at Loiv Pressures through the Passage of Electricity. 181 



quite clearly by the experiments described in my first paper 

 on this subject. And though the variation in X is far less 

 prominent in the tables of this paper, the presence of X in 

 the abscissa, X//>, brings the present observations into satis- 

 factory agreement, and reduces their discrepancies within the 

 limits of accountable errors. 



The errors of observation may be considered, in most cases, 

 to fall within about 5 per cent. It should be mentioned that 

 the experiments in Tables I. -IV. were made with a normal 

 laboratory temperature. The observations recorded in the 

 last four tables were taken last midsummer, the temperature 

 of the laboratory being from 22° to 26° C. As no correction 

 has been made for temperature, an error of 3 per cent, might 

 thus arise among the abscissas X/jt>D. 



The errors involved in AQ should not, as a rule, exceed 

 2 per cent. The errors involved in Ap are about the same. 

 The latter are due partly to errors associated with a McLeod 

 gauge j partly to the incomplete absorption of the water- 

 vapour by the phosphorus pentoxide. It was verified, how- 

 ever, that this absorption was practically complete and im- 

 mediate, so that Ap could be determined by observing the 

 fall of pressure as soon as the current was stopped. 



But though the errors in the values of Ap are in most 

 cases small, yet they increase as the pressure is reduced. 

 For the current when the pressure is about 1 mm. tends to 

 run down considerably. And so, if X is to remain sensibly 

 constant during the experiment, the current can only be 

 permitted to run for a short time. Thus AQ, and therefore 

 A p, is very small for the lowest pressures. 



Now the few points in fig. 3 which fall abnormally wide 

 of the mean locus, represent experiments made at a mean 

 pressure of less than 1*5 mm. Their failure to approach the 

 mean locus as closely as the rest must, therefore, be attributed 

 to errors of observation. It does not naturally or necessarily 

 suggest that for these low pressures the line of fig. 3 is less 

 accurately than for the higher pressures the true locus of 



points whose ordinates are -^ , r .{ ~ . This view is con- 

 1 pD 1 pDAQ 



firmed by the closeness with which other points, corresponding 



to the lowest pressures in the tables, approach the line. 



The equation 



A/. _pD :'>X 



AQ " 10" + 800 



may, then, presumably be considered as nearly true for the 

 lowest pressures of the tables as it is for the rest. 



