222 
MR. H. L. CAL LEND AR ON THE PRACTICAL 
was probably clue to an imperfection in the drying apparatus which I subsequently 
discovered. The exit tube from the last drying bottle was plugged with cotton and 
glass wool, to stop dust, so tightly that the air could not pass freely, and was connected 
to the apparatus by an old and rather rotten piece of black seamed rubber tubing 
through which damp air probably leaked when it was suddenly connected with the 
vacuum of the air thermometer bulb. The wdiole drying apparatus was, therefore, set 
up afresh before Series A.-iv. with new sulphuric acid U-tubes (prepared by Stas’ 
method) and seamless solid rubber tubing, and the joints made perfect with hot 
paraffin wax. The whole was tested, and found to be perfectly air-tight. 
Assuming that the anomalous value of 9 0 = 270H was due to the presence of water 
vapour, the error may be eliminated beyond 100° 0. by applying from this point 
onwards the usual value of the coefficient of dilatation for dry air, namely 6 0 = 272'90. 
A correction (t — 100)-f 7 - 8 „ has, therefore, to be added to the values of t given in the 
Tables A.-ii., A.-m. 
The values of pt require no correction, and are calculated, assuming the coefficient 
•003460. 
Thus reduced, the latter observations in Table V., p. 192, become the following 
Table A.-vi. 
N umber of observation. 
Temperature. 
DifFeren^e 
observed. 
Difference by 
curve. 
By air 
thermometer. 
By platinum 
thermometer. 
O 
O 
O 
o 
III., 9. 
268-5 
264-8 
3-7 
7-1 
I., 6. 
282-1 
277"5 
46 
8-1 
II., 1. 
356-1 
343-1 
13-0 
14-2 
III., 7. 
401-0 
382-1 
18-9 
18-9 
1 HI., 1. 
472-6 
447-1 
25-5 
27-5 
; hi., 8. 
5845 
542-6 
41-9 
44"6 
Mean error, 2’1° C. 
The broken line on Plate 13, fig. 9, is used for deducing the temperature t by air 
thermometer from the formula 
t—pt — 1*57 {(£/100) 3 — (^/100)}, 
the temperature pt being known by observation. 
It gives at once the difference {t — pt) in terms of pt as abscissa. 
It is readily constructed by measuring off from each point of the first curve, back¬ 
wards along the abscissa, a distance equal to one-tenth of the ordinate. 
By using a curve like this an accuracy of-j^-th degree Cent, at 600° maybe obtained 
in the relative values of t deduced from observations with a known platinum wire, or 
with platinum wires which have been compared with a known wire. The absolute 
