614 VITALITY AND EFFICIENCY WITH RESTRICTED DIET. 



of every shock with response or failure to respond was recorded. In 

 elaborating the records, the average between successive voltage 

 records when the subject responded in one case and in the next case 

 failed to respond, was arbitrarily considered a threshold determina- 

 tion. For example: if a man responded to a shock voltage of 104 volts 

 and had been responding to shocks of this strength, or stronger, and 

 then failed to respond to a shock of 100 volts, the threshold determina- 

 tion was considered the average of 104 and 100, or 102 volts. Usually 

 about 20 such determinations would occur in the records taken with a 

 subject in the test period of one evening. 



The average standard deviation for the individual series of determi- 

 nations on a single subject (see tables 177 and 178) is about 7.5 volts. 

 There are rather wide variations from this, that is, with Squad A they 

 range from 2 to 20 volts. Pea has the largest average standard 

 deviation (10.8 volts) of any man in Squad A. The smallest is con- 

 sistently found with Moy (4.3 volts), whose threshold was also the 

 lowest average for the group. Van and Sch, of Squad B, have standard 

 deviations of 11.6 and 11.3 volts, as averages for the reduction period. 

 Their thresholds were, however, much higher than that for Pea, 

 being 145 and 169 volts, as compared with 100 volts. The standard 

 deviation is usually about 6.5 per cent of the average threshold value. 

 Squad B show slightly less than this, particularly in the average for the 

 five normal experiments, the percentage for which is 4.8. The ex- 

 ceptional case, in the two squads, is that of Pea in Squad A, with an 

 average coefficient of variabiUty of 11.1 per cent for the low-diet period. 

 It is exceptional to find any other of the 25 subjects who comprised 

 these two squads showing a variability of as much as 10 per cent on 

 any date. The small average variabiHty of both squads in their first 

 experiment is due to the much smaller number of threshold deter- 

 minations that were made on these dates, when the method was not 

 the same as described in the paragraph above, but was identical with 

 that used in the normal series of 1917. The small variability within 

 any individual series of threshold determinations makes this measure- 

 ment compare very favorably with any sensory threshold measurement 

 with which we are familiar. 



The fluctuations of the two sets of results throughout the group 

 of experiments are shown in figure 116. First it may be noted that 

 the standard deviation and coefficient of variabiUty are practically 

 stationary from first to last and are at very nearly the same level for 

 both squads, tending to be a little smaller with Squad B. There is a 

 slight increase in the standard deviation for October 27, Squad A, 

 apparently due to unusually large deviations with Tom, Vea, 

 Pec, and Mon. Both squads show one marked depression (higher 

 threshold) during the period of the experiment. In the case of A this 

 occurred at the first reduction date, October 13, when the average was 



