792 TRUEBLOOD. 



root of the number of determinations, one finds 0°.027 as the deviation 

 of the mean. This is Httle more than nominal, for it is highly improb- 

 able that the large errors involved were all accidental. Accepting it, 

 however, the difference between the value of m as derived from these 

 measurements and the value obtained from the experiments with the 

 S-type of plug is about three times the deviation of the mean of the 

 results of the fxCp measurements — that is, about 2.5 per cent, of fx 

 itself. There does not appear to be any reason for suspecting that 

 the experimental work of Knoblauch and Mollier was such as to 

 involve a constant error in either direction. Their experimental 

 errors, as indicated by the values of Cp plotted on Fig. 5 of their paper, 

 were of the order of about one per cent, in the region of the plane here 

 involved. It happens that their Cp curve for 4 kgm./cm.^, the posi- 

 tion of which largely determines the values of Cp used in the calcula- 

 tions of fx, is for some reason drawn by them along the lower edge of 

 their band of experimental points in the temperature region within 

 which the fxCp measurements were made. If the curve had been here 

 so located as to pass through the mean position of their experimental 

 points, about one-third of the discrepancy of 2.5 per cent, noted above 

 would disappear. 



As has been already stated, the /xCp experiments were dropped 

 rather early in the present research in order to settle the heat-leak 

 question in the adiabatic work, and there is not sufficient experimental 

 evidence to yield any conclusion of consequence regarding a constant 

 error of either sign, or regarding the cause of the large experimental 

 errors, in them. ^Yith plugs A2 and A3, the effect of high steam 

 velocity in the cross-channel would be to make the measured values too 

 small. This effect ought to be absent from the results obtained with 

 plug A4, wdiich are, in fact, larger than with the other plugs. Varia- 

 tions in the provisions for mixing the steam produced no consistent 

 results. It was suspected at first that an outward heat-leak might 

 have been due to the fact that, in order to get heat from the heating- 

 coil into the steam, it had been necessary to make the temperature 

 of the coil considerably higher than that of the oil bath. Measure- 

 ments of the resistance of the manganin coil used with plugs A2 and 

 A3 seemed to indicate this, but the indication was not reliable, owing 

 to the small temperature coefficient of manganin. The matter was 

 more thoroughly tested in plug A4 by employing a heating coil of 

 stranded invar. This alloy has a high specific resistance and, for an 

 alloy, a high temperature coefficient.* The resistance of the coil, at 



* Mean coefficient, 30° to 170° C, 0.00145 per °C. 



