In making permanent temperature scales from these curves to attach to the existing galvanometer scale, intervals of 100° may be taken and marked opposite to the corresponding divisions on the existing scale. Each 100° may then be equally subdivided into as many parts as the length of scale permits, and numbered at suitable intervals. If the junction used yield a calibration curve departing greatly from a straight line, every 50° interval should be taken, or, if necessary, every 25°. In the examples given both curves are nearly straight lines in the working region, viz. 400° to 800° for the iron-constantan junction, and 500° to 1100° for the platinum-iridioplatinum. One precaution necessary in standardizing an indicator by this method is to ensure that the metals used are pure, as impurities lower the melting points. If ordered as “pure” from any dealer of repute, the metals will generally be found satisfactory. The common salt used should be the ordinary salt sold in blocks, and not a prepared table salt. A second precaution, when observing melting points, is to guard against a possible error due to the substance becoming “surfused” or “overcooled”; in which case the temperature falls below the ordinary freezing point before solidification commences. When freezing occurs, however, the temperature rises to and remains at the true melting point, and an increase of deflection following a gradual fall always indicates overcooling. The higher deflection then attained is the true freezing point. Antimony frequently overcools to 600° before freezing, but on setting rises to the correct figure—631°. All metals and salts are liable to overcooling occasionally. =Standardization by Measurement of E.M.F.=—It has been found, as the result of experiments, that the relation between the E.M.F. developed by a junction and its temperature—under constant conditions of the cold junction—may be expressed approximately by a formula as under:— log E = A log _t_ + B (Holman’s formula), where E = electromotive force in microvolts, _t_ = temperature in Centigrade degrees, and A and B are constants depending upon the junction. With certain junctions this formula may be applied over the working part of the scale with an error not exceeding 2° C., but with others the discrepancy is greater. In order to determine the constants A and B, it is necessary to measure the E.M.F. at two known temperatures, which should be chosen as far apart as possible in the working region. When these constants are known, a measurement of E enables the temperature _t_ to be found by calculation.