It is of advantage to make the tube t very thick, the hole through it very small, and to blow the sphere s very thin.  It is of the greatest importance that the sphere s be placed in the centre of the globe L.

[Illustration:  FIG. 14.—­FORMS AND PHASES OF THE ROTATING BRUSH.]

Figs. 14, 15 and 16 indicate different forms, or stages, of the brush.  Fig. 14 shows the brush as it first appears in a bulb provided with a conducting terminal; but, as in such a bulb it very soon disappears—­often after a few minutes—­I will confine myself to the description of the phenomenon as seen in a bulb without conducting electrode.  It is observed under the following conditions: 

When the globe L (Figs. 12 and 13) is exhausted to a very high degree, generally the bulb is not excited upon connecting the wire w (Fig. 12) or the tinfoil coating of the bulb (Fig. 13) to the terminal of the induction coil.  To excite it, it is usually sufficient to grasp the globe L with the hand.  An intense phosphorescence then spreads at first over the globe, but soon gives place to a white, misty light.  Shortly afterward one may notice that the luminosity is unevenly distributed in the globe, and after passing the current for some time the bulb appears as in Fig. 15.  From this stage the phenomenon will gradually pass to that indicated in Fig. 16, after some minutes, hours, days or weeks, according as the bulb is worked.  Warming the bulb or increasing the potential hastens the transit.

[Illustration:  FIG. 15.  FIG. 16.  FORMS AND PHASES OF THE ROTATING BRUSH.]

When the brush assumes the form indicated in Fig. 16, it maybe brought to a state of extreme sensitiveness to electrostatic and magnetic influence.  The bulb hanging straight down from a wire, and all objects being remote from it, the approach of the observer at a few paces from the bulb will cause the brush to fly to the opposite side, and if he walks around the bulb it will always keep on the opposite side.  It may begin to spin around the terminal long before it reaches that sensitive stage.  When it begins to turn around principally, but also before, it is affected by a magnet, and at a certain stage it is susceptible to magnetic influence to an astonishing degree.  A small permanent magnet, with its poles at a distance of no more than two centimetres, will affect it visibly at a distance of two metres, slowing down or accelerating the rotation according to how it is held relatively to the brush.  I think I have observed that at the stage when it is most sensitive to magnetic, it is not most sensitive to electrostatic, influence.  My explanation is, that the electrostatic attraction between the brush and the glass of the bulb, which retards the rotation, grows much quicker than the magnetic influence when the intensity of the stream is increased.