Roman arches are nearly always semicircles, or at least circular segments. The usual explanation for this is lack of imagination or lack of understanding of the exact nature of the stresses involved. I find this explanation highly insulting to engineers whose work still stands and functions today, after several thousand years of neglect. (One can only wonder if the work of those engineers who belittled the Roman engineers will remain standing even 1/10 of this amount of time.) It is also insulting to the intelligence of 500-750 years worth of very clever engineers, who had plenty of time and plenty of incentive to come up with more economical methods of construction. Any construction engineers who constructed arches by the thousands and tens of thousands got very, very, very good at this art, and developed extremely sophisticated tools to be able to construct their arches quickly & with the least amount of extraneous material. The Romans not only constructed single arches, but many of their buildings & bridges consisted of a series of arches, each depending upon one another for support, so that they could not have been built or functioned as isolated arches. This meant that they had to be built together, simultaneously. After a search of the internet, it would appear that the most common suggestion for how to build an arch requires the building of a template for the arch -- typically out of wood -- and then using that template to hold the bricks or stones in place until the arch was complete, after which the template could be removed. But this means that the template would have to be strong enough to hold all of the stones in place until the keystone at the top completed the arch and it became self-supporting. The "template" method is very general -- it can be used to build an arch of any shape -- so if someone goes to the trouble of building a template, then how come he doesn't utilize that template to build an arch that is optimized for some other feature -- e.g., a more perfect deflection of stresses, or a more elegant-looking arch? If the template method was so common, how come there wasn't a much greater variety of arches? We're talking about a _thousand years_ of arch construction here! Even dumb engineers eventually become bored & try other things over a period of 1,000 years. Another problem with the template method -- especially with an arcade series of arches -- are that all of the templates have to be in place for all the arches prior to the installation of all the keystones. So the builder can't even re-use the template from one arch to the next. An even more serious problem with the template method is the amount of wood required -- arched structures were built in many locations where wood was very scarce. Simply finding enough wood and transporting this amount of wood to the building site might have been almost as much work as cutting & transporting the stones and/or bricks. My conclusion is that the Romans were more clever than this, and that the circular arc arch was the _necessary_, rather than the _incidental_, result of their method of arch construction. What is a circular arc? It is the _rotation_ of a line segment of given fixed length around an axis, or pivot point. But it is that very _rotation_ that may provide the insight to how the Romans could have constructed their arches so quickly & economically. Consider a circular clock face with typical numbering of the hours 1-12. The top half of this clock face is semicircular in form, going from "9:00" to "12:00" to "3:00". Now consider such a "clock" with two equal-length "hands": hand #1 (the "left" hand), and hand #2 (the "right" hand). The "left" hand will stay in the range "9:00" to "12:00", while the "right" hand will stay in the range "12:00" to "3:00". We will try to keep both the "left" and "right" hands at approximately the same elevation at all times. Thus, when the left hand is at "9:00", the right hand will be at "3:00"; when the left hand is at "10:00", the right hand will be at "2:00"; when the left hand is at "11:00", the right hand will be at "1:00". The Romans must have realized that _if an arch was going to be able to stand up on its own after construction_, then the forces on each of the stones in the arch _during_ construction should approximate those _after_ construction. In particular, the stones further down in the arch -- e.g., those at "10:00" and "2:00" -- can only feel the forces just above themselves and just below themselves, so if those forces approximate those in the final arch configuration, they will be happy and won't fall down. So the arch construction could proceed as follows. Build the arch "piers" -- the vertical stones below the start of the semicircular arch. Once these piers are completed & both sides are at the same level (no small feat it itself), a straight horizontal wooden beam can be built across the span from pier to pier. (Most Roman arches have a little "lip" right at the top of these piers, which provides a very clear delineation between the pier and the arch above it. This "lip" is no accident, and comes in handy for a variety of uses, both during & after construction.) In the center of this horizontal wooden beam which spans between the tops of the piers, a rotating "axle" is built, and the two "hands" of the "clock" are constructed which will rotate/pivot on this axle. The angled stones ("voussoirs") forming the perimeter of the arch can then be subsequently placed at "9:00" and "3:00", "10:00" and "2:00", etc., using the left and right hands as braces. The key insight for the Romans was that _these hands could be moved/rotated_ during construction, and the already placed lower voussoirs (angled stones) would be content to stay without any bracing, since the forces above and below them would hold them in place. Once both hands had come together at the very top ("12:00") of the arch, the keystone could be put in place, at which point the axle, the two hands, and the supporting wooden span could all be removed. Note that each of the clock "hands" would only have to hold one of the stones at a time -- _not_ the entire arch -- and thus transmit only the force of at most 2 of the stones to the pivoting axle. Thus, the amount of wood required is substantially less than that required for the "template" method, and the "clock" method could readily be used on multiple arches in an arcade. If necessary, the end of each of the left & right "hands" could have a small circular _wheel_, which would ease the task of moving each "hand" upwards, as the little wheel would rotate along the most recently place stone instead of rubbing along it. Note that approximately the same technique could also be used for concrete construction, so long as the concrete hardened & became strong at the same rate as the construction proceeded. It appeared that this was indeed the case for Roman concrete, which used far less water than modern concrete, and was tamped/hammered into place to make sure that there were no air pockets ("voids").