Two sticks of quartz are introduced and caused to meet just in front of the inner cone--the hottest part of the flame. They are then drawn apart so as to form a fine neck, which softens and is bent in the direction of motion of the flame gases. When fusion is complete the neck separates into two parts, and a thread is drawn from each of them. By alternately lightly touching the rods together, and drawing them apart, quite a mass of threads may be obtained in two or three minutes, when the process should be stopped. If too many threads get entangled in the pins, one gives one's self the unnecessary trouble of separating them. On taking down the card it will be found that the threads have been caught by the pins; but the card now being laid black side upwards, the former easily slip off the points. Threads at least a foot long, and perhaps vastly longer, may be obtained by this method, and are extraordinarily fine. When I first read Professor Nichols' statement (Electric Power, 1894) as to the value of these fibres for galvanometer purposes, I was rather sceptical on the ground that the threads would tend to get annealed by being drawn gradually, instead of suddenly, from a place of intense heat to regions of lower temperature. Now annealing threads by a Bunsen makes them rotten. The threads being immersed in the hot flame gases could only cool at the same rate as the gas, and it was not--and is not--clear to me that annealing of the threads can be avoided. On the other hand, it may be possible that a thread cooled slowly from the first does not suffer in the same way as a cold thread would do when annealed in a Bunsen flame. Again the velocity of the gases is beyond doubt exceedingly high, so that the annealing, even supposing it to be deleterious, might not be carried very far. Threads drawn by this method and measured "dry," i.e. by mounting them on a slide without the addition of any liquid, turned out to have a diameter of about 1/20000 of an inch. I do not think I could manage to mount such fine threads without very special trouble. All the threads lying on the board, however, were found in reality to consist of three or four separate threads, and there is no reason why several threads should not be mounted in parallel, provided, of course, that they are equally stretched and touching each other. Equality of tension in the mounting could be secured by making one attachment good, then cementing the other attachment to the other end of the threads, and "drawing" the two attachments slightly apart at the moment the cement commences to set. This method may turn out to be very valuable, for, so far as I can see, the carrying power would be increased without an increase of torsional stiffness of anything like so high an order as would be the case were one thread only employed. On the other hand, the law of torsion could hardly be quite so simple, at all events, to the second order of approximations.