It will, of course, be understood, when it is said that bridge-building may be called a science, that it can only be so when in the hands of an engineer whose judgment has been matured by wide experience, and who understands that no mechanical philosophy can be applied to practice which is not subject to the contingencies of workmanship. There are many bridges which will stand the test of figures very well, and which are nevertheless very poor structures. The general plan of a bridge may be good, the computations all right, and yet it may break down under the first train that passes over it. There are many practical considerations that cannot be, at any rate have not yet been, reduced to figures. It is not enough that the strains upon each member of a bridge should be correctly estimated, and fall within the safe limits: the different members of the bridge must be so connected, and the mechanical details such, as to insure, under all conditions, the assumed action of the several parts. In fine, while we can say that a bridge that does not stand the test of arithmetic is a bad bridge, we cannot always say that a structure which does stand such a test is a good one. We often hear it argued that a bridge must be safe, since it has been submitted to a heavy load, and did not break down. Such a test means absolutely nothing. It does not even show that the bridge will bear the same load again, much less does it show that it has the proper margin for safety. It simply shows that it did not break down at that time. Every rotten, worn-out, and defective bridge that ever fell has been submitted to exactly that test. More than this, it has repeatedly happened that a heavy train has passed over a bridge in apparent safety, while a much lighter one passing directly afterwards has gone through. In almost all such cases, the structure has been weak and defective; and finally some heavy load passes over, and cripples the bridge, so that the next load produces a disaster. For the test of a bridge to be in any way satisfactory, we must know just what effect such test has had upon the structure. We do not find this out by simply standing near, and noting that the bridge did not break down. We must satisfy ourselves beyond all question that no part has been overstrained.