32 PROCEEDINGS OF THE AMERICAN ACADEMY 



As shown by the number deduced from this expression and given 

 in the table, the value of A II for A y = 0.1 per cent will be at least 

 2 mm. throughout the range of these measurements. Also when it 

 is considered that constant barometric errors, since they enter with 

 nearly the same magnitude and effect, but with opposite signs, in the 

 determinations of K and of y, become inappreciable in effect when 

 not amounting to several millimeters. That the value of A II is so 

 large will be seen, on inspection of the foregoing differential, to be 

 due in part to the small value of the differences of initial and final 

 pressures used in these experiments. 



An inspection of measurements 26 to 38 of Table VI. will show that 

 the average deviation in these determinations of K is 0.2 per cent. 

 The average deviation iu K which could come from the above-men- 

 tioned sources will be seen, however, to be less than 0.2, probably as 

 small as 0.1 per cent, since A''as measured is a function of p^, p^, and^^g 

 only. It thus appears pi'obable that the assumed precision in the in- 

 strumental measurement of the pressures is not far in erroi\ 



Among other sources of variable error are : — 



1. Irregular action of the aspirator causing more or less irregular 

 changes in the amount of pressure at the exit of the gas, i. e. corre- 

 sponding to pg. Fluctuations of this sort would of course be felt at 

 once on the gauge, but the corresponding adjustment of p^ would re- 

 quire some time, as it would require the transpiration of a definite 

 amount of gas from the p.^ space to thep„ space. By the use of the 

 regulator which I have devised for the purpose, and which has been 

 already described at page 5, it has been found feasible to keep the 

 pressure constant within a millimeter during a whole day's experiment- 

 ing, while the water pressure would run from 20 to 35 jDounds at fre- 

 quent intervals. And, as will be seen from an inspection of the values 

 of p„ given in the tables, the pressure during a series of half a dozen 

 experiments frequently remained constant within 0.2 or 0.3 mm. In a 

 single experiment a change of 0.2 mm. was rare, the average being 

 under 0.1 mm., including errors of reading by the cathetometer. It is 

 thus evident, I think, that this irregularity in p„, which is of the nature 

 of an accidental error, since it may be of any magnitude and of either 

 sign, would on the average introduce into ^ or K a deviation not 

 exceeding 0.2 per cent, and probably less than that, and is therefore 

 but little greater in its effect than other sources of accidental error 

 already mentioned. 



2. Owing to the want of perfect constancy in the temperature of 

 the oil-bath, and to a slight opportunity for irregularity in the heating 



