OP ARTS AND SCIENCES. 37 



tion was devoted to this branch of the work. The precision obtained 

 in the thermometry was doubtless greater than was necessary to cor- 

 respond to that obtained in other parts of the determination of y, espe- 

 cially below G0° ; but it is doubtful if a satisfactory elimination of 

 constant errors could have been obtained with any less care. 



The investigations on thermometry which I have made, and still 

 have under way, will be discussed here only in their relation to the 

 measurements given in this paper. The important problem of how 

 far the mercurial thermometei" is available as a convenient representa- 

 tive (proxy) for the air thermometer at temperatures up to 300° C., 

 and under what conditions and methods it shall be so used, does not 

 enter to any considerable extent into the work in hand, on account of 

 the comparatively small degree of accuracy required in my measure- 

 ments, even at the highest temperature (225°) used. The results 

 which I have already obtained in this direction I hope to supplement 

 by enough other material of the same nature to render the whole of 

 sufficient value to be given by itself at some future time. 



All the thermometric readings were taken with the bulb immersed 

 in the oil-bath around the second capillary to a depth of about two 

 inches. Above this point, the projecting stem was surrounded by a 

 thin-walled glass tube, closed at the bottom by a rubber stopper 

 through which the thermometer stem passed, and filled with water or 

 glycerine, according to the temperature of the oil. This stem bath 

 was stirred by a vertically moving stirrer, and its temperature was 

 determined at one or more points, according to the length of the 

 mercury column enclosed, by one or more auxiliary thermometers. 



The readings were properly corrected for calibration error (except 

 in the case of the Baudin normals, which were found by calibration 

 to be accurate to two or three tenths of a division), for temperature 

 of exposed stem, for true value of unit of scale (error of fixed points), 

 and for deviation from the air thermometer. Poggendorff' s correction 

 becomes unnecessary where the mercurial thermometer, as in this 

 work, is compared with the air thermometer direct, and all readings 

 are reduced to that standard. 



The thermometers used were respectively, one by Casella of London, 

 No. 32378, and two by Baudin of Paris, Nos. 7335 and 7789. The 

 following table gives the description of these, and of several others to 

 be referred to later. 



