No.eW ALCOHOL AND WHITE BATS 293 



the outside to the center required a rat to turn alternately 

 to the left and the right at successive doorways. A rat's 

 training was started at the age of 56 days, after prelimi- 

 nary feeding in the center of the maze on each of the 7 

 preceding days. Three successive trials a day were given. 

 After the first and second trials the rat was removed 

 from the center as soon as it had tasted the food (bread 

 and milk) which was always found there ; after the third 

 trial, it was allowed to eat for five minutes. This train- 

 ing was given for eight successive days. The observa- 

 tions were so automatic that there was practically no 

 possibility that the results were being influenced by an 

 unconscious bias on the part of the observer. In the 

 case of the treated rats the alcohol was given each day 

 following the trials in the maze. 



Results. — The average time per trial for each day of 

 the training of the different groups of rats is represented 

 in Fig. 1. The test rats, whether actually treated, or 

 the descendants of treated rats, are represented by the 

 broken lines, and their respective controls by the. solid 

 lines. The numbers of rats included in the different 

 curves, beginning at the left, are as follows: 55 treated 

 rats and 62 controls ; 46 tests and 48 controls ; 25 tests 

 and 25 controls; 8 tests and 20 controls. The broken 

 lines tend to lie above the solid lines. The tests tend to 

 give higher time averages than the controls, that is, the 

 tests took longer time to run a trial. The inferiority 

 shown by the treated offspring from treated parents 

 (fourth pair of curves), and by the untreated offspring 

 from untreated parents and treated grandparents (third 

 pair of curves) is of the same order of magnitude as 

 that shown by the treated animals themselves ; untreated 

 offspring from treated parents show less inferiority than 

 their own untreated offspring. Considering the signifi- 

 cance of the differences between the tests and controls 

 for each (1m v indopondontly, the following results are 

 found: the (linVi-fiiccs ])Ptwoen the tests and controls are 

 over three ihiwfi tlicir probable errors on five days in the 



