132 F. B. SUMNER, M. E. MCDANIEL AND R. R. HUESTIS. 



them, on the contrary (including one of the fullest years) give 

 little indication of such relations. 



More instructive are the results of combining these years into 

 three groups of two to three years each, and plotting the seasonal 

 cycle from the combined data for each of these groups (Fig. 3). 

 The resulting graphs require no further discussion. 



105 



95 



85 



Feb-Ap r 



~Jul Aug-Oct 



Nnv-Jan 



105 



85 



Jan-Mar 



Apr- J un 



Jul-Sep 



Oct-Dec 



FIG. 2. Sex ratio graphs obtained by combining monthly records into 

 " seasons " of three months each. The lower graph is based upon date of con- 

 ception, the upper upon date of birth. 



The data for the separate years were also grouped into 3- 

 month periods, and graphs plotted for inspection (not repro- 

 duced). In 6 cases out of 7, there was a well-marked fall from 

 the first to the second period, just as in the lot as a whole (Fig. 

 2). In 6 cases out of 7, likewise, there was a rise from the 

 second to the third period. In only 3 cases out of 7, however, 

 was there a fall from the third period to the fourth, one line re- 

 maining horizontal and the other three rising. (One of these 

 last is based upon only 19 individuals.) 



