~The Emission Spectra: Spark Spectra.~--The factors which tend to limit the value of the absorption spectra for analytical purposes, for the most part disappear when the emission spectra are employed. In the case of the spark spectra, indeed, great differences are observed according to the conditions and method of experiment; but the arc spectra are practically invariable under all conditions, and hence they constitute the ultimate test in all cases. The spark spectra are observed when one terminal--the cathode--of an induction coil is embedded in the oxides to be examined, and the discharge then passed. The discharge is also frequently passed between platinum poles partly immersed in a strong solution of a salt of the element under examination; a form of apparatus very suitable for this method of observation has been described by Sir W. Crookes.[186] The spectra so obtained are in a high degree characteristic, but they vary very considerably with the form and dimensions of the coil, the length and cross-section of the wires, the potential difference employed, and so on. An entirely new spectrum also is obtained in many cases by mere reversal of the current; under these conditions, a phosphorescent appearance is observed, the spectrum of which--reversed spark spectrum of de Boisbaudran--has been found in many cases to resemble the cathode luminescence spectra of Crookes. [186] _Proc. Roy. Soc._, 1903, ~72~, 295. ~The Arc Spectra.~--The final criterion of purity in the examination of a rare earth element is in almost all cases the arc spectrum. Since for some of the elements, especially in the yttrium group, the entire spectrum has not yet been accurately mapped out, spectra are generally observed frequently throughout the course of a fractionation; by this means, the separation can be followed by the disappearance of some lines, and the appearing or strengthening of others, and such examinations have led occasionally to the discovery of new elements (see, for example, under Separation of ytterbium earths, p. 205). Such determinations, however, require much time and extensive and complicated apparatus.