As to the third class of clutches, much of interest can be said for the benefit of those particularly concerned. “A friction clutch” to mechanics is a familiar term, since the name is applied to all pulley clutches, that grip on a smooth surface. Many of these clutches are a success for the purpose for which they are intended. The most common form used on machines where the requirements are similar to those of a cycle, is the “Roller.” The cycle experimenter nearly always strikes upon this clutch first, and with sufficiently good reason. It has been adopted in many arts, and is used in England upon tricycles in combination with cranks, with moderate success, but just here allow me to call attention to a cardinal difference in the requirements of a clutch as used on crank tricycles and successfully in the arts heretofore. In the crank-clutch cycles the clutch is used for the purpose of detaching the cranks from the spindle when the machine requires no driving, as in running down grade, but when once the clutch is gripped, it remains so till further power ceases to be required. Now, this is also just the action of all belt-pulley clutches, and between such action and that required in a lever-clutch cycle the difference is exceedingly conspicuous. In the crank-clutch cycle, as in other uses, the immediate solid grip is a matter of little concern; if a half turn of the parts takes place before clutching, it does little harm, since it is so small a fraction of the entire number of revolutions to be made before the grip is released. But if a grip is to be taken at every down stroke of the foot, as in a lever-clutch cycle, the least slip or lost motion is fatal. This incessant clutching action, together with the great weight the parts have to sustain, and the repeated concussion of one piece upon another under this weight, makes up a combination of disturbing elements which will cause mischief against which it is almost impossible to provide. In a form of roller-clutch I have tried, the inner frame or carrier is made loose upon a spindle. In the drawing herewith annexed we have first a spindle in the centre, then a little open space around it, and then the clutch frame _bb_, which is connected loosely, not rigidly, to the drum. By this arrangement the pressure is distributed evenly upon the three rollers _d_, _d_, _d_, outwardly at three points against the casing, and in no event is the work done by a single roller. This device worked as well as any of this class I have tried; but the patterns are for sale at a very reasonable price. The main trouble I found in this contrivance and all other roller clutches was, that the great pressure disintegrated the oil, making a paste that would cause the rollers to slip in spite of everything.