36 Messrs. C. W. Waidner and F. Mallory on 



The results of these comparisons can best be shown by 

 means of the accompanying curves. 



Fig. 4 shows the result of each of the independent com- 

 parisons of Professor Rowland's Baudin thermometers with 

 the platinum-resistance thermometer. Abscissae represent 

 temperature on the centigrade scale, and ordinates the 

 corresponding corrections that must be added to Rowland's 

 air-scale, as determined by the Baudin thermometers, to 

 reduce to the Callendar-Griffiths' air-scale. The almost 

 constant difference between the curves of May 18th and 

 May 24th suggests at once a constant error affecting the 

 entire series of one or the other comparisons. This difference 

 is, however, not of any great importance ; for in the 

 determination of the mechanical equivalent we are practically 

 only concerned with temperature ranges over the interval 

 5° to 35°, and these are practically identical on either curve. 

 After looking over the results, we have attributed these 

 differences to a slight error in the value of the temperature- 

 coefficient of the coils of the box, which, on account of the 

 high temperature of the room, could not be kept at 20° in 

 the comparisons of May 24th and May 21st. This conclusion 

 is further strengthened by the fact that the comparison of 

 May 28th, when the box was again near 20°, practically 

 coincides with that of April 10th. 



A resume of all the comparisons between the Platinum and 

 Baudin thermometers is also given in the table opposite. 



Fig. 5 gives the final correction curves for each of the 

 Baudin thermometers. The ordinates of curves I. (mean of 

 the individual comparisons shown in fig. 4) give the 

 corrections that must be added to Rowland's air- scale, as 

 determined from the Baudin thermometers, to reduce to the 

 Callendar-Griffiths' air-scale. To pass from the air-scale to 

 the absolute scale of temperature use was made of Rowland's 

 table xvii. (p. 114, Proc. Am. Acad.xv. 1879). In this way 

 curves II. were obtained which give the corrections to reduce 

 Rowland's absolute scale to the Callendar-Griffiths' air- scale. 



The results of the comparison of the platinum-resistance 

 thermometer with Tonnelot 11801, which had been standard- 

 ized at the Bureau International, are shown in curve A, in 

 which ordinates represent the corrections that must be added 

 to the Callendar-Griffiths' air-scale to reduce to the Paris 

 nitrogen- scale as given by Tonnelot 11801. The close agree- 

 ment of these two scales is a strong confirmation of the 

 accuracy of the platinum-air interpolation formula. Indeed 

 such close agreement must be partly fortuitous, as we have 

 certainly no right to expect so close an agreement, con- 

 sidering the difficulties of gas-thermometry. 



