OF ARTS AND SCIENCES. 29 



Fui'ther constant errors may be due to the following causes: — 



1. Impure mercury in gauges. Carefully redistilled mercury only 

 was employed. Any slight error from this source would be nearly 

 eliminated through the use of the same mercury in the determination 

 of both y and K. 



2. Imperfect drying of the gas. The gas was retained over concen- 

 trated sulphuric acid from fifteen minutes upward before being passed 

 through the drying tubes into the apparatus. These tubes were of 

 more than 0.5 meter in length, and of such size that their contents 

 would supply the apparatus with gas for upwards of a quarter of an 

 hour. They contained anhydrous phosphoric acid, of which the sur- 

 face was made greater by placing a portion of it upon a horizontal 

 partition of wire gauze. The apparatus was dried as thoroughly as 

 possible before the beginning of each day's observations, by repeated 

 careful exhaustions, and by the transjiiration of the dried gas during 

 several hours. 



3. Leakage into the apparatus in the portion where p^ was measured 

 would not occur, as this pressure was always equal to or in excess 

 of the external. Leakage into the space containing the gas under 

 p^ would be fatal to accuracy, and was prevented at the necessary 

 joints by placing as close together as practicable the ends of the 

 glass tubes to be joined, wiring the rubber tube used as a connector 

 firmly into place, and surrounding the tube with a bath of glycerine to 

 preserve the rubber from drying and cracking. This arrangement 

 showed upon careful tests no indication of leakage either of air or of 

 moisture ; and it is believed that such could not have occurred in 

 amounts sufficient to vitiate the results to any appreciable extent, in 

 view of the other precautions taken against accumulated moisture. 

 Leakage into the space whose pressure was jOj, unless absurdly large 

 in amount, could not cause error. 



4. More or less complete clogging of the capillary tubes which 

 would bring about any change in diameter was not indicated either by 

 the course of the direct experiment, or by the check measurement of 

 iiTmade at several times. 



5. Badly shaped connecting tubes would have introduced frictional 

 resistance of an uncertain magnitude, but the tubes used were designed 

 of suitable size to make the velocity through them so small as to re- 

 quire a motive pressure certainly less than 0.01 mm. 



6. There arises naturally a question as to how far the pressure at 

 which the gases are transpired alFects the viscosity, a point which is 

 discussed further at page 18 of this paper. That this effect is not 



