282 E. C. MACDOWELL AND E. M. VICARI 



the alcohol treatment. Brother-by-sister matings were made 

 throughout. The rats were given three trials a day for eight 

 days on a circular maze {training), and after thirty-one days 

 (during which time they were trained in our multiple-choice 

 apparatus) they were given four days more, twelve trials {reten- 

 tion). Sixty rats in this generation were trained, thirty-one 

 tests and twenty-nine controls. The records of ten of these 

 rats include one or more trials that were not successful; these 

 rats have been called 'failure' rats. No satisfactory method of 

 treating the successful records of these rats has been found; 

 accordingly, they have been omitted from most of the sum- 

 maries. In order to show that the results so obtained do not 

 depend upon the omission of these rats, they have been included 

 as well as excluded in the summaries based on the criterion of 

 time. 



1. During the period of rapid learning the test rats spent 

 more time than the controls in running their trials. In reaching 

 this conclusion, the averages of the tests have been compared 

 with the averages of the controls in each of the following group- 

 ings of the data: males and females separately and together in 

 each strain by itself and in all strains combined for each of the 

 following groups of trials: first half of training, secorid half of 

 training, omitting the first day, training, retention, training and 

 retention, and each day of training and retention by itself. In 

 the group of rats mcluding all strains and both sexes the averages 

 of the tests are higher in each of the eighteen groups of the trials. 

 The probable errors of the differences between the averages of 

 the tests and controls for the first half of training, omitting the 

 first day, training, training and reteiition, first, second, third, 

 fourth days of training and the fourth day of retention, show that 

 these differences are significant. The frequency distribution of 

 the ratios of all the test trials to corresponding points in an 

 assumed standard learning curve differs from the frequency 

 distribution of the ratios of all the control trials to the same 

 standard learning curve. WTien these two distributions are 

 compared by means of the x" test, it is found that the probability 



