572 MR EDMUND J. MILLS’S 
That the coefficient a="0001545 is too large in a particular case was 
noticed by the author of the article “ Schmelzpunkt ” in the Handwiirterbuch 
der Chemie, vii. 368; and he suggested that ‘000185 would be more accurate 
generally. This number does not differ much from one deducible for short 
exposures from my own formula. 
According to Movusson’s hypothesis,* the exposed capillary column of 
mercury is, together with the glass tube containing it, in the same predicament 
as a rod conducting heat applied at one of its extremities. He determines 
the correction as follows :—The thermometer is completely immersed in a 
large bath whose temperature changes very slowly, and a reading taken ; it is 
then, after an interval of say five minutes, raised N° out of the bath; and these 
operations are repeated. The mean of the first readings gives the actual 
temperature T of the liquid ; the mean of the second readings furnishes a lower 
number T’. The difference AT’=T—T?’ is thus known, and yields the constant 
AT’ , 
e=aN—y 
which implies the hypothesis in question. If in any given case a correction 
AT’ has to be made when T’ and N are observed, we have 
AT’=T(aN—e), 
an equation from which, by two observations, a could be calculated if required. 
WULLNER?t gave in the same year a very similar formula ; adding, however, 
an observation of the temperature of the top of the column. Neither of these 
expressions has hitherto been submitted to any extended experimental 
verification. 
It is plain that various expressions may equally well represent small total 
corrections. In this case it seems inaccurate to assume that a vertical 
thermometer, consisting of two materials heated below, and therefore causing 
an upward current of diminishingly warm air at the outer glass surface, is 
merely in the condition of a uniform rod heated at one end, and conducting 
heat in a medium at uniform temperature and free from convection. The 
experimental state of affairs is probably this. With an exposure of medium or 
considerable length, the current of heated air at the thermometer’s surface 
heats the instrument much faster, and higher up, than it could be heated by 
conduction ; the temperature of the middle of the exposed part must therefore 
be a close approximation to the mean temperature required. In the case of a 
short exposed length, conduction may play a more important part then con- 
* Pogy. Ann. cxxxiii, (1868) p, 316, 
+ Lehrbuch der Experimentalphysik, iii, pp. 295-298. 
