EXPERIMENTAL EVOLUTION. 119 



food when it entered the correct compartment (at the extreme right or left, as 

 the problem might be — the 'end-compartment'). The steps in the training 

 were these: Preliminary training, when the regular series of doors was opened 

 in successive trials, but the rat was fed upon entering any compartment (2 

 days, 20 trials) ; right-hand problem, when the rat was fed only when it entered 

 the open compartment at the extreme right (10 days, 100 trials); the same 

 problem with a different series of doors open (2 days, 20 trials); left-hand 

 problem, when food could be obtained only upon entering the open com- 

 partment at the extreme left (10 days, 100 trials); same problem with a 

 different series of doors open (2 days, 20 trials) ; after a month, the same problem, 

 with the same series of doers open as in the last step, to test retention (4 days 

 40 trials). Previously (Year Book 1919, 126) only the numbers of correct 

 first cho'ces and the number of wrong choices for the different parts of training 

 weie summarized, involving no comparison with the preliminary training, and 

 no comparison was made of the different claj's in the same step of training to 

 show the progress of learning. For the consideration of the progress of learn- 

 ing the trials have been grouped by twenties; that is, two daj^s are taken at a 

 time. The end-ccmpartments seem to be especially attractive, irrespective 

 of training, since about 70 per cent of the first choices are end-compart- 

 ments through all the training, although only 57 per cent of the open com- 

 partments are end-compartments. The nevv'' summaries are based on the 

 numbers of left end-compartments, and the numbers of right end-compart- 

 ments that were chosen first in every 20 trials of the whole training; besides 

 these, there have been studied the numbers of wrong choices, including end- 

 compartments in the same sets of trials. There were 21 test rats used and also 

 21 controls. 



"Figure 1 shows the percentage of the trials on which right and left end- 

 compartments were respectively chosen first, for each of the sets of 20 trials 

 in the whole training. The percentage for the test rats is shown bj^ the broken 

 lines, for the control rats bj^ the solid lines. Trials on which the right end- 

 compartment was chosen first are indicated by the heavy lines (both broken 

 and solid) and trials on which the left end-compartment was chosen first are 

 shown by the light lines. 



"The right-hand first choices (heavy lines in figure 1) will first be con- 

 sidered. Both the tests and controls increased the percentage of right end- 

 compartments in the right-hand problem. For the 5 points in the curve 

 showing the regular right-hand training the tests have fewer right-hand 

 choices, but when the different series cf open compartments was given Dn the 

 eleventh and twelfth days, the tests made more right-hand first choices than the 

 controls. The general superiority of the control percentages disappears 

 when it is observed that the tests started out in their prelimmary training 

 with fewer right-hand choices than the controls. If the slant cf the curves is 

 compared, it appears that the tests made more rapid and greater progress. 

 On the other hand, it may not rightly be claimed that the tests were actually 

 better than the controls, since the higher initial right-hand tendency of the 

 controls may have in itself limited the opportunity for rapid improvement. 

 The highest proportion of correct choices reached during the 120 trials of this 

 problem is little over half of the trials. In the left-hand problem the number 

 of right-hand first choices is immediately reduced in the first 20 trials. The 

 tests and controls are both lower than in their preliminary trials. In the 

 following sets of 20 trials there is no further reduction in the percentage of 

 right-hand choices; the controls remain about the same, but the tests chose 

 more right-hand doors than the controls, although the initial tendency of 

 the tests was to choose fewer right-hand doors. 



