[Illustration: _Lodge_ _Phil. Trans. 1893. A. Plate 31._ The mirror frame is pivoted on a vertical axis here Horizontal axis of frame Mode of mounting the semi-transparent mirror M so as to give altitude and azimuth movement to the reflected beam Details of brass plate supporting fourth mirror front, side and back views Back view shows the three slots in which the ends of the supporting screws rest giving a fine adjustment, the plate being supported by three rigid pushes and three elastic pulls. FIG. 14 Plan of optical frame with steel disk in position, and glazed drum to isolate them from the frame. G represents one of the panes of optical glass. Supports of telescope and collimator also shewn, and part of the fixing of the four mirrors 1.2.3.4., three of them let into recesses in the wooden frame, each mirror held by a brass plate supported by three finely cut screws against which it is pressed by the spring-bolts shewn M is the semi-transparent mirror West, Newman lith FIG. 14. _To face page 74._] The expense of the apparatus was borne by my friend the late George Holt, shipowner, of Liverpool. Fig. 16 exhibits something like the appearance seen in the eye-piece, with the interference bands on each side of the middle band, and with the micrometer wires set in position--each moved by an independent micrometer head. The straight vertical wire was usually set in the centre of the middle white band, and the =X= wire on the yellow of the first coloured band on one side or the other. The method of observation now consists in setting a wire of the micrometer accurately in the centre of the middle band, while another wire is usually set on the first band to the left. Then the micrometer heads are read, and the setting repeated once or twice to see how closely and dependably they can be set in the same position. Then we begin to spin the disks, and when they are going at some high speed, measured by a siren note and in other ways, the micrometer wires are reset and read--reset several times and read each time. Then the disks are stopped and more readings are taken. Then their motion is reversed, the wires set and read again; and finally the motion is once more stopped and another set of readings taken. By this means the absolute shift of middle band, and its relative interpretation in terms of wave-length, are simultaneously obtained; for the distance from the one wire to the other, which is often two revolutions of a micrometer head, represents a whole wave-length shift. In the best experiments I do still often see something like a fiftieth of a band shift; but it is caused by residual spurious causes, for it repeats itself with sufficient accuracy in the same direction when the disks are spun the other way round.