704 PROFESSOR THOMSON ON THE ELECTRO-DYNAMIC QUALITIES OF METALS. 
(§ 86) above: the temperatures of the baths can be changed with great rapidity, in 
consequence of the smallness of the quantities of oil which they contain ; and by 
watching the thermometers and adjusting the gas-burners, can be regulated as desired 
with great ease. I have found it not a small practical advantage to be freed from 
the necessity of bending the mean part of the conductor to be tested, and of making 
the arrangements to prevent irregular contacts and to keep the junctions and the 
thermometers in their proper positions immersed in the oil. When a rare metal is 
to be tested, or one, such as sodium or potassium, which cannot be kept in air, it will 
be of great consequence to be able to apply the tests to a little straight bar or slip 
only a few inches long, or to a small column filling a glass tube. 
95. For experimenting at low temperatures a modified apparatus was made, con- 
sisting of a double wooden box, each compartment, nearly a cube of 4 inches side, 
fixed to a common base with a space of about ^ inch between their sides, and a glass 
tube running through them and cemented at the apertures in the sides so as to hold 
water-tight and resist the action of acids which 
might be employed in freezing mixtures. The 
conductor to be tested and the thermometers 
are arranged in this glass tube, as in that of 
the other apparatus (§ 89 to 93) ; and while a 
freezing mixture is kept in one compartment, 
the other is either allowed to take the atmospheric temperature, or is heated by hot 
water or steam. 
96. The way of experimenting which I followed, was to raise the temperature of 
one bath until a deflection of the galvanometer-needle became sensible ; then to go 
on raising it, and letting that of the other follow, so that the two thermometers may 
indicate as nearly as may be a constant difference of temperatures ; and to watch 
the needle until a reversal is observed, or until the limit of temperature which the 
arrangement admits of is reached. 
As soon as a reversal is obtained, the two thermometers are allowed to sink until 
the needle begins to return from its reverse deflection. When it approaches zero 
the thermometers are kept from any rapid changes, but allowed to sink very slowly, 
with always the same difference, or at least with a quite decided difference of the 
same kind as that raised between them at the beginning. The last readings of the 
sinking thermometers which give a sensible deflection before the original deflection 
is recovered, several readings when the needle appears perfectly at zero, and the first 
readings when the needle is discovered to deviate again in the original direction, are 
carefully noted. The arithmetical mean of the temperatures of the two thermome- 
ters for each of these simultaneous or nearly simultaneous readings is taken; and it 
is generally found that the means derived from the readings taken when no deflec- 
tion can be discerned, lie within a fraction of a degree of the mean of the last sink- 
ing mean temperature of the junctions which show one deviation, and the first which 
shows the deviation in the other direction. The mean of either the readings which 
Fig. 17. 
