6 



SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. I07 



1936, and in figure 2 a graph of the mean march of temperature de- 

 partures for 20 days each, at Washington, associated with four cycles 

 of 6.6456 days each, occurring in that month. These data show a very 

 regular periodic march of about 14° amplitude, and, as nearly as the 

 mean of so few cases can indicate, of about 6.5 days period. 



16 



14 



12 



10 



m 



-s 



Fig. 2. — Separate data and mean march of temperature departures in Wash- 

 ington for four cycles of 6.6456 days each, occurring in March 1936. Abscissae, 

 days from zeroth day of cycle; ordinates, degrees Fahrenheit. (See table 2.) 



I have performed exactly similar operations on all the cycles of 

 6.6456 days in all months in all years from 1910 to 1945. All monthly 

 temperature data, so treated, agree to indicate the veridity of the 

 cycle of 6.6456 days. Comparing the results of individual months in 

 successive years, 1910 to 1945, there appears no progressive shift, 

 accumulating to as much as i day, from 1910 to 1945. But there do 

 appear in individual years phase shiftings of from i to 3 days from 

 the calculated positions. During any individual month, in any in- 

 dividual year, the four or five recurrences of the cycle usually return 

 approximately in the same phase, without shifting. But comparing 

 the results of one year with those of another, such phase shiftings as 



