262 PROCEEDINGS OF THE AMERICAN ACADEMY. 



Figure 7 shows the smooth curve that best represents the band of 

 Figure 6, translated back from "reduced" to ordinary units, both 

 Centigrade and Fahrenheit. This curve has proved useful in several 

 unexpected ways. For example, it will be made the basis of a dis- 

 cussion of the specific heat of very highly superheated steam in a 

 later paper (see page 292 of these proceedings). It has also made 

 certain cumbersome and uncertain computations in continuous flow 

 calorimetry unnecessary (see " Power," June 2, 1908, page 871). It 

 is hoped that the various scales of Figure 7 are open enough to make 

 the curve useful to others. 



All of the observations discussed in this paper have been examined 

 with considerable care, both arithmetically and graphically, for traces of 

 a systematic variation of the Joule-Thomson coefficient with pressure 

 at constant temperature, without success. If such a variation exists 

 even close to the saturation line, it is within the limit of error of 

 the data. 



Appendices. 



Discussion of Dodges Data. 



In Dodge's apparatus the low side chamber was protected against 

 loss of heat to its surroundings chiefly (although not wholly) by an 

 independently heated steam jacket made in one piece with the wall of 

 the chamber, and kept as nearly as possible at the same temperature 

 as the low side steam. Thermometers were placed in this jacket and 

 their temperatures recorded with the other routine data of each run. 

 As a matter of fact, the jacket temperatures usually ran somewhat 

 lower than the low side steam temperatures, so that some loss of heat 

 by conduction through the chamber wall was to be expected. The 

 high temperatures employed would also tend to make probable some 

 loss of heat by radiation. The possibilities were tested in six special 

 runs numbered 83 to 88, in which the partition between the high and 

 the low side chambers, with its orifice, was completely removed. It 

 was found that the low side thermometers in these tests did read 

 somewhat lower than the high side thermometers although there was 

 no throttling. The 27 observed differences can be fairly well repre- 

 sented by the empirical equation 



_ 12 (low side temp. — jacket temp.) + \ (high side temp.) 

 flow in lbs. per hour 



The forms of the two terms in the numerator were intended to cor- 

 respond to the two sorts of heat loss mentioned above. Corrections 

 corresponding to this formula were accordingly applied to the main 



