A PHYSIOLOGICAL CHARACTER. 129 



These correlations are small and of those for single month-periods 

 only three or four can be credited with statistical value, those for the 

 May, August, September, and October 1914 periods. Two of these 

 are positive and two are negative correlations. Five of the correla- 

 tions from March to September are negative and two of these are 

 significant, both in numerical value and in comparison with their 

 probable errors. One of the two positive correlations for the spring 

 and summer period is very small, but the other is larger and of 

 possible significance. The correlations for October-February are 

 all positive, but only one is significant. The correlation for the six- 

 month spring and summer period is -0.0193 ±0.0175; that for the 

 six-month winter period is +0.0442 ±0.0150. These very limited 

 data suggest that negative correlations are possibly general for the 

 spring and summer months, while positive correlations are general 

 for the late fall and winter months. The correlation for the year 

 April 1914-March 1915, while only +0.1369, is 12.2 times its prob- 

 able error. 



These data cover the year-period of most consistent shifts in the 

 composite reaction-time curves for the D. pulcx strains (figure 10d) 

 and probably show the changes in reaction-time associated with 

 season and temperature at its maximum for D. pulex. 



Correlations between temperature and reaction-time for S. 

 exspinosus were confined to separate correlations for the strains of 

 Line 757 covering in each case the selection data for 19 months from 

 October 1915 to April 1917, the period during which occurred the 

 greatest differences in reaction-time between the two strains. The 

 correlations were, for the plus strain (689 individuals), +0.2428 ±0.0242 

 and for the minus strain (683 individuals), +0.2363 ±0.0244. Each 

 value is large enough to have a real meaning and is approximately 

 10 times its probable error. 



Hence there is shown for both species, and particularly for 

 S. exspinosus, Line 757, some significant correlation between reaction- 

 time and the temperature of the water in the experimental tank. It 

 is a question how much of this correlation, if any, represents a real 

 direct temperature influence. If the temperature influence were 

 direct, one would perhaps expect within the limited range of tempera- 

 ture occurring in the tests that higher temperatures would be asso- 

 ciated with greater reactiveness and a lowered mean reaction-time, 

 which would be indicated by negative correlations. Since the 

 significant correlations are mostly positive, higher temperatures tend 

 to be associated with less reactiveness and a higher reaction-time. 



It is possible that the lowered reactiveness with higher tempera- 

 tures is really due to reduced oxygen in the water used in the experi- 

 mental tank. The water, when brought in the day it was to be used, 

 was always at a lower temperature than the experimental room. As 

 it warmed up, the excess gas in solution appeared as bubble?. The 



