Marcarino and Bassani have proposed the sodium liquid-vapour equilibri
um as a means to approximate the ITS-90 between 660 degrees C and 962
degrees C with an accuracy of about 10 mK. We have also measured the s
odium vapour pressure over this temperature range using a pressure-con
trolled heat pipe. Our results differ from those of Marcarino and Bass
ani by approximately 34 mK at 660 degrees C and 21 mK at 962 degrees C
with a linear temperature dependence of the difference between these
values. This relationship may be explained very satisfactorily by a di
fference in the impurity content of the two samples. We show that an a
dditional 200 x 10(-6) mass fraction of potassium in our sample as com
pared with theirs is sufficient to explain the discrepancy. As sodium
of sufficient purity to ensure repeatability of its use as a temperatu
re standard at the millikelvin level is not commercially available, it
appears necessary to characterize the pressure/temperature relationsh
ip for each sample prior to relying on a measurement of the pressure t
o determine the temperature of the heat pipe. The alternative is to re
ly on the heat pipe as an intercomparison device with the temperature
given by a calibrated thermometer.