If the stone is attached to the cap of the instrument, the table should be turned towards it so as to assure that the light passing into the instrument has actually traversed the stone. If little light enters through the opposite coign, a drop of oil placed thereon will overcome the difficulty (cf. p. 46). It is also necessary, for reasons mentioned above, to examine the stone in directions as far as possible across the girdle also. A convenient, though not strictly accurate, method is to lay the stone with the table facet on a table and examine the light which has entered the stone and been reflected at that facet. The stone may easily be rotated on the table, and observations thus made in different directions in the stone. Care must be exercised in the case of a faceted stone not to mistake the alteration in colour due to dispersion for a dichroic effect, and the stone must be placed close to the instrument during an observation, because otherwise the twin rays traversing the instrument may have taken sensibly different directions in the stone. Dichroism is an effective test in the case of ruby; its twin colours—purplish and yellowish red—are in marked contrast, and readily distinguish it from other red stones. Again, one of the twin colours of sapphire is distinctly more yellowish than the other; the blue spinel, of which a good many have been manufactured during recent years, is singly refractive, and, of course, shows no difference of tint in the dichroscope. Table VI at the end of the book gives the strength of the dichroism of the gem-stones. ABSORPTION SPECTRA A study of the chromatic character of the light transmitted by a coloured stone is of no little interest. As was stated above, the eye has not the power of analysing light, and to resolve the transmitted rays into their component parts an instrument known as a spectroscope is needed. The small ‘direct-vision’ type has ample dispersion for this purpose. It is advantageous to employ by preference the diffraction rather than the prism form, because in the former the intervals in the resulting spectrum corresponding to equal differences of wave-length are the same, whereas in the latter they diminish as the wave-length increases and accordingly the red end of the spectrum is relatively cramped.