BMCR 2007.05.32

Roman Siege Works

, Roman siege works. Stroud: Tempus, 2006. 160 pages, 8 unnumbered pages of plates : illustrations (some color), maps ; 25 cm. ISBN 0752428977 £18.99.

“The main objective of this book”, writes Davies in his introduction, “is to raise the awareness of the importance of field works to the success (or otherwise) of Roman siege operations” (8).1 It is difficult to see how such an objective could be accomplished without a general appreciation of Roman siege warfare. How else can we test D.’s contention that field works were important? Unfortunately, only the barest “summary of the general trajectory of developments” (133) is presented, almost as an afterthought, tucked away in the final conclusion. And worse, D. has selected as exemplars only those sieges during which field works were erected, silently passing over the others.

For him, the dilemma facing every Roman commander was “what manner of siege system should be employed” (37).2 But, very often, the Romans employed no “siege system” at all. For every Capua (212/211 BC where the town was surrounded by a double ditch-and-rampart system, there was a Leontini (214 BC where the besiegers simply forced the gates, or an Arpi (213 BC where the town was taken by clandestine escalade. For every Thala (108 BC encircled by a ditch and palisade, threatened by embankments, and battered by siege machinery, there was a Vaga (109 BC) or a Capsa (107 BC where the Romans built no “siege work elements” at all. The well-known blockade of Numantia (133 BC during which the site was ringed by a stone wall connecting a series of forts and camps, had been preceded by two failed attempts involving frontal assault (141/140 BC) and escalade (138 BC) but no “siege system.” The historical record, it seems, is at odds with D.’s basic premise, that structures were fundamental to siege warfare.

Be that as it may, D. has written a book about siege works, which he defines as “encompass(ing) all those structures and features constructed by an assailant for the purpose of undertaking operations (either directly or indirectly) against a defended centre” (7, author’s italics). He explicitly excludes siege machinery of every type, on the grounds that “their essentially mobile character and the fact that they were assembled rather than constructed, serve to distinguish them from the ‘fixed’ structural elements” (7). But such a narrow definition may be disputed on more than one count. Firstly, the ancients themselves drew no such distinction. The tasks of erecting an embankment, for example, and constructing a siege tower were both “works” ( opera or, in Greek, ἔργα), carried out by the same men under the direction of the same architecti or ἀρχιτέκτονες.3 Secondly, to restrict the concept of construction so that it cannot apply to siege machines is entirely arbitrary and unwarranted by the linguistic evidence.4 And thirdly, it makes little sense to limit the study of siege warfare to structures, while ignoring the machines which those structures were often designed to support. The reader is in no position to evaluate D’s ideas on the siting of artillery, for example (83-84), without a discussion of the form and function of ancient catapults.5

D. divides his book into eight chapters. He begins with an overview of “The Literary Sources” (ch. 1, 9-24), in which he introduces the reader to the main historians and technical authors. This is followed by “The siege in the context of Roman warfare” (ch. 2, 25-34), in which he discusses the circumstances under which a siege might develop. Here, the attempt to categorise sieges as either “planned” events or “reactive” events is unconvincing. D. exemplifies the “reactive” siege by arguing that the future emperor Tiberius “was compelled to besiege Dalmatian Andetrium (AD 9) to prevent Bato from using it as a secure base for guerrilla activity” (28), but similar reasoning dictated Scipio Aemilianus’ “planned” siege of Numantia (133 BC). There must always have been a degree of reaction, but this is not the same as being unprepared. When Caesar found the gates of Gomphi unexpectedly barred against him (48 BC his violent and sudden assault was preceded by thorough preparation in the form of camp-building and the rapid construction of ladders, shelters, and screens.

The rest of the book comprises separate chapters on a selection of D.’s siege work elements: “Preparatory works” (ch. 3, 35-44), “Blockade camps” (ch. 4, 45-62), “Circumvallation” (ch. 5, 63-95), “Assault ramps and siege mounds” (ch. 6, 97-116), “Mines” (ch. 7, 117-124), and “Miscellaneous engineering works” (ch. 8, 125-131). Under preparatory works, D. lists the securing of supply lines, the reconnaissance of the siege site, the erection of screening works, and ground preparation. However, his next element, the “blockade camp,” is less well conceived. He defines it as “a base from which the besieger might act to interdict supplies or reinforcements sent to a defended position or to prevent sorties or foraging efforts mounted from the same” (145). The archaeological examples from Nahal Hever certainly seem to fit this description, although, in the absence of dating evidence, their historical context is pure speculation. However, many of D.’s other examples are less apposite: Marcellus’s hiberna at Syracuse (213-212 BC Scipio’s camp at Cartagena (210 BC Caesar’s camps at Avaricum and Gergovia (52 BC); none of these was used to enforce a blockade. Of course, in siegecraft just as in regular campaigning, Roman armies routinely fortified a base camp, but it requires special pleading to justify the appellation of “blockade camp.”

D. believes that there was a “transition towards the adoption of circumvallation as the preferred vehicle of isolation” (46). The sieges of Agrigentum (262 BC) and Lilybaeum (250-241 BC) are surely the earliest definite examples of Roman circumvallation, although D. categorises them as “extended blockade camp systems” (46), claiming that, only later, did they undergo “conversion to a full system of circum/contravallation” (51).6 He highlights Capua (212/211 BC) as “the first real endorsement of the value of a well-organized circumvallatory scheme,” claiming that “there was a marked increase in the use of circumvallation” thereafter (64). But do the statistics really support this conclusion? There were two, perhaps three, circumvallations during the First Punic War (Agrigentum, Lilybaeum, and the possible example of Panormus, 254 BC and another two during the Second Punic War (Capua, and Scipio Asiaticus’ siege of Orongis, 207 BC). Thus, in the space of sixty years, Roman armies had utilised the tactic four or five times, as far as we know, whereas during the same period more than a dozen towns are known to have been taken by storm. Far from a “marked increase in the use of circumvallation” (64), the seventy-five years separating the sieges of Orongis and Numantia witnessed a strategy of investment only twice, at Ambracia (189 BC) and at Carthage (146 BC).7 Strangely, D. prefers to describe Scipio’s siege works on the isthmus at Carthage as a “quadrangular long fort” (47). However, it is clear that, in concept, this was simply a linear version of the Capuan circumvallation, designed to close off the isthmus with a double line of fortifications. D. states that “what might otherwise be thought of as a line of investment was given the character of a blockade camp” (53), but Scipio’s strategy was aggressive, with no blockading intent.8

D. cites Metellus’s ditch around Thala (108 BC) as proof that circumvallation “began to be viewed as a useful precursor to more direct approaches” (65). But this is apparently not the case. The next recorded instance comes fully a generation later, when Sulla used the tactic (passively, it should be noted) at Athens (87/6 BC) and Praeneste (82 BC which hardly indicates an accelerating trend, and it is not until the siege of Tigranocerta (69 BC) that we again recognise the aggressive use of circumvallation. D. jumps ahead to describe “the standard Caesarian siege approach, whereby assault preparations were put in hand immediately after the circumvallation had been completed” (65). Ironically, this analysis is not even accurate for those sieges that actually involved circumvallation, far less for Caesarian sieges in general. Only half of Caesar’s sieges involved a circumvallation at all, and most of these were played out as blockades, which rather contradicts the idea that “he regarded passive blockade as a waste of resources” (134). Not only is D.’s blueprint for the “standard Caesarian siege approach” flawed, but the theory that an assault habitually accompanied a circumvallation is mistaken; in fact, only Caesar’s sieges of Ategua (45 BC) and the town of the Atuatuci (57 BC) conform to this model.9

Addressing the subject of the siege embankment, D. believes that he can discern two different tactical functions, which warrant his subdivision into assault ramps and siege mounds. The former, he writes, are “raised to parallel the height of a defensive work enabling the passage of storming parties and the mounting of engines capable of effecting a breach” (145), while the latter are “raised to parallel or overtop the height of a defensive work allowing oversight of the defenders and the advantageous emplacement of artillery” (146). A fine distinction. Indeed, most of D.’s discussion is of “assault ramps.” However, he categorises Caesar’s embankment at Uxellodunum (51 BC) and “the two huge structures raised by Trebonius at Massilia (49 BC)” as “single function siege mounds” (99), and offers, as archaeological corroboration, the remains at Cremna. The case for such a subdivision is never fully argued, but D.’s reasoning appears to be the fact that the Cremna embankment “inclines gently upwards presenting an easy avenue of advance for the artillery (or the siege tower) that was positioned at its summit” (108). Of course, this is a perfect description of the “assault ramp,” which similarly requires heavy machinery to be manoeuvred along the apex. Indeed, the defenders of Cremna harboured no illusions about the approaching danger, as they laboured to buttress their walls against the expected battering attack.

Turning to mining, D. is obliged to bring in Greek and Persian examples to fill out a rather insubstantial chapter. Besides a lengthy description of the Persian tunnels at Dura Europos, he mentions Roman tunnelling only during Nobilior’s siege of Ambracia (189 BC Sulla’s siege of Piraeus (87/86 BC and Julian’s siege of Maiozamalcha (AD 363); at Avaricum, far from “seeking to bring about the subsidence of the enemy circuit” (118), the mining there was designed by the Gauls to destabilise Caesar’s embankment.10 Finally, D. employs a catch-all, miscellaneous category to encompass the brick tower erected by Trebonius at Massilia (49 BC the engineering yard identified by Adolf Schulten at Masada (here inexplicably termed a ” Baulager“), and the failed harbour mole at Lilybaeum (250 BC for which D. invents the “novel contrivance of artillery-delivered rubble as part of the infilling process” (126).11

In his conclusion, D. claims that “this concentration on the topic of siege works has allowed us to observe how Roman practice would appear to have varied over time” (133). But, as each chapter jumps from siege to siege in a desultory fashion, chronologically from Fidenae to Cyzicus, alphabetically from Agrigentum to Zama, the result is rather disjointed. Equally, the lack of an historical framework makes it difficult to identify any overall trends, and the emphasis on the small corpus of material remains unfortunately leads to a skewed picture of Roman siegecraft.12


1. I am unsure what D. means by his parenthetic “or otherwise.” Does he mean to imply that field works were important to the failure of Roman siege operations? And if so, how?

2. D. defines a “siege system” as “the general tactical deployment adopted by a besieger to achieve the reduction of a defended position comprising the totality of various disparate siege work elements” (146). The latter comprise “any structure or feature constructed by an assailant for the purpose of prosecuting operations (directly or indirectly) against a defended position” (146).

3. In his description of the siege of Ambracia (189 BC Livy illustrates the point well when he refers to Nobilior’s siege works as munimenta (“fortifications”) and his battering-rams as opera (“works”) (Livy 38.5.1); in Polybius’ version, the battering rams are likewise called ἔργα.

4. The example of Eretria (198 BC where “the surroundings offered timber in abundance for the construction of opera from scratch” (Livy 32.16.10), addresses D.’s contention that siege machines were not “constructed.” As these newly-built “works” subsequently demolished the walls, they are likely to have been battering rams and not some kind of siege structure.

5. For example, D. refers to “firing stations” for artillery, spaced along the circumvallation at Machaerus, but the platforms in question have a maximum depth of 2m, which is far too small for a standard arrow-shooting catapult.

6. D.’s use of the term circumvallation is confusing. No-one would dispute his initial definition, namely “any work of encirclement designed to ensure the complete investment of a target” (63). However, he then adds unnecessary complication by changing the definition “in those siege systems that deploy two distinct encircling lines” (145): “here, circumvallation may be taken to refer to the outward-facing barrier, whilst an inward-facing line should be distinguished as a contravallation” (63). This is a resurrection of the illogical scheme devised by Napoléon III to describe the remains at Alesia; cf. D.B. Campbell, Besieged. Siege Warfare in the Ancient World (Oxford, 2006), 192-195. For such double investments, the term “bicircumvallation,” coined by Peter Connolly, seems preferable: Greece and Rome at War (London, 1981), 292-293.

7. The opera which Livy reports circa Oreum (199 BC) may possibly indicate a circumvallation (Livy 31.46.14).

8. For the siege works at Carthage, see D.B. Campbell, Ancient Siege Warfare. Persians, Greeks, Carthaginians and Romans, 546-146 BC (Oxford, 2005), 40 plate G, with 63.

9. Details drawn from D.B. Campbell, Aspects of Roman Siegecraft, unpublished PhD thesis (University of Glasgow, 2002).

10. Caes., BGall. 7.22: ” [Galli] aggerem cuniculis subtrahebant. D. is perhaps confused by Caesar’s aperti cuniculi, which appear to be long sheltered corridors running along the Roman embankment; cf. Campbell, op. cit. (note 6), 132, following T. Rice Holmes, Caesar’s Conquest of Gaul (Oxford, 1911), 144.

11. There is no sign of this in Polyb. 1.47.4. Perhaps D. has been misled by Paton’s Loeb translation, which renders τὸ ῥιπτούμενον as “all they shot in.”

12. I noted the following errors. Diodorus Siculus comments on the origins of Roman siegecraft at 23.2.1, not 13.2.1 (8 and 137 n. 1); the Hellenistic period properly dates from the death of Alexander the Great, rather than “from the mid-fifth century BC” (8); for ” aide memoires” (9), read aides-mémoire; “Dio Cassius” is usually known as Cassius Dio (10, 13, et al.); “the compiler of the Scriptores Historiae Augustae” (13) should be the compiler of the Historia Augusta (since the scriptores are the fictional writers); Polyaenus (fl. AD 160) is by no stretch of the imagination a Hellenistic author and there is no reason to characterise the Romans as his “foreign enemies” (16); “Philip the son of Demetrius” (16) is usually known as Philip V of Macedon (son of Demetrius II), so the separate index entries should be combined (158); the “Amphictyonic investment (date?) of Cirrha” (16) (or Crisa, as it is called by Frontinus) may be placed within the bracket 595-585 BC, as it occurred during the First Sacred War; Philon of Byzantium belongs broadly to the later 3rd century BC, and it is only a theory that he was “writing in the 240s BC for the Ptolemaic army” (16); “Scipio Africanus the Younger” is usually known as Scipio Aemilianus (26); “M. Fulvius” (32, 35, et al.) is usually known as M. Fulvius Nobilior; “M’. Acilius Glabrio” (102) is elsewhere given the wrong praenomen (37) and his name is jumbled in the index (155); there was no “hermetic circumvallation” at Syracuse (52; also 64, 133); there is no reason to suppose that the siege towers at Lilybaeum “were earthfast rather than mobile structures” (52); “Pompeius Aulus” (59) is usually known as Q. Pompeius; it is debateable whether there are “practice works” at Woden Law (73); the captions to figs. 22 and 23 appear to have been transposed (75); Ammianus does not recommend “towers made of sun-dried brick or turf” for onagers (84); the skeleton at Dura Europos was found in Tunnel 1, not Tunnel 3 (121); note 15 to chapter 5 (presumably a reference to Cicero’s Ad familiares) is missing (140); the “Leuké” at Masada, consistently cited in its Greek form (81, 94, 101, 128), is absent from the index; and Labrousse 1966 (cited on 143 n. 17), Lammert 1932 (cited on 143 n. 3), and Shatzman 1989 (cited on 138 n. 5) are missing from the bibliography.