EP^FECT OF CONJUGATION 347 



conjugants; twenty-one days for the non-conjugants). The records 

 given are the numbers of fissions that have occurred during the two daj^s 

 ending on tlie date at the head of the column. (In only one case, for 

 the non-conjugants, in the column headed November 6, is the elapsed 

 period three days instead of two.) 



Each paix or split pair consisted of the two mated individuals a and h. 

 From each a and h the two sister lines x and y were kept in progress. 

 Thus from each pair there were derived four lines, ax, ay, bx and by. 

 But the lines from both a and b were kept throughout the experiment in 

 but few cases (IG in the pairs, 22 in the spht pairs). 



The lines from a and b were kept in separate moist chambers and 

 changed at different times, so that there is no opportunity for resem- 

 blance between them to arise through special similarity of treatment. 

 The two lines x and y, from a single individual, were however kept in the 

 two concavities of the same slide, in the same moist chamber, and 

 changed in succession. (This was for convenience in replacing one 

 from the other, but in repeating such an experiment, x and y should be 

 kept in separate moist chambers and handled separately; otherwise the 

 significance of any correlation between x and y is not entirely clear.) 



In working out constants of variation, the period August 27 to Sep- 

 tember 6 (twelve daj^s) was considered the 'first half for the conjugants; 

 August 28 to September 6 (eleven days) for the non-conjugants. The 

 second half for the conjugants included twelve days (September 8 to 18) ; 

 for the non-conjugants, ten days (September 8 to 16). 



The blanks left in the column under certain dates indicate that the 

 line in question died out on that date, and its place was supplied by 

 taking an individual from the sister line x or y, derived from the same 

 parent, that is, from the same a (or b, as the case may be). But a blank 

 in the final column of totals indicates only that the line in question did 

 not live independently throughout the experiment, but was supphed 

 from its sister line at some date, indicated as just set forth. 



In determining mean, standard deviation or coefficient of correlation 

 for any period or periods, only lines that lived independently throughout 

 that period are included. However, for the entire period, the few totals 

 included in parenthesis, in the last column, are employed also, since the 

 lines for which they stand coincided with another for only two or three 

 fissions at the beginning. 



In working out coefficients of correlation, for successive periods, it is 

 of course necessary to correlate any line with its real ancestral line, and 

 to do this it is necessary to pay careful attention to the blanks left in 

 certain columns and the replacement of certain lines which the}' indicate. 

 Thus, if in table 34 we wish to correlate the fissions in the first half of 

 the entire period with those in the second half, then Avhen, for example, 

 we enter the fissions for the second half of the time (September 8 to 18) 

 in line 8 by, we see that this second half descended partly from 8 bx; 

 there is no difficulty, however, in determining exactly how many fissions 

 occurred in the first half. We take in this case for the first half of the 

 period the sum of the fissions for 8 bx to September 2, plus those for 8 by 



