466 HENRY A. ROWLAND 



flame, but they stand very much above the air thermometer at 40. I 

 have since tried some of the Baudin thermometers up to 300, and find 

 that they stand Mow the air thermometer between 100 and 240 ; they 

 coincide at about 240, and stand above between 240 and 300. This 

 is very nearly what Eegnault found for " Verre Ordinaire." It is to be 

 noted that the formula obtained from experiments below 100 makes 

 them coincide at 233, which is remarkably close to the result of actual 

 experiment, especially as it would require a long series of experiments 

 to determine the point within 10. 



The comparison of thermometers also shows that all thermometers 

 in accurate investigations should be used as thermometers with arbi- 

 trary scales, neither the position of the zero point nor the interval be- 

 tween the and 100 points being assumed correct. The text books 

 only give the correction for the zero point, but my observations show 

 that the interval between the and 100 points is also subject to a sec- 

 ular change as well as to the temporary change due to heating. Of 

 all the thermometers used, the Geissler is the worst in this as in other 

 respects, except accuracy of calibration, in which it is equal to most of 

 the others. 



The experiments on the specific heat of water show an undoubted 

 decrease as the temperature rises, a fact which will undoubtedly sur- 

 prise most physicists as much as it surprised me. Indeed, the dis- 

 covery of this fact put back the completion of this paper many months, 

 as I wished to make certain of it. There is now no doubt in my mind, 

 and I put the fact forth as proved. The only way in which an error 

 accounting for this decrease could have been made appears to me to be 

 in the determination of ra in " Thermometry." The determination of 

 m rests upon the determination of a difference of only 0-05 C. between 

 the air thermometer and the mercurial, the and 40 points coincid- 

 ing, and also upon the comparison of the thermometers with others 

 whose value of m was known, as in the Appendix. Although the quan- 

 tity to be measured is small, yet there can be no doubt at least that m 

 is larger than zero; and if so, the specific heat of water certainly has a 

 minimum at about 30. 



One point that might be made against the fact is that the Kew stand- 

 ard, Table L, gives less change than the others. But the calibra- 

 tion of the Kew standard, although excellent, could hardly be trusted to 

 0-02 or 0-03 C., as the graduation was only to F. In drawing the 

 curve for the difference between the Kew standard and the air ther- 

 mometers, I ignored small irregularities and drew a regular curve. On 



