sufficient to say now that under these conditions equal lengths of filament of the same thickness—­in other words, bodies of equal bulk—­were brought to incandescence.  The three bulbs were sealed to a glass tube, which was connected to a Sprengel pump.  When a high vacuum had been reached, the glass tube carrying the bulbs was sealed off.  A current was then turned on successively on each bulb, and it was found that the filaments came to about the same brightness, and, if anything, the smallest bulb, which was placed midway between the two larger ones, may have been slightly brighter.  This result was expected, for when either of the bulbs was connected to the coil the luminosity spread through the other two, hence the three bulbs constituted really one vessel.  When all the three bulbs were connected in multiple arc to the coil, in the largest of them the filament glowed brightest, in the next smaller it was a little less bright, and in the smallest it only came to redness.  The bulbs were then sealed off and separately tried.  The brightness of the filaments was now such as would have been expected on the supposition that the energy given off was proportionate to the surface of the bulb, this surface in each case representing one of the coatings of a condenser.  Accordingly, time was less difference between the largest and the middle sized than between the latter and the smallest bulb.

An interesting observation was made in this experiment.  The three bulbs were suspended from a straight bare wire connected to a terminal of the coil, the largest bulb being placed at the end of the wire, at some distance from it the smallest bulb, and an equal distance from the latter the middle-sized one.  The carbons glowed then in both the larger bulbs about as expected, but the smallest did not get its share by far.  This observation led me to exchange the position of the bulbs, and I then observed that whichever of the bulbs was in the middle it was by far less bright than it was in any other position.  This mystifying result was, of course, found to be due to the electrostatic action between the bulbs.  When they were placed at a considerable distance, or when they were attached to the corners of an equilateral triangle of copper wire, they glowed about in the order determined by their surfaces.

As to the shape of the vessel, it is also of some importance, especially at high degrees of exhaustion.  Of all the possible constructions, it seems that a spherical globe with the refractory body mounted in its centre is the best to employ.  In experience it has been demonstrated that in such a globe a refractory body of a given bulk is more easily brought to incandescence than when otherwise shaped bulbs are used.  There is also an advantage in giving to the incandescent body the shape of a sphere, for self-evident reasons.  In any case the body should be mounted in the centre, where the atoms rebounding from the glass collide.  This object is best attained in the spherical bulb;