HABIT FORMATION IN THE DOG 31 



breathing will be noticed even by a blind animal. There are 

 other movements of like kind, which serve as auditory stimuli. 

 Now, there is only one reliable means of preventing disturb- 

 ance from these sources : Not only the experimenter but all others 

 should he outside the room in which the animal is u'orking. Roth- 

 mann's Diener or any bystander can disturb the dog as much 

 as could the experimenter, and indeed is more likely to do so, 

 as the experimenter if he is honest is apt to take greater pains 

 to be on his guard, fleeting this condition does not preclude 

 the experimenter's putting the animal through the proper reac- 

 tions in the first two or three days' work, although it is doubt- 

 ful if this is desirable; but as soon as the animal is left to dis- 

 criminate, the experimenter should leave the experiment room. 

 In control tests, even though he be in another room, it is highly 

 expedient that the experimenter should not watch the animal 

 while making his choice, but wait until the choice is made. This 

 will prevent the animals being given any auditory "unconscious 

 helps." In my later work I observed this precaution, with 

 satisfactory results. 



2. The order of the presentation of stimuli. This must be 

 varied irregularly. In a given series of say loo presentations, 

 the animal should be given 50 of each. In actual training work, 

 I have found, as have also Yerkes, Watson and others, that it 

 is usually unsafe to give the same stimulus more than three 

 times in succession, as a position preference is apt to become 

 established. In control tests, after the animal has really learned 

 to discriminate, the number of successive presentations of the 

 same stimulus can be safely increased. The best method I 

 have used in training work for determining the order of presen- 

 tation is the use of a well shuffled pack of cards ; allowing those 

 of one color to represent one stimulus, and those of the other 

 color the other stimulus. If more than three cards of the same 

 color appear in succession the extra ones can be laid aside until 

 needed to break into an overlong series of the other color, or 

 until the bottom of the pack has been reached. This method 

 assures the variation being irregular, and eliminates the possi- 

 bility of the animal's successful reaction merely to the order 

 of presentation. The probability of the coincidence of an order 

 so determined with the animal's established preference, is ex- 

 tremely slight. As has been suggested above, the experimenter 



