520 PROCEEDINGS OF THE AMERICAN ACADEMY. 



ever obtained have been given in the table, but not all these are 

 plotted. In the earlier work, A V was found both during freezing and 

 melting, that is, with increasing and decreasing pressure. In the 

 paper on mercury a detailed discussion is given, showing why the 

 values found during decreasing pressure are very likely to be in error. 

 None of these values are plotted. When it has not been specifically 

 stated in the table, the value has been obtained during* decreasing 

 pressure. Furthermore, all the points obtained with the kerosene as 

 the transmitting fluid below 0° have been discarded from the figure, 

 with the exception of the three low points to be mentioned later. 

 Nearly all the other points found with the kerosene below zero lie 

 above the curve given. 



At the highest pressures there is an entirely new source of error 

 which justifies the discarding of certain points. Before the explana- 

 tion was found, considerable trouble was given by the points apparently 

 lying on two distinct parallel curves. A succession of points on the 

 low curve, not very regular, was first found, reaching up to 55°. The 

 pressure was then pushed to 18,500 for an equilibrium point, and then 

 after that, more change of volume points were found at lower pressures. 

 These all lay on the high curve. Some of these points were determined 

 three times with different fillings of the apparatus, giving almost 

 identical results, and the measurements were extended back to 0° to 

 compare with the results with the other piece of apparatus, giving 

 approximate agreement here also. The high curve was then followed 

 out again to higher temperatures, giving good points until at 45° the 

 value jumped again to the low curve. The possibility of a new kind 

 of ice was ruled out by the regularity of equilibrium points. Explan- 

 ation was found in the viscous yield of the steel. This has the curious 

 property of going on at a uniform rate during the time occupied by an 

 experiment, not being asymptotic as one might expect. This yield 

 had been carefully looked for at the lower pressures and not found. 

 During the long course of the experiments so much practice had been 

 obtained that during these last measurements the readings of pressure 

 and piston displacement were made with almost clock-like regularity 

 every six minutes. The yield was so slow as to be imperceptible dur- 

 ing that interval of time, and the readings were made with such regu- 

 larity that the effect was not discovered, as it otherwise would have 

 been, by irregular points on the curve. Once the effect was discovered, 

 it was found entirely competent to account for all observed discrepan- 

 cies, since the melting occupies a fairly long time, between one and two 

 hours. Another point at 45° was then found, applying a correction for 

 the yield by noting its rate during the half hour before and after change of 



