Reference to fig. 2 will show that in order to realise this circuit in practice, one of the wires forming the couple must be used in the form of leads to the galvanometer. This can readily be done if the material of the wire is cheap; but if platinum or other expensive metal be used, and the galvanometer be some yards distant, the question of cost necessitates a compromise, and the circuit is then arranged as in fig. 3. The wires forming the hot junction are brought to brass terminals T T, from which copper wires lead to the galvanometer G. This arrangement results in three effective junctions, viz. the hot junction A to B; the junction A to brass, and the junction B to brass. It will be seen that the two junctions of copper to brass are in opposition, and cancel each other for equal heating; and the same applies to the galvanometer connections. A circuit thus composed of three separate junctions does not permit of a simple expression for the net E.M.F. under varying temperature conditions, and to avoid errors in readings care must be taken to prevent any notable change of temperature at the terminals T T in a practical instrument arranged as in the diagram. [Illustration: FIG. 3.—THREE-JUNCTION THERMO-ELECTRIC CIRCUIT.] A point of practical utility in thermo-electric work is the fact that if a wire be interrupted by a length of other metal, as indicated at C in fig. 3, no current will be set up in a circuit if both joints are equally heated, as the electromotive forces generated at each junction are in opposition. It is thus possible to interrupt a circuit by a plug-key or switch, without introducing an error; always provided that an even temperature prevails over the region containing the joints. Another useful fact is that if two wires be brought into contact, they may be fastened over the joint by soldering or using a third metal, without alteration of thermo-electric value, except in rare cases. Thus a copper-constantan or iron-constantan junction may suitably be united by silver solder, using borax as a flux, thus avoiding the uncertainty of contact which must always occur when the wires are merely twisted together. Welding, however, is preferable to soldering.