254 PROCEEDINGS OF THE AMERICAN ACADEMY. 



paper. What follows is based on a study of the original records, the 

 generous loan of which for this purpose is very gratefully acknowledged. 

 On his advice, the first 26 of his 92 runs were disregarded as prelimin- 

 ary, and 9 other runs were rejected, either because of experimental 

 mishaps, or because the log did not show satisfactorily steady condi- 

 tions. The data selected were corrected for probable radiation and 

 conduction losses in the way explained in the appendix of this paper 

 (page 262). 



Of the 47 selected tests, 14 were like those already discussed, except 

 that the temperatures were much higher, the high side steam being 

 superheated instead of saturated. The results of these 14 tests are 

 plotted in Figure 5. It will be noticed that in every case a smooth curve 

 through the low side points runs considerably below the corresponding 

 high side point, just as did Grindley's curves. In Grindley's case this 

 was because the entering steam carried water in suspension, the pres- 

 ence of which made the true total heat of the incoming mixture less 

 than its apparent total heat regarded as homogeneous saturated steam, 

 and dropped all the low side points onto throttling curves lower than 

 those on which they apparently belonged. A similar phenomenon may 

 be in evidence in Dodge's case, for although the incoming steam was 

 superheated, it may still have been carrying in suspension a part of the 

 water which had been sprayed into it for temperature regulation just 

 before it reached the high side chamber. 6 It must, however, be ad- 

 mitted that if this explanation is to account for the whole of the dis- 

 crepancy in Dodge's results, an extraordinarily large amount of water 

 in suspension must have reached the high side chamber — from one to 

 one and a half per cent of the whole weight present. It is therefore 

 probable that there is another source of error not yet discovered. 

 Nevertheless, if the high side points are disregarded and the low side 

 points are taken together in pairs as in Grindley's case, it is probable 

 that the resulting values of the Joule-Thomson coefficient will be 

 trustworthy. 



Each of the 14 runs was handled separately. It did not seem best 

 to take consecutive points together as in the other cases, because, at the 

 very high temperatures here dealt with, the temperature difference be- 

 tween consecutive points is much smaller than at lower temperatures, 

 and so an error in either observation would make much more difference 

 in the coefficient. Furthermore, the throttling curves are more nearly 

 straight in this range than at lower temperatures. The lowest point 

 of a run has therefore been taken with the point just beyond the middle 



6 See the work of Knoblauch and Jakob, Forschungsarb., 190G, 34, 109. 



