24 MEASUREMENT OP HIGH TEMPERATURES. 
temperature measurement so readily suggest themselves to the student 
of modern physics that special attention to them is. superfluous, whereas 
criticism of more searching value calls for special experiments. Such 
experiments, except in a few cases, I have not had occasion to repeat. 
Character of the measurements. — It is impossible to read the earlier 
memoirs on high temperature research without a feeling of uneasiness 
and disappointment. There is no lack of ingenious contrivances or 
of well-devised methods, but the results obtained are usually sadly at 
fault. In many cases no data for the absolute identification of the 
measurements made are discernible. In other cases not only do ob- 
servers, using different methods, fail to reach accordant results, but it 
is not unusual to find even skilled observers using the same method 
with errors in results as high as 10 per cent, for the same fixed high tem- 
perature datum, the boiling point of zinc. To secure certain facilities 
of manipulation Deville and Troost, at the outset of their researches, 
used iodine vapor as a gas for thermal measurement. This step must 
be regarded as a misfortune to science, and one which retarded the 
progress of high-temperature research many years. After the tendency 
of the iodine molecule to dissociate had been suspected, and the relative 
imperviousness of porcelain as compared with platinum air-thermometer 
bulbs had been clearly pointed out, the values of the boiling point of 
zinc begin to increase from the exceptionally low values of Becquerel 
(884°), and to decrease from the exceptionally high values of Deville 
and Troost (1,040° C.) over a total range of temperature of about 150°, 
until the final results of these observers (932° and 942°, respectively) 
agree to about 10°. Curiously enough, however, Weinhold, an observer 
of great assiduity and some experience, having made himself master 
of high-temperature measurement by the air thermometer methods, en- 
deavors to redetermine the value of the boiling point of zinc, and finds 
a value (1,035°) as high as the highest datum of Deville and Troost. 
Fortunately the subject has been rescued from this condition of vague- 
ness by the recent vigorous work of Violle, the results of which agree 
well with the mean data of Becquerel and of Deville and Troost. 
My chief object in giving this outline is to place before the reader the 
nature of the difficulties with which the problem of high temperature 
measurement is surrounded, and to indicate the diversity of the results 
reached even by the best of trained observers. Methods which in the 
hands of different investigators lead to data so widely different as the 
values just cited are not apt to inspire confidence. It is perhaps more 
for this reason than because of real difficulties of manipulation that the 
gas thermometer has been so little used as a standard of reference in 
high-temperature measurement. For the experimental operations are 
not necessarily more complex than those called for in some of the empiric 
methods of standardization— methods which have further burdened 
the unfortunate subject of high- temperature research with their own 
allotment of vagueness of principle and inaccuracy. 
(678) 
