THE STOP-WATCH AND THE ASSOCIATION TEST 175 



the prolongation of training the unreliability becomes greater, 

 as the reaction tendency becomes more fixed. When, in addi- 

 tion to the error due to the reaction tendency, and the irregular 

 errors due to incidental causes, is added the emotional effect on 

 the experimenter due to knowledge of significant words, we 

 may expect great difficulty in the stop-watch application of the 

 association tests. It is not improbable that in many cases a 

 significant set of results is obtained, not by actually measuring 

 reaction-times, but by lengthening the time-records where the 

 behavior (facial expression, tone of voice, etc.,) is suspicious. 

 This would account for the divergence between recorded times 

 for significant and non-significant terms in certain stop-watch 

 series being greater than are normally obtained by methods 

 more mechanically accurate. In other cases the experimenter 

 (according to his statements) correctly identifies the culprit 

 when the recorded times furnish no clue to one who has not 

 witnessed the test. I have found in my laboratory that in 

 many cases it is possible to pick out the guilty party from his 

 subsidiary reactions to the test-words without paying any at- 

 tention to the reaction-times and with little consideration of 

 the response words. This method is however so obviously suscep- 

 tible to the effects of prejudice and pre-conceptions that it ought 

 not to be substituted for the true association method. 



