30 PROCEEDINGS OF THE AMERICAN ACADEMY 



outside the limits of error of observation in this method is however 

 shown by the groups of measurements of K at various pressures as 

 given in experiments 26 to 38 on C0„. These show that for pressures 

 p^ — jOj, as given in the subjoined table, varying from 159.7 to 54.9 mm., 

 the mean values of K deduced differ at most by less than 0.14 per 

 cent, and the difference is also not progressive in either direction, but 

 irregular, the means for the extremes of pressure being precisely the 

 same. 



Pi—Pi- K. 



159.7 mm. 1.0387 



70.9 1.0373 



61.9 1.0375 



54.9 1.0386 



In addition to this, the close agreement of the results of the first 

 series on COg with those of the second series afford an important 

 check upon the latter, since the former were made with wholly differ- 

 ent measuring and heating apparatus, and at pressures from two to 

 three times as great. 



7. It should be here noted, that any unequal action at the entrance 

 of the various tubes which might vitiate results obtained by any direct 

 method are to a large extent eliminated by the differential method here 

 used, and would probably enter, if at all, as errors of the second order 

 of magnitude ; also, as elsewhere stated, that the surface friction or 

 slip of the gas over the walls of the tube would influence the results 

 only in so far as it changed with the temperature, and would therefore 

 probably be imperceptible. 



Variahle en'ors may enter from instrumental sources in readings of 

 the barometer or of the cathetometer. Errors of observation in the 

 measurement of the length of the mercury columns by use of the ca- 

 thetometer would rarely exceed 0.02 mm. ; but this amount is consid- 

 erably less than the pressure changes which occurred in the average 

 experiment. The barometer reading was also obtained with a devia- 

 tion of less than 0.02 mm. as far as concerns errors of observation. 

 To eliminate as far as possible progressive changes in jOp jo,, and p^ 

 during the limits of time required for making a complete set of obser- 

 vations at any given temperature, a pair or more of readings of each 

 was taken, symmetrically distributed in regard to time, so that the 

 means should correspond to about the same instant, and thus be as 

 nearly as possible equivalent to simultaneous observations. The order 

 of readings in the most complete sets would thus usually be barometer, 

 thermometer giving temperature t of bath, gauges A, B, C, temperature 



