DEPARTMENT OF EXPERIMENTAL EVOLUTION. 



125 



from normal parents, while in experiment II similar criteria indicated that the 

 normals were more successful than those from alcoholic parents. Data are 

 quoted in the following table to support this statement. There are five 

 methods of comparison: (1) the average time per trial of all the trials; (2) the 

 average time per trial for the last ten trials; (3) the average number of " perfect 

 trials" (4 seconds or less); (4) the average number of "perfect days" (3 trials 

 of 4 seconds or less); (5) the number of rats with three successive "perfect 

 days." 



Puzzle-box results for experiments I and II. 



"The multiple-choice apparatus has previously been described (Year Book 

 for 1915, page 131). As explained, the method consists of teaching the rats 

 to go for food to a door that bears a certain relationship to the other doors 

 opened ; and the correct door, in different trials, is always a different one. Now, 

 this is obviously a more difficult proposition than learning to run behind the 

 puzzle-box in order to open the door to the food chamber. Instead of basing 

 the test on time the results of the multiple-choice training are based on the 

 numbers of right and wrong doors that are chosen. For every trial the path 

 taken by the rat is graphically recorded on a separate slip of paper, and the 

 doors that are correctly or wTongly chosen are indicated thereon. This 

 method was adopted partly to form a check on the results obtained from the 

 puzzle-box; as such it has clearly demonstrated that the characteristics of a 

 rat that make for the speedy solution of the puzzle-box problem are quite 

 different from those called into play by the multiple-choice method. The 

 same group of rats described as experiment I were trained on the multiple- 

 choice apparatus. Three problems were presented, as well as a memory test, 

 on the multiple-choice apparatus. In problem I the opened door farthest to 

 the right was the correct one; in problem II the opened door farthest to the 

 left was the correct one; in problem III the second opened door from the left 

 was the correct one. The averages of the numbers of correct first choices 

 and the averages of the numbers of wrong choices in these three problems are 

 shown in the annexed table. There can be no question that the normal rats 

 are superior, from the standpoint of this test, to those from alcoholic parents, 

 thus reversing the results of the puzzle-box training. 



"Since the two methods described have offered different results, the impor- 

 tance of a third method is manifest. Accordingly, to give further evidence on 

 the relative abilities of the different rats, a circular maze of the Watson type 

 has been built, equipped with a camera-lucida device for obtaining accurate 



