima occur without change of phase in these inde- 

 pendent epochs. Thus we regard the periodicities found as having 

 reality and permanence. We ventured in November, 1930, to make a 

 forecast for 193 1 and 1932 of the probable march of solar variation.' 

 Thus far it has been well verified, although it called for solar-constant 

 values almost all the time since 1930 about one per cent above the mean, 

 notwithstanding that the values preceding the date of forecast for 

 several years had been prevailingly below the mean. 



It has been of great interest to us to note that several of the 

 periodicities found in solar variation are closely related to the sun-spot 

 period of ii^ years or 135 months. Thus, 68 months is within its 

 probable error one-half, 45 months one-third of 135 months. Again, 

 if we take a period approximately three times as long, or 400 months, 

 which is near the Bruckner period, 25 months is one-sixteenth, 21 

 months closely one-nineteenth, 11 months closely one-thirty-sixth, 

 8 months is one-fiftieth, and 7 months one-fifty-seventh of its duration. 



If we admit provisionally (subject to the findings of subsequent 

 years) that the solar variation is made up of the seven periodicities 



^ See Smithsonian Misc. Coll., vol. 85, no. i, fig. 3, I, 1931. 



'J.\Ul 



