when a trestle is situated near the middle of a sharp curve or near the apex of two heavy, rising grades, it would be bad judgment to assume a high velocity of train when finding the stresses due to centrifugal loading.

The combination of stresses in cantilever bridges and in arches is not so complicated as it is in trestles; but it is to be noted that the sections of members do not need to be increased because of erection stresses, unless such total stresses (including those from wind under an assumed probable pressure of ten (10) or fifteen (15) pounds per square foot) raise the intensities on the computed sections above those specified for a combination of the usual loads with wind.

In summing up stresses care must be taken to add only those that can act simultaneously, because some stresses can never occur together; for instance, live load and erection stresses in cantilevers and in arches erected by cantilevering, and live load and wind stresses in highway bridges. This word of warning seems almost unnecessary; nevertheless a careless computer is liable to sum up stresses that cannot act together, as the author knows from personal experience.

At the present time there is a division of opinion among bridge specialists concerning the combination of stresses of opposite kinds, some of them going so far as to ignore altogether the effect of reversing stresses. Until good reason is offered for making a change, the author intends to adhere to the method which he advocates in "Bridge Engineering." It is as follows: If the cause of the reversal be wind, the effect of reversion is ignored, because not only is there generally a long interval between reversals, but also the maximum wind stress on any piece is of infrequent occurrence. Reversals due to live loads combined with impact are divided into two classes; first, those which occur in succession during the passage of a live load over the structure, and, second, those which are caused by different loadings. In the first case, each of the two kinds of stress is to be increased by seventy-five (75) per cent of the other, then the section required for each combination is to be computed and the larger of the two adopted. In the second case the procedure is similar to that just described, except that the percentage to be added is fifty (50) instead of seventy-five (75). The author does not deny that it might be perfectly safe to reduce these percentages to fifty (50) and twenty-five (25), respectively, but he is decidedly averse to ignoring altogether the effect of reversion.

In any case it would require some exceedingly strong evidence to induce him to change his method of computing the number of rivets for connecting main members, viz., to add together without any reduction the two stresses of opposite kinds and proportion for the sum. Stress reversal is certainly harder upon the connecting rivets than upon the members themselves which they join.

After all, this controversy about the proper allowance for reversal of stress may amount to much ado about nothing, because, as pointed out in Chapter XI, the difference in the weights of metal in two continuous-truss