478 HENKY A. EOWLAND 



The apparatus for the comparison up to 100 C. is described in the 

 paper above referred to. The thermometers are totally immersed in 

 the water with their stems very near the bulbs of the air thermometers. 

 From 100 up to 250 an oil bath is used, the bulbs only being in the 

 oil, but the stems are heated to the same degree by being in contact with 

 a heavy copper bar, whose temperature is noted by separate thermome- 

 ters. 



The ordinary comparison is made with the stems of the thermometers 

 in a vertical position. Where they are used in a horizontal position a 

 correction will have to be made, and this correction will be determined 

 when it is so desired. When the comparison is made only to 40, we 

 can compare them in a horizontal position, but we cannot then insure 

 the same accuracy as when they are vertical, and it is never advisable to 

 use them in that position. 



Where desired, a study will be made of the changes of the zero point 

 as a function of the temperature to which it has been heated, and of the 

 time, but this study is not advised, as it does not lead to very valuable 

 results. 



Thermometers with metal, wooden or paper scales are generally too 

 poor to be worth comparison, and would often be spoiled by the immer- 

 sion in the water. Thermometers with metal caps of Geissler's form 

 are often injured, especially when heated to 250 C. Therefore, com- 

 parisons of thermometers of these classes will not be undertaken, ex- 

 cept in the case of standards long used for some particular purpose, or 

 in that of fine G-eissler thermometers. 



Three intervals for the comparison have been selected. 



A. Between and 40 for thermometers used for meteorological 

 observations, determination of the temperature of standards of length, 

 calorimetric determinations, and all purposes where extreme accuracy is 

 desired within that limit. To obtain the full value of such a compari- 

 son, thermometers should be graduated at least as fine as 0-1 C. or 

 0-2F. 



B. Between and 100 C. It is advised that the thermometers gent 

 be graduated at least as fine as 0-2 C. or 0-5 F. 



C. Between 100 and 250 for thermometers used by chemists in the 

 determination of melting or boiling points. Thermometers should be 

 graduated to 1 C. or 1 F. 



Three kinds of comparison will be made for each of the intervals 

 to 40, to 100, and 100 to 250, as follows: 



1st. Direct comparison with the air thermometer, and also a primary 



