88 PROCEEDINGS OP THE AMERICAN ACADEMY 



for these specimens, and the seventh column contains the values cal- 

 culated from these values. These values also satisfy the observations 

 above 100° for the given specimens. 



The table seems to show that between 0° and 100° a thermometer of 

 Choisy-le-Koi almost exactly agrees with the air thermometer. But 

 this is not at all conclusive. Regnault, however, remarks,* that be- 

 tween 0° and 100° thermometers of this glass agree more nearly with 

 the air thermometer than those of ordinary glass, though he states 

 the difference to amount to .1 to .2 of a degree, the mercurial ther- 

 mometer standing below the air thermometer. With the exception of 

 this remark of Regnault's, no experiments have ever been published 

 in which the direction of the deviation was similar to this. All ex- 

 perimenters have found the mercurial thermometer to stand above the 

 air thermometer between 0° and 100°, and my own experiments agree 

 with this. However, no general rule for all kinds of glass can be 

 laid down. 



Boscha has given an excellent study of Regnaidt's results on this 

 subject, though I cannot agree with all his conclusions on this subject. 

 In discussing the difference between and 100° he uses a formula of 

 the form 



T—t — — t (100 — 0, 



and deduces from it the erroneous conclusion that the difference is 

 greatest at 50° C, instead of between 40° and 50°. His results for 

 T — t at 50° are 



Choisy-le-Roi —.22 



Verre Ordinaire -}— .25 



Verre Vert H - *^ 



Verre de Suede — |— .56 



and these are probably somewhat nearly correct, except the negative 

 value for Choisy-le-Roi. 



With the exception of Regnault, very few observers have taken up 

 this subject. Among these, however, we may mention Recknagel, 

 who has made the determination for common glass between 0° and 

 100°. I have found approximately the constants for my formula in 

 this case, and have calculated the values in the fourth column of the 

 following table. 



* Comptes Kerulus, Ixix. 



