DISCRIMINATIVE RESPONSES TO VISUAL STIMULI 461 



after establishing control of other variables than vision, to use 

 the speed and accuracy of the work done under a given lighting 

 condition as an indicator of the appropriateness of the condition 

 to visual performance. 



Such a method, while simple in principle, demands the fulfil- 

 ment of certain requirements, which I may be pardoned for 

 enumerating for the benefit of non-technical readers : 



1. The work done under the different lighting conditions must 

 be of equal intrinsic difficulty. Otherwise, differences in speed 

 and accuracy might wrongfully be attributed to the influences 

 of differences in lighting. 



2. The work done under the different lighting conditions must 

 be distributed among them so that the effects of fatigue and 

 practice are evenly distributed. Otherwise, differential effects 

 of these variables may be erroneously interpreted as effects of 

 the differences in lighting; or they may serve to distort or to 

 obscure genuine effects of differences in lighting. 



3. The measure of the subject's performance must be obtain- 

 able independently of his own reports or description. This 

 point was mentioned in the comment on Cobb's work, but it 

 cannot be too strongly emphasized. 



4. A sufficiently large number of observations must be obtained 

 from each subject to insure that such effects as may be obtained 

 are not accidental. In other words, enough results must be 

 accumulated to justify statistical treatment, and the application 

 of standard measures of reliability to them. 



In the past ten years, the pages of American lighting journals 

 have been burdened by reports of tests which fail to satisfy 

 certain of the above requirements, especially the third and 

 fourth. In the judgment of the writer which at this point 

 accords with that of the best psychological authorities a proper 

 regard for economy of time would not justify one in reading such 

 reports farther than to verify the description of the defective 

 methods employed. 



In the beginning of the present work, several types of work- 

 material were examined and found unsatisfactory. The most 

 promising material was arithmetical, compiled by Prof. Knight 



