200 



These results indicate that the second method is feasible, and 

 that uncertainty in choosing begins at nearly the same absolute 

 stimulus- value whichever method is used. The relative ease of 

 learning for the chick is decidedly in favor of method No. 1. 

 Chick 1 failed to learn to discriminate when trained by method 

 No. 2, but established a perfect habit in 443 trials when trained 

 by method No. 1, despite numerous disturbing factors. Chick 2 

 learned by method No. 1 in 214 trials, as against 585 trials for 

 method No. 2. The results given in table 3 should not be taken 

 as indicating with accuracy where the bird's stimulus-threshold 

 lies, as the training was interrupted too soon. I did not wish 

 to risk spoiling the bird by frequent punishment for work on 

 the main problem. 



DIFFERENCE-THRESHOLD FOR DIRECTION 



The term "threshold" is properly used to designate the mean 

 between effectiveness and ineffectiveness of a stimulus or 

 difference between stimuli, in producing a response. In the 

 present discussion this mean is assumed to have been reached 

 when the stimulus-difference is effective in half the total of a 

 fairly large number of presentations or " trials." If the differ- 

 ence were wholly ineffective, the animal should be expected, on 

 the theory of probability, to respond correctly in 50% of a 

 large number of trials. If the difference were always effective, 

 100% of the responses should be correct. The stimulus-differ- 

 ence at which the percentage of correct choices is 75 is therefore 

 taken as the animal's " threshold." The accuracy of such a 

 determination is indicated by the consistency of the results. 

 The measure of consistency used herein is the mean variation 



