492 H. M. JOHNSON 



settings made during the lapses of attention are not necessarily 

 related to the subj ect's instructions or intention. In other words, 

 the setting of the instrument at a given instant does not neces- 

 sarily measure the subject's threshold. The method was thor- 

 oughly tried recently in the work on low oxygen at the Air Service 

 Medical Research Laboratory, and had to be discarded. 



Work-tests are unimpeachable in principle, and the difficulties 

 in the way of applying them are not insuperable. But the 

 preparation of an adequate amount of uniformly difficult material 

 is a task which many students would find appalling; and the use 

 of such material is an absolutely indispensable condition of 

 getting interpretable results. If the type of work chosen for 

 the test involves a high degree of muscular coordination, a very 

 long period of preliminary training is necessary. And, since the 

 time required to execute a coordinated muscular response may 

 be highly variable in comparison with the short time required 

 for simpler registration of the act of discrimination, such a work- 

 test will require the accumulation of a larger quantity of data 

 than most students would consider to be worth the expense. 



The reaction-time method requires a simple muscular response, 

 which is usually stabilized within a few weeks of training so that 

 further improvement is slow and gradual. The determination 

 of discrimination-time is probably sufficiently sensitive to show 

 the effects of any condition which tends to increase the difficulty 

 of sensory response. The price of definite results, especially if 

 the effect is small, is threefold : (1) adequate control of the external 

 conditions so that secondary stimulus-variables are not effective; 

 (2) adequate training of the subjects, to stabilize effects of 

 fatigue and practice, and insure an approximation to uniformity 

 in methods of observing and reacting; and (3) the accumulation 

 of a sufficient number of reactions to make the results reliable. 



The number of reactions necessary to demonstrate a differ- 

 ential effect depends, obviously on the magnitude of the effect 

 and the dispersion of the results. For example, the last 600 

 left-hand reactions of subject A, equally distributed among the 

 three lighting conditions described above, show a difference in 

 favor of condition Bi over DI which is 14 times the probable 



