ALCOHOLISM AND BEHAVIOR OF WHITE RATS 217 



temporary circumstances have occasioned this unusual behavior. 

 That just these rats and not others responded in this manner 

 may have been due to some particular sensitiveness, but the 

 alcohol treatment of the grandparents does not seem responsible 

 for any such hypothetical difference, since control rats as well 

 as tests have 'failures'; the numbers are four controls and six 

 tests, which, in such small numbers, is a very close approxima- 

 tion to equality. A second method of handling the data of the 

 rats that had 'failures' would be to omit the time spent on 'fail- 

 ure' trials, assuming that the 'failure' was due to circumstances 

 quite apart from the nature of the rat. But, as stated, this does 

 not eliminate the effect of the 'failure' upon the immediately 

 following trials. 



Perhaps the best way to include the 'failure' rats is to put 

 down a uniform time (ten minutes) for each 'failure' and treat 

 that as a successful trial; then, in averaging, count the number 

 of trials actually attempted. When a 'failure' occurred on the 

 first or second trials, there were no more trials attempted that 

 day. This method has been used for including the 'failure' 

 rats in the two main series of summaries of the time data, although 

 it is obviously not free from very real objection. In applying this 

 to the summaries based on the individual days, a further adjust- 

 ment was needed. Since the sum of the three trials for each 

 rat was used instead of the average per trial, it was necessary to 

 compensate in some way for the missing third trial when a 

 'failure' occurred on the second trial. This was done by increas- 

 ing the sum of the first trial and the 'failure' (ten minutes) by 

 one half of this sum. The two rats that lost a whole day through 

 failures on the first trials have not been used in these averages. 



Since no satisfactory method of including the data from these 

 'failure' rats has been discovered, the best method of meeting 

 the situation seems to be to omit these data. Accordingly, for 

 most of the criteria only the averages excluding these data are 

 given. But, in order to show that the results are not dependent 

 upon their exclusion, the averages for time, which may be con- 

 sidered the most important single criterion, are given both includ- 

 ing and excluding these 'failure' rats. Theoretically, both sets 



