ALCOHOLISM AND BEHAVIOR OF WHITE RATS 277 



number of trials before the first perfect trial for strains separately 

 and together. The 'plus' before each ratio sign indicates that 

 perfect trials appeared later in the training of the tests than in 

 the training of the controls. In every case the averages of the 

 tests are higher; that is, they took more trials before making 

 their first perfect trial. When all the rats in each group are 

 put together, the difference between the tests and controls is 

 significant as tested by the probable errors ( + 5.66 rb 1.66; 

 that is, 3.41 times the probable error). Two test rats and one 

 control had no perfect trial during all the thirty-six trials. In 

 calculating the mean number of trials before the first perfect 

 trial these three rats were given the total number (thirty-six). 

 Many more trials may have been required, however, before 

 attaining a perfect trial. If these rats are excluded from the 

 averages, the difference is still about the same (+5.09 ± 1.44, 

 which is 3.53 times the probable error). The frequency distri- 

 bution of the numbers of trials, from which these averages have 

 been calculated are given in figure 16. Besides having made 

 fewer perfect trials, the tests required more trials before they 

 made a perfect one. 



c. Time spent in runnirig perfect trials. A third criterion 

 involving perfect trials is the time spent on running them. In 

 figure 17 are shown the perfect trials classified according to the 

 time spent; the frequencies are in terms of per cents of the total 

 numbers of perfect trials. Thus the differences in the numbers 

 of perfect trials for tests and controls is removed and the differ- 

 ence observed between the curves is due alone to the fact that 

 the tests spent more time in running their perfect trials than did 

 the controls. The curves are based upon 102 perfect trials for 

 the tests and 163 for the controls; the average number of perfect 

 trials per rat in training and retention together is, for the tests 

 4.08, for the controls 6.50. The average time spent on a perfect 

 trial is, for the tests 8.48 seconds, for the controls 7.52 seconds; 

 making a difference of 0.96 ± 0.25, which is 3.84 times its prob- 

 able error and accordingly a significant difference. 



The tests made fewer perfect trials, required more training 

 before making the first one and took more time in running them 

 than did the controls. 



