260 



E. C. MACDOWELL AND E. M. VICARI 



the basis of each trial by itself. The averages for all the various 

 groupings of strains and sexes have been computed; the results 

 are in full accord with those obtained when the averages for each 

 day were compared. They are not presented because the addi- 

 tional evidence they would offer does not seem needed. 



d. What is the probability that the test data and control data 

 for distance are not random samples of the same population? From 

 the control averages of each of the twenty-four successive trials 



TABLE 12 



Standard deviations of the averages for each rat when all strains and both sexes are 



put together for each day of the training. Conventions as in table 11 



Training . 



Retention 



+92.0±51.8 

 +52.2±23.8 

 -h 7.7±19.0 

 + 6.1± 1.9 

 + 2.0±16.8 

 +15.5±14.1 

 -45.6±13.6 



- 1.3±13.8 



-18.7±10.5 

 +62.6±15.6 



- 4.1±11.1 

 +15.0±11.2 



DIFF./P. E. 



1.7 

 2.2 

 0.4 

 0.5 

 0.1 

 1.1 

 3.3 

 0.1 



1.8 

 4.0 

 0.3 

 1.3 



1 Omitting rat 1211. 



in training a curve was smoothed by interpolation; this was 

 taken as the standard of comparison in computing a ratio for 

 each trial of each rat. The distribution of these ratios above 

 and below the central point of 1.000, or unity, is shown in figure 

 12. Ratios at the left of the center are from trials less (shorter 

 distance) than the assumed standard; ratios at the right of the 

 center are from trials that were longer than the assumed standard. 

 The ratios for the tests are shown by the broken line; the ratios 

 for the controls by the sohd line. In the first graph the ratios 

 are classified in classes 1.699 in width; in the second graph the 



