'ARU8.J PORCELAIN AIR THERMOMETRY. 231 
•efers, of course, to apparatus of porcelain. In the fire-clay air ther- 
mometer the porosity of stem will doubtless be greater and the com- 
pensator corrective essential. 
Finally, it is seen that /3 measured correctly within 5 per cent, is sat- 
isfactory. Concerning both a and (3 much that has been said above is 
aere applicable (Cf. page 198.) 
Accuracy of the measurements made. — After this tolerably full analy- 
sis of possible errors, the degree of certainty with which the present 
data attest the excellence of this method of air thermometry and of 
pyrometric comparison employed may be satisfactorily discerned. It 
is expedient to use the graphic tabulation Figs. 41 and 42, in which all 
the data in question have been inserted. 
Let us consider the result for Bulb No. 1 first. The points belong- 
ing to the divers Series I to IV for Bulb I are marked with numerals; 
moreover, when temperature increases (gas on) the numeral is placed 
above the point ; when temperature decreases (gas off, furnace cooling) 
the numeral is placed below the point. In this way the amount of in- 
formation contained in the diagram is much increased. With respect 
to Series I the remark has already been made that but a siugle ther- 
mometer was used in measuring T x . Now, the temperature of a room 
in which furnace experiments are being made certainly differs in tem- 
perature by as much as 1° for a vertical height as great as B G, Fig. 
40 (150 cm ). Hence the exceptionally lateral position of the series of 
points marked "1" is easily accounted for. Nor was a correction ap- 
plied in these cases for permanent variation of the volume v . Again, 
in Series II an accident by which the thermo-couple was withdrawn 
from the air-thermometer bulb cooled the electrical apparatus abnor- 
mally. This occurred during the stage of decreasing temperature, and 
the lateral position of certain points marked "2" on the diagram is also 
accounted for. The remaining points are grouped in close proximity 
to a uniform locus. The maximum elongation of any of the points, 2, 
3, 4, in question is not greater than 10°, whereas, as a rule, this differ- 
ence is very much smaller. At the outset it is to be borne in mind 
that into this aggregate maximum discrepancy of 10° are crowded all 
the errors of the thermo-electric measurement, to say nothing of the 
errors incurred by the double graphic interpolation by which Fig. 41 was 
derived from the individual time series of observations of e and T. Of 
course, results of this kind are susceptible to great improvement if the 
observations are made by a number of observers instead of by one ob- 
server only, for in this way the time error may be eliminated, and 
observations may be made simultaneously. Quite apart from these 
considerations an error of 10° is easily incident to the method in its 
present stage of experimental development. However carefully the 
manometer may be screened from the furnace an error in T x of 0.2° 0., 
or even more, is not improbable; nor was the attempt made to measure 
V x with greater precision than 0.1 cc . This already accounts lor half of 
(SSry) 
