EFFECT OF LENGTH OF BLIND ALLEYS ON MAZE LEARNING 51 



The present experiment was devised to present conditions 

 which might test the efficiency of the " completeness of response " 

 theory outlined recently by the writer, suggesting a means of 

 learning based on the overlapping and thereby simultaneously 

 operative effects of successive stimuli. Identical mazes were 

 used for separate groups of animals, but they were so arranged 

 that their several cul de sacs could be conveniently varied in 

 length. By this means control groups of rats were run in mazes 

 differing only in the relative lengths of their cul de sacs, certain 

 of these being long for one group and short for the other. Two 

 modifications of such differences were used; the one pair of 

 control groups having all the blind alleys short in the one maze 

 and all long in the other, while the other pair of controls each 

 had some long and some short blind alleys making the total 

 length of blind alleys equal for both members of the pair. In 

 all, twenty -four rats were used. Groups of rats were interchanged 

 in the mazes, after the first problems were completed, so that 

 each problem was tried both by trained and by untrained animals. 



Detailed records were kept of the behavior of the animals. 

 Complete entrances, half way entrances, and beginning entrances 

 to cul de sacs were indicated; complete returns and returns to 

 blind alleys already passed were noted; and the direction of the 

 rat's movement on emerging from blind alleys, whether forward 

 or back, were recorded. The exact time for each trial was 

 kept but not used in the present report. 



we have: forward from the blind alley 3/4 + 1/4 = 1; into the blind alley 1/2 + 

 1/2 = 1; and return toward the starting place in the maze 1/4 + 3/4 = 1. This 

 gives equal exercise to the acts in all these three directions, on pure probability 

 laws, when returns are considered. But the animal is taken out of the maze only 

 at the food end of the trail; hence for each trial there must be one more forward 

 run at any given place in the maze than return runs. This gives the forward direc- 

 tion an advantage statistically of 1/4 runs for each trial over the entrance to the 

 cul de sac and of 1/2 over returns. This advantage is proportionately small where 

 many returns are made, as near the maze entrance, and large where this is not 

 the case, as near the food and later in the learning process at any given point. No 

 returns will be made to the last cul de sac, and when entrance to it has been elim- 

 nated none will be made to the one next to it; and so on. This condition, then, 

 affords a fine theoretical basis for explaining learning in the maze, and also for the 

 backward elimination of errors of entrance to blind alleys. There are, however, 

 serious flaws in this argument when given in support of frequency, either alone or 

 combined with recency, as the only principle operative in the learning of the maze. 

 Frequency and recency factors really operate against this explanation, rather than 

 in favor of it; for they favor the mere repetition of the choices first made at any 

 of these critical positions in the maze, and therefore the strengthening of the im- 

 pulses to enter blind alleys rather than their weakening. The force of this point 

 will be shown concretely in the paper now in preparation. 



