exactly the same. The circular arch was, therefore, adopted on account of its superior appearance.

Concerning the relations between the principal dimensions for arch bridges of various types but little can be said, for the reason that but little is known, because of the scarcity of such bridges in this country. In most cases the length of span and the rise are determined by the existing conditions at the crossing. For any given span, the greater the rise the less the effect of uniform load stresses, but the greater the effect of partial load stresses, and vice versa. Again, for any given span and rise, the arch depth does not affect the uniform load stresses materially, while it does so affect the partial load stresses; and as the latter are inferior in importance to the former, it results that the depth of an arch for economy of material will be very much less than the best depth for all ordinary truss of the same span. The arch depth, too, will depend upon whether the arch has fixed ends and continuous crown, hinged ends and continuous crown, or hinged ends and hinged crown. For the first type, a varying depth increasing from the centre to the ends is economic; for the second, a varying depth increasing from ends to centre is best; while for the third, a constant depth from end to end seems preferable. Again, the arch depth will depend considerably upon the style of web, i.e., whether it be plate, open-riveted, or pin-connected. The best depth or depths to adopt for any case should be given a special study, in making which Chapter VI of Prof. Howe's book will be found of great assistance.

In respect to the style of curve to adopt, whether circular, parabolic, or elliptical, the author's preference would generally be for the circular on account of its simplicity, although the parabolic might theoretically give better results.

In the author's opinion, a plate-girder arch should be made without hinges, an open-webbed riveted arch either with or without hinges, and a pin-connected arch with hinges. In the latter case, it is only the web members that should be pin-connected, for the chord members should be riveted up and fully spliced from end to end. There should be only a  single   system   of  cancellation   used   in  webs  of  arches,  so as to