Ancient artillery can be divided into two primary groups: torsion and non-torsion. Torsion engines derive their power from the twisting of a spring mechanism, generally made from a combination of animal sinew and hair. Non-torsion engines refer to those that derive their power from bending a stave made of wood or bone, much like a bow. Ancient artillery can be further divided by the type of projectile launched; wooden or metal bolts, or stone spheres, like cannonballs. However, various torsion and non-torsion engines made use of these projectiles, and a few engines had the ability to fire both. According to the size of the projectile and the type of power used, artillery can come in many shapes and sizes.

The ancient evidence regarding the development of ancient artillery comes from three primary sources: ancient written texts, artistic depictions, and surviving artifacts. The ancient written texts come primarily in the form of technical treatises. The compilation of treatises in Greek and Roman Artillery, by E. W. Marsden, includes the works of four authors: Heron, Biton, Philon, and Vitruvius. These collected works trace the development of Greco-Roman artillery from the earliest oversized crossbows to the impressive siege engines produced in the Hellenistic period and in the Roman Empire. The bulk of what is known and conjectured about Roman artillery is based on the four treaties commented and translated by Marsden.

The four authors of these treaties and the sources they used span the period from the early third century B. C. to the late first century A. D. However, the technical interest in most of these writings is centered on the period 350-270 B. C., when constructional developments in artillery were relatively rapid, and had the greatest impact on the technical advancement of both torsion and non-torsion engines.

The ancient treatises are invaluable for our understanding of the technical and mechanical aspects of Roman artillery. However, because of the attention they pay to technical details, the treatises sometimes fail to tell us about quite basic matters. In particular, they do not tell us the raw materials used in the building of artillery. The treatises would be an excellent source for making schematic diagrams or reporting trajectory angle of a certain engine, but they are not so helpful in finding the type of wood most commonly used to build engine frames, or the method for making the sinew springs utilized in torsion engines..

The ancient depictions of artillery come in the form of reliefs from various parts of the Roman world. Trajan's column (ca. A.D. 110) is a particularly valuable source of ancient evidence. It contains five separate depictions of various types of Roman artillery. The scenes in which the engines are pictures give us clues about the context in which ancient artillery was put to use. In addition, the nature of the column itself allows for a rough dating of each of the depictions. Other ancient depictions include the balustrade relief at Pergamum (third century B.C.), the tombstone of Vedennius , and a relief called "The Cupid Gem." These reliefs can be tentatively dated to the middle of the fourth century B.C. because they depict the use of washers in the frames, which had not been used until around this time. These depictions confirm the use of artillery in the Roman world. Several of the engines depicted correspond closely with the technical writings of ancient authors, thus strengthening our belief in the validity of the literary evidence.

The surviving archaeological components of ancient artillery consist mainly of metal portions of the frames, as well as both metal and organic remains of bolts. Quite a few of the actual projectiles, small objects made of stone, have also been recovered. Various artifacts have been found throughout the Roman world, and, like much of the artistic evidence, these finds are entirely consistent with the engines described by ancient authors.

The artifacts also confirm both the size of the engines and the size of their projectiles. The various types of bolt heads point to some specific uses of ancient artillery not discussed in the treatises or depicted in the reliefs. The combination of these several types of ancient evidence allows us to construct an accurate time line for ancient artillery. We can also accurately describe the technicalities of construction and the use of various engines.


The first ancient pieces of artillery were non-torsion bolt throwers. These engines are called gastraphetes ("belly-bow") by ancient authors, although names in ancient texts are used somewhat informally (Marsden 270). The name is fitting, because the gastraphetes was often loaded by being braced across the midsection (Marsden 21,23). For the most part, these engines are simply bows mounted on a wooden frame. The treatise of Heron is the primary ancient literary authority on this early engine.

The bow used in the construction of this gastraphetes was too strong for a man to throw by hand. Therefore, the user had to brace the engine against himself and draw the string into successive grooves until it was all the way back. Then, a bolt was inserted into a semi-circular channel between the rows of grooves. A trigger mechanism was used to release the force of the bow (Marsden 21,23).


Later, advancements were made on this earliest engine, with a goal of shooting a larger missile with increased force. This made it necessary to make the arms of the bow more powerful. This was accomplished by constructing the arms on a larger scale. As the scale of the engine grew, the engines had to be fitted in a frame and eventually mounted on bases, as they became too large to be handled individually. A more powerful pull-back mechanism also accompanies the introduction of increasingly more powerful arms. Instead of the manual drawing and rest method used in the earliest gastraphetes, a type of winch was used. The bow string was attached to a slider, which in turn was attached to a cord. The cord was winched around a revolving axle at the back of the engine.

supposed pulley system for pull-back mechanism

But, even with this powerful winch system, the increasing size and strength of the engines made it necessary later to recourse to a pulley system (Marsden 25, 27). This made the pullback yet slower. These advances in power and strength are covered thoroughly in both Heron's and Biton's treatises. However, Biton also describes a more advances non-torsion engine, which hurled rocks. This stone-throwing engine is the most advanced non-torsion engine before the move to torsion power in the Roman world (Marsden 270). The advancements made in this engine are in the pull-back system and in the frame. The winch system utilized here incorporated a pulley system into the frame if the engine, and that made faster and easier cranking possible (Marsden 181). Also, in possibly the most significant advance, Biton stressed the use of metal in the frame of the engine. Although he recommends simply the reinforcement of the wooden sections with iron or bronze, this is the first engine where metal is incorporated into the frame. This is the precursor to stronger, more durable torsion engines whose frames are largely metal.

Although this stone-throwing engine is the most advanced non-torsion engine known, it does not necessarily mark the point in the evolution of ancient artillery where torsion power began. Many non-torsion engines continued in use after the introduction of torsion power. The change from non-torsion to torsion power was gradual, and the use and production of both types of engines overlap.

Torsion engines were powered by two sinew springs. Not much is known about the actual preparation of the sinews, or what animal tendons were both strong and springy enough to power a torsion engine. Most experts agree that the springs were a combination of animal sinew and human hair. It is supposed that the hair somehow held the sinews together, though the weaving process is unknown. As would be expected, most experts believe that the sinews were attained from the more conditioned parts of animals, such as the shoulders of oxen or the ankles of horses. Archaeologically, this lack of information is due mainly to the fact that sinew, an organic material, would have decayed long before modern historians could observe its condition. Another factor is that very little was written about use and preparation of animal tendons as torsion springs.

Two wooden arms connected by a sinew (forming a bow string) were inserted into two sinew springs, one arm in each spring. The center of the bow string was attached to a trigger mechanism. The string was pulled back by a winch mechanism similar to those described above for non-torsion engines. As the string was pulled back, the arms were drawn as well. This drawing of the arms resulted in the springs being placed under greater and greater tension. This tension provided the energy to shoot the bolt or rock (Southern and Dixon 152).

The technical advances of the torsion engine were all centered in the type of frame used to house the two springs. Originally, in what is called the Mark I frame, the springs are housed individually, and are wrapped vertically all the way around the frame. This type of frame, though probably stronger than most non-torsion engines, limited the movement of the arms, because the springs were wrapped around the entire frame (Marsden 49). A later improvement solved this problem. By building frames with holes in the top and bottom through which springs could be wound, the artillery designers allowed for more movement of the arms and, therefore, more power. In these weapons, the springs are housed within the frame, with the exception of the portions which extend through the holes of the top and bottom.

Mark I: springs are housed | Mark II: frames with holes

individually and wrapped vert- | in top and bottom where springs

ically all the way around frame | can be wound


In the interest of obtaining greater power from a more complete twisting of the frame, later advances made use of a lever and washer system within the frame of the springs. In this type of engine, each spring was wound around a lever, often made of iron, and the levers in turn were mounted on washers. The turning of the levers drew the springs taut. This advance is important because it addressed the ability of a spring to retain its resilience over time. As with most other things, as time went by, the strings in these torsion engines became stretched, which either seriously compromised the power of the engine or demanded a new spring. However, with the introduction of the lever and washer system, one could twist the springs independently from their frames. Consequently, if a spring became stretched and weak, a few turns of the lever would make it taut once again. This made the torsion engine more economical and increased the lifespan of the springs.

The frame of the springs in the torsion engine continued to be the area of emphasis and technical advances until the time of the development described by Vitruvius in his engine called the Scorpion.

Vitruvius' Scorpion

Here, some improvements are made on the arms of the engine. Vitruvius recommends the use of curved and tapered arms, set at an angle to each other, rather than in an arch (Marsden 197). However, he does not say whether these arms are more efficient. They must indeed have been better, because this type of arm is used widely after its introduction in the Scorpion. This is supported by later engines, such as Heron's cheiroballistra, and the Pergamum relief , both of which include curved and tapered arms. In addition, we can make various assumptions about the improved performance of artillery that had curved arms by using comparative evidence. Think of a bow: a recurved bow is stronger and more accurate than a bow made from a straight piece of wood.

classic curved bow recurved bow

The logical assumption would be that the curved and tapered arms of the Vitruvian engine increased the power derived from the bending of the arms.

Over time, various small modifications were made to the arms, springs, washers, and frames of Roman engines to maximize their power and increase their spring life. However, the frame of the engine remained largely the same in style. But, over time, more and more metal plating and parts were being incorporated. Bronze was used primarily for plating the frames, such as corner plates, while iron was used in the various working components. The use of metal in the frame of artillery was a necessary step. As the power of these engines grew, so did the stress put on the engine frame during drawing and shooting. Metal is more suited to handle the shock of firing. Also, metal is not subject to heat and humidity, which would destroy and distort a wooden frame. Naturally, the parts of the engine under the most stress were the first to be enforced by metal. The culmination of the use of metal to frame engines is represented by Heron's cheiroballistra.

Heron's cheiroballistra

In this particular engine, the springs are stretched in two separate metal casings. A metal stud was attached the top of each of the field frames, to hold them together. Another stud was attached to the bottom of the field frames and the base of the engine, to hold the spring casings in place (Marsden 209).

Heron's cheiroballistra represents the most advances two-armed torsion engine used by the Roman army. However, the Romans made use of a one-armed torsion engine as well. This was the largest weapon used by the Roman army.

It is appropriately named the onager, or "wild donkey", because of its strong kick (Webster 242). The onager was a torsion engine, despite having only one arm. The construction of the onager differed somewhat from that of the typical two-armed engine, but many of the same devices were used. A sinew spring system was used in the onager, as in other torsion engines. Because it has only one arm, the onager needs only one spring, but that one spring is somewhat larger than those of two-armed engines. The onager spring is positioned horizontally, as opposed to the vertical springs in earlier engines. The winch system is essentially the same (on a larger scale) as that on two-armed engines but here, the winch pulls the arm down instead of back. There are also differences in the contents of which onagers and two-armed engines were used.


The onager was specifically designed for siege warfare (Southern and Dixon 158). It could be used to bombard enemy walls and artillery, or to hurl large stones or missiles at oncoming forces. Conversely, most two-arm artillery were used in open battle, against specific targets, but the larger ballistas may well have been used during sieges.


The remains of a late fourth-century A.D. arrow-shooting engine was found in the Roman fort at Orsova, Romania. The find consisted of two iron rings joined by two beams and an iron arched rod, forked at either end. The findings are structurally parallel to Heron's description of parts of the cheiroballistra, and it is possible to identify the joined rings as the spring frames. The arched rod, as described in Heron, was required to hold the top of the spring frames in position. This find is very important, because it shows the metal frames stressed in Heron's technical description of the cheiroballistra.


Iron "kamerion" used to hold the top of field frames in position from Orsova, Romania measuring 1.45 m.

Field frames found in Orsova, Romania. They consisted of two iron rings joined by two beams measuring 36 cm.

Additional spring field frames were found in two corner towers of the fourth-century fort at Gornea, Romania. These frames are smaller than the frames recovered at Orsove. However, they are believed to represent a version of Heron's cheiroballistra and a smaller torsion engine, the manuballista, as described by an ancient military author, Vegetius. Apparently, this type of engine was similar to the gastraphetes of Heron, but with torsion, as opposed to non-torsion power (Southern and Dixon 155).

Metal fittings from the frame of a third century A.D. engine were discovered at Hatra. This engine was found at the foot of a tower upon which it had originally been placed. The find consisted of cast bronze corner plates, torsion counter plates and washers, five rollers, and a nailed sheeting, which appears to have been attached to the front of the engine. The sheeting had two semi-circular holes cut into it, and we can infer that this was a two-armed engine. When comparing these findings to ancient literature, it appears that this was possibly a medium caliber stone-throwing engine such as the one described by Biton (Southern and Dixon 157).

Remains of artillery bolts have been recovered throughout the Roman Empire, especially the pyramidic iron heads. A particularly valuable site is Dura-Europus, where numerous socketed heads were found, including one bolt which retained its wooden shaft.

Bolt head with wooden fletchings from Dura-Europos, Syria

The shaft was made of maple, and had wooden vanes which apparently formed the fletchings of the arrow. The vanes were not inserted in one side of the shaft to allow it to be placed in the groove of an engine. Also, no rock is evident at the end of the shaft, because this is unnecessary due to the fact that the string is held by a trigger.


Trajan's Column has five separate scenes which show Roman artillery. These reliefs depict one engine in many different contexts. Relief 1 depicts an engine mounted on a wagon. This apparently was common practice, according to the ancient author Vegetius, who tells us that engines might be carried in wagons pulled by mules, and the relief supports this. This type of engine is referred to as the carroballista, and was designed to provide a mobile engine employable in battle situations. It is possible to identify the engine in this relief as related to Heron's cheiroballistra, because of the cylindrical spring casings. Relief 2 also depicts as engine mounted on a wagon drawn by donkeys. Again, the spring cases identify the weapon pictured as the cheiroballistra. Both of the engines mounted on wagons appear to be operated by a single person. The three remaining reliefs depict artillery which appear to be a part of some sort of fortification. Judging from their spring casings, these engines also appear to be cheiroballistra. This suggests that it was fairly common to use this type of engine to repel assaults, as well as in open-field situations. The engines in reliefs 4 and 5 are manned by two soldiers. This suggests that these engines were larger and more powerful than those mounted in wagons.

Reliefs from Trajan's Column

Another relief depicting artillery is found on the tombstone of Vedennius. This relief pictures a two-armed torsion engine. We appear to have the view from directly in front of the engine. There is some sort of carving within the spring frame, but this is most likely an artistic touch. This relief is often cited, as it confirms that round washers were used in Roman artillery (Marsden 53).

Tombstone of Vedennius

This is also the case with the relief "The Cupid Gem", which features round washers. Again, Heron's cheiroballistra appears to be the device depicted. However, this engine utilized a slightly more complex winch system. Heron does not describe the ratchet and pawl system depicted here, but this type of winch is described in the writings of Ammianus on the onager (Marsden 233). These scattered remind us of the sporadic and irregular spread of technology in the ancient world.

The Cupid Gem

The ballustrade relief at Pergamum is a military collage. The engine pictured has square washers, as opposed to the round washers in other reliefs. Also, this engine's torsion springs are twisted, so that we clearly have a lever and washer frame system. In addition. the arms of the engine are curved and tapered. This corresponds to the engine described in the treatise of Vitruvius.

ballustrade relief at Pergamum


There are many ways to describe ancient Roman artillery. By using literary texts, ancient depictions, and modern-day comparison, there is much to be learned about how people in the ancient world both defended themselves and attacked enemies, and how the artillery used evolved over time.