bar.  This feature is also shown in Fig. 14.  The bars, after being carried across the support, were inclined into the adjacent span and provided with a liberal, well-rounded hook, furnishing efficient anchorage and provision for reverse stresses.  This was at one end only, for—­to make matters worse—­the other end was a wall bearing; consequently, the benefit of continuity was denied.  The bent-up bars were given a double reverse bend, as already described, carrying them around, down, in, and up, and ending finally by clasping them in the hook of the horizontal bar.  This apparently stiffened up the free end, for, under the test load, its action was similar to that of the completely restrained end, thus attesting the value of this method of end-fixing.

The writer has consistently followed this method of reinforcement, with unvaryingly good results, and believes that, in some measure, it approximates the truth of the situation.  Moreover, it is economical, for with the bars bent up over the supports in this manner, and positively anchored, plenty of U-bars being provided, it is possible to remove the forms with entire safety much sooner than with the ordinary methods which are not as well stirruped and only partially tied across the supports.  It is also possible to put the structure into use at an earlier date.  Failure, too, by the premature removal of the centers, is almost impossible with this method.  These considerations more than compensate for the trouble and expense involved in connection with such reinforcement.  The writer will not attempt here a theoretical analysis of the stresses incurred in the different parts of this beam, although it might be interesting and instructive.

[Illustration:  FIG. 14.]

The concrete, with the reinforcement disposed as described, may be regarded as reposing on the steel as a saddle, furnishing it with a rigid jacket in which to work, and itself acting only as a stiff floor and a protecting envelope.  Bond, in this case, while, of course, an adjunct, is by no means vitally important, as is generally the case with beams unrestrained in any way and in which the reinforcement is not provided with adequate end anchorage, in which case a continuous bond is apparently—­at any rate, theoretically—­indispensable.

An example of the opposite extreme in reinforced concrete design, where provision for reverse stresses was almost wholly lacking, is shown in the Bridgeman Brothers’ Building, in Philadelphia, which collapsed while the operation of casting the roof was in progress, in the summer of 1907.  The engineering world is fairly familiar with the details of this disaster, as they were noted both in the lay and technical press.  In this structure, not only were U-bars almost entirely absent, but the few main bars which were bent up, were stopped short over the support.  The result was that the ties between the rib and the slab, and also across the support, being lacking, some of the