236 PROCEEDINGS OF THE AMERICAN ACADEMY. 



mensurate with the sensitiveness of the instrument, that it is impossi- 

 ble to draw a curve through them which shall be much more reliable 

 in the upper parts (the portion most frequently needed) than the 

 straight line. Although this statement is perhaps counter to others 

 that have been made on the subject, it is the author's opinion, arrived 

 at after prolonged experience with the instrument. 



In subsequent temperature measurements this curve is, of course, 

 used for interpolation in the customary manner ; that is, any deflec- 

 tion being observed, and at the same time the temperature c' of the 

 cold junction, the plot is entered with this deflection as abscissa, and 

 the corresponding ordinate of the curve is read off. This ordinate 

 will be approximately {t — c') the temperature of the hot minus that of 

 the cold junction. Hence the desired unknown temperature t is ob- 

 tained approximately by adding c' to this ordinate. , 



Proposed Method. 



Observations. — The scale is adjusted so that the spot reads zero 

 with the circuit open. The cold junction is kept at a temperature 

 about that of the room, and measured by a mercurial thermometer. 

 In careful work this temperature should be kept constant, or nearly 

 so, either at 20°, or, better, at 0° C. The other junction is tlien ex- 

 posed successively to two known temperatures, preferably boiling sul- 

 phur and melting copper (for reasons to be given), or, if temperatures 

 upwards of 1200° C. or 1300° C. are chiefly to be measured, to melt- 

 ing copper and melting platinum. Deflections and temperatures of 

 the cold junction are taken. 



Correction for Gold Junction. — If the cold junction is always kept 

 within about one degree of a constant temperature c°,e. g. 20° (or 

 0°), both in calibration and subsequent work, these observed deflec- 

 tions require no correction. If that is not the case, then the deflec- 

 tion should be corrected to obtain what it would have been had this 

 been done. This is readily effected as follows. Let a be the 

 change of deflection per degree at 20° C. This may be found nearly 

 enough by observing once for all the deflection with the cold junction 

 at some observed temperature, anywhere from 10° to 15°, and the hot 

 junction at an observed temperature from 30° to 40°. The deflection 

 divided by the temperature difference will, of course, be a. Then 

 if the observed temperature of the cold junction in a measurement 

 is, e. g., 16° instead of 20°, the correction would be (20 — 16) a = 4 a, 

 to be subtracted from the observed deflection, since that was ob- 

 viously too large. These corrections are small, and can easily be 



