OF ARTS AND SCIENCES. 7 



was placed around AB, the regulator at iV adjusted, and the transpira- 

 tion allowed to proceed until the gauges J^ and jP were nearly station- 

 ary. Then c was closed, and /"opened and adjusted so that the reading 

 of H slowly increased, whereupon c was again opened for the slow es- 

 cipe of the gas and the maintenance of a constant initial pressure p^. 

 The transpiration was allowed to proceed until £J, F, and R were suffi- 

 ciently near stationary, when readhigs of the mercury columns by 

 the cathetometer and of the thermometers were taken systematically. 

 After a complete set of readings had been thus obtained, the apparatus 

 was thoroughly inspected and a second set made : and so on until the 

 number was deemed sufficient, whereupon the temperature of Z^Cwas 

 changed or the pressure modified as desired. The results thus ob- 

 tained furnished, on reduction, the values of jOj, j^.^, p^, and t for the 

 computation of the ratio of the coefficient of viscosity at the higher 

 temperature to that at 0° C, the further requirements being the coa- 



stants ^^ =: K for the tubes, 7?^ and X^ R. and X,, being the radius 



and length respectively of the tubes AB and CD, as shown at page 9. 

 This value of K was found from measurements similar in all respects 

 to those just described, except that the tubes AB and CD were both 

 surrounded with ice at the same time. 



The apparatus was also arranged to place DC in steam. When 

 left at night for subsequent use, an outward pressure was created in 

 all parts of the apparatus. 



EXPEUI.MENTAL RESULTS. 



The results will be here given as they were for convenience classi- 

 fied during the progress of the work, viz. with dry air in five series, 

 with carbonic acid in two series. The method of reduction of the 

 results is given at page 2G, and in my first paper. 



Ah: 

 First, Series. — This consists of the twenty -one measurements made 

 in 1876 as given at page 49 of the article referred to, and gives as a 



mean result of the ten measurements of ^ as shown in Table I. 

 below, the value y = ^ = 1.234, with an average deviation* of 0.044, 



* Let aj, a.2, ag, . . . On, be a series of measurements of tlie same quan- 

 tity, and let m be their arithmetical mean; tlien ^i — m = d^, a.j, — m=:(L,, 

 n^ — m = il^, &c., will bo tiie deviation of tliose mcasurfiiients from tlioir 



