DISCRIMINATIVE RESPONSES TO VISUAL STIMULI 493 



error of the difference; and a difference in favor of condition 

 BI over B 2 which is 13 times its probable error. This number 

 of reactions is evidently enough to demonstrate the direction 

 and the relative certainty of these differences. The same data 

 show a difference in favor of condition B% over condition D 1 of 

 3 times the probable error of the difference. This means that 

 in a large number of repetitions of the experiment a reversal of 

 the effect would be expected once in each 22 times. Whether 

 this difference is sufficiently reliable is of course a matter of 

 opinion. I should consider it desirable to quadruple the number 

 of data, since if the effect is persistent this procedure would 

 halve the probable error and would increase nine hundred fold 

 the probability against the difference being due to chance. 



Obviously the most economical procedure is to tally the results 

 daily as they are accumulated; summate and average them from 

 time to time, and obtain the constants of dispersion and reli- 

 ability; and stop work under any compared conditions when the 

 results show a sufficiently definite effect, as judged by the degree 

 of precision which may be demanded. 



The number of reactions necessary for definite results from 

 a trained subject is of course far smaller than from an untrained 

 subject, as the effect of training is to enable the subject to reduce 

 the dispersion of his results very greatly. The most economical 

 method therefore requires the subjects to be thoroughly trained 

 before the external variables whose effects are to be compared, 

 are introduced. 



CONCLUSION 



I wish to acknowledge the technical assistance of Mr. George 

 Hathaway in the experimental work; and the aid and counsel 

 of a number of colleagues, especially of the following: Dr. W. 

 Weniger, in planning the electrical system; Captain P. W. Cobb, 

 in the photometry of the various surfaces and in criticism of 

 the paper; Dr. C. F. Lorenz and Dr. A. G. Worthing, in the 

 selection of flashing lamps and in valuable criticism of technical 

 methods; and Captain J. E. Coover, who acquainted me with 

 the valuable abbreviations of statistical procedure which he has 



