A PHYSIOLOGICAL CHARACTER. 151 



were relatively large. The single test series, however, quite fails to 

 lend support to a difference in reactiveness between the two strains, 

 and this case likewise remains doubtful. 



Line 795 clearly shows a lack of selective effect during most of 

 the experiment. The last month's data is very suggestive of signi- 

 ficant reaction-time differences, but the data are too slight in amount 

 to receive serious consideration. 



A mutation possibly occurred in the minus strain of Line 740, 

 causing it to be more reactive to light. The case was critically 

 examined (page 89) and the conclusion was reached that the unusual 

 reactiveness of a few broods of this strain was probably due to 

 peculiar local environmental conditions. 



None of the above possible cases of differences in reactiveness 

 between the two strains of a line is well enough supported by the 

 evidence to be considered proven. The cases best supported by the 

 evidence are in Lines 689, 719, and 711. In Lines 689 and 719 the 

 differences are in the reverse of selection. With Line 719 the 

 difference was slight during the latter portion of the experiment. 

 Line 689 was lost before enough evidence was obtained to convince 

 one that the difference was necessarily genetic. Line 711 shows con- 

 sistent differences in accordance with an effect of selection, but, like 

 Line 689, was lost before enough evidence was obtained to convince 

 one that the difference was certainly genetic. 



With Line 757, on the other hand, a marked effect of selection is 

 attested by every significant bit of evidence which the data for this 

 line provide. The divergence in mean reaction-times began to appear 

 after the first 3 months at the beginning of the experiment (for which 

 the two strains had the same reaction-time means). For the second 

 (two-month) period the means differed by 25 22.82 seconds, and 

 the means for the following two-month period differed by 72 17.78 

 seconds, a difference of statistical significance (4.05 times the 

 probable error), in spite of the small numbers and consequently large 

 probable error. This is followed by 6 two-month periods, during 

 ^ which the differences are all in the same direction and 3 of which 

 (even for the shorter periods) are of large statistical significance, 

 though the differences of 55, 63, and 70 seconds were not of 

 statistical value because of large probable errors. For the two-month 

 period June- July 1914 the minus strain was the more reactive and 

 the difference was 132 29.79 seconds, a difference 4.4 times the 

 probable error. For the following two-month period the difference 

 is only -2 seconds. From this period on, all the 16 differences (by 

 two-month periods) indicate greater reactiveness in the plus strain. 

 With 3 exceptions the differences were large and of statistical value, 

 even when considered by two-month periods. One of these smaller 

 differences, which was not of statistical value, occurred within 3 



