COMPLEX NEURAL ARCS. 99 



measurements make it smaller. When the spring and the current are 

 carefully adjusted the latency of the break may be even less than 0.2cr. 

 The total latency of the Harvard signal without friction is not over 

 1.50-. This seems to be constant within the errors of measurement for 

 currents such as were used in our experiments. We were particularly 

 gratified at this showing of the Harvard marker. We started to use it 

 because of its availability. We continued to use it because of its excel- 

 lence. From these various records of the latency of our apparatus, it 

 appears that the actual latency of the word-reactions will really be about 

 37 a less than the recorded value. This total error, however, will vary 

 from word to word, but will be relatively constant for similar initial 

 sounds. The instrumental variation is much smaller than the unit 



H 



10 



FIG. 16. Voice-reaction key. 



of our measurements. In no event can it be understood as constituting 

 a bias for or against alcohol days. We have entered into this careful 

 analysis of 'the instrumental errors to guarantee as far as instrumental 

 accuracy is concerned that the drug effect, though relatively small, 

 indicates a real physiological difference occasioned by the administra- 

 tion of alcohol. 



EXPERIMENTAL PROCEDURE. 



Position of the subject. For the word-reactions the subject was seated 

 at position I (fig. 1), as in the knee-jerk and memory experiments. The 

 back of the seat was raised so that the subject sat upright with adequate 

 support at his back. The leg was freed from the knee-jerk apparatus. 

 The left hand held the voice-reaction key lightly but firmly against the 

 upper lip, as per standard instructions. The left arm was supported 

 either on the table or on a rest which was held in the subject's lap. 



Stimuli. A standard set of 24 words was used throughout the year. 

 In every word-reaction experiment the entire set of 24 words was 

 reacted to. Since the reaction time for reading varies directly with 

 the length of the word, as shown by Cattell, 1 and by Erdmann and 

 Dodge, 2 an arbitrary word-length of 4 letters was adopted. All the 



'Cattell, Phil. Stud., 1885, 3. p. 313. 



2 Erdmann and Dodge, Psychologische Untersuchungen liber das Lesen, Halle, 1898. 



