The trouble with a coherer is that when once it becomes conductive it remains so unless the filings be shaken apart. Lodge therefore arranged for the tube to be continually struck by clockwork or by a mechanism like that of an electric bell. Marconi effected a further improvement by making the current passing through the coherer control the striking mechanism, so that the latter is normally quiet but administers one or two taps at just the right moment. Sir Oliver Lodge and Dr Muirhead devised another detector which, though quite different in form, is really much the same in principle. A steel disc with a sharp knife-like edge is made to rotate above a vessel of mercury. The edge just touches the mercury but no more. On the top of the mercury there floats a thin layer of oil, a bad conductor. Now as the disc revolves it picks up on its edge a film of oil, which it carries down into the mercury. The film adheres so tightly that it prevents the moving disc from actually touching the liquid metal. Thus, under normal conditions, the two are electrically insulated from each other by the film of oil and no current can pass from mercury to disc. Oscillations, however, caused by incoming electric waves, are able to break through the oil film and so bring disc and mercury into contact, whereupon the current flows. The constant movement of the disc restores the oil-film as soon as the oscillations cease. The reason why these detectors act as they do is not quite understood. One suggested explanation is that the oscillating currents heat the particles and so partially weld them together. Another is that adjacent particles become charged as the plates of a minute condenser, and so are drawn tightly together as the plates in an electrostatic voltmeter are drawn towards each other. Supposing that the original non-conductivity of the loose filings be due to the film of air which may surround them, either of these things would account for the film being broken or squeezed out, resulting in better contact and improved conducting power. But both suggestions seem to be contradicted by the fact that if the pieces in contact be of certain substances the coherer works the opposite way. Under those conditions the conductivity is normally good, but the influence of the incoming waves causes it to become bad.