1 16 BRIDGMAN. 



the width of the band of scattering was greater than the 5% usually 

 found. The magnitude of the scattering varied with the specimen, 

 as it might be expected to. 



In setting up the first sample, the ground of the heating element 

 was made to the sample itself. It has been previously explained that 

 this introduced an effect due to the Peltier heat, which is unusually 

 large for this metal. The mean of readings with two directions of the 

 heating current should eliminate this effect. With the other two 

 samples an independent ground was used, and the effect disappeared. 



It is to be expected that if there is any error due to heat leak along 

 the thermo-couple wires that it will be especially large for this metal, 

 whose own thermal conductivity is so small. In order to test this 

 point, duplicate runs were made on the third sample, first with the 

 ordinary copper-constantan couple, and then with a couple of " therlo" 

 constantan. Very nearly the same results were found, showing the 

 adequacy of the precautions taken to prevent leak along the couple 

 wires. 



The third sample gave the most regular results; probably some 

 accidental twist or bend in the wires made them less likely to be 

 displaced under pressure than those of the other samples. The 

 results with this sample are reproduced in Figure 10. 



The results found with the different samples are as follows : ^ 1 , 

 -38.8% for 12000 kg.; ^2, -38.8%; #3 (copper-constantan couple) 

 —37.3%, and ^3, (therlo-constantan couple) — 35.5%. Mean 

 —37.8 %. Because of the low conductivity of bismuth the correction 

 for the transmitting medium is large, amounting to 13.8%, and is in 

 the direction to make the true effect more negative than the observed 

 effect. 



The pressure coefficient of thermal conductivity to be deduced 

 from the above measurements is — O.O43I. It is to be noticed that 

 this is negative, and also that it is abnormally high. The abnormal 

 sign agrees with the pressure effect on electrical conductivity, which 

 also decreases under pressure. I found at 30° the average pressure 

 coefficient of electrical resistance up to 12000 kg. to be — O.O4212. 

 Bismuth was not among the metals measured by Lussana, so there 

 are no previous results for comparison. 



Antimony. Measurements were made on four samples, all by the 

 longitudinal flow method. The material was obtained from the J. T. 

 Baker Chemical Co. It was supposed to be especially pure, although 

 I have no analysis. Antimony from the same source was formerly 

 used by the Bureau of Standards to give a fixed melting point, but 



