A PHYSIOLOGICAL CHARACTER. 123 



general trend is very nearly that of a straight line. This general 

 form constitutes intrinsic evidence of the reliability of the curves 

 as a whole. 



The curve for Line 757 plus (figure 15) starts higher than the 

 level attained by the Line 740 plus strain for the same period. -This 

 is probably significant. The two-month period curve for the Line 757 

 plus strain held very closely to a level near 800 seconds for the first 9 

 months of selection. The curve for the plus strain of Line 740, 

 while prevailingly at a lower level for these periods, fluctuated from 

 672 seconds to 876 seconds. The mean difference between the 

 plus strains of Lines 757 and 740 for this nine-month period is 

 39 12.36 seconds, 3.15 times the probable error. Hence there is 

 evidence that at the start the plus strain of Line 757 was less re- 

 active than the plus strain of Line 740. 



During the first 31 months (to February 1915) of selection with 

 Line 757 the plus strain in general had a higher reaction-time than 

 the plus strain of Line 740 (table 48 and figure 15), which tends to 

 strengthen one's conviction that at the start the plus strain of Line 

 757 was less reactive than the plus strain of Line 740. But from 

 February 1915 the fairly consistent differences between the plus 

 strains of Lines 757 and 740 are striking and indicate that the plus 

 strain of Line 757 had become the more reactive of the two. Of 14 

 two-month periods the plus strain of Line 740 was the more reactive 

 only twice. One of these differences is slight (16 41. 5 seconds), 

 the other is large (246. 6 34. 2 seconds); but this difference arises 

 from the coincidence of an exceptionally low average for Line 740 

 plus (the lowest, by 104 seconds, of any two-month average for this 

 strain) and an exceptionally high average for Line 757 plus. 1 Aver- 

 ages by two-month periods are obtained from too few individual 

 reaction-time records to be highly reliable when considered singly, 

 so, while this difference is large, it must be considered in relation to 

 the other averages and evaluated accordingly. For the other 12 

 two-month periods of the latter half of the experiment with those 

 two lines, the plus strain of Line 757 was the more reactive as com- 

 pared with the plus strain of Line 740; the differences were from 22 

 to 240 seconds, and were respectively 3.6, 1.0, 0.7, 9.2, 1.6, 2.7, 

 3.1, 2.6, 1.7, 3.7, 1.4, and 6.4 times their probable errors (see table 

 45). That is to say, even when considered by shorter (two-month) 

 periods, 12 of 14 differences indicate that the plus strain of Line 

 757 was more reactive than the plus strain of Line 740, although, 

 with their large probable errors due to relatively small numbers, 

 only 7 of these differences are statistically significant. 



lr The mean for the December 1916-January 1917 period for the plus strain of Line 757 

 was obtained from the reaction-time records of only 36 individuals of 3 broods. The corre- 

 sponding mean for the 740 plus strain resulted from 41 individual reaction-time records. These 

 numbers are scarcely more than half the usual numbers for two-month periods. 



