The effect was always zero, however, when spurious results were eliminated; and it is clear that at no practicable speed does either electrification or magnetisation confer upon matter any appreciable viscous grip upon the ether. Atoms _must_ be able to throw it into vibration, if they are oscillating or revolving at sufficient speed; otherwise they would not emit light or any kind of radiation; but in no case do they appear to drag it along, or to meet with resistance in any uniform motion through it. Only their acceleration is effectual. In the light of Larmor's electron theory, we know now that acceleration of atoms, or rather of a charge upon an atom, necessarily generates radiation, proportional in amount to the _square_ of the acceleration--whether that be tangential or normal. There is no theoretical reason for assuming any influence on uniform velocity. And even the influence on acceleration is exceedingly small under ordinary circumstances. Only during the violence of collision are ether waves freely excited. The present experiment, however, has nothing to do with acceleration: it is a test of viscosity. An acceleration term exists in motion through even a perfect fluid. [Illustration: FIG. 12. General view of whirling part of Ether machine, with pair of steel disks, and motor. _To face page 72._] [Illustration: FIG. 13. General view of optical framework--sustaining mirrors, telescope, and collimator--to surround the disks of the Ether machine. Compare fig. 11. _To face page 73_.] The conclusion at which I arrived in 1892 and 1893 is thus expressed (p. 777 of vol. 184 _Philosophical Transactions of the Royal Society_): "I feel confident either that the ether between the disks is quite unaffected by their motion, or, if affected at all, by something less than the thousandth part. At the same time, so far as rigorous proof is concerned, I should prefer to assert that _the velocity of light between two steel plates moving together in their own plane an inch apart is not increased or diminished by so much as the 1/200th part of their velocity_." That was the conclusion in 1893; but since then observations have been continued, and it is now quite safe to change the 1/200th into 1/1000th. The spin was sometimes continued for three hours to see if an effect developed with time; and many other precautions were taken, as briefly narrated in the _Philosophical Transactions_ for 1897. The following illustrations give an idea of the apparatus employed.