was in immediate contact with the optical frame. The girders at first rested on the same gallows as supported the frame (as shown in the figure, page 767, just referred to), but this was ultimately found to be bad, because of a torque received from the whirling air and transmitted to these piers, which conveyed some trace of it to the frame. So while the frame was still supported on its independent piers, and the whirling machine was still clamped to its massive stone altar on the rock, the drum which received and screened the blast was now separately supported by special uprights from the floor (on which people did not walk during an observation), and this, on the whole, was an improvement. Any torque effect, however minute, being of a reversible character, was peculiarly dangerous, for it might easily have been mistaken for a result of the kind that was being looked for.
This memoir shall be abbreviated by the omission of all the careful sets of readings taken during this period, a record which occupies seventy pages of the laboratory note book; for it must be admitted that, although representing a good deal of work, they fail ultimately to show the air effect; and this probably for the reason that any effect of that magnitude would be certainly masked by the residual slight disturbing causes present.
The only thing I will record is a plotting of one of the larger spurious shifts (obtained before the foundation was inspected and altered) to illustrate its typical lagging character. The dots in this case represent individual readings, not averages of setting, and they incidentally show the kind of setting which is possible at high speeds through all the cover-glasses, with the light three times round, and when the steadiness was by no means perfect. The process was as follows :—
The micrometer wires were set, the single vertical wire in the middle of the middle band, and the X wire on the yellow of the first band to the left; or else vice versâ. Both wires were read, at gradually increasing, and then at decreasing speeds, and the results plotted on the right-hand side of the two diagrams (figs. 5 and 6), so as to show (a) the shift of the middle band due to strains and slight communicated tremors, (b) the change in the scale of wave-length due to concertina action. Then the brushes of the dynamo were reversed, and another spin taken in the opposite direction, and then the readings taken which are plotted on the left-hand side of the two diagrams. The total maximum shift was about 14th of a band on this occasion.
The following averages of a set of readings taken in July, 1893, may also be quoted :—
| With disks stationary, the middle band read | 94.4 | each division being | 161th | of a wave-length. |
| Disks revolving 3000 a minute, the middle band read | 94.4 | " | 159 | " |
| Disks stationary again, the middle band read | 95.6 | " | 167 | " |
| Disks revolving 3000 a minute in the opposite direction, the middle band read | 95.5 | " | 177 | " |
| Disks stationary again, the middle band read | 92.3 | " | 160 | " |
