PATTERN-DISCRIMINATION IN VERTEBRATES 201 



of the percentage of correct choices for the daily series of trials 

 from the average percentage of correct responses for the total 

 number of trials at a given stimulus-difference. 



In much of the past work on the special senses of animals 

 carried on by means of training methods, the experimenters 

 have trained the animals to respond to a given difference between 

 the stimuli, and have then reduced the difference until the 

 animals cease to discriminate. When a break-down occurs, the 

 experimenters frequently discontinue the training after a few 

 additional trials, and take the stimulus-value or stimulus-differ- 

 ence at the breakdown as the animal's " threshold." In some 

 earlier work I followed this procedure, having observed that in 

 some instances further training confused the animal and caused 

 it to sulk. In the present work I have endeavored to find the 

 degree to which such confusion can be overcome by continued 

 training. 5 The results obtained on the chicken show a training- 

 effect which is important, though not relatively large; while the 

 final results obtained on the monkey yielded a threshold of a 

 different order of magnitude from the stimulus- difference at 

 which the first break-down occurred. As appears in table 1, 

 in the records from August 22 on, the monkey showed a strong 

 tendency to break down at a stimulus-difference of 8° to 10°; 

 but he overcame this tendency after continued training, and 

 eventually yielded a threshold in the neighborhood of 2°. Had 

 I given the monkey a considerably larger number of trials at 

 the stimulus-differences of -13°, " -10°, +10°, +8° and +7°, 

 after the full effect of training had been obtained, the average 

 percentages of correct responses could be plotted against the 

 stimulus-differences on a consistent and fairly smooth curve. 



The results of Monkey 2 are summarized in table 4, and 

 those of Chick 2 in table 5. In compiling the results for each 

 stimulus-difference, I observed the following procedure : Having 

 obtained the average percentage of correct responses for the 

 total number of trials, I took the average percentage of correct 

 responses for each daily series, and obtained the mean variation 

 of these daily averages from the average for the group. Since 

 20 trials compose the standard daily series, I weighted the 



5 My adoption of this course is the result of an extensive discussion carried on 

 with Professor Knipht Dunlap some two years ago, regarding the validity of the 

 discrimination-method, as the latter is usually employed. 



