226 Mr Griffiths, On Thermometric "Fixed Points:' [Nov. 23, 



caused was always constant ; but glass varies greatly in composi- 

 tion and its solubility probably depends so largely on the nature 

 of the substances previously in solution that the effects may be 

 very variable. We should thus be compelled to use platinum or 

 gold vessels, and the heavy expense entailed would alone suffice to 

 put the freezing point of "pure water" out of court for all prac- 

 tical purposes. Granted that these difficulties could be overcome 

 there arises a question as to the influence of the contained gases. 

 Assuming the accuracy of the ordinarily-accepted constants we 

 find that the freezing point of water if saturated with oxygen 

 and nitrogen would be lowered about 0°*002 C, and it is difficult 

 to see how this cause of error is to be eliminated. Under such 

 circumstances the experimenter who diminishes the causes of error 

 only departs the further from what we may call the practically 

 accepted standard, and additional refinements may but introduce 

 greater confusion unless some general agreement is arrived at. 



An observer who is in the habit of standardising thermometers 

 usually repeats his experiments under similar conditions, and find- 

 ing his results consistent in themselves is apt to conclude that his 

 standard is identical with that of others. 



The following is an illustration of the danger arising from such 

 a conclusion. 



I have been in the habit of standardising my own platinum 

 thermometers by immersing them in a large copper vessel contain- 

 ing about 15 lbs. of finely-powdered ice and ordinary distilled 

 water. This vessel is separated from a second copper one by a 

 layer of Norwegian felt nearly 2 inches in thickness. The thermo- 

 meters are immersed to a depth of 8 or 10 inches, and, where they 

 pass into the vessel, are surrounded by felt. The indications of 

 these thermometers are unaffected by the stem temperature, and 

 under these circumstances it appeared probable that the tempera- 

 ture indicated could not be appreciably affected by the temperature 

 of the laboratory. I have recently been using a platinum thermo- 

 meter presenting certain novel features which I believe tend to 

 greater accuracy, also the coils of the resistance-box consist of 

 naked wires immersed in rapidly-stirred oil, and thus the great 

 impediment to accurate work, viz. uncertainty as to the tempera- 

 ture of the box coils, is eliminated. On four different occasions 

 this thermometer was placed in the ice, as above described, and its 

 resistance on these respective occasions was as follows : 



1293-968 

 •966 

 •963 

 •966 



Now in this thermometer the value of -=- (at ) = 5*073, and 



