452 PROCEEDINGS OP THE AMERICAN ACADEMY. 



verified. This verification is obviously accomplished by immersing the 

 instrument into a mixture of cracked ice and water and adding the dis- 

 solved substance until the desired point is reached and maintained. An 

 analytical determination of the concentration of the solution then gives 

 at once by comparison with the standard curve the true value of the tem- 

 perature which corresponds with this concentration, and hence the error 

 of the thermometer. For example, supposing that the error of the 

 graduation of an uncertain thermometer at reading —1.513° is desired, 

 the thermometer is immersed in ice and water, and hydrochloric acid is 

 added until this point is reached. The solution is then analyzed and 

 found to be 0.4085° normal. From the standard curve given below it is 

 quickly found that this concentration should give a reading of 1.498°, 

 therefore the thermometer at this place was clearly 0.015° in error. 

 The practical value of the method obviously depends wholly upon the 

 accuracy with which the curve comparing concentration and temperature 

 reading has been determined. This accuracy depends, first, on the accu- 

 racy of the standard thermometer; second, upon the care in adjusting of 

 the freezing point ; and third, on the accuracy of the analysis. 



The thermometer used in the present investigation was one carefully 

 standardized by the Bureau International des Poids et des Mesures in 

 Sevres for this express purpose. It was made by Baudin, and is num- 

 bered 15275. Its scale reads as far as 12° below zero. Because our 

 results depend upon this single thermometer, they must be considered as 

 only preliminary, for, no matter how careful the calibration, the results 

 of a single thermometer are somewhat uncertain. Nevertheless this 

 thermometer is undoubtedly so much better than those usually used that 

 our results have practical value, even although they cannot be considered 

 as quite the final ones. In the near future we shall hope to take this 

 matter up again with several similar thermometers, obtaining an average 

 value which will be more nearly certain than the present one. 



With regard to the care in adjusting the freezing point, it is enough 

 to say here that all reasonable precautions were used. Two series of 

 determinations were conducted : one in a tall beaker, air-jacketed with 

 another beaker, which was in turn surrounded by a mixture of ice and 

 solution of almost exactly the same temperature as the mixture within 

 the inner beaker. The other series was conducted in a large Dewar 

 silvered vacuum jacketed tube, about ten centimeters in internal diameter 

 and thirty centimeters long. This was the more convenient form of 

 apparatus. These two series gave essentially the same results. The 

 thermometer was left in ice for twenty-four hours in every case before it 



