472 AUSTIN RALPH MIDDLETON 



in two lines, three times; and in one line four times. When the 

 differences between the number of generations produced by the 

 corresponding lines during the whole ninety days were ascer- 

 tained it was found that in not a single case had the slow line 

 produced more generations than its fast one. And in only two 

 lines was the excess of the fast over the slow probably too small 

 to be significant (fast line ten produced only five more gener- 

 ations than slow line ten, and the two lines number fourteen 

 produced the same number of generations). Furthermore dur- 

 ing each ten-day period the total number of gerierations produced 

 htj all the thirty fast lines was much larger than that produced by 

 the slow lines, and the average number of generations per line per 

 ten-day period was uniformly greater for the fast than for the slow 

 lines. Also the per cent of the difference in proportion to the 

 total number of generations produced by both sets was calcu- 

 lated and found to be remarkably uniform; these percentages 

 are shown in the extreme right hand column of table 6. Finally, 

 the average number of generations per line per ten-day period 

 for each set of lines is plotted as a polygon in figure T-a, giving 

 a graphic representation of that phase of table 6. 



Figure 8 gives the curve of the daily differences between the 

 average number of generations produced by each fast and each 

 slow line. On only three days during the whole ninety days 

 of this balanced selection experiment was this difference too 

 small to be significant (on the eighth and sixteenth of February, 

 1914, both sets of lines produced the same number of generations 

 and on February 18, 1914, the difference was 0.03 in favor of 

 the slow set). The average per day for each fast line, including 

 the three instances just cited, was 0.251 generation greater than 

 each slow line. From table 6 and from figures 7 and 8 it is there- 

 fore evident that, when measured by the test of balanced selec- 

 tion, the eighty days of opposite selection had produced a differ- 

 ence of fission rate between the two sets of lines that is heritable. 

 Furthermore, analysis of the daily records shows that this result 

 is not due to the chance isolation of a 'mutating' line in either set 

 of lines and the subsequent development of the thirty lines of 



