I9I4.] AN ARTIFICIAL HISS. 325 



An ideal experimental series would now have consisted of equal 

 numbers of mouth-sounds and whistle-sounds arranged in haphazard 

 order. But the experimenters could not thus accurately reproduce 

 the whistle-sound. If the series chanced to call for several mouth- 

 sounds in succession, then the setting of the mouth could be main- 

 tained; but if mouth and whistle alternated, or if a single mouth- 

 sound were to be given among several whistle-sounds, there was 

 need of readjustment and possibility of initial failure. It seemed 

 best that whenever the mouth-sound was obviously wrong, and 

 when (as sometimes happened) the squeeze of the bulb was too 

 light or too heavy, the experimenter should say: " Don't count that! 

 Repeat ! " and should simply try again. This procedure gave the 

 observers a certain advantage ; but we thought it preferable to a 

 voluntary change or a further haphazard determination of the 

 stimulus. The more successful of our two experimenters, Mr. 

 Stephens, was obliged, even at his highest level of practice, to re- 

 peat the mouth-sound in 12 to 15 per cent, of the trials. 



There was a further complication. The observers sat at a 

 distance of not more than i m. from the source of sound; they 

 declared that, if they were to judge discriminatingly, they must 

 be as near as possible; and they tended to lean or move in toward 

 the experimenter. The preliminary experiments showed that, under 

 these conditions, their judgment might be influenced by secondary 

 indications — the direction of the sound, the noise of breathing, of 

 setting the mouth, even of the squeeze of the bulb, the noise of 

 preparatory movements in general. Hence they were informed be- 

 fore the regular series began that these indications were not to be 

 relied upon, but that the experimenter might in any given test make 

 misleading preparations. In fact, 50 per cent, of the mouth-sounds 

 were accompanied by noises of bulb and table, and 75 per cent, of the 

 whistle-sounds by noises of breathing and mouth-setting. The nu- 

 merical results and the introspective reports prove that this ruse was 

 successful. The observers based their discriminations, for the most 

 part, upon the temporal course and the " size " of the stimuli ; the 

 sound was judged to be " whistle " if it was hard, clear-cut, abrupt, 

 and to be " mouth " if it was fluctuating, " trembly," soft, diffuse. 

 Sometimes pitch was referred to (whistle higher), and sometimes 



