EFFECT OF LENGTH OF BLIND ALLEYS ON ^L\ZE LEARNING 5 



After pointing out, successfully to the writer's mind, the difficulty 

 in the way of Thorndike's principle of " satisfiers," he contends 

 that there is no immediate connection backwards between the 

 obtaining of food and the elimination of errors. Watson attempts 

 on the basis of the probability doctrine, suggested in another 

 relation by Stevenson Smith, to show how frequency alone may 

 suffice in the acquiring of maze habits. He argues that an 

 animal, having started along the maze path A, has an equal 

 chance on coming to a ciil se sac X, all other factors equal, either 

 of taking B, the true path beyond the bhnd alley, or of going 

 into X; that on returning from X, in case of the wrong choice 

 having been made, it again has an equal chance of taking B. 

 It thus has a probability of 3 4 (or 1 2 + 1 2 of 1 2) of keeping 

 the right path. 



If no other factor than frequency operates in such a case we 

 should expect an animal to continue entering the cut de sacs 

 indefinitely; for on turning back from any point toward the 

 starting place in the maze the same law must apply. The 

 chances are again 1 2 that any cul de sac passed will be entered, 

 and 3/4 that the animal will continue in its general direction, 

 now toward the starting point in the maze. In a maze with 

 several blind alleys, each of which has a chance of 1 4 of turning 

 any rat reaching it back toward the maze entrance, the proba- 

 bility would be verv' slim that the animal would at the first 

 trial reach the food. The returns would therefore tend to fix 

 the habit of entering cul de sacs as strongly as that of going 

 toward the food. Mere probability explains truly enough how 

 the animal gets to the food each time, but that is not the problem 

 of learning; it does not explain how it happens that on the whole 

 the second trial is better than the first, the third better than 

 the second, and so on. Frequency based on probability does 

 not bring such a result : it fails utterly to explain learning, even 

 in the simple case of the maze.** The real issue has been over- 

 picked at random, instances occur in numerous places of \nolations of the principle 

 stated. A detailed presentation of these instances will be reserv^ed for a later 

 article, as proper attention to them here would lead us too far away from the main 

 purpose of the present paper. Instances are ver^' frequent when the animal takes 

 certain blind alleys entirely contrary to the expectations based upon either fre- 

 quency or recency or of both combined. 



'^ This statement is based on actual data of a supposed case of a rat in a maze 

 of six cut de sacs whose " choice " at each bifurcation of the trail is determined by 

 the flipping of a coin. After considerable data by this method has accumulated — ■ 

 after most any number of trials — it becomes very evident that if the frequency 



