ALCOHOLISM AND BEHAVIOR OF WHITE RATS 



275 



6. Comparison of the tests and controls on the basis of perfect 



trials 



a. Number of perfect trials. The error data provide another 

 set of comparisons, namely, the trials on the point of the scale 

 of errors, those with no errors, or perfect trials. Table 19 

 presents the averages of the number of perfect trials for each 

 rat for training and retention, with the strains separately and 

 together. In this table the plus sign is given to the ratios when 

 the tests' averages are lower than the controls'. 



TABLE 20 

 Stavdard deviations of the numbers of perfect trials when all strains and both sexes 

 are put together; based on 'completes' and 'incompletes.' The differences and their 

 significance in terms of their probable errors are given. The plus signs indicate 

 that the tests are less variable than the controls. The tests are significantly less 

 variable than the controls in the number of perfect trials they made in training; 

 in retention the difference lies in the same direction and is very nearly great enough 

 to be considered significant 



The ratios indicate that the tests had fewer perfect trials than 

 the controls; the females in strain A make the only exception. 

 Combining the strains removes all minus signs; that is, the tests 

 in all cases have fewer perfect trials. The differences between 

 the averages, when the sexes and strains are combined, favor 

 the controls; for training, the difference is +1.28 ± 0.40, which 

 is 3.07 times its probable error; for retention, +1.62 ± 0.40, 

 which is 4.05 times its probable error. These differences are fully 

 significant. In figure 15 the distribution of the numbers of 

 perfect trials of each rat are shown for training and retention. 



Compared by the standard deviations of the numbers of perfect 

 trials, the controls are more variable than the tests (table 20). 

 The difference is 3 times the probable error for training and 2.53 

 times for retention. 



This is the only criterion that gives a significantly greater 

 variability for the controls. But obviously this has a different 



