726 PROFESSOR TAIT ON THERMAL AND ELECTRIC CONDUCTIVITY. 
round and round, until it was very uniformly heated; being occasionally 
turned end for end. It was found that when the bar was not heated above 
200° C., but little oxidation was produced during the time required for the 
heating. -When it was necessary to raise the temperature higher, the nature of 
the effect on the surface was described by its colour, which was noted and 
compared with the effect found to be produced on different parts of the corre- 
sponding long bar by its more gradual heating. It would be very easy to burn 
a mixture of gas and air, and so to a great extent get rid of the possibility of 
smoking the surface, but practically it was found that no insuperable difficulties 
were introduced by taking the ordinary coal-gas. But, for a reason presently 
to be mentioned, the short bars had always to be raised to a temperature much 
higher than that at which the readings of the thermometers commenced. Thus 
all my results must necessarily be a little too large, as the cooling was in every 
case observed on a bar more oxidised than the portion of the long bar which 
had the same temperature. 
§ 10. With reference to the estimation of the true temperature of the bulbs 
of the thermometers from the readings of a variably heated stem, the great 
difficulty experienced was one felt by Forbes also—one which he endeavoured 
to get rid of by detaching arbitrarily a column of mercury, and throwing it up 
into the little bulb at the top of the thermometer, thus working from an 
arbitrary zero. Dr Batrour Stewart told me it was almost impossible to get 
trustworthy results from the thermometer so treated, and I determined to take 
my chance of the insufficient heating of the column of mercury in the thermo- 
meter, which was not directly immersed in the mercury in the holes in the bar. 
I do not think very much error can be introduced by this, for the following 
reasons. If we calculate for a temperature of 250° C.—which is nearly the 
highest-used in the greater number of the experiments—the utmost error that 
can be introduced in the indications of the thermometers used is somewhere about 
10°C. That is to say, the highest temperatures were read at the most 10° less 
than they would have been if the whole thermometer had been exposed to the © 
same temperature. This correction of 10° at 250° diminishes at lower tempera- 
tures, and increases at higher nearly as the square of the excess of temperature 
above the freezing point. But as the same thermometers, or exactly similar ones, 
were employed, under precisely* similar conditions, in the short bars as in the 
long ones, the difference between the corresponding errors in the two associated 
experiments must have been at most a fraction of a degree even at the higher 
temperatures, The numerical results, therefore, are stated in terms of the 
temperatures so read, and these involve (from this cause) an error in defect, of 
* Jan. 13, 1879. In spite of the contents of § 11*, now added, this is nearly true of my experi- 
ments, for the highest of the thermometer readings in the cooling bars were not used in the 
calculations. 

