38 INJURY, RECOVERY, AND DEATH 



by the pressure and the weak electric currents to which 

 it is subjected during the experiments. 



The solutions were made with all possible precautions. 

 The salts used were the purest obtainable. The distilled 

 water was, as a rule, twice distilled from quartz or 

 glass, 20 using cotton plugs in place of cork or rubber 

 stoppers in the distilling apparatus. The first and last 

 parts of the distillate were discarded. 



The reaction of the solutions is of great impor- 

 tance. Unless otherwise stated it was close to neutrality. 



It may be desirable to add a word of explanation 

 regarding the treatment of results. Most of the curves 

 here presented are time curves in which each point 

 represents the average of several experiments. In such 

 curves it is desirable (as indicated on page 68) to average 

 times (abscissa}) rather than resistances (ordinates). 



The probable error of the mean has been calculated 

 in all cases by Peter's formula and expressed as per 

 cent, of the mean. 21 Since, however, space is lacking to 

 present all the data, a general idea of the accuracy of 

 the results may be given by saying that there is no point 

 on the curve whose probable error of the mean exceeds 

 a certain per cent, of the mean. 



The temperature was controlled in short experiments 

 so that the fluctuations did not amount to more than 

 2C. In longer experiments (lasting several days) 

 greater fluctuations were unavoidable, but the effect of 

 these was minimized by starting all the experiments of 

 a series at the same time so that the fluctuations affected 



30 Water distilled from a copper still should never be used. 



* Thus, if the observations are 99, 104, 102, 97, 100, 96, 103, 101, 98, 

 the mean is 100, the sum of the deviations 20 and the probable error of 

 the mean 20 (.0332) = .664 which is .664% of the mean. 



