Retreats and Withdrawals in Republican Roman Warfare

Retreats and Withdrawals in Republican Roman

Warfare Daniel Morgan, BA (Hons) (Newcastle)

May 2020

This research was supported by an Australian Government Research Training

Program (RTP) Scholarship

A thesis submitted in fulfilment of the requirements for the degree of Doctor of Philosophy in Ancient

History and Classical Langugages

Retreats and Withdrawals in Republican Roman Warfare 2020

2

Statement of Originality

I hereby certify that the work embodied in the thesis is my own work, conducted under

normal supervision. The thesis contains no material which has been accepted, or is being

examined, for the award of any other degree or diploma in any university or other tertiary

institution and, to the best of my knowledge and belief, contains no material previously

published or written by another person, except where due reference has been made. I give

consent to the final version of my thesis being made available worldwide when deposited in

the University’s Digital Repository, subject to the provisions of the Copyright Act 1968 and

any approved embargo.

Signature: D. Morgan Date: 5 May 2020

Daniel Morgan

University of Newcastle


3

Acknowledgements

I would first and foremost like to thank my primary supervisor, Dr. Jane Bellemore, for her

direction and assistance during all stages of my studies, and especially for her assistance with

this thesis. I wish I had a sestertius for each gratuitous or ill-considered question I bombarded

her with. Even so, I received a considered and useful response to each and every one. She has

given me a great deal of feedback on numerous drafts of this thesis, and I would not have

been able to do it without her guidance. I also wish to extend my sincerest thanks to Dr.

Elizabeth Baynham, who has been my co-supervisor for the latter stages of my work. She has

also given me valuable feedback that has significantly improved this thesis. I further wish to

extend my thanks to the many other History and Ancient History staff that I have had fruitful

discussions with over the years. Finally, I wish to express my gratitude towards the

University of Newcastle itself, where I have been studying since 2011. The University has

been very supportive throughout my studies and I owe this instititution a significant debt of

gratitude.

Naturally, any remaining errors in this work are entirely my own responsibility.

Sincerely,

Daniel Morgan


4

Table of Contents

Abstract.......................................................................................................................................

Introduction: How Battles and Campaigns Ended in Roman Republican Warfare....................

Preamble............................................................... ........................................................

Recent Developments in the Study of Roman Battles and Tactics...............................

Modern Views on How Battles and Campaigns Ended................................................

What are Retrograde Manoeuvres, Retreats and Withdrawals? .................................

How Often Retreats and Withdrawals Occurred.................................. ......................

Overview of Key Sources.................................. ..........................................................

Terminology and Complications in Identifying and Distinguishing Types of

Retrograde Manoeuvres.................................. ................................... .........................

Purpose and Scope of the Thesis.................................. ..............................................

Thesis Structure............................. .................................. ...........................................

Method.................................. ............................. ........................................................

Translations............................ .................................. ..................................................

Conclusion............................. .................................. ...................................................

Chapter One: Retreat and Withdrawal: Why and Where To? ..................................................

Introduction............................ .................................. ..................................................

Part A: Why Armies Retreated.................................... ................................................

Retreats Due to Defeat in Battle.......................................................................

Retirements from Indecisive Battles................................................................

Disengagements at Nightfall.................................. ........................................

Disengagements Due to Poor Weather.............................................................

Feigned Retreats.................................. .................................. .........................

Summary of Part A.................................. ........................................................

Part B: Where Armies Sought Refuge.................................. .......................................

High Ground.................................. ..................................................................

Camps.................................. .................................. .........................................

Walled Settlements.................................. ........................................................

Summary of Part B.................................. ........................................................

Part C: Retrograde Movements over Longer Distances.................................... ..........

Summary of Part C.................................. ........................................................

7

8

9

11

14

18

19

20

30

38

39

41

43

43

44

45

46

46

49

51

55

56

60

60

60

62

72

74

75

79


5

Conclusion................................................................ ...................................................

Chapter Two: The Impact of Pursuit on Retreating or Withdrawing Forces...........................

Introduction............................ .................................. ..............................................

Types of Pursuits.................................. .................................. ..........................

How Pursuit Could Produce Disorder.................................. ........................................

Considerations Against Pursuit.................................. ............................................

Conclusion.................................. .................................. .................................. ...........

Chapter Three: Roman Retreat Tactics, Part I: Ratio and Combined Arms.............................

Introduction............................ .................................. ...................................................

Factors of Success in Retreats and Withdrawals...........................................................

Physical Order in Retrograde Movements.................................. ..............................

Legionary Training, Tactics and Reserve Lines..........................................................

The Roles of Subordinate Officers.................................. ...........................................

Equites, Velites et Auxilia.................................. .................................. ......................

Conclusion.................................. .................................. ...........................................

Chapter Four: Roman Retreat Tactics, Part II: Virtus, Consilium and Scientia Rei

Militaris..................................................................................................................................,

Introduction.................................. ..................................

The Roman Imperator.............................. ..................................

The General’s Tactical Duties..................................

Generalship in Retreats from the Battleline and Into a Fortified Position

Generalship During Retirement From a Fortified Position and On the March

Conclusion..................................

Chapter Five: The Development of Roman Retreat Tactics from the Middle to Late Republic

(218-46 BC)

Introduction............................ ..................................

Retreats in the Punic and Macedonian Wars (218-168 BC)

Retreat Tactics from 149 to 102 BC..................................

Retreat Tactics in the Late Republic............................

Conclusion

Chapter Six: Disastrous Withdrawals in the Late Republic and Early Empire.......................

Introduction.................................. ..................................

The Battle of Carrhae, 53 BC

80

82

83

84

89

93

97

99

100

101

104

109

117

122

126

128

129

130

133

136

142

148

150

151

153

161

169

182

187

188

189


6

Marcus Antonius’ Parthian Campaign, 36 BC...............................................................

The Clades Variana, AD 9.................................. .....................................................

Conclusion.................................. .................................. .......................... ................

Conclusion: Retreats and Withdrawals in Roman Warfare.................................................

Summary of the Thesis.................................. ............................. .......................

The Relationship of Retreats and Withdrawals to Roman Tactics and Strategy....

Retreats and Empire.................................. ............................ .................................

Suggestions for Further Study....................................................................................

Appendix A: Table of Basic References to Retreats and Withdrawals...............................

Appendix B: Table of Retreats and Withdrawals According to Tactical Context................

Appendix C: Table of Routs and Flights from Battle..........................................................

Appendix D: Detailed References to Retreats and Withdrawals by Chronology, Location

and Result

Reference List.........................................................................................................................

198

209

220

223

224

229

231

234

236

237

239

240

247


7

Abstract

Roman battles and campaigns have been studied for centuries, but one aspect of the subject

that has received less attention than others is the fate of defeated armies. Accounts of major

disasters in which enormous numbers of men died are, of course, well known, and these have

received the most attention from modern scholars examining the outcome of defeat in ancient

battles. However, there is significant evidence in the ancient sources that many armies which

were defeated in battle were not destroyed, but were able to escape their enemies relatively

intact. Such orderly retreats and withdrawals, and the role they played in ancient warfare,

have yet to be examined in detail.

This thesis will examine retreats and withdrawals in various battles and campaigns, and

analyse how they affected the tactics and strategies of Roman armies in the Middle to Late

Republic. Subjects for examination will include the prevalence of these manoeuvres, the

reasons they were carried out, the manner in which they were conducted, and their impact on

Roman military success and failure. It will be demonstrated that retreats and withdrawals

occurred commonly enough to have had a major impact on the outcomes of battles and

campaigns, and that the conduct of these manoeuvres, as well as planning for them,

significantly shaped Roman tactics and strategies. The Romans deliberately trained and

prepared for retreats and withdrawals, and they developed the sophistication and complexity

of their tactics for these manoeuvres until they could carry them out with unparalleled

organisation and skill. By the first century BC, this was a significant factor in their military

success. It will thus be shown that even a heavily defeated army was not necessarily doomed

to slaughter, and that a major aspect of Roman tactics and strategy has been overlooked in

modern studies.


8

The principles of strategy and tactics, and the logistics of war are really absurdly simple: it is

the actualities that make war so complicated and so difficult, and are usually neglected by

historians.

(Field Marshall Lord Wavell)

Introduction

How Battles and Campaigns Ended in Roman Republican

Warfare


9

Preamble

On 24 June 1812, l'Empereur des Français Napoleon I invaded the Russian Empire with an

army of 685,000 men. Napoleon preferred to avoid a long and costly campaign and sought to

destroy the Russian army in a decisive battle, and he aggressively manoeuvred the separate

divisions of his enormous army to create such an opportunity.1 He fought a major and

successful action at Smolensk in August, but the Russians successfully retreated, denying

l'empereur a decisive victory.2 As the French continued their advance, the Russians continued

to withdraw, extending Napoleon’s campaign into the winter months.3 On September 7th,

Napoleon brought the Russians to battle at Borodino. The French were victorious in the

exceptionally bloody engagement that followed, but the Russians successfully retreated

again, continuing to avoid decisive defeat.4 There was some tension in the Russian leadership

over this strategy of withdrawal, but events would soon vindicate its exponents.5

Napoleon reached Moscow, but the Russians did not offer peace as he expected, and he

hesitated.6 On October 19th, with the onset of winter and no apparent prospect of decisive

victory, the French began their long withdrawal from Russia.7 In the following weeks, lack of

supplies, harsh weather, and incessant Russian attacks caused the collapse of Napoleon’s

army. The French remnants reached safety in late November, but of the gargantuan forces that

invaded the Russian Empire in June, less than 20,000 would ever fight for Napoleon again.8

Napoleon’s first Grande Armée, the most experienced he would ever command, had been

destroyed.9

1 Lieven, 2016, 131, 173, 215. 2 Ibid., 166-70. 3 Ibid., 172. 4 Ibid., 209-14. 5 Ibid., 133-6, 209-11, 525-6. 6 Ibid., 251-2. 7 Ibid., 256-7. 8 Ibid., 282. 9 Ibid., 3, 527.


10

Napoleon hereafter commanded less experienced troops, and his power was broken the next

year in the Battle of Leipzig.10 Although his ultimate defeat was a close-run thing, the

successful withdrawals carried out by the Russian army in 1812, and the impact of

Napoleon’s own catastrophic retreat, were significant historical events.11 Perhaps retreats and

withdrawals are less exciting than glorious cavalry charges, gruelling sieges and major

conquests, but the Napoleonic campaign of 1812 demonstrates that in the right circumstances,

such manoeuvres could be of decisive military, and historical, importance.

Differences in scale notwithstanding, if one tries to think of a comparable story from the

ancient world, few examples may come to mind. Perhaps one would be the famous March of

the Ten Thousand. This event, in which Xenophon led a contingent of mercenaries from the

heart of the Persian Empire in 401 BC, to safety on the western shores on Anatolia in 399 BC,

was recorded in detail in Xenophon’s Anabasis. The achievements of Xenophon and his

soldiers, and their memorable cry ‘θάλαττα! θάλαττα!’ (The Sea! The Sea!) are rightly

famous (Xen. Anab. 4.7.24). Even so, the long–term political consequences of this event were

not momentous. Did retreats and withdrawals ever assume significant importance in ancient

warfare and history? If ever they did, one would expect to find evidence of such in the

extensive and well–known military history of Rome.

Consider the Battle of Carrhae, 53 BC. In this engagement, a Parthian army using novel

tactics crushed seven legions led by Marcus Licinius Crassus, who was one of Rome’s great

men at the time (cf. Plut. Crass. 23-27.4). Following the battle, the Roman army attempted to

withdraw to Armenia, but disintegrated en-route (Crass. 27.5-31.7). Crassus himself died in

this chaos (31.5) and overall, 30,000 men were lost (31.7). These events brought about a

lapse of Roman dominion in the east and inititated decades of violent competition between

Rome and Parthia for control of the region.12 It also upset the fragile balance of Roman

politics, indirectly leading to the final round of civil wars that destroyed the Republic.13

10 Ibid. 527. 11 Ibid., 527. 12 Sampson, 2015, 169-72. 13 Goldsworthy, Caesar, 379, 451


11

As Sampson has pointed out, it was in the failed withdrawal from the stricken battlefield at

Carrhae that defeat turned to catastrophe.14 The failed withdrawal at Carrhae thus assumes

major significance. Of course, Carrhae is well known as a major Roman defeat, yet it is the

battle itself that one thinks of first, and not the disastrous withdrawal that followed. This is

only one example of how the role of retreats and withdrawals has been overlooked in

Republican Roman warfare. There are many others. The following thesis will seek to correct

this imbalance.

Recent Developments in the Study of Roman Battles and Tactics

The analysis of ancient battles and campaigns is an old scholarly field, but although the

mechanics of ancient combat are well studied, one can ask whether all aspects of this subject

been examined with equal weight. Traditionally, there has been a great deal of focus on

Roman tactics by orthodox, military historians such as Delbrück who pioneered a method

known as Sachkritik to analyse battles.15 In this method, the known capabilities and practices

of modern armies were applied to counterbalance the exaggerations of ancient sources.

Comparative analysis is useful in studying battles and campaigns, but not all of its exponents

have utilised the same finesse as Delbrück, and this approach fell out of favour.16

Over the past few decades and partly in response to the perceived stagnation of the traditional

approach to battle studies, there has been a great deal of new work on Roman battles and

campaigns, particularly as a result of the introduction of John Keegan's (1976) ‘face–of–

battle’ approach to battle analysis into the field of ancient warfare.17 This approach utilises

knowledge of ancient weapons and armour and modern human psychology to examine the

experience of participants in battle, and thereby draw out details that are not explicitly stated

in the sources.18 This approach has since been applied by numerous experts to ancient history

and the study of Roman warfare.19 One of the first ‘face–of–battle’ analyses in the field of

Roman military history was MacMullen’s ‘The Legion as a Society’, published in Historia in

1984, which utilised modern studies of combatant psychology to draw inferences about the

14 Sampson, 2015, 147. Sampson (17) states that the withdrawal was a greater disaster than the battle itself. 15 Wheeler, 2001, 171. Cf. Delbrück, 1920. 16 Wheeler, 2001, 171. 17 See Gilliver, 2007, 130–6; Cagniart, 2011, 92–3. 18 Kagan, 2006, 15, 17–8 19 Eg. Goldsworthy, 1996; Sabin, 2000, 2007; McCall, 2001; Daly, 2002; Cagniart, 2011; Anders, 2015, etc.


12

motivations of Roman soldiers in battle, arguing in particular that combatants were motivated

mainly by concern for their close comrades, as well as by the fear of the enemy, just as

argued by Keegan.20 However it was not until 1996 that Keegan’s ideas were fully introduced

to the study of Roman warfare by authors such as Lee and Goldsworthy.21 Lee examined the

motivations of Roman soldiers in combat, and Goldsworthy’s significant work, The Roman

Army at War: 100 BC – AD200, combined a traditional, tactics–based analysis of Roman

warfare with Keegan’s ideas on combat motivation and psychology. The resulting shift in

emphasis away from a fixation on large–scale manoeuvres and tactics towards the

psychological and smaller scale aspects of battle has generally been a positive development.22

Several important further works have been produced from this new school of Roman military

study, including Sabin’s profound article ‘The Face of Roman Battle’ in 2000, McCall’s

excellent analysis of Roman and Italian cavalry in 2001, Daly’s detailed study of Cannae in

2002, and Levithan’s brilliant study of siege warfare in 2013.23 In particular, Sabin’s article in

2000 gave a new perspective on Roman combat, providing a mechanical explanation for the

length of most Roman battles by pointing out that men needed to rest regularly, both

physically and psychologically, between short bouts of intense fighting.24

However, there has been just as much excellent work on Roman battles and warfare that has

not relied upon the ‘face–of–battle’ approach. One example is Lendon’s thought–provoking

analysis of Roman infantry tactics and psychology in Soldiers and Ghosts: A History of

Battle in Classical Antiquity.25 Another is Phang’s 2008 book, Roman Military Service:

Ideologies of Discipline in the Late Republic and Early Principate, which analyses and

explains disciplina militaris that was characteristic of the Roman army. In terms of battle

mechanics, Zhmodikov published an article in 2000 that, along with Goldsworthy’s The

Roman Army at War, helped to inform Sabin’s groundbreaking article of the same year.26

Zhmodikov convincingly argued that Roman legionaries spent as much or more time

engaging in missile combat than they did in hand–to–hand fighting. He thus provided a

20 MacMullen, 1984. This article relies upon Marshall (1947) and Shils & Janowitz (1948) to make his psychological arguments. As will be touched on below, these are unstable foundations. 21 Cf. Goldsworthy, 1996; Lee, 1996. 22 Wheeler, 2007, 187. 23 Cf. Sabin, 2000; McCall, 2001; Daly, 2002; Levithan, 2013. 24 Sabin, 2000, 17. 25 Lendon, 2005, 312-5. 26 Cf. Sabin, 2000, 11–2.


13

purely tactical and source–based explanation for the long duration and lulls of Roman battles,

without reliance upon the psychologically–oriented ‘face–of–battle’ approach. Slavik built

again on this idea in 2018, by elaborating that it was not legionaries who performed most of

the missile combat, but light infantry such as the velites. This also offered a new explanation

for the poorly–understood Roman line–relief system that historians have been debating since

the field’s inception.27

There has also been direct criticism of the ‘face–of–battle’ approach. Everett Wheeler, for

instance, published a convincing article in 2001 arguing against the foundations of the

approach. Wheeler rightly points out that the World War Two–era studies of Marshall and

Shils & Janowitz are scientifically unsound and cannot be used as a basis for understanding

the psychology of Roman soldiers in battle.28 Wheeler finds Marcus Junkelmann’s alternative

reconstruction of early Imperial Roman battle mechanics to be more compelling than the

better known model of Goldsworthy.29 Also, Kimberly Kaga has argued convincingly against

relying entirely on a Keegan–esque soldier’s point of view in analysing battles. She advocates

a more holistic approach combining the combatant’s view with that of the traditional

commander’s point of view, taking into account not only the importance of the combatant’s

fear of enemy weapons, but also tactics and manoeuvres, in shaping the motivations of

combatants and courses of battles.30 An emphasis on the individual experience has perhaps

overlooked the nuances of group behaviour in battle, or what is referred to by Wheeler as the

‘automaton’ approach.31 As such, there is still ample space for the study of tactics and

manoeuvre in shaping the outcomes of battles and campaigns.

27 Slavik, 2018, 168. Direct evidence from the sources regarding line–relief is limited (Sabin, 2007, 431). A recent re–assessment of line–relief has been carried out by Taylor (2014, 320–2). There is extensive debate over how this system worked (eg. Junkelmann, 1986, 242–3; Daly, 2002, 60–3; Lendon, 2005, 181; Cagniart, 2007, 92–3; Sage, 2008, 69–76; Koon, 2010, 85–6; Taylor, 2014, 320). 28 Wheeler, 2001, 173. On Marshall, cf. Spiller (1988). On Shils & Janowitz, cf. Bartov (1992) and Rush (2001) which provide alternative explanations for the sources of small group morale in World War Two. 29 Wheeler, 2001, 171. Cf. Junkelmann, 1983, 236–46; Goldsworthy, 1996, 12–38, 116–247. 30 Kagan, 2006, 19–22, 107–8, 189–91, 200. 31 Wheeler, 2001, 170, 174.


14

Modern Views on How Battles Ended

While the application of the ‘face–of–battle’ approach to the study of ancient warfare has

brought a new emphasis on the experience of soldiers and small units in battle, most battle

analyses still place greater emphasis on the course of the battle than on what happened to

combatants once the battle was over.32 In Greco–Roman culture, control of the battlefield was

considered the mark of a victory.33 This is well–recognised, but what of manner in which the

enemy removed themselves, or were removed from the battlefield?

The approach of ‘face–of–battle’ analyses is to divide battle into distinct ‘phases’, and the

terminal phase of battle under this schema is the rout, in which the defeated army's line is

broken and many or all of the defeated soldiers turn and flee, often suffering heavy casualties

in the process.34 A prominent example of this emphasis on the rout as the usual end to an

battle is found in The Roman Army At War, where Goldsworthy briefly mentions the

possibility that battles could end indecisively (ie. with retreats), but assumes that most battles

ended with a catastrophic breakdown by one side.35 Thus, while the ‘face–of–battle’ approach

has brought some new attention to the matter of how battles ended, the subject has not been

examined in extended detail but only as part of broader analyses on how battles worked.

When the subject has been dealt with, it has been assumed that routs are the usual way in

32 Eg. Goldsworthy, 1996, 222–7; Sabin, 2000, 11–13, 17, 2007, 413–6; McCall, 2001, 16–9, 53–79; Daly, 2002, 196–200, passim; Cagniart, 2011, 92–3; Anders, 2015, 293–298. The vast majority of pages in these analyses are dedicated to the battle itself, with much less space dedicated to how battles ended. This trend is not necessarily specific to 'face–of–battle' studies (eg. Hoyos, 2011, 71–2). As another example, the promisingly named How Fighting Ends: A History of Surrender (2012), with a chapter on Ancient Rome by Loretana de Libero (29–38), does not examine retreats or routs, but focuses specifically on surrender. 33 Control of the battlefield, which for the Greeks would include erecting a trophy, and for the Romans would at least involve a count of the bodies, was also a literary convention that usually marked the end of a battle account (Lendon, 2017, 155). 34 Division of battle into phases: Keegan pioneered this approach, and it has been utilised in slightly different forms by numerous Roman military scholars (Goldsworthy, 1996, 171–247; McCall, 2001, 53–77; Daly, 2002, 166–202; Anders, 2015, 268–291; etc). Too schematic: Kagan, 2006, 22, 198. In the introduction to The Roman Army at War Goldsworthy (1996, 5) outlined the flaws of attempting to impose a ‘neat order’ onto the ‘chaos’ of battle, as an argument for using the ‘face–of–battle’ approach as opposed to traditional tactics–focused accounts, and yet at times 'face–of–battle' proponents can be just as guilty of attempting to impose order onto chaos (also see Cagniart, 2011, 92). Thus, while this method has been criticised as too schematic, it has stimulated some interest in the matter of what happened to soldiers when their army was defeated. Face–of–battle studies emphasising the rout: Goldsworthy, 1996, 222–7, 2011, 184–5; Daly, 2002, 198–201; Sabin, 2007, 413–6; Cagniart, 2011, 92–3. 35 Indecisive victories: Goldsworthy, 1996, 224, 245. Routs: Goldsworthy, 1996, 205–6, 222. Goldsworthy's newer works continue to emphasise the rout as the main way in which Roman–era battles ended (2011, 184–5).


15

which battles ended.

A brief overview of this is given by Philip Sabin: 36

The usual outcome of major land battles in [the Republic] was that one side suffered a

clear (and often overwhelming) defeat. The first priority of troops in the losing army

was to escape their vulnerability to one–sided slaughter ... Even if they had a line of

retreat, infantry in particular would be desperately exposed to pursuing cavalry in open

terrain. The only way for a broken army to escape such a catastrophe was to seek

shelter within a nearby fortified camp ...

Sabin says no more of retreats, but continues to examine the disproportionate casualty

numbers of victorious and vanquished armies in the sources, making the following

assessment:37

... the overall picture that emerges from the sources is one in which the victors usually

get off far more lightly than the vanquished ...

The impression is given that (organised) retreats, in which the defeated army did not suffer

heavy casualties, were unusual and when they did occur, they were only made towards a

fortified camp. This assessment by Sabin is not unusual.38 Gilliver states in a similar fashion

that the ultimate purpose of facing an enemy in a line of battle was to force them to turn and

flee, and then continues:39

... this was when the majority of casualties occurred ... For Roman troops on the

defensive this should have been the point at which, if they were able, they made for

their camp or a nearby defended city, or retreated to high ground in a close formation.

36 Sabin, 2007, 413–4. 37 Ibid., 416. 38 See Van Wees, 2004, 191; Southern, 2006, 196–7; Sabin, 2007, 430–2; Wheeler, 2007, 211–2; Cagniart, 2010, 92–3; Hoyos, 2011, 72–3; Anders, 2015, 295. 39 Gilliver, 2007, 138.


16

Elsewhere, Catherine Gilliver has given more weight to the possibility of retreat as a way in

which battle could end:40

It was when the ranks had been broken up that retreat and flight were most likely ...

Ordered retreat was possible for disciplined troops, and they might be pushed into

renewing the fight, even successfully, by strong leadership. But retreat could swiftly

turn into flight, and panic and wholesale flight rather than withdrawal in formation

were more likely for both Roman and non–Roman troops.

As can be seen, however, Gilliver emphasises that while ‘ordered retreat’ was possible in the

event of defeat, ‘panic and wholesale flight ... were more likely ...’41 Other scholars ,

including Sabin and Goldsworthy, have also inferred that significant ‘retrograde manoeuvres’

could be made by ancient armies, but this is generally depicted as part of the normal ebb–

and–flow of battle.42 Gilliver goes further by pointing out that an orderly retreat was one way

(other than a rout) that a battle could be brought to an end. She points out that Roman troops

‘on the defensive’ could retreat ‘to their camp or a nearby defended city’ or ‘to high ground in

a close formation’.43 This idea occasionally appears in the work of other scholars. For

example, in Daly’s analysis of the Battle of Cannae, he mentions that the Carthaginian centre

conducted a ‘controlled withdrawal’ under Roman pressure, and Anglim et al. mention that a

Roman commander could withdraw from battle, if possible, but these manoeuvres have not

been explained in detail.44 Some scholars of other periods and places have also touched upon

the possibility of retreat from battle.45 Modern accounts of Roman battles occasionally

mention fighting retreats but they are never examined in any depth.46 Even Gilliver states that

40 Ibid., 134. 41 Ibid., 134. Gilliver provides the following references for the assertion that flight was likely than withdrawal: Caes. BCiv. 3.94; Tac. Ann. 2.17, Agr. 37; Dio 75.7. 42 Sabin, 2007, 412. Sabin explains that infantry battle–lines could withdraw significant distances without being decisively defeated, but he does not examine the prevalence of these manoeuvres nor their central role in many battles. Goldsworthy (1996, 224, 226–7) points out that the losing side in close combat could retreat and armies could mutually separate, but nonetheless emphasises the rout as the normal manner in which a battle ended 43 Gilliver, 2007, 138, cf. BAfr. 85, B Hisp. 31, BAlex. 40. 44 Daly, 2002, 190–1; Anglim et al., 2013, 52. 45 For example, the disastrous fighting withdrawal conducted by the Byzantine army at Manzikert in AD 1071 examined by Haldon (2008, 178–80). Hanson (1989, 35–6) has also mentioned the possibility of a ‘fighting withdrawal’ in hoplite battle. 46 Fighting withdrawals mentioned: Hammond, 1984, 46; Goldsworthy, Fall of Carthage, 210, 283; Sampson, 2015, 147. Not mentioned: Goldsworthy, Fall of Carthage, 243, contra Livy 30.18; Goldsworthy, 2011, 183, contra App. B Civ. 2.80; Keppie's brief sketch (1984, 43) of the Battle of Pydna overlooks the Roman


17

retreats from battles occurred only rarely, and that chaos and mass flight were the usual

products of an army’s battlefield defeat.

Analysing warfare is, of course, not only the study of battles. An examination of the

manoeuvrings during Caesar's Ilerdan campaign of 49 BC (Caes. BCiv. 1.61–84)

demonstrates that a campaign could be won decisively through manoeuvre instead of

fighting, particularly by cutting off supply lines.47 As will be argued, both tactical retreats and

strategic withdrawals were a major part of this Ilerdan campaign.48 As such, focusing on

battles while ignoring the role of grand manoeuvre and supply lines cannot sufficiently

explain the outcome of engagements between armies. Withdrawals outside of battle, in

addition to retreats from battle, must thus also be considered to gain a full picture of the

reasons for the success of an army in Roman warfare.

Given this state of affairs, there are numerous questions about retreats and withdrawals in

ancient warfare that need to be asked. Is it true that a vanquished army would almost always

suffer disproportionately in the event of defeat? If not, how was such a disastrous scenario

avoided? When armies did retreat, did they only do so towards a fortified camp, as Sabin and

others have described? How were these manoeuvres performed, and how did they shape the

outcomes of battles and campaigns? To answer these questions, and provide a more balanced

view of how battles (and campaigns) ended in Republican Roman warfare, it is necessary to

carry out a dedicated study of Roman retreats and withdrawals. A detailed study of this topic

has not been done before, and so the opportunity exists to gain new insights into Roman

tactics and warfare by doing so.

withdrawal completely. 47 Roth, 1999, 333. 48 See Ch. 5, 168–72.


18

What are Retrograde Manoeuvres, Retreats and Withdrawals?

In modern (United States) military parlance, a retreat is a type of ‘retrograde operation’,

broadly defined as ‘an organised movement to the rear of or away from the enemy’ which

‘can be forced by enemy action or executed voluntarily’.49 A retrograde operation ‘can be

used to avoid decisive combat under unfavourable conditions, or to save forces for decisive

action elsewhere’, and the main reason for conducting such an operation is ‘to improve a

tactical situation or to prevent a worse one from occurring.’50 For reasons of style, throughout

this thesis I will mainly use the terms ‘retrograde movement’ or ‘retrograde manoeuvre’,

rather than ‘retrograde operation’. Either of the former terms will utilise the above definition,

and thus refer to a movement away from the enemy, for some tactical or strategic purpose.

These two terms will be used interchangeably and as umbrella terms for retreats and

withdrawals as a whole.

It is also necessary to distinguish between retreats and withdrawals, because, as will become

clear, these two types of retrograde manoeuvres often occurred under differing circumstances

and had somewhat distinct characteristics. Retreats will be loosely defined as retrograde

movements in which an army moved a short distance directly away from an enemy army,

usually as an immediate defensive reaction to tactical defeat or the threat of such (eg. the

Romans retreated from the battle). Withdrawals were retrograde movements over a longer

distance, often taking the form of a march from one defensible position to another, or to

another area of operations entirely (e.g. Hannibal withdrew from Latium). The distinctions

between the two will become clearer throughout the thesis, but the reader must also maintain

a certain flexibility of mind in distinguishing between retreats and withdrawals, for they also

shared similar characteristics and are not always easily distinguished.

I will occasionally use other terms for the purposes of variety. I will at times use ‘retire’ as a

synonym to describe either a deliberate retreat or withdrawal, and ‘disengage’ as a general

synonym for a retreat from battle. On a case–by–case basis, I will refer to some disorderly

retrograde movements simply as flights or routs. Most importantly, I will utilise the phrase

49 Field Manual 7–20, The Infantry Battalion, Chapter 5: Retrograde Operations. Retrieved 5 Mar 17 from http://www.globalsecurity.org/military/library/policy/army/fm/7–20/Ch5.htm. Note that in this particular document, the withdrawal is referred to as a ‘retrograde operation’, but I will utilise the term ‘withdrawal’. 50 Ibid.


19

‘retreat tactics’ throughout the thesis, particularly in later chapters, as a general term to

describe the methods by which armies carried out successful retreats and withdrawals.51

How Often Retreats and Withdrawals Occurred

Of the questions posed above, perhaps the easiest to answer is, how often did retreats and

withdrawals occur? The question is an important one, as it determines the relative importance

of retrograde manoeuvres in studying Roman battles. It is also relatively straightforward to

answer in a broad sense, by identifying and counting the examples of retrograde manoeuvres

found in the ancient sources.

As both Sabin and Gilliver have said, the usual result of a battle is understood to have been a

clear defeat for one side, followed by a panicked rout.52 Comparing the quantity of source

references of routs to retrograde movements does not necessarily provide evidence to dispute

this view. For instance, there are approximately 60 examples of Roman retrograde

movements that will be examined or analysed throughout this thesis, and these are by no

means all of the examples to be found.53 Depending on how one defines a retreat or

withdrawal, more examples could no doubt be found.54 However, for the purposes of a rough

comparison, the terms fuga and receptus respectively appear 617 and 24 times in Livy, and

113 and 14 times in the Caesarian corpus.55 Evidently, there are many more references to

routs than to orderly retrograde movements.

There are, however, literary factors that likely shaped this disproportionate number of

references to routs and flights following battles. First, it is not unlikely that ancient writers

preferred to describe battles or skirmishes that ended with routs rather than retreats, because

the latter were more decisive, more exciting, and more entertaining for the intended readers.

51 I will not use the term ‘withdrawal tactics’ or anything similar, but in the interests of brevity and simplicity, I will combine discussion of the tactics of retreats and withdrawals under the one term. As will be seen, there were also a number of general similarities in the methods by which the Romans safely retreated and withdrew, so that a single term for the variety of tactics used in retrograde movements is not wholly inappropriate. 52 Gilliver, 2007, 134; Sabin, 2007, 413–4. 53 See App. A. More examples can be found in accounts of Classical and Hellenistic Greek battles (cf. Conclusion, fn. 827). 54 For instance, the phrases e castra/castris and ex castra/castris, describing an army moving out of their camp, appear hundreds of times in Livy and other authors. Some number of these would likely qualify as withdrawals. 55 These numbers are drawn from the PHI Latin Database.


20

Also, Roman military narratives were often structured to ultimately end in Roman victory.56

This could have produced a bias against reporting all Roman defeats that actually occurred,

or downplaying the extent of certain defeats, while emphasising the chaos of non–Roman

defeats.57 Further, the Roman literary trope of disorderliness among non–Roman armies may

also have factored into the presentation of their defeats.58 Of course, it is impossible to know

what writers did not include in their histories. But the possibility must be considered that

retreats, withdrawals, and a variety of uneventful disengagements occurred more often than

recorded by the ancient authors.

There are enough examples of retreats and withdrawals to state that they were not uncommon

in Republican Roman warfare, although they no doubt occurred less often than routs. Even if

a retrograde manoeuvre did not always occur during or following a battle, a knowledgeable

general will surely have considered the possibility of such. Therefore, even when retreats or

withdrawals did not occur they must often have shaped tactics and strategy in some way. The

modern consensus that battles usually concluded with a rout rather than by a retreat is thus

technically correct, but retreats and withdrawals certainly occurred commonly enough to have

been a significant factor in generalship and tactics.

Overview of Key Sources

Literary evidence will be the main evidence utilised throughout the thesis. To discuss in any

detail how retreats and withdrawals are described by ancient authors, it is first necessary to

examine these authors and their aims, style and authenticity. I will utilise contemporary or

near–contemporary writers of military experience whenever they are available, but this is not

always possible. Other sources of information, such as archaeological evidence or inferences

drawn from military equipment, will very occasionally be utilised where appropriate.

However, detailed literary sources are central to developing an understanding of retrograde

manoeuvres and their role in ancient battle. The main sources I will utilise in this thesis are

Polybius, Caesar, Sallust, Livy, Plutarch and Appian.

56 Clark, 2014, 210, 213. 57 Clark (2017, 192, 207–8) has also argued that Roman authors such as Livy used defeats as a narrative technique to create drama, as easy victories were less entertaining, but they also minimised the scale of defeats, or paired them with later (not always authentic) victories. 58 Riggsby, 2006, 102–3, 105.


21

Polybius is the only detailed contemporary source for military affairs in the Middle Republic,

and most fortunately, he was a formidable analyst and was well experienced in military

affairs.59 Citing his own experience, he considered personal research and political/military

experience to be essential prerequisites for the reliable presentation of history, and held a

scientific and factual approach to historical enquiry.60 He is one of the most important

historians for the Republican Roman army.61 His work is, of course, not free from (pro–

Roman) bias. For instance, his work is slanted in the favour of his Roman patrons, the

Scipios, and he had an ulterior motive in his work of discouraging Greek states from rebelling

against Rome.62 In effect, Polybius became uncritical when writing about the Scipios, and

attempted to paint the Roman army as near–invincible.63 He also had a tendency to overwrite

otherwise logical accounts of cause–and–effect with an air of inevitability.64 However,

Polybius demonstrates a genuine interest in the arrangements and movements of armies in

battle, and goes out of his way to describe matters of deployment, terrain and manoeuvre.65

He certainly records a number of Roman defeats.66 Due his biases, the possibility must be

considered that Polybius exaggerated the orderliness of some retrograde movements he

describes, but there is nothing in his style or purposes that suggest such movements would

have been totally fabricated.67 His accounts of retreats and withdrawals may therefore be

trusted.

59 Hoyos, 2005, 215–6. 60 Polybius explains much about his own approach to history in Book XII (4c–5, 12.1–3, 22.6, 25b, 25e, 25g). That said, Polybius was also a firm supporter of traditional heroic values such as courage, honour and glory, and would assess historical figures according to these qualities (Eckstein, 1995, 54). 61 Sage, 2008, 2. 62 On Polybius’ purposes, see; Walbank, 2009, 277; Seager, 2013, 247–8; Sommer, 2013, 317–8; Thornton, 2013, 214. Cf. Polyb. 1.1–2. Relationship with the Scipiones: Southern, 2006, 20. Dissuade rebellion: Erskine, 2013, 245. 63 Davidson, 2009, 126–7. On his presentation of the Roman army, see Erskine, 2013, 231, 234–5, 242–5. 64 Davidson, 2009, 133–4. 65 Lendon, 1999, 284. Cf. Polyb. 11.22–3, 18.28–32. Polybius has a particular interest in the general’s ability to impose order over the chaotic forces of battle (Eckstein, 1995, 193), which will be useful in later chapters for assessing individual cases of generalship. 66 Polybius studiously records the numerous Roman defeats at the hands of Hannibal: cf. Polyb. 3.72–4, 3.83–4, 3.105, 3.113–6. 67 Walbank’s commentary will be of particular importance for interpreting Polybius (cf. Gibson and Harrison, 2013, 1–2).


22

For the Late Republican period, Caesar is an important contemporaneous source. Caesar was

a witness and participant in most of the campaigns detailed in his commentaries, and wrote

on military matters in unusual detail.68 As he wrote his books yearly, his memory of the

events he describes will still have been relatively fresh.69 Despite some criticism that he

presented events in a manner favourable to himself, considering that Caesar wrote of

contemporary events for an experienced audience, it would have been difficult for him to

falsify many military details and he is surprisingly honest about his own failings.70 Caesar

does have a tendency to depict his opponents as less capable than himself and his men, but

these depictions are not outside the realm of possibility.71 He does minimise the role of his

subordinate officers in accordance with his literary aim of self–publicisation.72 However, he

also provides valuable detail on the tactical roles of centuriones, who often feature in his

battle narratives.73 In battle accounts, Caesar viewed morale factors (ie. virtus and animus) as

more important than the mere arrangement of forces in deciding the outcome, and so he is

vaguer about matters of terrain than Polybius.74 With that said, Caesar was a far more

experienced general than was Polybius, and so Caesar’s emphasis on psychology in battle

must be seen as providing valuable and unique insight into Roman battle mechanics.75 As

Lendon has said, Caesar is ‘uniquely suited’ for the study of battle accounts and

understanding the reasons for the success of Roman tactics.76 Considering the

contemporaneousness of his writings, his own extensive military experience, his relatively

realistic style, and the fact that he describes a significant number of retreats and withdrawals,

68 Lendon, 1999, 276. Like other authors, Caesar’s popular purposes forestalled the need to give technical explanations of military matters (Edwards, 1917, xvii), but Caesar’s extensive military knowledge and personal presence at events he describes still leave us with some uniquely descriptive accounts of Roman battle tactics: for instance, Caesar’s description of a planned retreat at Dyrrhachium (BCiv. 3.45–6), or of the withdrawal of Afranius and Petreius’ army at Ilerda (BCiv. 1.64) 69 Kagan, 2006, 111. 70 Fuller, 1965, 55. 71 For instance, the Gauls are always depicted as inferior to the Romans in virtus (Riggsby, 2006, 101–3). Note that Caesar attributes the superior virtus of his troops to his own superior qualities as a general (Riggsby, 104). Also, Caesar’s depiction of Pompey as an incompetent at Pharsalus (Caes. BCiv. 3.94) also seems unlikely in light of the latter’s experience, but it would be hard to argue that Pompey had actually led well (Goldsworthy, 2007, 523). 72 Welch, 1998, 85–7; Kagan, 2006, 193; Goldsworthy, 2007, 7. Note also that Caesar excuses reverses suffered by his army: for example, at Gergovia, Caesar blames the heavy casualties on the failure of his men to retreat as ordered (BGall. 7.51). However, it is not implausible that this actually occurred (Gilliver, 2005, 70). 73 Welch, 1998, 85–7. 74 Lendon, 1999, 324–5. 75 Ibid., 323, 325–6. 76 Ibid., 276.


23

Caesar must be considered as the most detailed and useful source for the purposes of this

thesis, if only for the first century BC.

It must be added that only two of the commentarii, the Bellum Gallicum and Bellum Civile,

were actually authored by Caesar. Nonetheless, the author(s) of Bellum Alexandrinum,

Hispaniense, and Africum were still experienced officers if not necessarily eyewitnesses to

the events they described, and they utilised Caesar and other eyewitness participants as

sources for their narrative.77 These sources certainly contain a number of detailed accounts of

battles, featuring several retreats and withdrawals.78 Also, the Bellum Hispaniense appears to

have originally been written by a relatively junior officer, and possesses a unique focus on the

experience of common soldiers in battle.79 These sources are, therefore, still of significant

use, even if their reliability must be considered less than that of the genuine writings of

Caesar.

Sallust was a Late Republican writer, expounding a theme of steady Roman moral decline

since the defeat of Rome’s great enemy, Carthage.80 One particular work of his, the Bellum

Jugurthinum, provides a detailed account of a Roman retreat through Numidia in 106 BC

(Sall. Jug. 97–8), which will be of central interest to this thesis. In assessing the reliability of

this account, it can be noted that Sallust was an experienced politician and observer of

political events from the time of Sulla to that of Caesar, and as his battle accounts are

generally detailed, it reasonable to presume that he depicted battle and tactics in a realistic

manner.81 His military career was neither long nor distinguished, but he will have been

knowledgeable enough on Roman tactics to describe them with basic accuracy.82 On the other

hand, he was not a participant in the Jugurthine War, for he wrote in the triumviral period and

77 Gaertner (2017, 275–6) has argued that the composer of Bellum Alexandrinum, Bellum Hispaniense and Bellum Africum was none other than Aulus Hirtius, author of Book VIII of the Bellum Gallicum, utilising the notes and writings of Caesar and other participants. An authorship verification system has also convincingly shown that the author of the Bellum Alexandrinum.was probably Aulus Hirtius (Kestemont et al, 2016, 93–95). 78 BAlex.40, 76; BAfr. 15–8; BHisp. 31. 79 There is extensive debate over who wrote this book (Milne, 2009, 125–40), even if Aulus Hirtius may have composed it (Gaertner 2017, 275–6). However, it can be argued that it gives a soldiers’ eye view of events. 80 Connolly, 2009, 183; Roller, 2009, 218. Sallust wished to expose the corruption of Rome’s traditional aristocracy and the degradation of Rome’s adversarial political system (Jug. 5.1–2). 81 Batstone, 2010, xi–xix. 82 On his short military career, see Paul, 1984, 1. Sallust will have at least received the militaristic upbringing of any aristocrat (Goldsworthy, 1996, 169–70)


24

he had few sources available for the military narrative he provided.83 It is likely that he

personally contributed a number of details to his military narratives.84 In regard to Marius

and Sulla (the key Roman figures in this battle), Sallust viewed neither with great favour due

to their alleged lust for power.85 However, my analysis of this battle account will focus on the

tactics and military decisions of the generals rather than their characterisation, and Sallust’s

personal disapproval does not appear to have significantly tainted the account in this regard.86

All said, Sallust had the knowledge to depict Roman tactics authentically. As he provides the

only detailed example of a fighting retreat from the period of the Jugurthine War, he is vital to

understanding the development of Roman retreat tactics from the second to the first centuries

BC.

Livy wrote much of his work in the Augustan period, and provides the fullest account of the

Republic’s political and military history.87 His key purpose, particularly in his presentation of

the Second Punic War, was to demonstrate the importance of morality in Roman history,

emphasising the impact of ethics and character on the outcome of events.88 This approach

appears less scientific than, for instance, Polybius’ analytical approach. Even so, throughout

Livy’s battle accounts, his keen interest in the motivation and psychology of participants in

events causes him to provide insightful and authentic assessments of the morale of

combatants in battle.89 He had a less scientific but also less schematic view of tactics than

Polybius.90 In this regard, Livy’s explanations of Roman tactics are closer to those of Caesar,

and are not invalid simply because his approach differs to that of Polybius.91

83 Paul, 1984, 2. His was perhaps the first detailed treatment of the war, but he may have had the memoirs of at least two participants, P. Rutilius Rufus and Sulla, to draw upon (Paul, 1984, 3–5). Considering Sallust’s glowing and frequent depiction of Sulla as a soldier (Jug. 96, 100.2, 101.4–9, etc.), it seems likely that Sulla was at least one of his sources. 84 Paul, 1984, 4–5. 85 Connolly, 2009, 190. 86 Marius and Sulla are depicted favourably throughout the battle itself (Sall. Jug. 97–9). 87 Cf. Hoyos, 2005, 217–8: Livy wrote 142 books over a period of about 45 years, covering Roman history from the founding of the city to the death of Drusus I in 9 BC. 88 Levene, 2010, 391–2. 89 Walsh, 1961, 168, 172. 90 Levene, 2010, 311–16. 91 Cf. Lendon, 1999, 323, 325–6.


25

However, despite the historical breadth of his work and the fact that he was a Republican

author, the quality of his history, particularly the accuracy of his military accounts, has been

sharply criticised.92 Walsh argues that Livy was not always critical in his use of sources, he

often utilised one source only for a particular event, and he is argued to have had a poor

understanding of tactics, resulting in some confused narratives.93 Livy was also content to

fabricate military detail for literary effect, with the result that many of his battle accounts are

riddled with tropes.94 Walsh judges that, while Livy did make some efforts to present a

truthful narrative, he was fundamentally a writer first, and an historian second.95 With that

said, others take a more forgiving view. According to Koon, Although Livy has a tendency to

produce clichéd battle accounts, the sheer volume of these among his enormous work

provides a good overview of what was generally considered to have been usual in Republican

battles.96 Hoyos points out that Livy transmits an enormous amount of otherwise–lost

material from annalistic sources.97 Hoyos also states, contra Walsh, that Livy did not rely

only on a single source, but created his own narrative blend based on at least two accounts.98

Lendon also argues (again contra Walsh) that Livy was certainly an historian, not a

rhetorician, and was read as such.99 These views credit Livy with far more industria in his

work that does Walsh.

Therefore, Livy remains a valuable source, but several things must be kept in mind when

relying on his narrative. It must be noted that there is no evidence of Livy possessing military

experience, despite the constant wars during his lifetime.100 On this account, Polybius and

Caesar’s depictions of Roman tactics might be considered more trustworthy. However, Livy

is essential to this thesis particularly for the sheer number of retreats and withdrawals he

describes.101 In assessing these, one must of course be critical. For instance, some retreats or

92 Walsh, 1961, 157–8, 168, 198–201. 93 Ibid., 142, 144–5, 157–8. Walsh (109) further argues that Livy was capable of distorting or suppressing facts to create certain impressions of individuals that suited his literary and moral purposes. 94 Ibid., 218. 95 Ibid., 287. 96 Koon, 2010, 86. 97 Hoyos, 2005, 217. 98 Ibid., 217. 99 Lendon, 2010, 57–8. 100 Walsh 1961, 4; Hoyos, 2005, 217–8. 101 Considering his literary purposes and flexible historical techniques, one might wonder if he had a tendency to report battles with clearer outcomes, thereby excluding accounts of indecisive battles that might have qualified for study in this thesis. Then again, one cannot be too upset by the great volume of material he does provide.


26

withdrawals that Livy describes could have been tropes to fill gaps in his sources or explain

indecisive battles. One must also keep in mind how Livy’s drive to establish morality as a

central explanation for Roman success could impact on his narratives.102 In general, Livy

adhered closely to the military details of his sources, and so the reliability of his presentations

of tactics much depended on those original sources.103 For periods covered by Polybius

(particularly the Second Punic War and Macedonian Wars) he is extremely useful, as he may

be read in comparison with Polybius.104 Several of the Livian accounts utilised throughout

this thesis fall into this category. Where no comparative account is available, however, I will

take the view of Koon, that even where a battle account does not appear to be based on a

reliable source or even accurately presented, the tactics that Livy describes are still indicative

of standard procedures in battles of the Augustan period.105 This approach can be used even

with battle accounts that cannot have had reliable sources, such as those describing events in

the Early Republic.106

Plutarch’s Lives are also a central source of information in latter chapters of this thesis. In

particular, the Lives of Aemilius, Crassus and Antony will be utilised heavily in Chapters Five

and Six. In general, Plutarch was non–contemporary to the events he describes, writing in the

late first and early second centuries AD. He was a biographer rather than an historian, and

thus focused more on anecdotes and events that revealed facets of his subject’s character,

rather than focusing on great events themselves.107 As he did not always possess reliable

biographical information to draw upon, he was not averse to drawing out or making up the

necessary details to complete his narrative.108 However, where his sources can be determined,

he does appear to have used the work of a variety of historians, and particularly valued high

quality sources such as Thucydides and Polybius.109

102 Cf. Connolly, 186, 188–9. 103 Erdkamp, 2006, 538–9. 104 Hoyos, 2005, 218. 105 Koon, 2010, 86. 106 Eg. Livy 4.39.4–9, 6.24. 107 Geiger, 2014, 293; Schettino, 2014, 417. 108 Geiger, 2014, 300. 109 Schettino, 2014, 425–6, 429.


27

The Lives mentioned above describe key battles that will be subjects of case studies in

Chapters Five and Six, and so it is mainly necessary to determine Plutarch’s reliability in

these specific accounts. Regarding Aemilius, Plutarch must be utilised to assess retreat tactics

at the Battle of Pydna, 168 BC, as several key sections are missing from Livy’s account, so

that Plutarch provides the only full account.110 Plutarch evidently consulted Polybius, as well

as the eyewitness Scipio Nasica (cos. 138) to compose his account of Pydna, which is a good

sign towards the reliability of Plutarch’s depiction of tactics during the battle.111 Plutarch is

likewise a key source for the Parthian campaigns of Crassus and Antonius, and while his own

sources for these campaigns are unknown, there were some to draw upon and so it was not

necessary for him to fabricate these accounts.112 Plutarch’s account of the Battle of Carrhae,

53 BC, appears preferable to that of Cassius Dio, which was written later and appears less

logical at points.113 Regarding Plutarch’s interpretation of events, an emphasis on personal

virtues is clear in Antony, so that he presents Antonius quite favourably throughout the

account of the latter’s retreat through Media in 36 BC.114 Conversely, bias against Crassus

stands out in Plutarch’s account of Carrhae, perhaps a product of the malign influence of

Cassius.115 Plutarch depicts Crassus tragically, as a victim of fortuna and the treachery of

easterners.116 In Plutarch’s view, this was a result of Crassus’ greed and ambition.117

Plutarch’s moralistic and biographical aims are thus evident in Antony and especially in

Crassus. That said, in assessing Crassus and Antony, Tarn thought quite highly of Plutarch as

an historian.118 Evidently, it is necessary to examine Plutarch’s battle accounts closely for

110 Regarding Livy, the Latin cuts off at postremo omnes, et cum praesidio ... omitting important parts of the battle (44.40–1). 111 Polybius’ narrative is lost, but Plutarch had evidently read it (Plut. Aem. 19.4). He also utilised Poseidonius, who claimed to be an eyewitness (Aem. 19.7). Of course, in using Polybius, Plutarch will be transmitting a highly favourable view of Aemilius Paulus Macedonicus, who was closely associated with Polybius’ Roman patrons, the Scipios, and was actually the father of Scipio Aemilianus. Cf. Davidson, 2009, 126–7. 112 Tarn (1951, 50), argues that an unnamed Mesopotamian Greek was probably Plutarch’s source for both Crassus and Antony. However, Pelling (1988, 221) argues that Quintus Dellius is Plutarch’s main source for Antony. 113 Sampson, 2015, 122–3. 114 Pelling, 1988, 221; Geiger, 2014, 297–8. 115 Cf. Ward, 1977, 1–3. Whether Plutarch viewed a protagonist favourably or unfavourably was influenced by his sources (Schettino, 2014, 427–8). Cassius’ possible influence over Plutarch’s narrative of Carrhae will be explained in more detail in Chapter Six (‘The Battle of Carrhae, 53 BC’), but suffice to say that Cassius was the most senior survivor of this disaster, and so any account of events he gave will have been influential, even though he abandoned his general to die (Plut. Crass. 29.4). 116 Geiger, 2014, 293. See Plut. Crass. 29.5, 30.3. 117 Schettino, 2014, 425. 118 Tarn, 1951, 51.


28

evidence of moralistic and biographical details shaping the tactical narrative, but he is still an

important source for the battles in question, and in fact preferable to the available

alternatives.119

Appian was a Greek historian and a man of significant practical experience who wrote in the

second century AD.120 His enormous Roman History was divided into twenty–four books,

organised both chronologically and according to the names of the peoples whom the Romans

were fighting at a given time.121 Within the surviving works, he provides a number of

accounts of retreats and withdrawals.122 However, there are conflicting views on his

reliability. According to Carter, Appian was more judicious in his use of sources than was

Livy, not simply relying on one source but taking a more critical approach and including his

own assessments of events.123 On the other hand, Hoyos considers his work to be of highly

uneven quality.124 Appian stands out as notably ignorant in regards to geography, which raises

questions about his credibility in tactical matters.125 For some reason, Appian’s account of the

Second Punic War appears to have utilised neither Polybius nor Livy (perhaps in favour of

uncritically–utilised annalistic sources), and is riddled with errors and incredible details.126

Fortunately, Appian is more reliable on the wars of the second and first century: he appears to

have utilised Polybius to create a believable account of the Third Punic War, and his account

of the Iberian wars is coherent.127 Indeed, he is the main surviving source for the Iberian wars

in the second century.128 Further, Appian’s account of the Battle of Magnesia, 190 BC is

actually preferable to Livy’s more confused account.129 In this thesis, the majority of

references to Appian are from this more reliable second and first century period. These

accounts may thus be viewed with some basic credence, and assessed with the usual amount

119 A more detailed analysis of Plutarch’s depiction of Crassus and Antonius will appear in Chapter Six (198–9, 208–9). 120 Carter, 1996, ix–x; Richardson, 2000, 1–2. Appian’s personal life is not well understood but he appears to have had a reasonably successful political career in Alexandria and to have later been an important Imperial official, possibly advocatus fisci, and potentially later a procurator. It is at least clear that Appian was experienced in politics. 121 Richardson, 2000, 2–3. 122 App. Hann. 41, Syr. 6.35, Pun. 103, Hisp. 82, 88, BCiv. 1.90, 2.80, 3.69, 4.128. 123 Carter, 1996, xxxii–xxxiii. 124 Hoyos, 2005, 218. 125 Richardson, 2000, 6; Hoyos, 2005, 218. 126 Richardson, 2000, 4; Hoyos, 2005, 220. 127 Richardson, 2000, 1, 5; Hoyos, 2005, 218. 128 Richardson, 2000, 1. 129 Appian probably relied upon Polybius for his account (Walsh, 1961, 161–2).


29

of scepticism. As will be seen, Appian’s accounts of retreats and withdrawals in the second

century are the only such accounts of any detail for this period. They are therefore invaluable

for the purposes of examining the development of Roman tactics between the well

documented Second Punic War, and the various civil and foreign wars of the first century BC.

The Imperial era saw the production of several extant military manuals, including the Latin

Strategemata (Stratagems) of Frontinus, and the Greek Strategikos (the General) of

Onasander. These manuals are not highly technical, but fall into the broad category of general

advice on strategy and tactics with frequent citation of historical examples and

illustrations.130 Both date to the early Empire, with Onasander dedicating his work to Quintus

Veranius, a consul in AD 49, and Frontinus serving as suffect consul c. 73 AD and again in 93

AD.131 Frontinus has provided us with the only surviving Latin military manual from this

period, and fortunately, his work includes a short chapter (Strat. 2.13) on retreats entitled ‘On

Retreating’ (De Effugiendo). That said, Frontinus’ work is not intended to be a definitive art

of war, but rather a catalogue of examples.132 It is therefore mainly of use as a source of

examples, rather than as a detailed source on the mechanics of retreats and withdrawals.133

The extent to which the advice contained in the Strategemata and Strategikos was really used

by Roman generals may be debated, but it must at least be noted that Frontinus was a man of

military experience.134 As for Onasander, his work was as much one of philosophy as

tactics.135 Like Frontinus, he is a useful source for drawing out general precepts of Roman

tactics in early Empire, and showing what the Romans (or at least the authors) considered to

be good sense. Even so, neither these nor any other manual provide a basis for detailed case

studies, and so they are not heavily relied upon in this thesis.136

130 Campbell, 1987, 13. 131 Ibid., 13–4. 132 Formisano, 2017, 130. 133 Campbell, 1987, 14–5. 134 Ibid., 27–8. He was not, however, immune to errors and anachronisms (Bell, 1965, 407). 135 Campbell, 1987, 13–4; Formisano, 2017, 145. 136 It would be remiss not to mention Vegetius, a well-known Late Imperial source, whose De Re Militari is referenced below on a very few occasions, but not in support of any key arguments. He is simply too late a source to fall within the scope of this thesis and, like Frontinus and Onasander, does not provide the narrative material that would support a detailed case study.


30

The vast majority of this thesis will focus on the Middle to Late Republican periods, using

the above sources. I will only occasionally use other sources to give supporting or contrasting

evidence. However, Chapter Six will conclude with a case study of the clades Variana, which

occurred in AD 9. The Conclusion will also incorporate a brief discussion of Roman retreat

tactics and their further development in the Imperial period. For these purposes, I will later

make some use of key sources for the principate, particularly Tacitus and Cassius Dio, but

also Velleius Paterculus and Josephus. These sources naturally have differing strengths and

weaknesses, and I will touch on them as necessary in Chapter Six and the Conclusion.

Terminology and Complications in Identifying and Distinguishing Types of Retrograde

Manoeuvres

There are many references to retrograde movements in the ancient sources.137 Having

examined the differing styles and experiences of the writers themselves, we may turn to these

accounts, and examine how different authors presented retrograde manoeuvres. The key

question for the purposes of this thesis is, how can retrograde manoeuvres be consistently

identified and distinguished?

Latin sources utilised a variety of terms to describe retreats and withdrawals. Such terms

included recipere (regain, recover) or se recipere, se convertere (to turn back, reverse) referre

(bear back, bring back), abire (go away, depart), discedere (divide, separate), subducere (to

lead away, withdraw), abscedo (move away, withdraw) and digredi (separate, go apart).138

Frontinus also used a variety of terms.139 The crux of the matter, however, is how

systematically these terms were utilised in the narrative sources. To throw light on this

subject, the writings of Polybius, Caesar and Livy will be contrasted and compared.

137 See App. A. 138 There are many examples, and more will be considered below. A handful of such now follow. recipere: Caes. BCiv. 1.46 (facilis est nostris receptus datus ... ac tutiorem nostris receptum dat), 3.45–6 (recepissent ... receptui); BAlex 40 (pugnans in orbem se recepit). Se recipere and se convertere: Caes. BCiv 1.46. Referre: Caes. BGall. 1.25 (pedem referre); Livy 30.18.12 (gradum sensim referentes ordines et tenorem). Abire: Sall. Jug. 98.5.2. Discedere: Caes. BCiv. 3.41 (discessisse). Subducere: Livy 25.36.2 (subducit); Tac. Ann. 2.46 (subduxisset). Abscedo: Tac. Ann. 12.33 (abscessus). Digredi: Livy 10.35.2 (digredientes). 139 Frontinus uses several different verbs to describe retreats, including abire (2.13.3) digredi (2.13.7), evadere (2.13.4), effugere (2.13.1–2) and recipere (2.13.7). He also uses the term regressus (Strat. 3.10). That said, as Frontinus’ work is a catalogue of examples rather than a technical military manual (Formisano, 2017, 130), it is not a decisive indicator of how military men described retrograde manoeuvres.


31

To begin, let us consider Polybius. As mentioned, Polybius was technical and precise in his

presentation of military matters, and as he was utilised by Livy as a key source, he provides a

good point of comparison.140 It is therefore worth examining how consistently he applied

terminology for retreats and withdrawals. As an example, Polybius’ description (18.25.4–7)

of the ebb–and–flow of the battle at Cynoscephalae in 197 BC is easy to interpret:

Flamininus, seeing that…his left was being forced back, some of them having already

perished and others retreating slowly (ἀναχώρησιν), his only hope of safety lay in

his right, hastened to place himself in command there ... The Macedonians now,

having no one to give them orders and being unable to adopt the formation proper to

the phalanx… did not even wait until they were at close quarters with the Romans, but

gave way thrown into confusion and broken up (διασπασθέντες ἐνέκλιναν) by the

elephants alone.141

Polybius’ description is relatively straightforward. He distinguishes clearly between the

orderly retreat (ἀναχώρησιν) of the Romans being pressed by the phalanx, and the

confusion of the Macedonians (διασπασθέντες ἐνέκλιναν) when attacked by the

elephants. The direct description of the positions and movements of elements of the opposing

armies makes it relatively easy to visualise what was happening. This clear description is

typical of Polybian battle descriptions.142 In regards to terminology, Polybius repeatedly uses

the terms ἀναχώρησις, ὑποχωρέω or ἀποχώρησις to describe retreats from battle.143

These appear, therefore, to be technical terms.144

140 Cf. Lendon, 1999, 284; Hoyos, 2005, 217–8. 141 Trans. W. R. Paton, 1922–7. 142 Recall Polybius’ overriding interest in the imposition of order over the chaos of battle (Eckstein, 1995, 193). 143 Polyb. 1.19.4 (ὑπεχώρουν), 1.27.8 (ὑπεχώρουν), 2.68.4 (ὑποχωρεῖν), 2.68.8 (ἀναχώρησιν), 2.68.9 (ἀναχώρησιν), 2.68.10 (ἀναχώρησιν), 5.86.3 (ἀποχώρησιν). 144 Onasander (Strat. 27) also uses ὑποχωρέω to describe retreats. This work was as much one of philosophy as

tactics (Formisano, 2017, 145), but it is still reasonable to assume this was a common Greek term for retreat.


32

However, Polybius also uses ἀναχώρησις and ἀποχώρησις to describe withdrawals, over

longer distances and in marching formation.145 For example, in describing the aftermath of

the Battle of Raphia in 217 BC, Polybius uses one term (ἀναχωρήσας) to describe

Ptolemy’s retirement from the pursuit into his own camp (5.86.1), and a different term

(ἀποχώρησιν) for both Antiochus’ flight from battle and his long–distance withdrawal from

Syria entire (5.86.8). The consistency of Polybius’ terminology for describing retreats and

withdrawals demonstrates that he did not distinguish between these types of retrograde

manoeuvres. Conversely, throughout this thesis, a distinction is made between retreat and

withdrawal, and so this distinction cannot be made purely by relying on terminology as

utilised by Polybius.

If Polybius is relatively clear in his own method of describing retrograde manoeuvres, Livy

presents different challenges. Consider his description (Livy 27.2.8–10) of a battle fought

between Hannibal and Marcellus in 210 BC, near Numistro.146

A new battle and fierce suddenly flamed out of a conflict now grown spiritless,

for the combatants were now fresh in spirit and in body. But night parted them

(diremit pugnantes) with victory undecided. On the next day the Romans stood in

line from sunrise until late in the day. When none of the enemy came out against

them, they gathered spoils at their leisure, carried corpses of their men into one

place and burned them. On the following night Hannibal broke camp (mouit

castra) silently and marched away (abiit) into Apulia.147

There are several distinct manoeuvres described here. The first is a mutual retreat when night

fell over the battlefield (diremit pugnantes), the second is Hannibal’s movement from his

camp (mouit castra), and another is Hannibal’s withdrawal into Apulia (in Apuliam abiit).

These manoeuvres are stated matter–of–factly, with no description whatsoever of any of the

distinct movements.148 And yet the manner by which the armies retreated from the battlefield,

145 Polyb. 3.74.6 (ἀπεχώρησαν), 3.84.13 (ἀπεχώρησαν), 11.24.7 (ἀναχώρησιν). 146 It must be noted here that this battle was likely not historical, but an annalistic embellishment (Erdkamp, 2006, 549–51). For examining Livy’s use of terminology, however, it is not necessary that this event occurred at all. 147 Trans. Frank Gardner Moore, 1943. 148 This is typical of many battle narratives: see Goldsworthy, 1996, 171; Sabin, 2007, 399–400.


33

and by which Hannibal withdrew from camp into Apulia, must surely have differed in each

case. Certainly the fact that Hannibal abandoned spoils on the battlefield (Livy 27.2.8)

implies that his was the beaten army, yet Livy does not describe any difference, relative to the

Romans, in the manner of his retreat from the field. In retreating, each army must have

maintained some sort of order or formation. Then, Hannibal must have organised his army

into a marching formation and withdrawn from the camp in some orderly manner. Finally, for

the march into Apulia, he must have made dispositions in case the Romans attacked him, as

shown later when the pursuing Romans caught up with him (Livy 27.2.11–2). In this one

short extract, Livy describes different movements, occurring from the battleline into a camp

by day and then from the camp to more distant territory by night. Thus, Livy demonstrates

that retreats and withdrawals occurred under differing circumstances, but he also utilises a

variety of terms to describe these manoeuvres.

When describing retreats from battle, Livy uses multiple phrases, particularly se recipere,

signum recipere and referre pedem, among a handful of others.149 When describing

withdrawals over longer distances, however, Livy uses verbs of movement such as abire,

movere, subducere and redire.150 The distinction is not always clear, such as in Livy’s usage

of another term, subducere, which could describe either a retreat or a withdrawal.151 For the

most part, however, Livy does distinguish between short retreats from battle and tactical

movements over longer distances.

It is instructive to compare Livy with Polybius. Livy at times utilised Polybius as his main

source for battle descriptions, and according to Erdkamp, he generally adhered faithfully to

Polybius’ terminology and technical details.152 For example, consider Polybius and Livy’s

accounts of Cynoscephalae and Cannae. In the former case, Polybius (18.25.4) uses the term

ἀναχώρησιν to describe the retreat of the Roman left flank, while Livy (33.9.6) uses the

149 Livy 4.39.6 (se reciperent), 6.24.1 (pedem rettuluit), 6.24.3 (inclinavit aciem), 6.24.4 (non recipebat se), 25.19.5 (signum receptui, reducti), 28.13 (turmatim abibant), 28.15.6 (rettulere pedem), 28.15.9 (se reciperent), 30.18.12 (referentes), 32.12.2–3 (recipere se, ad receptum), 33.15.11 (rettulere pedem), 33.18.18 (vertunt), 33.18.19 (receperunt sese in castra), 37.38 (recipere se), 38.25.12 (cedere sensim), 42.59 (cedentes sensim), 44.42.2 (excessit), 44.42.3 (abibant). 150 Livy 25.19.6 (abit), 25.19.8 (redit), 27.49 (abire), 32.12.9 (agmine petunt), 28.15.16 (castra movet), 33.18.18 (vertunt). 151 Eg. Livy 23.26.10, 25.36.1–2, 29.1.14, 44.37.2. 152 Livy did make changes to his source material, but it was the ‘drama and psychology’ of the account that interested him, not the technical details (Erdkamp, 2006, 531, 534, 538–9).


34

phrase pedem referentes to describe the same phenomenon. At Cannae, when the

Carthaginian centre fell back under the weight of a massive Roman attack, Polybius (3.115.5)

described this retrograde movement with the term ὑπεχώρουν, while Livy (22.47.6) again

used a variation of referre (referentibus). It thus appears that referre is at least one term that

Livy utilises as a technical term for retreat. Of course, as seen above, he uses several other

terms. Overall, Livy is more varied with his terminology than Polybius, which creates greater

ambiguity, but he also appears to distinguish more clearly between verbs that describe

retreats, and verbs of marching and movement for withdrawals. For the purposes of this

thesis, Livy’s style is more helpful than a strict adherence to a small number of specific

technical terms that do not distinguish between short retreats from battle, and longer

withdrawals.

Caesar, for his concise Latin and extensive military experience, might be expected to have

been particularly clear in describing retreats and withdrawals. Consider then his description

of how his men escaped a battle on unfavourable ground at Ilerda, in 49 BC (Caes. BCiv.

1.46):

When they [Caesar’s cohorts] had fought in this way continuously for five hours, and

our men were being grievously harassed by superior numbers, having spent all their

missiles, they draw their swords and, breasting the hill, charge the cohorts

[Pompeians], and after laying a few low, they force the rest to retreat (sese

convertere). When the [Pompeian] cohorts were thus pushed close up to the wall, and

to some extent driven by terror to enter the town, an easy withdrawal was allowed our

men (facilis est nostris receptus datus). Our cavalry, however, on each flank, though it

had been stationed on low–lying ground at the foot of the cliff, yet forces its way with

the utmost valour to the ridge, and, riding between the two lines of battle, allows our

men a more convenient and safer withdrawal (commodiorem ac tutiorem nostris

receptum dat).153

153 Trans. A. G. Peskett, 1914.


35

Caesar describes two different retrograde movements. The first is an involuntary retreat (se

convertere) by the Pompeians after the charge by Caesar’s men, with some chaos implied by

the fact that the former could no longer pressure Caesar’s own men. Caesar then describes

more controlled retreats by his own men, and twice uses the same verb (recipere) to describe

them.154 It is far from certain from Caesar’s account that the Pompeians did not maintain

some semblance of order themselves, even if they were reluctant to re–engage. As such, in

this instance, there is no clear practical distinction between se convertere and recipere. All

that is clear is that these terms both describe relatively short–distance retrograde movements,

or retreats. Caesar, therefore, could utilise a variety of terms to describe similar or even

identical movements, as did Livy.

That said, Caesar, like Polybius, had a tendency to utilise the same terms repeatedly to

describe retreats, particularly recipere or se recipere.155 Much less commonly he also utilised

signa inferre, reducere, and pedem referre (among other phrases) to describe retreats from

battle.156 When describing withdrawals over longer distances, Caesar did utilise verbs of

movement such as proficiscere and educere.157 However, he also applied the term se recipere

to organised long–distance retrograde movements that, for the purposes of this thesis, are

classified as withdrawals.158 Thus, Caesar goes even further than Polybius in having a single

favoured term for retrograde movements (se recipere), and applies this term to retrograde

movements over both short and long distances. A semblance of this terminology is visible in

Livy, who also uses se recipere, but amongst a larger variety of phrases. As such, se recipere

is clearly a quasi–technical term for an organised retreat, but despite his use of relatively

consistent terminology, Caesar’s use of such terminology does not always clearly distinguish

between short–distance retreat and long–distance withdrawal.

154 A. G. Peskett has translated se convertere as ‘retreat’, and recipere as ‘withdrawal’, perhaps with the implication that the term ‘retreat’ implies an involuntary retrograde movement, while ‘withdrawal’ is a more controlled manoeuvre. 155 Caes. BGall., 6.40 (se recipere), BCiv.. 1.45–6, 2.25 (se receperunt), BCiv. 2.41–2 (se recipiunt), 3.46 (receptui, se recipientibus, ad recipiendum, se receperunt) 3.94–5 (receptus and recipere), BAfr. 17 (recipere), 18 (recipientes, se reciperent, se recipiunt); BAlex.40 (se recepit). 156 Caes. BGall. 1.25 (pedem referre), BCiv. 2.41–2 (signa inferri, sese conferunt), 3.46 (reduci, reducti); BAfr. 18 (se contulerunt). 157 Caes. BGall. 5.27 (eductos ex hibernis), 5.31 (ex castris proficiscuntur), BCiv. 1.78 (ex castris proficiscuntur). 158 Caes. BGall. 5.37 (se recipient), BCiv. 3.97 (se recipere). Caesar does the same with subducere (BGall. 1.22.3, 24.1).


36

As is evident, clearly distinguishing between retreats and withdrawals through the use of

terminology is not straightforward. Polybius, Caesar and Livy all use specific terminology for

describing retreats from battle, although Livy uses a wider variety of phrases. This is not so

surprising, considering his intention to create dramatic battle narratives as opposed to more

concise and technical ones. Polybius and Caesar are similar to one another in that they use a

smaller pool of terms to describe retreats and withdrawals, but Caesar has a single term (se

recipere) that he overwhelming prefers. Caesar and Livy both differ from Polybius in that

they usually distinguish between retreats from battle and movements over longer distance,

although Livy is actually more consistent in his application of different types of verbs for

retreat and withdrawal. Unsurprisingly, Caesar and Livy share some terms, particularly se

recipere and pedem referre, demonstrating that the Romans did have a small pool of quasi–

technical phrases to describe retreat from battle. This was not so much the case, however,

with long–distance withdrawals, to describe which Caesar and Livy use a variety of verbs of

movement. This may have been a general pattern in Latin sources.159

There are two further issues. First, if an account leaves a retreat assumed but unstated, such

as in the case of stalemate, or the retreat is not clearly distinguished from a flight, what is to

be made of it? There were clearly a variety of ways in which battles could end. For example,

Livy (27.2.8–10) might describe a series of different retrograde movements in the one

incident, and Caesar (BGall. 7.51) might describe a combination of retreat and flight,

particularly if a retreat broke down under pressure. Caesar (BGall. 1.24–7) could also

describe a flight in which the vanquished troops rallied and counterattacked, and yet Livy

(27.2.8–10) might recount a retreat in which troops attempted to make a stand, but were

beaten and forced into flight. Finally, as seen above (Livy 27.2.8–10), if a battle was a

stalemate, then the armies must have effected a retreat from their fighting positions, but this

is usually not stated clearly and is certainly not often described.160 It is thus not always

straightforward to determine how exactly an army moves from the sources. For the purposes

of this thesis, it is necessary to simply assume that, where a battle is described to have ended

with a stalemate, one or both armies retreated from the location of the fighting.

159 In describing withdrawals, Frontinus also uses phrases that describe marching, or returning to a march, but which in this context imply withdrawal or to continue a withdrawal (cf. Front. Strat. 2.13.6–7). 160 Other accounts of battles being broken up by night or weather show the same lack of description regarding how the armies separated (Caes. BGall. 1.50, Livy 4.39, 8.1, 9.23, 32, 10.12, 27.2, 28.15, 38.40, 44.4; App. Hisp. 82).


37

The second matter is the question of whether the battle tactic of feigned retirement, which

was a deliberate retrograde movement to set up a surprise counterattack, should be included

in the discussion of this thesis. There are several examples of Roman forces carrying out such

manoeuvres.161 In the sense that a feigned retirement was an orderly retrograde movement in

relatively close proximity to the enemy, even if it gave the appearance of flight (Caes. BGall.

6.8), it was in fact a retreat. However, feigned retreats were pre–meditated battle tactics

carried out with the intention of continuing battle or launching an ambush.162 Even so, as

noted above, just as some examples of unplanned retreats ended in flight, some resulted in an

army standing its ground in a better physical position and some ended in active

counterattacks.163 Some retreats were also less orderly than others, and retrograde

manoeuvres that might best be described as genuine flights could also end with the fleeing

troops standing their ground once more or counterattacking.164 Thus, while the intention

behind feigned retreats may have been different to that of an involuntary retreat, the outcome

was sometimes the same. There is precious little to distinguish a feigned retreat from a

voluntary retreat made to improve an army’s physical position. If the physical movements

were similar, then distinguishing between them carries little meaning. As such, although they

fit uneasily into the definition retrograde movements due to the potential appearance of

disorder and their pre–meditated nature, feigned retreats will be examined as a form of

retreat.

To summarise the preceding discussion, Greek or Latin terminology cannot be purely relied

upon to define a retreat or withdrawal, and so the determination of whether an account

describes a retreat or withdrawal will be made on a case by case basis. Withdrawals will be

mainly distinguished from retreats according to their physical characteristics, including the

distance moved by the army and whether or not a marching formation was adopted. Orderly

retrograde manoeuvres will be generally identified as distinct from flights or other similar

phenomena according to their evident or implied levels of orderliness, direction and

command, on a case by case basis. Stalemates will generally be assumed to have ended with

retreats from battle, even where this is not stated. Feigned retreats will also be considered,

161 Eg. Caes. BGall. 6.7-8; Sall. Jug. 50; Livy 6.24, 29.34; Onas. Strat. 21.9. 162 One of few Roman examples of such tactics is Titus’ Labienus’ use of a feigned retreat in Gaul, 53 (Caes. BGall. 6.7–8). This case will be explored in detail in Chapter One (56–8). 163 See Ch. 1, 60, 80–1. 164 See App. B.


38

even where they gave the appearance of flight. Above all, it is important for the reader to

keep an open mind when considering the use of the terms ‘retreat’ and ‘withdrawal’ in the

following study. Only when examining specific examples in context does it become possible

to distinguish clearly between different kinds of retrograde manoeuvres, but based on the

information available from the sources, it is not always possible to draw such a neat

distinction.

Purpose and Scope of the Thesis

Since retreats and withdrawals have not received due attention in modern scholarship, this

thesis will primarily explore how the conduct of retrograde manoeuvres affected the tactics of

soldiers, officers and generals, and vice versa. This will shed light on what was generally to

have been considered good tactical practice and good generalship. The purpose of this thesis

is to study retreats, withdrawals and their associated tactics in Republican Roman warfare, in

order to determine the overall impact of these things on Roman tactics and generalship and

thereby to provide a more nuanced overall view of the role played by retrograde movements

in terminating battles and campaigns.

The scope of the thesis will be limited in several ways. Firstly, while there were many

scenarios in which armies might have retired with or without fighting, only retrograde

movements in proximity to the enemy will be examined. In contrast, strategic movements in

which armies manoeuvred for advantage, without tactical considerations regarding a nearby

enemy army, will not. The main focus of the thesis is retrograde movements as a method for

ending battles, or for re–establishing an army in a better position for future fighting. These

distinctions will be made clearer in Chapter One. In terms of chronology, the thesis will

adhere mostly to the Middle and Late Republican period, when the Roman army was active

in major wars and battles, and for which a multitude of sources exist to describe these events.

Finally, although the Romans are the main subjects of the thesis, there will be some analysis

of how non–Roman armies behaved in similar circumstances. In essence, however, the thesis

will endeavour to answer one central question: how did retreats and withdrawals impact on

the tactics and strategies used by the Romans during the Middle and Late Republic?


39

Thesis Structure

The overall structure of the thesis is intended to firstly outline the most general observations

that might be made of retreats and withdrawals, and then gradually to elaborate upon the

more detailed specifics of how these worked and their relevance to Roman Republican

warfare and tactics. Chapter One of the thesis is entitled ‘Retreat and Withdrawal: Why and

Where To?’ The main purpose of this chapter is to establish the contexts in which retreats and

withdrawals were carried out, so that the two different types of retrograde manoeuvres may

be more easily distinguished and also so that their impact on military strategies and tactics

may be examined in more detail in following chapters. I will firstly examine the reasons that

armies usually retreated, including defeat in battle, stalemates, the onset of darkness, and

other scenarios. I will then examine where Roman armies generally aimed to seek refuge

when carrying out a retreat, including fortified camps, high ground and walled settlements,

comparing and contrasting the value of these different types of refuge in order to explain why

the Romans chose one over the other in varying scenarios. Finally, I will examine why and

how withdrawals, which were retrograde movements but carried out over longer distances,

were conducted in order to demonstrate the similarities with and differences from retreats. In

the conclusion I will summarise the typical scenarios in which Roman armies retreated or

withdrew, where they preferred to retreat or withdraw to, and the salient differences and

similarities between retreats and withdrawals.

In Chapter Two, ‘The Impact of Pursuit on Retreating or Withdrawing Forces’, the issue of

retreats and withdrawals will be considered from the perspective of the victors. The varying

reactions of armies when their enemies retreated will be examined, particularly in cases

wherein the victors mounted a pursuit or attack, in order to establish the impact of enemy

action of the success or failure of retreats. Firstly, different types of pursuit will be examined,

in order to establish what usually happened when one side retreated. I will then explain how

pursuit could turn an orderly retreat into a disorderly flight in differing circumstances,

including short retreats from battle toward a place of refuge, and longer withdrawals when

immediate refuge was unavailable. Finally, I will assess the reasons that one army might have

decided against pursuing another. I will then conclude by summarising what usually occurred

when an army retreated or withdrew, and how this affected the conduct of these retrograde

movements.


40

Chapter Three, ‘Roman Retreat Tactics, Part I: Ratio and Combined Arms’, is the first of two

chapters analysing the methods by which the Romans retreated or withdrew successfully, and

focuses on the training, organisation and (junior) leadership of the Republican Roman army.

The chapter will open with a brief examination of two accounts of successful retreats,

showing some common factors. Some of these factors will already have been identified in

Chapters One and Two, but others will be examined in more detail. It will be argued that

‘order’ was an important factor, and I will provide a detailed discussion of what order was,

and how Roman training, tactics and leadership contributed to maintaining order. In terms of

leadership, I will focus on the roles of senior soldiers and certain types of officers, including

centuriones, principales, veterani, and more senior officers such as tribuni and quaestores.

Finally, the tactical roles of light infantry and cavalry will be considered in order to give a full

picture of the Roman army’s use of combined arms during retrograde manoeuvres. The

conclusion will summarise the characteristics of successful retreats and give an outline of the

methods and tactics that Roman forces could use to retreat successfully.

Chapter Four, 'Roman Retreat Tactics, Part II: Virtus, Consilium and Scientia Rei Militaris'

considers the role of the general in a retreat or withdrawal. First, a brief and general overview

of the role and responsibilities of the Roman general prior to and during battle will be given.

Following this, the role of the general will be examined in circumstances more specific to

retreats, including retreats from the battleline and into a fortified position, retreats out of

fortified positions, and marching withdrawals over longer distances. The chapter will then

conclude by summarising the discussion and outlining the relationship between generalship

and the tactics of successful retreats.

Chapter Five, ‘The Development of Roman Retreat Tactics from the Middle to the Late

Republic (218–46 BC)’ is an in–depth analysis of how the Roman army changed and

improved its retreat tactics over time. The chapter will examine these developments

throughout three general periods: the late third to mid second century BC, the mid to late

second century, and finally the first century BC. Numerous examples from each period will

be utilised along with a handful of detailed case studies to demonstrate the tactics that the

Romans utilised to carry out successful fighting retreats in each period. The relationship

between gradual changes in the Roman army’s organisation and tactics during these periods

and the Romans’ conduct of retreats will be explained, culminating in the highly sophisticated


41

retreat tactics utilised during the campaigns of Caesar and others in the mid–first century BC.

Having assessed these developments by the mid first century BC, the final chapter,

‘Disastrous Withdrawals in the Late Republic and Early Empire’, will examine how even the

most advanced Roman retreat tactics failed under certain conditions. The chapter features

three extended case studies. The Roman withdrawal following the defeat of Crassus at

Carrhae in 53 BC, Antonius’ withdrawal during his Parthian campaign of 36 BC, and the

movements of the Roman army in the clades Variana of AD 9 will all be examined in detail.

The relative success of Antonius’ withdrawal from Parthia in 36 BC will be juxtaposed with

Crassus' retrograde movements following the Battle of Carrhae in 53 BC. The complex

evidence for the clades Variana will then be analysed, and the event characterised as a

withdrawal ‘gone wrong’. Further discussion of these and other events will throw light on the

tactical and strategic limitations of even the most sophisticated retreat tactics.

The Conclusion will summarise and draw together the preceding analyses and arguments to

demonstrate the role played by retreats in Republican–era warfare, as well as the impact of

retreats and withdrawals on contemporary tactics and generalship. I will then be able to give a

clear assessment of the role played by retrograde movements in terminating battles and

campaigns in Republican Roman warfare. Following this the continuing role of retreats in the

Imperial period will be briefly examined and the thesis brought to an end with an outline of

the relevance of this thesis to current scholarship and suggestions for further study.

Method

As a result of limitations in applying the ‘face–of–battle’ approach to this thesis, and due to

the limited evidence for combatant experiences in ancient warfare, my key methodology will

be re–examination of the ancient evidence for retreats and withdrawals, looking for common

factors as well as differences in tactics.165 I will begin with the selection of a number of

typical and relatively detailed examples, and the context and content of each will be

considered in order to make salient deductions that will contribute to the overall argument.

This approach will be mainly utilised in Chapters One to Four, where a large number of

examples and accounts are considered in turn. By the end of Chapter Four, a general picture

will have emerged of what tactical and leadership factors contributed to the success or

165 Cf. Wheeler, 2001, 170–4; Lendon, 2004, 443, 447.


42

otherwise of a retrograde manoeuvre.

In Chapters Five and Six there are several larger case studies. There I will utilise a similar but

deeper type of analysis, in which I attempt to explain in detail why specific retreats turned out

the way they did, even where this is not necessarily clear from the sources. I will consider in

more detail the context and content of the sources in order to answer questions that had been

raised by the success or failure of these retreats, and the sources’ depiction of such. I will

carry out these case studies with reference to the factors that contributed to successful

retreats, as identified throughout Chapters One to Four.

I will borrow from the ‘face–of–battle’ approach the general emphasis on the importance of

combatant psychology in determining the course and outcome of battle, incorporating the use

of examples, but I will avoid taking a combatant’s eye–view of affairs. Modern studies of

combatant psychology, a feature of the ‘face–of–battle’ approach, will be utilised in a very

limited fashion, with the main evidence for matters of morale and motivation being drawn

from the ancient sources.166 In line with the approaches of Wheeler and Kagan, the

importance of tactics and manoeuvre will be emphasised in determining the outcome of

battles, rather than only the combatants’ fear of their opponents’ weapons.167 In the detailed

case studies, I will utilise a more traditional approach to military history, emphasising the

analysis of terrain and the manoeuvre of units, in conjunction with the matters of morale in

the ‘face–of–battle’ approach, to avoid overemphasising the perspectives of either the

individual or the commander. However, the nature of the evidence, written as it was by the

officer class for a similar audience, will necessitate a slightly greater emphasis throughout the

thesis on the roles of generals, and their decisions as to how and why retreats should have

been carried out.

166 The limitations of applying Keegan’s approach to ancient warfare have already been considered, but as pointed out by A.B. Lloyd (1996, 194), the psychological reactions of ancient combatants were perhaps not very different to their modern equivalents. 167 See Keegan, 1976, 87. Cf. Wheeler, 2001, 170, 174 and Kagan, 2006, 19–22, 189–91.


43

Translations

All original Latin and Greek is drawn from Loeb Classical Library editions, sometimes in

hard–copy but mostly from where I have had access to them online. Translations, if not my

own, are credited to the translators. This material is almost entirely drawn from either

lacuscurtius.edu or perseus.edu. For the most part, the translations of Polybius, Caesar,

Sallust and Plutarch are taken from lacuscurtius, courtesy of Bill Thayer, and Livy from

perseus, courtesy of the site editor Gregory Crane and his staff.

Conclusion

In this chapter, it has been argued that retrograde movements have been overlooked by

modern scholars as a factor in ending battles and shaping tactics in ancient warfare. Modern

scholars have assumed that ordered retreats were carried out only occasionally, and that when

they did occur, they were directed towards fortified camps or some other safe location, but

there has been no detailed study of how often, or by what means retreats were carried out, let

alone how they affected the strategies and tactics of armies in the Middle to Late Republic.

Withdrawals have received just as little attention, or perhaps even less. As such, the purpose

of this thesis is to study retreats, and the tactics associated with the successful conduct of

retreats, to shed new light on contemporary Roman tactics and generalship. The

argumentation of the thesis will aim to answer one primary question: how did retreats and

withdrawals impact on the tactics and strategies used by armies during the Middle and Late

Roman Republic? The following chapters will seek to answer this question systematically.


44

But seeing that both their flanks had been forced off the field, they despaired of giving help

there and of returning to their camp ... They kept, however, in close order and retired on

Placentia, being not less than ten thousand in number.

(Polyb. 3.74.5–6)

Chapter One

Retreat and Withdrawal: Why and Where To?


45

Introduction

On the bank of the river Trebia in 218 BC, a Roman army suffered a crushing defeat to

Hannibal Barca. Much of the Roman army was killed or scattered in panic, but even while

this was occurring, the Roman centre had broken through their Carthaginian opposites and

pressed forward to relative safety. These ten thousand men were able to make an orderly

withdrawal to the nearby town of Placentia (Polyb. 3.74). Why did they do this, and what else

could they have done?

The first step to analysing retreats and withdrawals in detail is to contextualise their place in

military operations, tactically and strategically. To do this, it is necessary to explain when and

why they occurred, and to place them into a physical context by identifying and examining

the various places to which armies retreated and withdrew. It has been pointed out in the

Introduction that retreats are assumed to have been carried out usually towards camps, or

sometimes towards high ground or cities, but this aspect has not been studied in detail.168 It

has also been mentioned that retreats will generally be distinguished from withdrawals

according to the distance moved and the formation adopted by the army. This chapter will

seek to expand upon these matters by examining the reasons why armies retreated or

withdrew, where they sought refuge, and elaborating on the distinctions between retreats and

withdrawals.

The main purpose of this chapter is to establish the contexts in which retreats and

withdrawals were carried out, and to clarify how different types of refuge, and the distance

from such refuge, affected an army’s retrograde manoeuvres. The analysis will be made in

three parts. In Part A, I will examine several different reasons for why armies carried out

retreats. I will begin with the most obvious scenario, retreats from battle as a result of tactical

defeat. I will next consider more varied circumstances that caused armies to desist from battle

or to retreat, such as stalemates, nightfall and poor weather. Finally, I will examine feigned

retreats, and their use by the Romans. Then, in Part B I will analyse a number of accounts to

show the various types of refuges to which armies could retreat, particularly high ground,

camps and walled towns or cities. Finally, in Part C I will examine scenarios in which an 168 While Sabin (2007, 413–4) only mentions camps, Gilliver (2007, 138) is more detailed, pointing out that armies could retreat towards camps, high ground, or cities. However, she only gives a handful of sources for this statement (cf. BAfr. 85; BHisp. 31; BAlex. 40).


46

army needed to withdraw a longer distance in search of a safer position, and distinguish

between such manoeuvres and retreats. In the conclusion, I will establish how the Romans

generally carried out retreats and withdrawals in response to particular situations, and the

reasons that they chose particular places of refuge. This discussion will ultimately aid in

defining retrograde manoeuvres and clarify the differences between retreat and withdrawal.

Part A: Why Armies Retreated

Retreats Due to Defeat in Battle

I have argued that battles commonly ended with retreats, in particular, as battle usually ended

with a victor and a loser, it is straightforward to imagine that one side might retreat after

having been defeated and forced back from its battleline. Sabin sums up the matter as

follows:169

The first priority of troops in the losing army was to escape their vulnerability to one–

sided slaughter ...

Sabin also points out that there was some natural ebb and flow in the way the lines pushed

each other back and forth during a battle, but what he describes above is a complete

retirement from the battlefield as the defeated army sought to escape its enemies.170 The

ramifications of defeat on the battlefield could be bloody for the losing side, and so armies

tried to avoid this fate by retreating if defeat seemed likely or assured.171 Consider the

following extract from Appian’s account (BCiv. 4.128) of the Second Battle of Philippi, 42

BC, as the legions of Brutus fought a losing battle against the Caesarians:

169 Sabin, 2007, 413. 170 Ibid., 412. 171 Ibid., 413, 416. The majority of accounts of retreats paint this scenario, in which an army is clearly beaten on the battlefield and deliberately attempts to move away from the enemy (see App. B).


47

Finally, the soldiers of Octavian ... pushed back the enemy's line (τὴν φάλαγγα

τῶν ἐχθρῶν ἐκίνουν) as though they were turning round a very heavy machine

(ὥσπερ τι μηχάνημα τῶν βαρυτάτων ἀνατρέποντες). The latter were driven

back step by step, slowly at first and without loss of courage (τὸ ὀπίσω βάδην ἔτι

καὶ μετὰ φρονήματος). Presently their ranks broke and they retreated more rapidly

(ὀξύτερον ὑπεχώρουν), and then the second and third ranks in the rear retreated

with them (συνυποχωρούντων), all mingled together in disorder (ἀκόσμως),

crowded by each other and by the enemy, who pressed upon them without ceasing

until it became plainly a flight (ἔφευγον).172

As Appian describes, the initial retreat by Brutus’ men was gradual and orderly, but became

more disorganised as the infantry attempted to move more quickly and the lines intermingled,

until they were clearly in flight. Appian describes the breakdown of this retreat as a gradual

process rather than a sudden implosion of morale or order. This account makes clear that

Roman forces in this period were capable of conducting orderly retreats, even as victorious

enemies continued to press upon them. Although the forces of Brutus were driven into their

camp thoroughly beaten, and later surrendered en masse (App. B Civ. 4.128), they were not

broken and slaughtered in the field. Their retreat, therefore, gave them the chance of survival,

and so was to some extent a success. However, it was by no means a voluntary manoeuvre,

but was forced by the pressure of the victors, and it must also be observed that even such an

orderly retreat could still collapse quickly.

Did an involuntary retreat always signify the end of a battle?173 In Caesar’s account of the

Battle of Ilerda, 49 BC (BCiv. 1.45), he describes how his men were repulsed in fighting on

high ground near the strategic town of Ilerda:

172 Trans. Horace White, 1899. 173 In some cases it obviously could, such as in the example above. The Spartans were heavily beaten when forced to retreat from a hill by the infantry of Antigonus III at Sellasia in 222 BC (Polyb. 2.68). The Ptolemiac general Nicolaus was also defeated at the pass of Porphyrion c.218 BC due to losing control of the high ground (Polyb. 5.69).


48

... Caesar exhorts his men and leads the Ninth Legion to their support. He checks the

foe who are pursuing our men with insolent daring, and compels them again to turn

and retreat to the town of Ilerda and halt beneath the walls (rursusque terga vertere

seque ad oppidum Ilerdam recipere et sub muro consistere cogit). But the men of the

Ninth Legion, carried away by zeal in their desire to repair the loss received, rashly

pursuing the flying foe too far, get into unfavourable ground (temere insecuti longius

fugientes in locum iniquum) and approach close under the hill on which the town of

Ilerda was situated. When our men wished to retreat (recipere) from this position, the

enemy in turn kept pressing them hard (ex loco superiore nostros premebant) from the

higher ground.174

The Pompeians were repulsed from one position, but established a new fighting line on the

slope leading up to Ilerda. The fact that the Pompeians were able to reform their battleline

implies that there was time and space for them to do so, further implying that Caesar’s men

were not pursuing so closely that the former were unable to manoeuvre and reform. The

Pompeians may have been able to retreat into the town itself if they so wished.175 As such,

their retreat from one hill to another appears to have been a tactical decision, carried out with

the deliberate intent of continuing the battle. In this case, the Pompeians clearly did not

retreat voluntarily (rursusque terga vertere seque ... recipere), and yet their retrograde

movement allowed them to continue fighting with genuine success.

As such, involuntary retreats in response to the advance of a stronger army certainly

occurred, just as Sabin and others have described. However, it is also evident that an

involuntary yet orderly retreat from the battleline could not only prevent mass slaughter, but

could even present the opportunity for a beaten army to renew the fighting. The chance of

avoiding a rout or slaughter, and perhaps even of re–engaging successfully, is evidently the

main reason that an army would retreat from battle. Defeat in battle was by far the most

common reason for a retreat, and such incidents will be examined in greater detail in the

following chapters.

174 Trans. A. G. Peskett, 1914. 175 Standing and fighting in front of fortifications could actually make it more difficult to retreat into the camp if it became necessary (cf. App. Mith. 42–4).


49

Retirements from Indecisive Battles

A retreat might occur if one army was defeated in battle, or if an army wished to improve its

tactical position, but what happened when a battle was indecisive or interrupted by natural

phenomena, such as darkness or weather?176 In the case of an indecisive battle, both armies

had to retreat from their fighting positions at some stage.177 However, it is less obvious

whether both armies withdrew simultaneously, or one before another. Also, the reasons for

retreat in the case of a stalemate cannot be fully explained without establishing why the battle

reached a stalemate in the first place. To clarify these matters, two examples of stalemated

battles ending in retreats will be examined.

In Livy’s account (10.35.1–3) of a battle in Apulia during the Third Samnite War, 294 BC, he

describes action between a Roman army led by Marcus Atilius and a Samnite force that was

besieging the Apulian town of Luceria:

The ... consul, Marcus Atilius ... was marching, at the head of his legions, towards

Luceria, which he had heard was being besieged by the Samnites, when the enemy

met him at the Lucerine frontier. On this occasion rage made their strength as great as

his, and the battle was one of shifting fortunes and doubtful issue

(proelium uarium et anceps fuit), yet its outcome was more discouraging to the

Romans, both as having been unaccustomed to defeat, and because, as they were

retiring (digredientes) from the field, they could see, even better than during the actual

engagement, how much their side had got the worst of it in killed and wounded.178

176 Darkness: Caes. BGall 1.50; Livy 4.39, 9.23, 32, 10.12, 27.2, 38.40, 44.4; Plut. Crass. 27.3, 28.1; App. Hisp. 82 Weather: Livy 8.1, 28.15. It should be noted here that battles did not always take place even after armies drew up their acies, as armies would commonly face each other as a form of demonstration, then withdraw without fighting, eg. Livy 28.14, 37.38. Goldsworthy (2011, 175) discusses this phenomenon briefly. However, retreats following such non–engagements will not be discussed below. 177 In circumstances of great tension, armies could remain in battle positions overnight. The Persians did exactly this at Gaugamela in 331 BC (Arr. Anab. 3.11). As Arrian points out, this will have fatigued the soldiers greatly and so armies must have generally have avoided doing such a thing. 178 Trans. B. O. Foster, 1926.


50

As pointed out by Oakley, this account of the indecisive fight is sparse, the only real detail

being that the Romans came off worse.179 Livy does not describe it as a retreat as such, but

his use of the verb digredi suggests that he considered it somewhat organised and not a flight.

This is suggested further by the fact that the battle was indecisive (uarium et anceps), giving

the Romans no reason for flight. Why there was a stalemate is not clear, but Livy mentions

(10.35.3–4) that after retreating the Romans spent an anxious night in their camp, leaving

open the possibility that the battle was left undecided due to encroaching darkness or that the

battle was of long duration and that the combatants were tired. Both armies appear to have

been psychologically exhausted in its aftermath and were reluctant to continue fighting the

next day (10.35.3–36.4). Casualties seem to have been heavy.180 Thus, the reason for the

stalemate appears to have had something to do with the battle having been particularly

arduous and extended, so that both armies retreated in a state of exhaustion without

interrupting one another. Whether due to exhaustion, casualties, the time of day or a

combination of these things, the apparent reluctance of the combatants to continue fighting

was the clearest reason for the retreat.

It is easy to imagine embattled and stressed soldiers would often have wanted to retreat from

fighting, but the decisions of generals must also have played some role in the outcome. For

example, consider Livy’s account (25.19.1–8) of a battle fought at Capua between Hannibal

and the Roman consuls of 212 BC. The Romans had been ravaging the Campanian region

when Hannibal arrived (25.18.1), and the latter drew his army up with the intention of beating

the Romans decisively (25.19.1–2). Hannibal ordered a cavalry attack against the Roman

infantry, which was then being hard–pressed by missiles, but the Romans counterattacked

with their own cavalry, so that the cavalry of both sides were engaged (ita equestre proelium

erat), while, presumably, the infantry were left unengaged (25.19.3–4). However, an

unidentified Roman army then made an appearance (29.19.4–5):

... the distant sight of the Sempronian army, commanded by Gnaeus Cornelius, the

quaestor, inspired in both armies the same fear (parem metum) that fresh enemies

(hostes novi) were approaching. As if by agreement (velut ex composito) the signal for

recall (signum receptui datum) was given on both sides, and marching back to the 179 Oakley, 2005, 362. Livy builds the account more around the fear each side had of each other in the aftermath and the divine intervention that brought final victory to the Romans (Livy 36.11–12). 180 Livy 10.35.2, 10.35.7.


51

camp (reductique in castra) they separated on almost even terms.181

It is clearly stated by Livy that both sides retreated at the same time. A signal for the retreat

was evidently given by both armies. This must have been ordered by the respective generals.

There is no suggestion that the retreat was anything but orderly, no doubt helped by the fact

that the infantry appears to have been unengaged, and it must have occurred quickly, before

the approaching Sempronian army was identifiable. The reason for the retreat according to

Livy was that both armies (or rather their commanders) feared that the unidentified army

could attack them while they were otherwise engaged, and so they called off the battle rather

than gamble that the newcomers were friendly. It is clear from this example that in some

circumstances generals could call off a battle already in progress, and that this could happen

almost by mutual consent when the opposing commanders shared the same fears.182

Disengagements at Nightfall

According to Sabin, fighting could be ‘ended indecisively by nightfall’, and although this

occurred ‘rarely’ as battles seldom lasted so long, armies would typically disengage upon the

onset of darkness.183 According to Gilliver; ‘If flight were not impeded, it would normally be

continued for as long as possible, until natural obstacles or nightfall made further pursuit

impossible.’184 These two scholars thus express the view that battles usually ended at

nightfall and that darkness prevented pursuit, but we still need to examine the questions of

when, why and how armies retreated at night.

181 Trans. Frank Gardner Moore, 1949. 182 For a similar example, see Polyb. 3.103–5. 183 Sabin, 2007, 399–400, 413. Also see Polyb. 35.1. Battles did often last several hours (Goldsworthy, 2011, 184), and so must have typically started earlier in the day (eg. Caes. BGall. 5.35). In 58 BC, the German king Ariovistus attacked Caesar's camp for several hours, and withdrew easily at nightfall as there was apparently no threat of pursuit from the Romans (Caes. BGall. 1.50). 184 Gilliver, 2007, 138.


52

A detailed example is found in Sallust’s account of a Romano-Numidian battle fought in 106

BC.185 This battle was fought between Gaius Marius and the African kings, Jugurtha and

Bocchus. The Africans had caught the Romans by surprise near nightfall while the latter were

on the march.186 Sallust describes the battle as follows (Jug. 98.1–5):

Marius was neither frightened nor less confident than before, but with his bodyguard

of cavalry ... he went from place to place, now succouring those of his men who were

in difficulty (laborantibus suis succurrere), now charging (invadere) the enemy where

they were pressing on in greatest numbers. He directed the soldiers by gestures (manu

consulere), since in the general confusion his orders could not be heard. And now the

day was spent, yet the barbarians did not at all relax their efforts, but thinking that

darkness would favour them, as the kings had declared, they attacked with greater

vigour. Then Marius, adapting his tactics to the situation and wishing to provide a

place of refuge for his men (atque uti suis receptui locus esset), took possession (inter

se occupant) of two neighbouring hills ... he himself gradually rallied his scattered

forces and the enemy were in no less disorder, and then led them all at the double

quick to the hill. Thus the kings were compelled by the strength of his position to

cease from battle.187

Marius managed to organise his forces and retreat to higher ground from which his army

could better defend themselves (Sall. Jug. 98.1–2), but the process of organising this

withdrawal was seriously hindered by the darkness and the scattered dispositions of Marius'

army, and it took him some time to carry out the retreat (Jug. 98.4). Marius was assisted by

the fact that the Africans were also in confusion due to the darkness (98.4), and he

successfully retreated to improve his tactical position, and after this the Africans withdrew

from the fight. It is striking that the Africans, who chose to attack at night and were familiar

185 Sall. Jug. 97–8. Sallust is not clear on the site of the battle, only pointing out it was not far from Cirta (Sall. Jug. 101.1, 102.2). 186 Sall. Jug. 97.3. Attacking so close to nightfall was clearly a calculated risk on the part of the Africans, as the confusion caused by darkness affected the Africans as much as it did the Romans (Jug. 98.4). Paradoxically, in taking possession of strong positions on nearby hills, it was Marius that made better use of the ground (98.3). However, it is important to note that Jugurtha apparently gave battle just before nightfall with the intention that his army could escape more easily if the battle went awry (Front. Str. 2.1.13). It is also striking that the Romans were marching so late in the evening, which Paul (1984, 239) was a deliberate attempt to avoid the heat of the day. 187 Trans. John C. Rolfe, 1931.


53

with the ground, still fell into disorder.188 Even though Marius was a very capable general, it

took him significant time and effort to organise his army and move it onto the nearby hills

(98.1–4). The process of reconnoitring the hills for water, forming a plan, ordering the

cavalry into a reserve position nearby, and gradually gathering his forces could have taken

hours, as Marius had no control over the bulk of his army but was forced to move between

groups of men and personally instruct and lead them (98.1). The Romans experienced all of

these difficulties even though part of the battle was carried out in light of dusk.189 It can be

observed that, although it was evidently not impossible, fighting at night was difficult and

managing an organised retreat in such circumstances was a hard task, even for a good

commander with experienced troops.190

Consider too Livy’s account (4.39.4–9) of the Battle of Verrugo allegedly fought in Latium in

423 BC between the Romans under Gaius Sempronius Atratinus, and the Volsci.191 A long

and difficult battle between the Romans and Volsci had escalated to the point that some

Roman equites had dismounted to fight as pedites (4.38.2–5). The dismounted troopers,

commanded by a decurion named Tempanius, ended up separated from the legions (39.1–3),

and were forced to withdraw to a hill and defend themselves there.192 Meanwhile the main

Roman and Volscian armies continued to fight until darkness fell:

The battle did not end till nightfall. Neither did the consul relax his efforts anywhere,

but kept the enemy engaged (hostem tenuit) as long as there was any light. Darkness

put a stop to the indecisive struggle (nox incertos diremit), and the terror in each camp

was such, in consequence of men's ignorance of the outcome, that both armies,

abandoning their wounded (relictis sauciis) and a good part of their baggage, retreated 188 Sall. Jug. 98.4. Fighting at night was evidently disorientating (Goldsworthy, 1996, 159–60), even if one was familiar with the ground. 189 Marius could not have given signals by gestures (manu consulere) if there was not some light (Sall. Jug. 98.1). 190 Not all of the Roman troops were experienced, but some are said to have been veterans, while the recruits appear to have conducted themselves well (Sall. Jug. 98.5: Denique Romani veteres novique ... et ob ea scientes belli, si quos locus aut casus coniunxerat, orbis facere ... ) 191 According to Ogilvie (1965, 592), the bare facts of this defeat could be grounded in fact but the story of the battle and the specific roles of Sempronius and Tempanius are grounded in family legend. Themes have been borrowed from the defeat of a later Sempronius at Trebia in 218 BC (cf. Livy 21.52–4), and the dismounting of the cavalry is from Cannae (22.49.3). The retreat of the dismounted cavalry to a hill is another trope, but it is exactly so because it is logical behaviour for men in retreat, as will be shown below. 192 Livy 4.39.5. Ogilvie (1965, 593–4) suggests the real name of the cavalry commander should be Tampanius. There is memorial evidence of his existence (Ogilvie, 1965, 592).


54

to the nearest hills (se in montes proximos reciperent), as though defeated.

Nevertheless the mound was besieged till after midnight. But when word was brought

to the besiegers that their camp was abandoned, they too supposed that their side had

been defeated, and every man fled (fugerunt) where his panic led him in the

darkness.193

Livy does not describe a retreat from battlefield to camp, leaving it to be assumed that this

occurred. He does, however, distinguish between the retreat (se ... recipere) of both armies

from their camp to nearby hills, and the flight (fugere) of the Volscians still attacking

Tempanius and his men. The most striking feature of this account is the confusion in which

both the Romans and Volsci operated. Both armies, having retreated to their camp in the usual

manner, then withdrew to nearby high ground, believing themselves defeated. Although Livy

uses recipere to describe the withdrawal, it was evidently a disorganised and panicked affair,

as the Romans even abandoned equipment and wounded in their camp (4.39.9). It was not

until daylight that anybody, in this case Tempanius and his cavalrymen, could make sense of

what had occurred. As such, darkness created such uncertainty for all involved that the level

of danger an army was in at any time was not clear, which could easily stimulate panic, and a

retreat or withdrawal by night could evidently lead to equipment and men being abandoned.

Retreating or withdrawing in darkness was, therefore, dangerous and many commanders will

have been reluctant to give orders to do so.

Simply put, it appears to be true that battles were rarely long enough to still be continuing by

nightfall, and that only a minority of battles ended due to the onset of darkness, as reported

by Sabin. The onset of night usually did end any extended battles, probably because armies

easily became disordered at night and commanders will have sought to minimise the risk of

panic.194 Night was dangerous to both victors and vanquished, and so it is also true that an

army retreating at night would not always be pursued, as Gilliver has said.195 However, the

Romans appear to have been driven by practical considerations in deciding whether or not to

193 Livy 4.39.5–7. Trans. B. O. Foster, 1922. Valerius Maximus (3.2.8) also provides an account of this battle, which supports the broad point that the armies withdrew (abierunt) after nightfall ended the battle (nocte dirimente), with victory not clearly decided. 194 Cf. Caes. BCiv. 1.67. Caesar even discourages foraging at night (BCiv. 1.59). 195 Gilliver, 2007, 138. Note the Parthian fear of operating by night (Plut. Crass. 28.1; Dio 40.24.2). They did not pursue the broken Roman army after the Battle of Carrhae, 53 BC, until daylight returned (Plut. Crass. 28.1).


55

operate at night, for, contra Gilliver, there are numerous examples of such occurring.196

Frontinus (Strat. 2.10) recommended using night to one's advantage, for instance by attacking

an enemy camp while the enemy celebrated a victory. It must be noted that most examples of

Romans attacking, pursuing or retreating by night are from the Late Republic, when Roman

armies and generals were generally more efficient than their Middle Republican

predecessors.197 Of course, even in cases where operating at night was advantageous, there

was significant risk involved. Regardless of the period, then, armies will have been reluctant

either to retreat or pursue by night, although skilled generals with more experienced troops

were more likely to attempt such things. Armies retreated at night only because they had to,

fearing a greater disorder by continuing to fight in darkness than might be caused by

escaping.

Disengagements Due to Poor Weather

Finally, there are a handful of accounts of battles being affected by especially poor

weather.198 According to Livy (28.15.11) in 206 BC at Ilipa, after the Carthaginians had been

forced into their camp and the Romans began to attack it, battle ceased because of a

downpour so heavy that the Romans struggled to return to their own camp (vix in castra sua

receperint se). Livy also records (8.1.4) that a still–undecided battle between the Romans and

Volsci in 341 BC was broken up by a storm (cum tempestas eos neutro inclinata spe

dimicantes diremisset), but he goes into little further detail. In his account of the clades

Variana in AD 9, Dio (56.20.3, 21.3) states that heavy rain and wind made it difficult for the

Roman legionaries to stand or to move securely, or to use their weapons effectively. On the

other hand, bad weather was a literary trope that could be used by authors for several

196 Eg. Caes. BGall. 1.26, 7.81–2, 88, BCiv. 1.51; BHisp. 9; Sall. Jug. 99; Livy 25.23, 30.6; Plut. Crass. 27.2–8; App. B Civ. 1.92–3. It has been suggested that the Romans disdained night fighting, as they did not consider it honourable (Gilliver, 2007, 138; Dawson, 1996, 144.). However, Caesar's pursuit of the broken Gauls at Alesia as late as midnight (Caes. BGall. 7.88) shows that the Romans could launch pursuits at night if they wished. There are other examples of fighting continuing past nightfall, such as Caesar's attack on the Helvetian camp at Bibracte in 58 BC (Caes. BGall. 1.26), or Sulla's attack on the Marians at the Colline Gate in 82 BC (App. B Civ. 1.92–3). In both cases, the fighting was carried on because it was advantageous to the attackers and a decisive result was within reach. There are several further examples of the Romans conducting attacks by night (Caes. BGall. 7.81–2, BC 1.51, BHisp. 9; Livy 25.23, 30.6). Fundamentally, the Romans in this period appear to have held pragmatic attitudes towards night fighting even after difficult battles in that it was to be avoided in general, but might be done if it was practical and advantageous. 197 Cf. Goldsworthy, 2011, 48. 198 Livy 8.1.4, 28.15.11; Dio 56.20.3, 21.3.


56

purposes and may not have necessarily represented the real reason that a battle was ended.199

Östenberg has argued that the Romans could see natural events such as bad weather to be

active antagonists of the Roman military, which could contribute to or cause defeat.200

However, as Östenberg points out, weather is not an adequate lone explanation for any given

defeat.201 The evidence on the subject is limited, but it is not hard to imagine that the

decreased stability and visibility in bad weather could have increased the risk of an army

becoming disordered. In the light of Roman literary techniques, the evidence is not sufficient

to state that rain or wind could physically stop combatants from fighting, but it could

certainly hinder them, as at Kalkriese (Dio 56.21.3–4), and thus contribute to a calculated

decision by the commander to retreat.

Feigned Retreats

As touched on previously, feigned retirements fit uneasily into the definition of a retrograde

movement, as they were a pre–meditated battle tactic rather than a method of escaping

unfavourable engagement, and they could appear to be disordered even if they were not.

However, by virtue of the fact they were carried out deliberately, it can be argued that they

were organised retrograde manoeuvres, as distinct from a genuine rout. As they could appear

so similar to other retreats from battle and produce similar outcomes, the following

discussion will briefly examine the Roman use of these tactics.

A detailed example is given by Caesar, who recounts a feigned retirement by his legate Titus

Labienus in Gaul, 53 BC (BGall. 6.7–8). Labienus had been attacked by a significant force of

Treveri while isolated with only a single legion (6.6–7). Labienus responded by carrying out

by marching his army out of its camp and away from the enemy, mimicking the uproar of a

panicking army (6.7–8):

199Walsh, 1961, 200–1; Murdoch, 2008, 107; Östenberg, 2017, 253–9. For example, in the case of Ilipa, the weather could have been a patriotic addition by Polybius (11.24.7–9), who was probably Livy’s source, to explain Scipio’s failure to decisively beat Hasdrubal Gisco on the battlefield. 200 Östenberg, 2017, 253–9. 201 Ibid., 259.


57

The rear of the column had scarcely moved clear of the entrenchment when the Gauls

... proceeded to cross the river without hesitation and to engage on unfavourable

ground. Labienus had surmised that this would happen, and to entice them all to his

side of the river he pretended to march as before ... He sent the baggage a little

forward and packed it on some rising ground ... he commanded the troops to wheel

towards the enemy and deploy into line, and, detaching a few troops of cavalry to act

as baggage–guard, he disposed the rest of the cavalry on the flanks … When [the

enemy] unexpectedly saw men whom they believed to be in retreat (fugere) advancing

against them in attack formation, they could not even sustain the assault, and at the

first charge they were put to rout and sought the nearest woods.202

As Labienus faced the prospect of fighting a battle against significant enemy forces (magnis

copiis) and only had a single legion with him (cum una legion), he carried out a feigned

retreat in order to improve his tactical position, with the deliberate intent of fighting a

successful action from that better position (Caes. BGall. 6.7). The manoeuvre was pre–

planned and highly organised, featuring deceptive measures such as feigned confusion

(strepitu et tumultu) in breaking up the camp and giving the appearance of flight (fugae

similem).203 The Romans fought only once they reached a stronger position from which to do

so.204 Evidently, this Roman strategy was highly successful. This case thus demonstrates the

potential utility of carrying out a deliberate retreat to allow an army to fight from a stronger

position. The fact that the retreat was not forced by the enemy’s physical attacks, but was

already planned, gave greater opportunity for the Romans to re–establish themselves in a

better position relative to the enemy. However, Labienus had the luxury that he had not yet

engaged in battle. This deliberate retreat, or more specifically, feigned retreat, was therefore a

premeditated tactic rather than a reaction to enemy success in battle.

202 Caes. BGall. 6.8.1–7. Trans. H. J. Edwards, 1917. 203 Caes. BGall. 6.7.8, 6.8.1. 204 Caes. BGall. 6.8: Labienus counterattacked once the Gauls had reached unfavourable ground (... flumen transire et iniquo loco committere proelium ...)


58

Were feigned flights effective battle tactics? Roman literature shows that the Romans were

certainly well aware of such tactics and the threat they could pose to disorderly pursuers.205

The effectiveness of the tactic was a product of the combination of the physical disorder of

pursuing troops, and the psychological shock of a counterattack by an enemy that the

pursuers believed had been defeated.206 These manoeuvres appear to have been a particular

specialty of peoples who had an affinity with mounted warfare.207 Cavalry were most

effective at carrying out these manoeuvres as they could outpace pursuing infantry easily, and

so it was possible for cavalry to disperse in flight (rather than retire gradually) and reform in

time to counterattack a disorganised pursuer, just as the Numidians did to Roman infantry at

the Battle of the Muthul in 108 BC (Sall. Jug. 50). Highly–skilled cavalry could even perform

this manoeuvre against enemy cavalry, such as Scipio and Masinissa did against the

Carthaginians at Utica in 203 BC (Livy 29.34). Feigned retreats were, therefore, very

effective with cavalry, and so were more readily utilised by peoples such as the Numidians

and Parthians than by the more infantry–focused Romans.

The most important question is whether or not the Romans used such tactics with any

regularity. Frontinus (Strat. 3.11) provides several examples drawn from the Greeks of how,

through guile and trickery, an infantry force could make a feigned retreat work, and

Onasander (Strat. 11) emphasises that any major retreat by an enemy, particularly in

unknown country or on disadvantageous ground, should be held suspect. It was, therefore,

not unknown for infantry–based forces to use such tactics.208 The Romans, however, could be

dismissive of such tactics, considering them unmanly.209

Further, there are no Roman examples, even including that of Labienus given above,

describing infantry making a feigned retreat directly from battle, then turning and counter–

attacking their pursuers, as cavalry could do. In fact, in the case of Labienus, he will have

required quite significant distance between himself and the Gauls to depart his camp, put

205 Sall. Jug 50; Livy 6.24, 29.34; Onas. Strat. 21.9. 206 Effects of disorder: Lendon, 1999, 282–5. Shock: Onas. Strat. 21.9, cf. Strat. 29.34. 207 For instance, the Numidians (Sall. Jug. 50; Livy 29.34) and Parthians (Goldsworthy, 1996, 64–5). 208 It could be argued that the centre of Hannibal's army conducted a sort feigned withdrawal at Cannae in 216 BC (Polyb. 3.115). 209 Dawson, 1996, 144. When withdrawing from Gergovia in 52 BC, Caesar was keen to ensure his escape did not resemble a flight (ne … similis fugae videretur: Caes. BGall. 7.43.6). Also see Livy 5.27.5–8.


59

aside the baggage, and reform a battleline.210 The presence of a steep river between himself

and the enemy army (Caes. BGall. 6.7) will have been a significant factor in giving the time

and space needed to complete these tasks.

Why was this so? The danger of a feigned retreat was that it had to give the appearance of

disorder, or even flight, for an enemy to pursue quickly and rashly, which could create the

sort of disorder among the enemy that a counterattack could capitalise on. It was one thing

for an army to retire prior to engaging in battle, and return to surprise a relaxed enemy (Onas.

Strat. 11). However, when in very close proximity to the enemy during a regular engagement,

there will have been little to distinguish feigned disorder among retreating troops from real

disorder, particularly if a strong pursuit was made. This is yet one more reason that an

infantry force, such as that of Labienus in Gaul, will have required a good amount of space

between themselves and the enemy in order to perform a feigned retreat as a deliberate

stratagem.

There is little evidence then that Roman armies performed these types of retreats regularly. If

an army had sufficient time and space to retire and reorganise itself prior to engaging the

enemy, there could be some opportunity for a pre–battle retreat, or a feigned flight. However,

feigned retreats were more easily and safely performed by cavalry than by infantry, and they

did not suit the measured and orderly tactics of the Roman legions. For infantry the dangers

of retreat, unless it were absolutely necessary, were significant. It will have been safer for a

Roman army in a battleline to stand its ground or, if necessary, to conduct a conventional,

orderly retreat to a safe position, and at any rate it will have been viewed as more honourable

to defeat an enemy in battle without resort to trickery.

210 Cf. Caes. BGall. 6.7–8. How long it took the Gauls to marshal their relatively decentralised forces, approach the river, and then cross it, is unknown but it was clearly long enough for Labienus to march a short distance, move his entire baggage train onto a nearby hill, and rapidly redeploy his army as the Gauls approached. Perhaps the Gauls slowed their pursuit once they saw the Romans reforming, or perhaps Labienus had cut his timing very finely, but as the Romans were able to raise a shout, throw pilum, and advance, the Gauls must have been very close.


60

Summary of Part A

There were several main reasons that retreats could occur. The most common was as a result

of defeat in battle. Such retreats may not have been voluntary, but were a deliberate attempt

to avoid the chaos and mass casualties of a rout, and could even allow an army to re–engage

with greater success. In the case of an indecisive battle, whether this was a result of

exhaustion, casualties, or the caution of the general, one or both armies might also have

retreated. Such retreats could occur simultaneously if opposing commanders shared similar

fears, and tired and fearful combatants on both sides may have been relieved to desist from

further fighting. Also the onset of nightfall, when it occasionally occurred during an ongoing

battle, was another trigger for retreat. Fighting at night could create chaos and panic, and so

the typical reaction of both armies upon nightfall was to retire from combat. Even so, the

Romans did occasionally use night as an opportunity for successful action, particularly if the

army and its general were confident and skilled. Even more rarely, poor weather could also

motivate one or both armies to retreat, as decreased stability and visibility increased the risk

of chaos and panic. Finally, it was possible for a Roman army to perform a feigned retreat or

simulated flight as a battle tactic to lure the enemy into a weak physical position, from where

they could be defeated in a counterattack. However, this was dangerous for an infantry–based

army, as a simulated flight could easily become a real one under enemy pressure, and at any

rate, some generals may not have viewed such tactics as honourable. As such, feigned retreats

were not commonly performed by Roman forces. In general, the Romans would retreat only

when they were forced to by the enemy, or due to the fear of disorder if fighting were to

continue.

Part B: Where Armies Sought Refuge

High Ground

As pointed out by Sabin, the first priority for a beaten army ‘was to escape their vulnerability

to one–sided slaughter’, and that the only way for a ‘broken army to escape…was to seek

shelter within a nearby fortified camp ...’211 Gilliver is slightly more expansive, stating of

Roman soldiers in retreat, ‘if they were able…made for their camp or a nearby defended city,

211 Sabin, 2007, 413.


61

or retreated to high ground in a close formation.’212 Was reaching high ground, a camp, or a

nearby city therefore the standard objective of a fighting retreat? How did these sorts of

positions help a retreating army, and what factors contributed to an army’s choice of one

place over another?

As noted, one option for a beaten army or one that feared defeat was to retreat and renew

battle from a better position, and higher ground could provide just such a position.213 There

are numerous examples of retreating armies using high ground to protect themselves from

pursuit or to continue fighting from a stronger position.214 A handful of these examples will

be examined below in order to show the value of high ground to a retreating army.

Firstly, as Plutarch’s account (Aem. 20.5–6) of the Battle of Pydna in 168 BC reveals, when

the Roman army was initially repulsed by the Macedonian phalanx, it retreated uphill directly

away from the phalanx towards their own camp.215 The pursuing phalanx followed, until

eventually the Macedonian infantry was drawn onto uneven ground and became disordered,

allowing the Romans to make a decisive counterattack (20.7–10). The Romans had only

needed to retreat a short distance before the hill and uneven ground disrupted the phalanx.216

The hindering effect of the slope on their pursuers allowed the Romans to reform a fighting

line and counterattack. In this case, the socii in the front line were retreating involuntarily and

the yet–to–be–engaged legions were deliberately withdrawing along with them, but all the

lines of the army used the high ground to support themselves in retreat. The retreat to high

ground, perhaps inadvertently, was a decisive factor in the eventual Roman victory.

The Romans were not the only ones who could use high ground to even the odds. In 58 BC,

the Helvetii faced Caesar’s army at Bibracte (Caes. BGall. 1.23–6) Caesar’s army was first

deployed on high ground, and the Helvetii were thrown back from their attack lower down

the slope, but the Celts then retired to high ground (1.25), turning the tables on the Romans

who were then forced to attack uphill themselves, which Caesar implied was very difficult.217

212 Gilliver, 2007, 138. 213 Eg. Caes. BGall. 6.7–8, BCiv. 1.45; Sall. Jug. 98.1–2; Livy 4.39.5. 214 Saving an army from pursuit or destruction: Polyb. 3.40, 3.86, 11.24; Caes. BGall. 1.25; BAlex 40; Livy 27.41–2, 28.15.8–11; Sall. Jug. 51, 98; App. B Civ. 1.90. Re–engaging: Xen. Hell. 3.5.19–20; Caes. BGall. 2.23, BCiv. 1.45. 215 Hammond, 1984, 40. 216 Ibid., 40. 217 Caes. BGall. 1.25–6. The length of the battle betrays its hard–fought nature: Nam hoc toto proelio, cum ab hora septima ad vesperum pugnatum sit, aversum hostem videre nemo potuit (BGall. 1.26).


62

When a Pontic army rushed uphill at Caesar’s forces at Zela in 47 BC, the imperator and his

men were caught totally by surprise, yet they repelled the attack decisively (BAlex. 74–6).

High ground could thus confer a major tactical advantage to the defenders.

As a consequence, many armies will have been reluctant to attack uphill for fear of this

occurring, even in pursuit of a retreating enemy.218 High ground was advantageous to

retreating troops for several reasons. Fighting from high ground made things more physically

difficult for the attackers, as they had to move against gravity and fight taller opponents. The

slope also had a double psychological impact, making the defenders more confident while

intimidating the attackers.219 There was, however, the potential drawback that an army

retreating to higher ground could be surrounded and besieged, forcing those on the hill either

to fight their way out or to surrender.220 Even so, high ground gave reasonable hope of

success to an army situated there, and conversely, potential attackers would think twice

before attacking or pursuing uphill. As a result, high ground was a logical place for a

retreating army to seek refuge, and was commonly utilised as such.

Camps

It was also possible that a retreating army would retire to its camp.221 This is not surprising,

as in this period camps were habitually positioned, by the Romans and others, near

battlefields.222 It is worth considering if it was more or less preferable for an army to retreat

to a camp, or to high ground, and under which circumstances one might be more appropriate

than the other. Also, while it is well known that the Romans routinely constructed camps

while on campaign, it is an open question whether these were specifically positioned for the

purposes of protecting an army in retreat. The following discussion will consider these

matters, and examine the role of camps in Roman retreat tactics.

218 BAlex. 40. In Sallust (Jug. 98), when Marius managed to withdraw his army onto higher ground, the enemy refused to engage further (Ita reges loci difficultate coacti proelio deterrentur). 219 These things are routinely emphasised in battle accounts by both Greek and Latin sources: see Xen. Anab. 4.8.10–4, 6.5.12–20, Hell. 3.5.19–20, 5.4.42–5; Polyb. 11.22, 18.22, 25; Caes. BGall. 2.23, 5.35, 6.40, 7.51, BCiv. 1.44–5; BAlex. 40; Livy 5.38, 27.18, 28.2, 32.12, 38.21, 39.31; Sall. Jug. 50–1. 220 Polyb. 3.86; Caes. BGall. 6.40, BCiv. 3.95–8; App. Hisp. 88. 221 Eg. Livy 4.39.5–6, 29.19.4–5; App. B Civ. 4.128. The Romans at Pydna did withdraw uphill, but also directly towards their camp (Plut. Aem. 20.5–6, cf. Hammond, 1984, 40). 222 Sabin, 2007, 414.


63

To answer the given questions, the defensive value of a camp must first be considered. It has

been suggested that the Roman–style marching camp was not designed to act as a defensive

position.223 However, they did have some defensive value and could at any rate be further

fortified.224 In particular, it has been argued that some extra fortification was a normal

procedure for Roman battle camps, as opposed to marching camps that were not necessarily

intended for defence.225 Roman siege camps, such as those excavated at Numantia,

demonstrate a formidable combination of natural and constructed defences.226 Hellenistic

armies, for their part, did use their camps as defensive positions when necessary.227 The

positioning and fortification of a camp was subject to an individual general’s inclinations, so

that some camps will have been better protected than others.228 Their size and layout will also

have varied according to the local topography and composition of the army.229 However, a

well–fortified camp was an exceptional defensive position, as demonstrated by the ability of a

few Roman survivors of the disaster at Atuatuca in 54 BC to fend off the enemy once they

had fled back into their hiberna (Caes. BGall. 5.37). Considering that camps could offer such

protection to a retreating army, and that camps were usually positioned near battlefields, it is

unsurprising that there are many examples of armies using camps as safe havens in the event

of defeat.230 Under what circumstances then, were camps more or less preferable to high

ground, and what was the conventional Roman preference when retreating?231

223 Southern, 2006, 193–4. 224 Goldsworthy, 1996, 112. Caesar (BGall. 5.40) describes the fortification of a Roman camp. To elaborate, Roman camps were generally surrounded by a trench with the soil piled up to form a ‘low rampart’ that was topped with palisades (Keppie, 1984, 37–8; Sage, 2008, 135–7). To further fortify such a camp was relatively simple, though labour–intensive: the trench could be deepened, the rampart could be made higher and reinforced, and towers could be constructed (Keppie, 1984, plate 9). 225 Goldsworthy, 2011, 33; Southern, 2006, 193–4. This was probably more common in the Republican era (Southern, 2006, 192). Vegetius (3.8) specifically recommends that stationary camps (as opposed to briefly occupied marching camps) should be well–fortified. 226 Dobson, 2008, 400–5. 227 Polyb. 11.22, 24; Livy 32.12, 37.43; App. Mith. 44. 228 Note the multiple instances where Carthaginians (Polyb. 11.24; Livy 25.13–4, 39, 28.15) or Hellenistic armies (Polyb. 5.73; BAlex 30–1, 76; Livy 31.41; App. Mith. 42–4, 50) were forced to defend their camps, though with mixed successes. Philip V in particular appears to have made a habit of building very defensible camps on steep ground (Polyb. 18. 20; Livy 32.12). One might be tempted to conclude Roman camps were generally better constructed and defended than those of their opponents (Keppie, 1984, 37–8), and a good Roman general might go to some effort to place their camps in a strong location if there was an enemy nearby (Caes. BGall. 2.17–8). However, evidence of certain debacles (eg. Livy 41.3) shows that the Roman camp did not always prevent disaster. Some generals may not have fortified at all (Sage 135–6). 229 Dobson, 2008, 121. 230 Polyb. 1.57, 18.19; Caes. BGall. 6.40; Livy 10.20, 23.46, 25.37, 27.42, 28.15, 29.26, 37.21, 42, 39.30. 231 Where the destination of an army retreating from battle is specified, there are a similar number of examples of retreats to camps and hills (see App. B).


64

To address this question, the defensive value of camps and high ground must be compared.

Consider the fate of the Romans under Gnaeus Scipio Calvus (cos. 222) at Ilorca, 211 BC.

Prior to this battle, the Carthaginians under Hasdrubal Gisco had destroyed a Roman army

under Publius Scipio (cos. 218), and Gnaeus was in retreat to avoid a similar fate (Livy

25.34–5). Unable to escape pursuing cavalry, Gnaeus chose to withdraw his army (subducit)

to a hill (in tumulum) that, according to Livy, was not high enough to provide significant

protection (non quidem satis tutum) and was sine munimentum (25.36.1–2, 4). Gnaeus

attempted to fortify the hill with a makeshift rampart of baggage, but after some initial

hesitation, the Carthaginians cleared away the baggage and stormed the hill, annihilating the

Roman army.232 Livy emphasises that Gnaeus found the position unsuitable for defence as the

ground was too hard and bare for making any sort of trench or rampart and was not elevated

enough to significantly hinder the enemy ascent (25.36.4–7). Even though Livy refers

(25.36.11) to the Roman position as a camp (castra), it was barely worthy of the name. It is

therefore clear that high ground only offered a strong defensive position if it was high enough

to hinder the enemy’s ascent, or could at least be fortified by a trench and rampart. A fortified

camp thus provided a much stronger defensive position than a hill such as the Romans

attempted to defend at Ilorca.

However, in comparing camps and hills as places of refuge, it must be emphasised that some

examples of higher ground were much more impressive than others and could offer

advantages over a camp. As discussed above, a battleline drawn up on a very steep hill was so

intimidating that most armies would refuse to attack it at all, and if they did, they would

almost certainly be soundly defeated.233 Further, retreating onto a hill was a simple act of an

army moving backwards up the hill, while maintaining its battleline. Conversely, for an army

to enter a walled camp while under attack could be very difficult, because the army might

have to squeeze through relatively narrow entrances, while under attack.234 An army that

retreated to its camp but wished to avoid being attacked at its entrance could choose to stand

and fight outside the ramparts, but if they were beaten in such a place they could be trapped

and annihilated, just as occurred to a Pontic army at Charonaea in 86 BC.235 Finally, an army

232 Makeshift rampart: Livy 25.36.4–7. Appian (Hisp. 16) reports a tower, but his account is so dissimilar as to be discounted. Initial hesitation: Livy 25.36.8. Clearing away the baggage and taking the hill: Livy 25.36.9–11. 233 Eg. Caes. BGall. 1.25; BAlex. 75–6. 234 Pol 14.4–6; Livy 25.39; App. Mith. 44. 235Appian heavily criticised the Pontic general, Archalaus for forcing his men to fight outside the camp after a


65

could be cut off from its camp by enemy forces during battle, or the camp could even be

taken while its parent army was in the field. In such situations it would be more logical for an

army to retreat towards an undefended hill, in contrast to fighting to reach or retake its

camp.236 Therefore, there were situations in which a steep hill that offered significant

defensive value could be a safer option for a retreating army than its camp.

There were other potential issues with retreating into a camp. Even once a retreating army

was inside, an army battered by defeat in the field could still struggle to defend its camp, as

many defenders could be wounded or demoralised, and the attackers, energised by the

promise of a decisive victory.237 Perhaps for these reasons, camps could be (and in many

cases were) taken by storm, potentially leading to slaughter among the soldiers trapped inside

their own fortifications.238 Even an army that was inside its camp might be enticed to leave,

retreating elsewhere, so rendering them vulnerable to attack once more.239 Finally, for

‘barbarian’ armies that populated their camps with civilians, so as not to expose their families

to danger by retreating to such a camp, they might retreat elsewhere.240 Considering these

issues, it is questionable whether camps were always superior to high ground as places of

refuge.

defeat, disastrously trapping them between the enemy and the wall of the camp, as opposed to sheltering in the camp itself (App. Mith. 42–4). 236 Cut off: Caes. BCiv. 2.42. Lost to the enemy: BCiv. 3.95; Livy 32.12; App. Hann. 41. When the Romans were defeated in the Battle of Nicopolis, 48 BC, they retreated to the nearby hills (BAlex. 40), even though there was a camp somewhere nearby (BAlex. 39). The Romans, being surrounded, may not have been able to retreat towards the camp, and the hills also offered a complete retreat from the enemy rather than the possibility of being trapped inside a camp. 237 Sabin, 2007, 414. Note the inability of the Seleucids at Magnesia to defend their camp from Roman attack, despite fierce resistance, after their overwhelming defeat in the line of battle (Livy 37.43). In the Battle of the Nile in 47 BC, Caesar managed to storm the well–defended Ptolemiac camp (BAlex. 31). Also see Caes. BCiv. 3.67, 95; BAlex. 76; Livy 31.41, 33.15, 38.23, 238 Polyb. 5.73, 14.4–6; Caes. BCiv. 3.67, 95; BAlex. 76; Sall. Jug. 99; Livy 24.16, 25.14, 39, 31.41, 33.15, 37.43, 38.22–3, 40.31; App. Mith. 44, 50. 239 Sall. Jug. 38; Caes. BCiv. 3.97–8; B. Afr. 85. According to Appian (Hann. 41), at one point Hannibal used a stratagem wherein he attacked a Roman camp with a small (but very loud) force and had some of his men yell orders in Latin to compel a Roman army to leave their camp and take refuge on a nearby hill, allowing him to attack them as they withdrew from the camp. The Roman general Fulvius Flaccus (cos. 211 BC) saw through the deception (Hann. 42) and managed to keep his army in the camp, defeating Hannibal's attack. An army retreating on the march, whether from a camp or in other circumstances, was particularly vulnerable (see Ch 2, 86–7). 240 At Bibracte in 58 BC, after the Helvetii retreated from their initial assault on Caesar, they retreated to a hill (Caes. BGall. 1.25) even though there was an encampment close by (BGall. 1.26). This encampment, however, was filled with their families (1.29). The sheer size of this encampment probably also prevented the Helvetii from putting it on a hill or fortifying it properly, or perhaps they did not have time.


66

That said, even if it were accepted that steep hills were better destinations of retreat, such a

position had to be readily accessible. A camp could be constructed and fortified in virtually

any location that had access to building material, including in the plains of Mesopotamia, but

finding a hill was not always possible.241 Conversely, camps were routinely positioned in the

vicinity of a battlefield by the Romans and many others in this period, and so were likely

more commonly available as a place of safety than steep hills.242

Further, it was not uncommon for camps to be positioned on hills, whether to provide a good

view of the surroundings or specifically to increase the camp’s defensibility.243 An army

could, therefore, retreat towards its camp while also gaining the advantages of higher ground

as it did so, just as the Romans did at Pydna.244 At any rate, it is reasonable to assume that a

fortified camp in a place that was already naturally fortified (such as a steep hill) was more

difficult to assault than the hill alone.245 For example, Livy (23.26.9–11) records that in 216

BC the Carthaginian general Hasdrubal Barca, while positioning a camp in Spain, halted on a

steep hill (in collem satis arduum) that was even naturally defended by a river

(flumine etiam obiecto), and yet he considered neither the incline nor the river to be sufficient

defence, and so constructed an earthwork (nec aut colli aut flumini satis fidens, castra vallo

permunit). Unsurprisingly, it is evident that a fortified camp on a hill was a stronger defensive

position that either a bare hill or a camp on flat ground. A good general will surely have

positioned his camp on a hill as a matter of habit.246

In comparing camps and hills as places of retreat, there are further examples worth

considering. Caesar (BGall. 6.32–41) records an incident in 53 BC in which one of his

legates, Quintus Tullius Cicero, having been left in command of a single legion to guard the

army’s baggage in a camp (6.32), was attacked unexpectedly by Germans (6.37) while a large

group of men was out foraging (6.36). These foragers, perhaps numbering around two

241 Plut. Ant. 38.3, 39.6. On the wood in that area cf. Ant. 38.2. 242 Sabin, 2007, 414. 243 Xen. Hell. 3.5.19–20; Polyb. 11.24; BAlex. 76; Livy 28.17, 32.12, 38.21, 26–7. Also see Dobson, 2008, 15, Fig. 9; Bellón et al., 2016, 104. High ground also gave a good view of the surroundings. Vegetius (3.8) specifically makes the recommendation that camps be situated on high ground. Also see Southern, 2006, 191. 244 Plut. Aem. 20.5–6. On the terrain and position of the Roman camp at Pydna, see Hammond, 1984, 40. 245 Polyb. 18. 20; Caes. BGall. 2.17–8; Livy 32.12. 246 Caes. BGall. 2.18. In excavations of the potential battle site of Baecula, 208 BC (cf. Polyb. 10.39; Livy 27.18), Spanish archaeologists have found that the Romans and Carthaginians both constructed camps at different times on a strategic piece of high ground in the area (Bellón et al., 2016, 73, 104).


67

thousand legionaries with numerous (magna ... multitudo) camp followers (calones), were cut

off from their camp by the Germans.247 The camp was still nearby, but even closer was a hill

that was not blocked by Germans (6.40). The calones immediately attempted to seek refuge

on this hill (in proximum tumulum procurrunt), but were thrown back by a German attack and

so joined up with the heavy infantry (6.40). As follows:

Some of [the legionaries] proposed to form a wedge and break through speedily, as

the camp was so near at hand, feeling confident that if some part were surrounded and

slain, yet the remainder could be saved. Others proposed to take post on the ridge

(alii, ut in iugo consistant) and all face the same risk together. This course was not

approved by the veterans ... [who] broke through the midst of the enemy and came

into camp (in castra perveniunt) all safe to a man. Close behind them in the same

onrush came the camp–followers and the cavalry, who were saved by the valour of

soldiers. But the party which had taken post on the ridge, still utterly untutored in the

practice of war, had proved unable either to abide in the place they had approved and

defend themselves on higher ground, or to imitate the vigour and the speed which

they had seen to be of assistance to others; indeed, in the attempt to retire into camp

(in castra recipere) they came down to unfavourable ground ... Part of the soldiers ...

arrived, though they did not expect it, safe in camp; part were surrounded by the

natives and perished.248

The veteran component of the soldiers immediately chose to make for the camp, and this

group and those who followed them were successful. Those who retired to the hill were

forced to abandon this position and rush for the camp regardless, and many died doing so. It

is clearly Caesar’s judgement, through the medium of the veteres ... milites, that this was the

ideal course of action. That there was serious risk in this course of action is clearly admitted

by the veterans (et si pars aliqua circumventa ceciderit, at reliquos servari posse confidunt).

However, the alternative course of action, fighting from the hill, is shown by Caesar to have

been a mistake not only through the course of events, but through the fact that it is the

calones and inexperienced men (nullo etiam nunc usu rei militaris) who recommended it. It is

striking that the veterans recommended the first course of action, and not the easier route of 247 Numbers of men: Caes. BGall. 6.36. The Roman cohorts, being in the midst of a hard campaign, were probably not at full strength. Cut off by Germans: BGall. 6.39. 248 Trans. H. J. Edwards, 1914.


68

fighting from the hill, which was readily accessible and was clearly more defensible than the

flat ground (iniquum ... locum). It is thus the judgement of Caesar that soldiers who were in

danger should have striven to reach their camp, even if they had to fight their way through a

ring of enemies to get there.

Was Caesar’s view that the camp was the ideal place of retreat for a beaten army

representative of conventional Roman military wisdom? In his fictionalised account of the

Battle of the Allia, Livy (5.38.1) specifically criticised the Roman officers for not setting up a

camp or fortified position in the case of retreat:

There the tribunes of the soldiers, without having selected a place for a camp or

fortified a position to which they might retreat (… non loco, castris ante capto, non

praemunito vallo quo receptus esset) … drew up their line with the wings extended to

prevent being outflanked by the numbers of the enemy; yet could not stretch their

front as wide as his, though they thinned it till the centre was weak and scarce held

together. There was a little eminence on the right which they decided to occupy with

their reserves, a measure which, though it was the beginning of their panic and flight,

was also the sole salvation of the fugitives.249

Debate about Livy’s military credentials aside, and he was certainly not a military expert in

any way comparable to Caesar, but even so Livy (or his source) considered the positioning of

a camp, or the fortification of some position for retreat, to be a standard procedure in

preparation for a battle and castigates the tribunes for not having done so.250 Note that Livy

comments on the presence of high ground near the Roman battleline, but does not suggest

that this should necessarily have been the location for the camp. He approves of the decision

to post reserves there, because the strength of that position gave relief to the rest of the army,

even though the latter simply used the opportunity to flee (5.38.3–7). Livy thus suggests that

it was common knowledge that a general should construct a camp near a battlefield

specifically for the purpose of giving a place of refuge in the case of retreat.

249 Trans. B. O. Foster, 1924. 250 On Livy as a military authority, Walsh (1965, 4, 157–8) considers Livy to completely lack military experience or nous, whereas Koon (2010, 23–6) considers him sensible and perhaps even experienced. Either way, he can provide some insight into what was considered contemporary good sense.


69

Tacitus also provides a telling account (Ann. 2.46) of a battle between the German leaders

Arminius and Maroboduus in 17 AD. The two armies fought a battle with an unclear outcome

and separated, so that although another battle was expected (sperabaturque rursum pugna),

Maroboduus withdrew his camp to some nearby hills (castra in collis subduxisset), and this

apparently demonstrated his defeat (id signum perculsi fuit). Tacitus (Ann. 2.46) also implies

that this move was cowardly (id signum perculsi fuit), as Maraboduus’ men abandoned him

soon after this. According to Tacitus then, moving one’s camp away from its position near the

battleline, to higher ground, signified defeat. He thereby implies that the camp was the proper

place to retreat after a reversal, and that high ground was a secondary option.251

One last example will further illustrate this point. At the Battle of Thapsus, fought in 46 BC

between Caesar and a Pompeian faction led by Metellus Scipio (cos. 52), there is a detailed

account of the flight of Scipio’s army after its total defeat by Caesar’s forces (BAfr. 85.3–5):

Meanwhile Scipio's forces, now thrown into utter confusion, were in wholesale retreat

(toto campo fugientibus) in every sector of the field, and Caesar's legions promptly

pursued (consequi) them without giving them any respite in which to pull themselves

together (colligendi). When the fugitives reached the camp they were making for

(perfugerunt), with the object of making a recovery there and defending themselves

once more, and of trying to find someone to lead them – someone to look up to, under

whose authority and command they could carry on the fight; when they got there and

perceived that there was nobody guarding it, they forthwith discarded their armour

and beat a hasty retreat to the royal camp (in regia castra fugere contendunt). This too

on their arrival they saw to be in the hands of the Julians. Abandoning all hope of

salvation, they now halted on a hill (in quodam colle consistent) and gave the military

salute by lowering their arms.252

As can be seen, when Scipio’s forces were beaten and were forced from the battlefield, they

immediately fled to a camp, but they found that it was undefended (or perhaps even

occupied). Discarding their armour (but not yet their weapons), thereby demonstrating that

they had little intention of engaging in further offensive action, they chose to flee to another

251 In the Second Battle of Philippi, Brutus also retreated first to a camp and then, once his defeat was abundantly clear, he occupied a nearby hill (App. B Civ. 4.128–9). 252 Trans. A. G. Way, 1955.


70

camp, apparently perceiving this to be their best hope of refuge. When they found this camp

occupied, only then did they make their way to some nearby high ground, and from there they

attempted to surrender. It is thus abundantly clear in this case that the Romans preferred the

defence of camps over high ground.

In addition to the views of the elite, it is also worth considering the motivations of milites.

Naturally a man in danger would want to seek safety wherever it seemed best, hence the

debate between veterans and novi milites that Caesar described above (BGall. 6.40). Shame

may have been a motivator for beaten soldiers to return to camp, for as demonstrated by the

case of Maroboduus in AD 17 (Tac. Ann. 2.46), abandoning one’s camp to the enemy was a

sign of decisive defeat. Importantly, camps contained an army’s baggage, including soldiers’

personal possessions and the army’s plunder, not to mention military equipment, medical

supplies and food.253 The desperation with which soldiers could act to protect their plunder

and possessions is recorded by both Caesar (BGall. 5.33) and Livy (10.36.12–5). Thus,

soldiers will have often been intensely motivated to return to their camps and defend them

after a defeat in the field, rather than abandon their valued belongings and plunder to the

enemy.

Livy (44.39.1–5) transmits a speech by Aemilius Paulus that is particularly revealing of

Roman attitudes towards the military camp. Just prior to the Battle of Pydna, 168 BC,

Aemilius Paulus gave the following justification to his consilium for taking time to establish a

camp before giving battle:

Oh yes, you say, we should have had a battle-line undeployed and disorderly; but our

camp was fortified, our water-supply found, the way to it secured by placing guards,

and reconnaissance made all around. Or did our men have nothing but the bare ground

on which to fight . . .? Your ancestors regarded a fortified camp (castra munita) as a

haven (portum) against all the mischances of an army, whence they might go out to

fight, and whither they might find shelter from the storm of battle. Therefore when

they had walled in their camp with fortifications, they used also to strengthen it with a

powerful garrison, because one who had been stripped of his camp, even though he

had won the battle, would be considered the loser. Camp is the shelter of the

253 Roth, 1999, 79–84.


71

conqueror, the refuge of the conquered. How many armies meeting with none too

favourable fortune in battle, after being driven (compulsi) within their rampart (intra

vallum), have in their own good time, sometimes after only a moment, sallied forth

and routed the victorious enemy? This abode is a second home (patria altera) for the

soldier, its rampart takes the place of city walls (moenibus) and his own tent

(tentorium suum) is the soldier's dwelling and hearthside (domus ac penates). Should

we have fought like nomads with no abode, so that, whether conquered or conquerors,

we might return (reciperimus) —where?254

Two key themes are evident in Livy’s rhetoric. First, the camp provided an important tactical

function in providing refuge to a beaten army, sheltering the army as it recovered its strength

for a renewed attack, or providing a safe haven for the victors to recuperate. Second, the

camp held psychological significance to the miles, who viewed the camp as his country

(patria), and his tent (tentorium) as home (domus). The latter theme is significant. Although

the idea is transmitted through the medium of Livian rhetoric, in the absence of the soldier’s

point of view, it may be taken to represent a conventional soldier’s view. It has already been

shown that a Roman soldier would balk, even panic, at the prospect of losing his possessions

to the enemy (Caes. BGall. 5.33). In the context of a campaign away from country, home and

family, it is not difficult to believe that Livy conveys a fundamental truth about the emotional

attachment of Roman soldiers to their camp. If given the option, soldiers would surely have

preferred to return to their tent and belongings than a bare hill.

Thus, while there were certainly a variety of situations in which a steep hill was a better place

for an army to retreat to than a camp, it appears that at least for the Romans, the standard

procedure in the event of a battlefield defeat was to retreat to a camp rather than an

unfortified hill. Caesar, Livy and Tacitus all imply that a beaten army should retreat to its

camp if possible, and that while high ground could provide refuge if necessary, it should not

have been the first choice for retreating troops.255 Indeed, Livy directly states that it was

common sense for a camp to be constructed specifically for the purposes of giving refuge to

an army in the event of retreat (5.38.1). In general, Roman armies would retreat to hills only

254 Trans. Alfred C. Schlesinger, 1951. 255 Camps were at any rate often constructed on higher ground (see Dobson, 2008, 15, Fig. 9; Bellón et al., 2016, 104). Cf. Xen. Hell. 3.5.19–20; Polyb. 11.24; BAlex. 76; Livy 28.17, 32.12, 38.21, 26–7.


72

when they did not have a camp or could not reach it.256 Otherwise, most Roman armies will

have retreated to their camp if defeated, unless prevented by some other factor such as being

cut off by the enemy (BAfr. 85.3–5), or due to confusion at night (Livy 4.39.4–9).

Walled Settlements

For the Gauls fighting against Caesar, who did not have the Roman habit of building fortified

camps near battlefields, retreating to the oppida that dotted the landscape was not uncommon

after battlefield defeats.257 Greek and Hellenistic armies did the same when necessary, for, as

Polybius (4.12) points out in his account of the Battle of Caphyae in 220 BC, a defeated army

was much better off retreating to nearby towns than fleeing in open ground.258 It was,

therefore, possible for an army to retreat into a walled settlement after defeat in battle.

In the case of the Romans, at Ilerda in 49 BC, the Pompeian army was driven back by an

enemy charge and retreated into the town of Ilerda itself although they also had a fortified

camp nearby (Caes. BCiv. 1.42, 46–7). Also consider the Battle of Munda, fought in 45 BC

between Caesar and some leading Pompeian generals (BHisp. 31). After the Pompeians had

marched out from Munda, a fortified town, they encamped in a strong position on high

ground, either adjacent to or inside the town itself (28). Facing Caesar with a regular line of

battle, they were defeated and put to flight, and fled directly into the town (31-2). Even if the

Pompeian camp was inside Munda, it is evident that it was the town itself, rather than the

camp or higher ground, that provided the most significant protection to the beaten army.259

The defensive value of a walled town or city implied by this situation is a strong argument in

favour of an army deciding to retreat to such a position. And so it appears to have been

commonly true that a retreating or fleeing army, if unable to seek refuge on a nearby camp or

256 A large group of Roman cavalry did so after the catastrophic ambush at Trasimene, 217 BC, wherein the Romans had not established a camp because they were attacked on the march (Polyb. 3.86). Also see Polyb. 2.34, Sall. Jug. 98; App. Pun. 103. At Bibracte in 58 BC, Caesar had intended to seize a particular hill prior to battle, but was informed that the Helvetii reached it first, so Caesar retired (subducit) to the nearest height and drew up his battleline (Caes. BGall. 1.22.3). However, this was only an emergency expedient, and Caesar established a new camp as soon as it was practical (BGall. 1.22.4). 257 Eg. Caes. BGall. 2.12, 7.68. Also note the frequent Gallic sallies from these places: BGall. 7.22, 24, 47–51, 80–5. 258 For another example the Spartans, when defeated at Olynthus in 381 BC, fled in multiple directions to nearby cities (Xen. Hell. 5.33). 259 Ita ipsis Liberalibus fusi fugatique non superfuissent, nisi in eum locum confugissent ex quo erant egressi ... ex fuga hac qui oppidum Mundam sibi constituissent praesidium ... (BHisp. 31–2).


73

hill, would go towards a defensible town or city.

Other than following a defeat in a regular battle, there was one more scenario in which a

walled settlement featured, that is when an army being besieged in a town or city could sally

out to attack its enemy and then return within the walls.260 In the Battle of Thapsus, for

example, the garrison of Thapsus sallied out to assist the Pompeians during the battle (BAfr.

85). Livy (2.48.5–6) similarly claims that the legendary Veientes, besieged by the early

Romans, did exactly this c. 479 BC. The Gauls too, fighting against Caesar in 52 BC,

habitually sallied out from their oppida to fight besieging enemy forces: at Avaricum (Caes.

BGall. 7.22, 24), Gergovia (BGall. 7.47–51) and Alesia (7.80–5), after which they retreated

back behind the walls when beaten. As such, armies routinely retreated back into a city or

town if they had sallied out to fight besiegers.

Finally, there are a number of accounts that do not explicitly state that retreating armies made

their way either to high ground or to any sort of defensive position.261 In most of these

examples, however, the authors simply leave the place of retreat unstated. The reader of these

accounts will have been expected to infer the place to which armies retreated, to a camp or a

hill, or failing that, a walled settlement, in the event of defeat.262

Nevertheless, Appian (Hisp. 88) describes a skirmish near Pallantia in 134 BC, in which he

specifically records that a retreat was made by some Roman cavalry to flat ground, with no

further information given:

Scipio ordered Rutilius Rufus, a military tribune … to take four troops of horse and

drive back the assailants. Rufus followed them too sharply when they retreated and

darted up the hill with the fugitives. When he discovered [an enemy ambush hidden

behind the hill] he ordered his troops not to pursue or attack … but to stand on the

260 Levithan, 2013, 219 261 Polyb. 11.17, 18.25; Livy 30.18, 37.38, 42.59, 44.42; App. Hisp. 82, 88. 262 For example, Polybius (18.25) mentions a retreat by part of the Roman army at Cynoschephalae, 197 BC, but gives little further detail. However, the Roman camp was in the area (Polyb. 21.4), their retreat must have been in that direction. Livy’s account (30.18.12–4) account of the defeat of Mago Barca in 203 BC does not mention a camp or hill to which the Carthaginian army retreated after being beaten, and yet the army was able to remain safe overnight (30.19.1), implying that the army did have safe place (either a camp or hill) to remain in this interval.


74

defensive with their spears presented to the enemy and merely ward off their attack.

Straightway Scipio, seeing that Rufus had exceeded his orders, and fearing for his

safety, followed with all haste. When he discovered the ambush he divided his horse

into two bodies and ordered them to charge the enemy on either side alternately,

hurling their javelins all together and then retiring, not to the same spot from which

they had advanced, but a little further back each time. In this way the horsemen were

brought in safety to the plain.263

It is not clear which cavalry Scipio ordered to split into two bodies: his own, Rufus’ or both?

Scipio presumably gave visual orders to the beleaguered troopers on the hill, but he will have

more easily given complex orders to men actually accompanying him, and so it is more likely

that Scipio brought and used his own cavalry with which to effect the fighting retreat. As for

the question of why the cavalry merely retreated to flat ground, it must be considered that the

reason Scipio ordered a cavalry attack in the first place is because the enemy were harassing

Roman foragers, as there was a camp nearby (Hisp. 87–8). The cavalry could thus easily seek

refuge there once reaching flat ground.

However, by virtue of their speed, as long as retreating cavalry could outpace their pursuers

there may have been no need to seek immediate refuge when in retreat. In another example,

Livy (37.38.1–4) records a retreat by some defeated Seleucid cavalry in 190 BC, in which

they attempted to use a river to delay their pursuers. This did not work particularly well,

precisely because they did not outpace their pursuers. Their final destination, which they

must have reached, is left unmentioned. As their parent army was four miles away (Livy

37.48.2), a relatively insignificant distance for cavalry, this is where they will almost

certainly have gone. Infantry caught in a similar position would certainly not have been able

to escape so easily.

Summary of Part B

It is evident that there were a number of places to which an army could seek refuge following

a retrograde movement. The common factor was that these places had to be defensible, so

that the retreating forces were either safe from or better able to resist further enemy attacks.

263 Trans. Horace White, 1913.


75

There appears also to have been a simple hierarchy in the sorts of positions that retreating

armies preferred to seek out. As a general rule, whether or not the place to which an army

retired is specified by a writer, a retreating Roman army will have gone to their camp, or high

ground if they could not make the camp. If neither of these options were available and

suitable, a walled settlement could also serve. Armies could attack from such positions with

greater confidence, and retreat back into them when necessary.

Part C: Retrograde Movements over Longer Distances

What if there was no refuge particularly close by, or if an army's defensive position was not

sufficiently protective and a more distant position was needed? Let us consider scenarios in

which an army was required to withdraw over a longer distance to find a defensive position

or safe location.

An army cut off from its camp and any nearby defensive position would obviously have had

to seek shelter elsewhere, as occured after the Battle of the Trebia, 218 BC. After the two

armies engaged, an ambush by Hannibal caused both Roman flanks to collapse (Polyb.

3.74.1–2) while the Roman centre was able to break through the enemy line, but the legions

in the centre became separated from the rest of the army and cut off from their camp (Polyb.

3.74.3–9):

... those in the van ... despaired of ... returning to their camp, afraid as they were of

the very numerous cavalry and hindered by the river and the force and heaviness of

the rain which was pouring down on their heads. They kept, however, in close order

and retired on Placentia, being not less than ten thousand in number. Of the remainder

the greater part were killed near the river by the elephants and cavalry, but the few

infantry who escaped and most of the cavalry retreated to join the body I just

mentioned and with them got safely into Placentia. The Carthaginian army, after

pursuing the enemy as far as the river, being unable to advance further owing to the

storm, returned to their camp.264

264 Trans. H. J. Edwards, 1922–7.


76

In this case, the beaten Roman army could not have merely moved its battleline a few

hundred metres to a camp or hill. Placentia was approximately 15 kilometres away from the

Roman battleline at the Trebia.265 It must be noted that the army will almost certainly have

needed to form a column to cover this distance, particularly as it probably had to go back

across the river Trebia to reach Placentia.266 There was some high ground to the south, yet not

only was the Carthaginian camp and army between the Romans and this terrain, but

considering the absolute superiority of Hannibal’s army after its massive victory it is likely

that the Carthaginians would have finished the Romans off if they had merely retreated to

high ground. Despite the distance to Placentia, it may well have been the closest defensible

position for the Roman army when withdrawing.267

There are two salient points to note from this example. First, as the Roman army was unable

to find safety nearby it had to withdraw a significant distance to do so. Second, the distance

the Romans had to move will have necessitated a change in formation. An army retreating

directly from battle may have only needed to move a few hundred metres to reach safety, and

they could have done this by simply moving their battleline (acies) backwards.268 For the

Romans at the Trebia to most easily reach a destination that was several kilometres away, it

will have been necessary to organise themselves into a less unwieldy marching formation

(agmen). There are numerous examples of Roman armies behaving in the same way.269 For

instance, after the Pompeian army was heavily defeated at Pharsalus in 49 BC, some officers

planned to withdraw the army to a town many kilometres away.270 The two observations

drawn above can therefore be applied to Roman armies in a number of other situations.

265 Walbank, 1957, 398. 266 Ibid., 407. According to Livy (21.56.2), this group of Romans had been forced into a square formation (in orbem) before their withdrawal, but considering they had to cross the river (which Livy overlooks: 21.56.3–4) and then march a significant distance to Placentia, it is unlikely they remained in this formation once away from the main enemy force. Even if they did, a square formation was still better suited to moving long distances than a battleline. 267 A few survivors made it back to the camp, according to Livy (21.56.5), but no significant forces were apparently able to do so. According to Appian (Hann. 7), survivors made their way to both Placentia and Cremona. 268 For example, the Roman battleline at Pydna (168 BC) met the Macedonian phalanx only 400 metres from the Roman camp (Plut. Aem. 18.9), although Plutarch implies that this was fairly close. 269 Eg. Caes. BCiv. 1.64, 79–80; Livy 28.15.16–16.3. 270 The Pompeians retreated first to their camp (Caes. BCiv. 1.94–5), but when this camp was stormed, they retreated again to nearby high ground (1.95. After this they finally made plans to withdraw to Larisa (1.97), which was about 30 kilometres away (Carter, 1993, 216). In the event, Caesar forced them to surrender before they could make this withdrawal (BCiv. 1.97–8).


77

Even if the Romans had safely retreated from initial danger to the closest place of refuge,

other problems might have arisen. There are numerous examples of armies withdrawing from

camps as a result of enemy action.271 There were several reasons for this. A camp was not

necessarily proof against enemy attack, and if it was stormed it could spell disaster for those

sheltering inside, particularly if they could not withdraw elsewhere.272 A general might,

therefore, wish his army to escape a victorious enemy entirely.273 The reality or threat of the

arrival of superior enemy forces could also trigger a withdrawal from a previous defensible

position.274 A general could also have simply moved his camp to exercise the army and

maintain discipline (Onas. Strat. 9). Food and supply shortages were particularly serious

threats, as any army trapped in its camp for any reason would eventually face this

insurmountable problem, no matter how defensible the camp was. In general, a group of men

or an army that was unable to forage for water, fodder and firewood would run into severe

difficulties within a few days.275 As mentioned above, an army stranded on high ground could

run into the same difficulties.276 Even an army that had retreated or withdrawn into a larger

town or city could eventually suffer from supply shortage, although this would usually have

taken longer period of time.277 There are thus several scenarios in which an army might have

withdrawn from an otherwise defensible position, with logistical shortage (or the threat of

such) being a particularly straightforward motivation for doing so.278

271 Eg. Caes. BGall. 2.10–1, BCiv. 1.61–4, 2.43; BAlex. 31; BHisp. 6; Livy 22.16–7, Livy 28.15.12–16.8, 36.46.7. 272 Eg. Caes. BGall. 3.5, 25–6, 6.37–8; BAlex 76; Livy 32.12.1–7, 38.23. Also see Sabin, 2007, 414. 273 Livy 28.15.12–16.8, 36.46.7. 274 Caesar provides a debate on the wisdom or otherwise of retreat in such a scenario (BGall. 5.26–31). Also see Caes. BGall. 6.7–8. 275 An army needed water on a daily basis (Roth, 1999, 36, 155). When the Pompeian army at Ilerda had been cut off by Caesar from water, fodder and firewood, it surrendered within a few days (Caes. BCiv. 1.84.1). Also, horses and baggage animals could themselves only work for a few days without fodder (Roth, 198, 307) and required grazing every five to seven days (Engels, 1978, 154–155). An army could only operate further than 80 to 100 kilometres from a supply base by grazing their animals (Roth, 1999, 129, 198, 201), which would necessitate the army being able to safely leave its camp. 276 Eg. Polyb. 3.86; Caes. BGall. 6.40, BCiv. 3.95–8; App. Hisp. 88. 277 De Souza, 2007, 455–8. 278 Also note the surrender of a Spartan contingent on Sphacteria in 425 BC, which resolutely suffered heavy casualties for some time before surrendering, in part due to being exhausted by their lack of food (Thuc. 4.35–6).


78

In many cases, Roman armies withdrew from within a fortified position.279 This is revealed

by the Pompeian army under Afranius and Petreius during the Ilerdan campaign of 49 BC

(Caes. BCiv. 1.61–84). Following a laborious engagement with Caesar’s forces, the

Pompeians had retreated into Ilerda (BCiv. 1.46–7). At some stage they then transferred to

their nearby camp (1.59.1), where Caesar was eventually able to cut them off from foraging,

prompting Afranius and Petreius to withdraw their forces from the region (1.59–61). The

Pompeians marched for around 12 hours before they were overtaken and forced to encamp,

albeit on higher ground, but over seven kilometres short of their intended destination (1.64–

5). Both Caesar and his opponents then departed from their positions to head for some

strategic ground nearby, which Caesar reached first, bringing the Pompeians’ march to a halt

(1.66–70). As the latter were still having trouble foraging in this new position they debated

marching to Tarraco around 100 km away (1.70), but they finally decided to return to Ilerda,

still over 30 km away.280 Caesar harassed them heavily as they marched, eventually forcing

the Pompeians to encamp in a position where they could not forage nor find water, and where

they eventually surrendered (1.80–4). From this, it is clear again that an army in withdrawal

might have had to move significant distances, and do so quickly. To do so the men would

have needed to move in a column. The relationship between withdrawals and shortage of

essential supplies is also made abundantly clear in the case of Ilerda, as the Pompeians were

willing to withdraw and march their armies as far as necessary to secure a position from

which they could forage and fight effectively. A short retreat from battle would not have

served their purposes.

Evidently, an army might have withdrawn more than once to find safety. After their defeat at

the Battle of Carrhae, 53 BC, the Romans were not safe in their camp and so they withdrew

by night to the nearby town of Carrhae (Plut. Crass. 27.5–6), a number of kilometres away.281

They were not even safe there, and so the bulk of the remaining Romans withdrew again

towards Armenia, although they did not do so successfully (Crass. 29–31). Over a century

later, in AD 66, Cestius Gallus launched an ill–prepared attack against Jerusalem in an

279 See App. B. As Roman armies typically remained in camps whenever they were static (Goldsworthy, 1996, 113), and a general might have withdrawn his army from a camp to march elsewhere for many reasons, it is no surprise that there are so many references to armies moving e castra/castris or ex castra/castris (the PHI Latin Database records hundreds of such). 280 Caes. BCiv. 1.78. On the distances see Carter, 1991, 193, 206. 281 Note that it took a small, fast group of 300 cavalrymen several hours to reach Carrhae from the battlefield (Plut. Crass. 27.2, 5–6).


79

attempt to rapidly extinguish a Jewish rebellion, but supply shortages forced him to withdraw

first to a camp at Gabao, and then several kilometres further towards Beth–Horon, all the

while suffering heavy casualties.282 The Romans were not even secure in Beth–Horon, and so

they withdrew again by night to Syria, leaving behind a rearguard and their baggage (2.551–

4). Withdrawals, therefore, were utilised as many times as needed, and toward a variety of

destinations, for an army to reach safety.

Was it always absolutely necessary for a Roman army to withdraw towards high ground or a

fortified position? Appian (Hisp. 81–2) for example describes the retreat of a Roman army,

commanded by Aemilius Lepidus Porcina and Junius Brutus Callaicus, from Pallantia in

northern Spain in 136 BC. It became necessary for the army to withdraw from its siege of

Pallantia due to supply shortages, but the withdrawal was ordered suddenly and performed in

a disorderly manner (82). Seeing an opportunity, the Pallantines harassed the Romans on the

march and inflicted significant casualties until nightfall, when the Roman army became

halted out of sheer exhaustion, and the Pallantines retreated (82). Although there is no

indication here of the Romans’ destination, since the Pallantians were little threat to the army

in a direct engagement, the Romans only needed to go somewhere that remedied their supply

situation. In a scenario in which a Roman army did not fear the attack of an enemy, but only

suffered from supply problems, it will not have been necessary to seek out a strongly

defensive position, only to find a place where the army could replenish its supplies. It is thus

a striking difference between retreats and withdrawals that, while retreats had to be carried

out towards a specific place of refuge due to the proximity of the enemy, withdrawals could

have vaguer destinations, depending on the level of threat posed by the enemy and the

general's intentions.

Summary of Part C

There are several criteria that distinguish withdrawals from retreats. The first was that

withdrawals occurred over longer distances, necessitating a march in a column of some kind

rather than in an unwieldy battleline. A withdrawal also did not necessarily require as specific

a destination as a short retreat from the battlefield, although sometimes they did. An army

could also withdraw repeatedly over a variety of distances to various locations in order to

282 Joseph. BJ 2.544–50; Gichon, 1981, 57–9, 62.


80

reach a satisfactory place of safety. Finally, withdrawals did not always follow battles, but

could occur for a wider variety of reasons beyond defeat in battle, such as supply shortage. A

withdrawal might follow a retreat, or take the place of one, if an army was unsafe, or

logistically or strategically handicapped, in an initial place of refuge.

Conclusion

An army might have retreated from the battlefield for a variety of reasons, defeat in battle

being the most common.283 Armies would also, in general, disengage if night fell during a

battle, if an engagement was indecisive, or if the weather were particularly poor, but these

were not particularly common scenarios. Armies could also simulate flight or a disorganised

retreat to lure their enemies into an ambush or surprise counterattack, but the time and space

required to reorganise an army after implementing such a tactic meant that this would be

difficult for infantry to do safely. Such practices were easier for cavalry than infantry, and

were not particularly favoured by the Romans. Also, retreats as a result of defeat in battle

could be either deliberate or involuntary, for the purposes of avoiding further fighting or

enabling re–engagement in a more favourable location.

In terms of where an army could retreat to, Gilliver is correct in her statement that an army in

retreat could move towards a camp, high ground or a city, but she has not fully explained that

these places did not necessarily hold equal defensive value, depending on the situation. The

general rule, at least for the Romans, was that an army defeated in battle would retreat to a

fortified camp or failing that to a nearby steep hill, and finally if such places were not

accessible, then an army might have tried to make its way to the closest defended

habitation.284 In choosing the location to which an army would move, the commander will

have also had to balance its defensibility with its accessibility. The proximity of a position

was obviously always an important factor, but an army retreating in order to re–initiate battle

in a more favourable position would have been best served by retreating onto higher ground

where it could maintain its battleline, rather than retire behind fortifications (Caes. BCiv.

1.45), yet an army wishing to avoid further fighting would seek the safety of fortifications

283 Gilliver, 2007, 134. 284 Cf. Gilliver, 2007, 134. There is no immediately apparent reason that Hellenistic or Carthaginian commanders, who also regularly built fortified camps near their chosen battlefields, should not also have followed this general rule.


81

first and foremost (BAfr. 85). Several factors, including the tactical situation, the proximity of

different types of refuge, and the intentions of the retreating forces, thus determined the

destination.

A withdrawal had some different characteristics to a retreat due to the increased speed and

distance that the army moved as well as its formation, generally in a column (agmen) rather

than in a formed battleline (acies). In both retreat and withdrawal, an army would move

towards the closest defensible position, but if the initial position could not provide long–term

safety, an army would aim to withdraw to another defensible position, even if this was a

significant distance away. Tactically, while an army would usually retreat due to defeat in

battle, the reasons for an army to withdraw were more varied, with logistical shortage being

particularly common. This would force an army to prioritise a location with ample supply

rather than one that was simply the most defensible. However in the broadest sense, retreats

and withdrawals were variants of the same thing: a retrograde movement by an army in order

to improve its tactical or strategic situation, or to prevent a worse one from occurring. The

context of the retrograde movement, and the distance that it was necessary to retire, would

determine its character.


82

But the men of the Ninth Legion, carried away by zeal in their desire to repair the loss

received, rashly pursuing the flying foe too far, get into unfavourable ground and approach

close under the hill on which the town of Ilerda was situated. When our men wished to retreat

from this position, the enemy in turn kept pressing them hard from the higher ground.

(Caes. BCiv. 1.45)

Chapter Two

The Impact of Pursuit on Retreating or Withdrawing

Forces


83

Introduction

At the Battle of Ilerda in 49 BC, Gaius Julius Caesar clashed with a Pompeian army near the

eponymous town, which was well–fortified and situated on a steep hill (Caes. BCiv. 1.43–5).

Caesar’s men attacked and pushed forward up this hill, where their enemies stopped and

fought, so that the Caesarian soldiers found themselves fighting in a disadvantageous position

from which they could not safely retreat, as the enemy pressed forward aggressively when the

Caesarians attempted to do so (BCiv. 1.45). Why exactly were Caesar’s men unable to retreat

from this position, and what were the potential ramifications of retreat against such an

aggressive enemy?

The emphasis of the preceding chapter was on when and to where armies ideally retreated or

withdrew, but how did the behaviour of the opposing army shape the tactics, and chances of

success, of a retrograde movement? If it was logical for one army to need to retire, it would

surely be just as logical for its opponents to take advantage if possible. It is, therefore,

necessary to consider how enemy action could interfere with the conduct of retrograde

movements in order to develop a complete understanding of these manoeuvres.

In this chapter the reactions of armies when their enemies retreated or withdrew will be

examined, in particular any pursuit, since enemy action necessarily had an impact on the

success or failure of retreats. Firstly, different types of pursuit will be examined, in order to

establish what usually happened when one side retreated. I will then explain how pursuit

could turn an orderly retreat into a disorderly flight in differing circumstances, including

short retreats from battle toward a place of refuge, and longer withdrawals when immediate

refuge was unavailable. Finally, I will assess the reasons that one army might have decided

against pursuing another. I will then conclude by summarising what usually occurred in the

case of a retreat or withdrawal, and the burdens this placed upon the retiring army.


84

Types of Pursuits

Pursuits occurred commonly, as displayed by the number of retrograde movements that failed

under enemy pressure.285 There were a variety of scenarios in which a victorious army, or

one in a position of advantage, could pursue its enemy. The most common scenario, in which

an army retreated directly from battle, would give the victors a clear opportunity to follow up

their success.286 Appian (BCiv. 4.128) depicts this in the Second Battle of Philippi, 42 BC. In

this scenario, the army of Brutus engaged with that of the Caesarians Octavian and Antonius,

and, after a hard–fought struggle, the former was pushed back (τὴν φάλαγγα τῶν ἐχθρῶν

ἐκίνουν). At first they retreated in an orderly manner (τὸ ὀπίσω βάδην ἔτι καὶ μετὰ

φρονήματος), but they were pursued by the Caesarians until the three lines of their triplex

acies became too compressed for order to be maintained, and under further pressure from the

Caesarians, they broke and fled (ἔφευγον), following which the Caesarians advanced even

to the gates of their camp. So, a retreating Roman army, even if it maintained good order,

might still be effectively pursued. The genuine importance of order in maintaining a

successful retreat will be touched on later, but suffice to say that it was not of itself proof

against pursuit.

Even if an army retreated successfully to a fortified camp, the actions of pursuers might still

have had an impact. At the Battle of Pharsalus, 48 BC, after the Pompeians had been routed

from the line of battle, they fled into their well–guarded camp (Caes. BCiv. 3.94), but

Caesar’s men attacked the camp, and, despite significant resistance by the camp’s defenders,

they drove the Pompeians from their ramparts and out of the camp entirely, forcing them to

retire again to some nearby hills (BCiv. 3.95). During the Battle of the Nile in 47 BC,

Caesar’s forces attacked the heavily fortified camp of Ptolemy XIII, were able to capture the

ramparts and begin killing those inside, causing such panic that large numbers of the enemy

threw themselves to their deaths from the rampart while others drowned in the Nile, including

Ptolemy himself (BAlex. 31). There are many other examples of camps being stormed by

excited victors.287 In essence, although armies commonly retreated into camps after a defeat,

285 See App. A. 286 Livy (10.36.3) even notes in instance in which two armies were afraid to retreat for fear that the enemy would advance. 287 Polyb. 5.73, 14.4–6; Caes. BCiv. 3.67, 95; BAlex. 76; Sall. Jug. 99; Livy 24.16, 25.14, 39, 31.41, 33.15,


85

it was also not uncommon for armies to be forced to withdraw from camps as a result of

continuing enemy action.288

So, not only was an army also vulnerable to attack while retreating from the field and to its

camp, as was an army that halted its retreat at the walls of its camp, where it could be trapped

there and destroyed, but any army might find no refuge even inside its camp.289 In addition, a

retreat into or withdrawal out of walled settlements in the presence of the enemy might have

led to further difficulties. For example, after the defeat of the Demetrius at the Battle of Gaza,

312 BC, his army became stuck attempting to enter the city and in the confusion, the

victorious Ptolemiac troops were able to get into the city and take possession of it (Diod. Sic.

19.84–6–8). Two usual methods of seeking safety after a defeat in battle, retreat to a fortified

camp or town, could thus be disrupted by determined pursuers during every stage of the

necessary manoeuvres.

In the case that an army that was forced to withdraw out of a town or city due to an enemy

attack or siege, they will have had to contend with the attackers or besiegers.290 Caesar’s

account (BCiv. 1.25–8) of Pompey’s withdrawal from the port city of Brundisium in 49 BC

highlights several methods that a besieger could use to prevent easy withdrawal from a town

or city. To deny escape by sea, Caesar constructed piers and dams to narrow the waterway,

and placed rafts fastened by anchors as obstacles where the water was deeper, and

constructed causeways from which withdrawing ships could be attacked (1.25). Caesar also

surrounded the landward side of the city, where legionaries with siege ladders were ready to

enter the city if an attempt to withdraw by sea was undertaken (1.28). The construction of

extensive trench–works, ladders and siege towers was standard Roman conduct in a siege and

the Romans could build siege–works of huge scale when necessary.291 It will, therefore, have

been no easy matter to escape from such a situation. In the event, Pompey did escape, but

through equally elaborate preparations. He constructed fighting ships to deal with the

37.43, 38.22–3, 40.31; App. Mith. 44, 50. 288 Eg. Caes. BGall. 2.10–1, BCiv. 1.61–4, 2.43; BAlex. 31; BHisp. 6; Livy 22.16–7, 28.15.12–16.8, 36.46.7. Some of the reasons armies might have done such a thing have been touched on in Chapter One (77–8). 289 Vulnerable: Pol 14.4–6; Livy 25.39. Trapped and destroyed: App. Mith. 44. 290 At Cenabum in 52 BC, when the town’s Gallic defenders attempted to slip out at night to avoid Caesar, but were detected by the Romans, who then stormed the town and overtook the Gauls trying to escape (BGall. 7.11.6). Also see Caes, BCiv. 1.25–8; Sall Jug 76.1; Plut. Luc. 19. 291 Levithan, 2013, 6. Also see Caes. BGall. 7.68–9; BCiv. 3.42–5; Joseph. BJ 7.9.


86

maritime fortifications, and on the land he built an array of obstacles and fortifications to

delay pursuers, after which he embarked his army in two stages, the bulk initially and in

silence and the rest quickly by night so that there was little time for Caesar’s men to react

(27–8). Pompey was able to do all of this with twenty cohorts, perhaps just enough men to

provide the necessary labour but not so many that the army was too unwieldy in the act of

escape. The efforts of Pompey to escape emphasises the fact that withdrawal out of a town or

city could be immensely difficult, if the enemy was engaged in preventing it.

If an army could not defend itself in a town or city, did not have the resources or ingenuity to

engineer an escape, and could not fight their way out, their situation could be desperate. In 72

BC, during Lucullus’ campaigns against Pontus, the Romans laid siege to Amisus and made a

surprise attack while the defenders were changing the guard on the ramparts (Plut. Luc. 19.1–

4). The defenders escaped only by setting fire to the city. More disastrously, in 108 BC when

Metellus Numidicius laid siege to a Numidian fortress at Thala, the Numidian king slipped

out of the city by night but the men of the garrison were unable to do the same, and so they

chose to retreat to their citadel and set it aflame, with themselves and much booty still inside

(Sall. Jug. 75–7). It is clear then that attacks on an army inside defended fortifications, and

pursuit if the enemy attempted to withdraw from their position, could make escape all but

impossible, or only achievable with drastic measures.

An army that did escape from a camp, town or city, given its need for speed, would form a

marching column (agmen) rather than move in a slow and unwieldy battleline (acies).292 One

advantage of withdrawing over open ground is that the marching army had enough space to

move in an orderly but rapid manner, but pursuers, particularly cavalry, will also have had

freedom of movement.293 During the Carthaginian withdrawal following their defeat at the

Battle of Ilipa, 206 BC, Hasdrubal Gisco decided to retire from the relative safety of his camp

to avoid being besieged and destroyed (Livy 28.15.12–6). The Carthaginians managed to slip

away during the night and moved as quickly as they could, but they were delayed by

harassment from Scipio’s cavalry, enabling the Romans to overtake them the next day, and

who proceeded to destroy the haggard and disordered Carthaginians.294 The impact of the

292 Eg. Caes. BCiv. 1.64, 79–80; Livy 28.15.16–16.3. 293 Diod. Sic. 19.84.6. Also see Gilliver, 2007, 138; Sabin, 2007, 413. 294 Livy 28.16.3–6. Livy expresses a tone of disapproval for Hasdrubal allowing his army to become disordered, so that it could not defend itself effectively when the legions arrived. This is probably the only place to


87

Roman cavalry pursuit stands out in this episode.295 Cavalry played an important role too in

one of Rome’s greatest ever defeats, the Battle of Carrhae in 53 BC. The disaster featured

two withdrawals after the battle itself in which the slow Roman infantry took enormous

casualties from pursuing Parthian cavalry.296 An army withdrawing over long distance was

thus highly vulnerable to pursuit, most especially from cavalry.

The importance of cavalry in pursuing a withdrawing enemy cannot be overemphasised. At

the Battle of Ilorca in 211 BC, the Romans were attempting to withdraw away from

victorious Carthaginian forces (Livy 25.34–5) but were unable to disengage from pursuing

cavalry (25.35.8–9). As the Romans were heavily delayed, by the time night fell they had not

gone far and were forced to halt in a weak defensive position (25.36.1–4). They attempted to

fortify and defend this position but were totally overrun (25.36.5–13). Leading up to this

event, the Romans were intent on retiring they were unable to do so due to the delay caused

by enemy pursuit. Similarly at Ilerda in 49 BC, when the Pompeian forces were withdrawing

they were so delayed and even disrupted by pursuing Caesarian cavalry that they were forced

to encamp in a sub–optimal position far from water supplies, which at length forced them to

surrender (Caes. BCiv. 1.81–5). In this second case, persistent attacks by cavalry during a

withdrawal threw a powerful army into serious disorder, even in the face of serious attempts

to counter the pursuing cavalry (BCiv.1.79.5–80.1). As such, it was not always possible for an

army to improve their tactical situation when withdrawing, whether by re–establishing a

fighting line, taking a strong defensive position or escaping their enemy entirely, if they could

not escape pursuers, and cavalry were particularly proficient at delaying and disrupting a

withdrawing army.

realistically find fault. On the other hand, having been defeated heavily in a major battle, and then suffered a number of desertions, there is no reason why offering battle to the Romans all over again, once the latter caught up, would have helped the Carthaginians in any way. Speed was therefore of the essence. 295 Scipio had a cavalry advantage, probably because he had defeated the Carthaginian cavalry during the main battle (Polyb. 11.21; Livy 28.13.6–10). Many of the cavalry may have deserted in the immediate aftermath of the defeat, however (Livy 28.15.14–5), which may explain the difficulty the Carthaginians subsequently had in covering their withdrawal against Scipio's cavalry and skirmishers. Cavalry pursuit also prevented the successful withdrawal of experienced Roman forces at Ilerda (Caes. BCiv. 1.79–80). 296 Plut. Crass. 28.1–2, 31.7. These withdrawals will be examined in detail in Chapter Six.


88

These considerations would have weighed on a commander who was trying to decide

whether or not to abandon a fortified position. Caesar (BGall. 5.26–38) described a disaster

near Atuatuca in 54 BC, in which two of his legates, Sabinus and Cotta, had to make just

such a choice. At the beginning of Caesar’s account, the Romans were physically safe in a

well–fortified camp, but a secretly rebellious chieftain, Ambiorix, came to the Roman camp

in friendship and insidiously explained that they would be besieged and destroyed by

Germans if they remained there, with the intention that he would ambush them after they left

the fort (BGall. 5.27). Caesar describes a debate between the two legati and other officers on

whether they should take the advice of a Gaul and make a forced march from their hiberna,

or if they should remain behind the walls (28–31). Sabinus argued that to remain would risk

besiegement by superior forces, in which situation he thought that the army would perish

either by hunger or the sword (5.29), while Cotta and other officers countered that the

hiberna was proof against enemy attack and they held extensive food supplies, so that

remaining in place would be safer than risking attack in the open. Sabinus prevailed, and

Caesar relates that the army put his plan into action with the foreshadowing phrase,

longissimo agmine maximisque impedimentis (5.31). Events proved Sabinus fatally wrong,

unfortunately for the entire Roman force (5.32–7). Caesar thus provides arguments for and

against remaining in a camp and suffering siege, and forced march in the open.297 According

to Caesar, it was safer for an army to remain behind fortifications and risk starvation than to

march into a certain ambush. Such was the danger conceived by Caesar in being attacked

while on the march.

There are several examples of Roman columns in withdrawal suffering major defeats when

caught on unfavourable ground. The Romans allegedly lost 20,000 men in a Boian ambush in

216 BC.298 The famous clades Variana AD 9 saw a comparable number of Romans perish as

they attempted to march out of hostile territory in Germany (Vell. Pat. 2.117.1). And in AD

66, in a defile near the Judean town of Beth–Horon, a withdrawing Roman army was all but

destroyed by lightly–equipped Jewish rebels (Joseph. BJ 2.540–555). As such, an army

297 Caesar’s personal view on the best course of action can be identified with Cotta’s, as demonstrated by the fact that the latter is supported by the senior centurions (primisque ordinibus) in making his argument (Caes. BGall. 5.30). Note that a fragment of the army was able to make their way back to the camp, where despite their small numbers they were able to defend it from initial attacks (5.37). 298 Polyb. 3.118. This episode may not classify as a withdrawal, as Postumius was campaigning aggressively (cf. Livy 23.24.6–13; Front. Strat. 1.6.4) but the Roman formation was the same as if they were withdrawing: they were certainly not expecting to be attacked whilst in agmine.


89

forced to withdraw over a long distance was highly vulnerable to a prepared enemy. For this

reason, armies sometimes attempted to move at night, when they were harder to detect and

potential pursuers caught unawares without time to follow.299 However, as seen in the case of

Ilipa (Livy 28.15.12–6), retiring by night did not guarantee escape, for night operations could

cause serious disorder among the withdrawing forces.300 The confusion of night operations,

evident in Sallust’s Romano-Numidian battle account (Jug. 97–8), will have been amplified

when armies were carrying out extended withdrawals. While there were numerous other

stratagems a general could attempt to deceive the enemy and allow a safe retreat or

withdrawal, nothing was necessarily proof against a vigilant enemy (Caes. BCiv. 1.65–6). It is

no wonder then that, in 54 BC, when rebellion flared up throughout Gaul and Caesar’s

hibernae were besieged by superior enemy forces, Caesar preferred that his legati stay put.301

Pursuit could thus take the form of an infantry advance, a cavalry attack, or a combination of

the two. An army in a losing engagement may have had no chance but to retreat, but as

shown, in doing so they became vulnerable at all stages of their retreat, particularly when

entering or leaving a fortified position. An army that has previously won an engagement

could wreak havoc on withdrawing enemies, for in open ground cavalry became a potent

weapon of pursuit, capable of causing harm in their own right and, even more decisively,

slowing a withdrawing army enough that the main body of the victorious army could arrive

and attack.

How Pursuit Could Cause Disorder

As observed, victors could and did attack retreating or withdrawing forces in a variety of

scenarios. A number of examples demonstrate that a particularly successful pursuit could

cause an orderly retrograde movement to disintegrate, such as occurred at Philippi in 42

BC.302 If an orderly retreat or withdrawal broke down, the only alternative was a disorderly

retreat or even flight, which were quite similar in practice (App. Hisp. 82). In the following

discussion, I will explore in greater detail how pursuit could cause an otherwise–orderly

299 Cf. Ch 1, 54–5. 300Also see Caes. BCiv. 1.67.4; Plut. Crass. 28.1–2. 301 The case of Sabinus and Cotta has been mentioned above. During the same uprising, when the legates (Quintus) Cicero and Titus Labienus faced with similar choices, they remained in their camps and successfully defended them (Caes. BGall. 5.40, 51–2, 57). 302 App. B Civ. 4.128. Cf. App. A.


90

retrograde movement to break down into panic and flight, focusing mainly on retreats from

battle, where most routs occurred.303

The first matter to consider is what caused an army to physically recoil from its enemy in

battle. Caesar emphasised the importance of impetus, particularly the vis exerted by attacking

troops and also the ability of the 'receiving' troops to resist (sustinere or ferre) this force, or

pressure, in determining the outcome of an infantry battle.304 Sabin notes that armies could

‘push one another back over significant distances’ during battle,305 but it is unclear from the

sources whether the pushing or pressure applied was always physical, or metaphorical (that

is, psychological).306 It is, therefore, worth considering both the physical and psychological

impact of attack and pursuit on retreating soldiers. Caesar conceived pressure as a result of

physical forces, including missiles and the weight of superior numbers.307 To add to this

however, Sabin’s influential model of ancient battle argues that soldiers in combat

occasionally needed to establish a ‘safety distance’ to physically and psychologically recover

between bouts of intense fighting.308 It is in this context that the concept of ‘pressure’ makes

the most sense. As retreating troops will have been tired from previous fighting, and probably

demoralised (Xen. Anab. 6.5.15–17), they needed to be able to establish this safety distance

and yet attacks by an emboldened enemy could strip them of such, inflicting on them

increasing psychological and physical strain.

303 Cf. App. A. 304 Throughout this thesis, the phrase ‘pressure’ has been used repeatedly to describe the force of attacks on retreating retreats. It is a convenient phrase, utilised to translate the physical metaphors of weight and pressing that are commonly used by ancient authors to describe combat, particularly by Greek authors (Lendon, 1999, 281, 285–6, 289–90. Eg: Polyb. 11.24, 18.24; Livy 33.15). Latin tradition: Lendon, 1999, 285–6, 295–6, 316, 325–6. Balance between force and resistance: Lendon, 286–7. Caesar also emphasises other factors that would affect the vitality of combatants, such as wounds and exhaustion, which reduced the vis of combatants (Lendon, 287). 305 Sabin, 2000, 6. 306 Lendon, 1999, 281, 285. 307 Ibid., 286–7. Missile weapons: See Zhmodikov, 2000, cf. Slavik, 2018. Missile weapons exerted vis according to Caesar (Lendon, 1999, 286, cf. Caes. BGall. 5.43, BCiv. 2.6; BAlex. 20). 308‘Safety distance’ is a concept emphasised by Philip Sabin, who has persuasively argued that fighting in many battles of this period was more intermittent and took longer than had previously been supposed (Sabin, 2000, 14–5). Sabin argues that safety distance was a physical space that combatants needed to maintain between themselves and the enemy during stages of battle, in order to physically rest and psychologically recover from the stress of the fighting (Sabin, 14–6). Sabin also argues that the effectiveness of flank and rear attacks may have been partly due to the fact that the victims were crowded together, depriving combatants of this safety distance (16). This idea can be applied to the fighting retreat, in that combatants who must have been already demoralised by the events that had led to the need for them to retreat, not to mention the manoeuvre itself, continued to be attacked and prevented from establishing their safety distance when it was needed most.


91

On the other hand, Thorne suggests that the pressure in battle accounts is figurative, or at

least psychological.309 The ‘face–of–battle’ approach has shown that psychology was central

in determining whether soldiers fled in battle.310 Goldsworthy, a leading face–of–battle

proponent, has observed that at a critical stage of battle, combatants are held in a tenuous

balance between advance and flight.311 Any sudden shock, such as the death of a general or a

flanking attack, could easily alter this balance and trigger a general flight.312 An army might

also have fled due to a gradual build–up of stressors, such as its suffering heavy casualties

and from exhaustion.313 Both modern and ancient evidence show that the flight of even a

handful of men could trigger a rout.314 Ancient writers, particularly Xenophon and Caesar,

repeatedly emphasise the importance of courage and confidence in determining whether or

not a rout would occur.315 It thus appears to have been a general rule of human psychology in

battle conditions that routs could occur easily and spread quickly. The psychological aspect is

thus central to explaining why a losing army would attempt to physically separate itself from

the enemy.

309 Thorne, 2007, 220. 310 Psychology is fundamental to the ‘face–of–battle’ approach, which argues that fear of the opponent’s weapons was the basic concern of soldiers in battle (Keegan 1976, 70, 100, 303; cf. Kagan, 2006, 15). This combatant–centric approach to explaining the results of battles has its critics (cf. Wheeler, 2001, 170–4; Lendon, 2004, 443, 447), but a fundamental argument of Keegan’s, that an army was in the end a mass of fearful individuals for whom the fear of death was a primary motivator, is an important one. According to A.B. Lloyd (1996, 194): ‘... we must not forget our common humanity with our ancient forebears. In many crises their psychological reactions were not particularly different from our own.’ 311 Goldsworthy, 1996, 280, 286. Also see Sabin, 1999, 13–4; 2007, 431. Certainly, ancient accounts of routs emphasise the suddenness with which armies could panic and flee (see App. C). Caesar observed this phenomenon on numerous occasions (see BGall. 6.8, 7.27–8, 51, BCiv. 2.34, 3.69; BAlex. 31). 312 According to Livy (10.29.1), for an army to lose their general was quae res terrori alias esse solet. Cf. Xen. Hell. 5.3.5–7; Livy 30.18.13; App. B Civ. 3.69. Also see Lee 1999 211; Sabin, 2007, 432. Even the perception of a disaster could have the same psychological effect as a real one (Livy 41.18; Plut. Pyrrh. 17). 313 Note the wavering of Caesar's right wing at the Sabis after it suffered heavy casualties (Caes. BGall. 2.25). In the Battle of Ilipa in 206 BC, Polybius (11.24) also emphasised the exhaustion of the Carthaginians. On the Battle of Zela in 47 BC, the author (BAlex. 76) also mentioned the heavy enemy casualties as they were pushed back. It is argued by Goldsworthy (1996, 224–7), however, that it was penetrating the enemy's fighting line rather than the accumulation of casualties that was key in triggering a rout, but it is easy to see how one could cause the other, whether in a pitched battle or a retreat. 314

Ancient evidence: Caes. BGall. 2.25, BCiv. 1.45; Joseph. BJ 5.2.5. Modern evidence: an American soldier present at the Battle of Camden, 1780, said of a sudden American rout: ‘The cause of [flight] I cannot tell, except that everyone I saw was about to do the same. It was instantaneous. There was no effort to rally, no encouragement to fight. Officers and men joined in the flight.’ (Garret Watts, quoted in Spring, 2008, 232). A study of American soldiers in action in 1944 has similarly shown that panic was highly contagious, so that the flight of even a single man could trigger a wider rout (Rush, 2001, 316). 315 Lendon, 1999, 290–5, 295–9, 301–3. Xenophon and Caesar agree on the irrationality of panic (Lendon, 295–7) and the importance of tactics in influencing morale (also see Kagan, 2006, 188–91), with Xenophon emphasising stratagems (Lendon, 293–4) while Caesar emphasises terrain (Lendon, 302–3).


92

Armies in retreat will have been equally vulnerable to psychological shocks or accumulated

stress. Xenophon admits (Anab. 6.5.15–17) that retreat was inherently damaging to an army’s

morale, and so an army retreating from a defeat in battle would be actually more

psychologically vulnerable. Changes in the defeated army’s tactical situation, such as moving

over unfavourable ground or the closing of an intended escape route, could readily spark

panic.316 From the ‘face–of–battle’ view, the failure of retreats in these situations was due to

the fear that orderly retreat was no longer possible, and so men fled to avoid the weapons of

the enemy since their formations no longer appeared to offer sufficient protection. As soon as

even a few men believed that they were safer in flight than in formation and started to slip

away, others would follow and an orderly retreat would disintegrate into a rout.

To return to the physical aspect of the pressure experienced by a losing army, Koon has

argued that the gradual pushing back of lines, which Sabin believes to have occurred

naturally in ancient battles, was probably a result of the repeated re–establishment by

combatants of their ‘safety distance’.317 However, an army in retreat that was being heavily

pursued by emboldened enemies may not have been able to re–establish this ‘safety distance’

at any stage, and insufficient space to manoeuvre or some sort of physical disorder could

cause a withdrawal to become confused and turn into a rout.318 Battlefield manoeuvring was

difficult at the best of times, and this would be more difficult under the psychological strain

of defeat or imminent defeat, not to mention the need for physical self–defence against the

advancing enemy.319 During a retreat, the chances of a major psychological shock occurring,

such as the death of a senior officer, can only have increased if a safety distance could not be

established. Thus, by stopping beaten fighters from resting physically, pursuit also caused

psychological strain to build and increased the chances that the retreating army would suffer

some shock or fright that would spark a general rout.

316 At Mantinea (207 BC) when the mercenaries of Machanidas discovered their avenue of escape has been blocked, a rout ensued (Polyb. 11.17). The rout of the Spartans at Sellasia in 222 BC (Polyb. 2.68) was due to poor ground, but it was this poor ground that made their escape route untenable. Poor terrain, making escape difficult, also contributed to a rout of Macedonians in 198 BC (Livy 32.12). The rout of a Roman consular army c. 281 BC is described as being both due to locus iniquus and the attack of enemy reserves (Livy 6.24). 317 Koon, 2010, 85. 318 Livy 33.18; App. Syr. 6.35, BCiv. 4.128. Also note Polybius' criticism (2.33) of Flaminius' deployment with the back of his army to a river in 223 BC. Regarding this criticism by Polybius, cf. the discussion of 'safety distance' below (ref. Sabin, 2000, 14–6). 319 Goldsworthy, 1996, 25, 57, 176–77. Psychological strain: cf. Spring, 2008, 64–65.


93

Considerations Against Pursuit

Considering the potential impact of pursuit on a retiring army, one could assume that a

victorious army would almost always pursue its enemy. However, there are numerous

examples of a defeated army retreating from the field with limited or even no pursuit by the

enemy.320 In the following discussion, examples in which victorious armies or generals chose

not to pursue will be analysed in order to draw out the reasons for doing so.

Chapter One has discussed several scenarios in which one or both armies retreated without

interference.321 On other occasions both armies might have retreated at the same time, or one

allowed another to retreat, as a precaution due to the looming advance of unknown forces

(Livy 25.19.3–4). At the Battle of Geronium, 217 BC, the Roman magister equitum,

Minucius (cos. 221), having advanced against Hannibal Barca, found himself the victim of a

Carthaginian ambush:

... the Roman light infantry were forced off the field by the numbers of the

Carthaginian horse, and, falling back on the legions, threw them into confusion, while

at the same time, on the signal being given to those [Carthaginian forces] lying in

ambush, they appeared from all directions and attacked, upon which not only the

Roman light infantry but their whole army found itself in a most perilous position. It

was now that Fabius, seeing the state of matters and seriously fearing a total disaster,

came up in haste with his own army to assist. On his approach the Romans again

plucked up courage, although they had now entirely broken their ranks, and collecting

round the standards retreated and took refuge under cover of Fabius' force after losing

many of their light–armed troops, but still more of the legionaries and the very best

men among them. Hannibal, being afraid of the legions, which, quite fresh and in

admirable order, had come to the help of their comrades, abandoned the pursuit and

brought the battle to a close.322

320 Xen. Anab. 6.5.25–32; Hell. 6.5.13–4, 7.5.24–5; Polyb. 3.74, 84; Caes. BGall. 7.51, BCiv. 1.44, 6; Sall. Jug. 103; Livy 25.37, 26.10–1, 27.49. 321 To recall, in the case of an indecisive battle, both armies might have disengaged without further fighting. Nightfall could not only break up an ongoing battle, but the darkness could prevent pursuit (Ch 1, 54–5). Particularly harsh weather could have a similar effect (Ch. 1, 54–5). 322 Polyb. 3.105.3–7. Trans. H. J. Edwards, 1922–7.


94

In this case, there is a commonality in the reason for Minucius’ retreat and Hannibal’s

restraint of pursuit. That is, the Romans retreated because they were thrown into disorder, and

Hannibal refused to pursue because he saw that the reinforcing Romans were in good order. It

appears that Hannibal pressed the attack keenly while the Romans appeared confused, but

once he was faced with regular and orderly formations, he decided to desist from further

fighting. Polybius (3.105.8) relates the opinion that, had Fabius not arrived, the Romans

would have suffered a ‘total disaster’. As such, the opportunity for pursuit was presented by

the disorder of Minucius’ army, but the opportunity closed due to the Romans’ reinforcement

with ordered formations. Hannibal made the decision not to pursue because he feared that the

Roman reinforcements might counterattack if a pursuit was made.323 This decision had real

grounding, for there are various examples of pursuits ending badly for the attackers.324 A

fundamental risk of pursuit was that if pursuers became disordered the enemy could possibly

counterattack, and, therefore, some victorious generals deliberately restrained pursuit by their

excited soldiers.325 The presence of reserves was thus a key factor in a general’s decision to

avoid further fighting and allow a retreat, as occurred at Geronium.

Now consider again Livy’s description of the indecisive battle between Hannibal and the

Roman consuls of 212 BC, in which both armies retreated from fear of an unidentified army

appearing on the horizon.326 As the heavy infantry had yet to engage, there was nothing that

had thrown either side into disorder, and since there was nothing that indicated that one side

was victorious over the other, there was little physical opportunity or psychological

motivation for pursuit. In other words, in this case the presence of order among the retreating

enemy (or the absence of disorder) provided no opportunity for pursuit, and the threat (if not

the reality) of enemy reserves again proved an important factor in preventing interference

with the enemy retreat.

Consider again part of Caesar’s account (BCiv. 1.46) of his organised retreat at Battle of

Ilerda, 49 BC, after his forces had found themselves fighting on an uphill slope from which

they could not retreat safely:

323 Onas. Strat. 22.3. Also see Caes. BGall. 1.24.6. 324 Xen. Hell. 3.5.19–20, 4.2.22–3, 5.3.5; Caes. BGall. 8.12, BCiv. 3.46; App. Hisp. 58, Mith. 85. 325 Polyb. 10.31; Livy 25.37.13–5. Onasander (Strat.. 27) points out the dangers of disorder in pursuit. 326 Livy 29.19.4–5. Cf. Ch 1, 49–51.


95

When they had fought in this way continuously for five hours, and our men were

being grievously harassed by superior numbers, having spent all their missiles, they

draw their swords and, breasting the hill, charge the cohorts, and after laying a few

low, they force the rest to retreat (reliquos sese convertere cogunt). When the cohorts

were thus pushed close up to the wall, and to some extent driven by terror to enter the

town, an easy withdrawal was allowed our men (facilis est nostris receptus datus).

Our cavalry, however, on each flank, though it had been stationed on low–lying

ground at the foot of the cliff, yet forces its way with the utmost valour to the ridge,

and, riding between the two lines of battle, allows our men a more convenient and

safer withdrawal (commodiorem ac tutiorem nostris receptum dat).327

In this case, the Pompeians did not pursue Caesar’s men from the slope in front of Ilerda even

though they themselves held the advantage of high ground. Evidently, they were

psychologically cowed (propter terrorem) by Caesar’s charge, and probably by the

exhaustion of a long struggle (pugnatum continenter horis quinque). Caesar’s cavalry also

loomed and so threatened any would–be pursuers. Thus, pursuit was prevented by both

psychological and tactical factors, especially the previous repulse suffered by the Pompeian

infantry, and the presence of Caesar’s cavalry, who were fresh reserves capable of

counterattacking any pursuers.

Other than fear of orderly formations and of the introduction of reserves, there were other

reasons why an army might have declined to pursue a retreating enemy. As noted, high

ground gave a physical and psychological advantage to a retreating force, and it was

physically difficult and even perilous for pursuers to attack uphill. As a result, a number of

victorious generals either did not pursue a defeated enemy uphill, or failed to do so

successfully.328 A defended camp might also have deterred pursuit.329 Thus, if a defeated

army was very close to its place of refuge, the victors perhaps refrained from pursuit, so as

not to risk casualties by engaging a retreating enemy close to a strong defensive position.

327 Trans. A. G. Peskett, 1914. 328 Eg. BAlex. 40, 74–6; Sall. Jug. 98; Plut. Aem. 20.5–10. Also see Xen. Anab. 4.8.10–4, 6.5.12–20, Hell. 3.5.19–20, 5.4.42–5; Polyb. 11.22, 18.22, 25; Caes. BGall. 2.23, 5.35, 6.40, 7.51, BCiv. 1.44–5, BAlex. 40; Livy 5.38, 27.18, 28.2, 32.12, 38.21, 39.31; Sall. Jug. 50–1. 329 Cf. Ch. 1, 62–72.


96

At Trebia (Pol 3.74) and Trasimene (Polyb. 3.84), large numbers of Romans managed to

retreat and then withdraw unmolested because they had broken through the enemy battleline

and were no longer directly opposed by a large body of troops. In the retreat of the

argyraspides at Gabiene, as they had already routed their opposing phalanx and a formed

phalanx, they had little to fear from cavalry.330 At the Battle of Metaurus, 207 BC, the

victorious Marcus Livius Salinator chose to let some Carthaginian troops escape (Livy

27.49), allegedly because his men were exhausted, and he wished the broken enemy to spread

news of their defeat. Such examples show that the decision not to pursue might arise from

one or more reasons, or in the case of Trebia, that a victorious army may not be presented

with a clear opportunity for pursuit.

For all of these reasons against pursuit of a retreating enemy, it is hard to imagine that victors

in battle would easily remain passive and allow their enemies an easy escape. It is telling that

there are many examples of retreats and withdrawals beginning in an orderly fashion but then

breaking down into flights under enemy pressure.331 There were many reasons for pursuing a

retreating enemy, in particular to defeat them decisively.332 In addition, the sense of imminent

victory will have stimulated the victors, and the impact of this was sharpened by the fact that

retreat had a demoralising effect on those doing the retreating.333 Since confidence was often

more important for intimidating the enemy than actual strength, even an army that was

physically superior could invite attack by retreating.334 Caesar (BGall. 6.8.1), in describing

the motivations of the Gauls choosing to follow Labienus’ feigned retreat in 53 BC, states

330 Diod. Sic. 19.43.1–5. They were also experienced and elite troops (Bosworth, 2002, 138–40). 331 See App. A. 332 Firstly, there were material advantages to a decisive victory, for if more enemy combatants were killed, more plunder and trophies would be available for the victors (Harris, 1979, 103–4; Hoyos, 2007, 67; Eckstein, 2009, 85–9). The victors also achieved greater 'bragging rights' for participating in a more decisive victory (eg. Sall. Jug. 53. Also see Lendon, 2005, 185–91; McDonnell, 2006, 70–1). There may also have been a fear that a retreating army could escape to fight again another time (cf. Onas. Strat. 22.3: ‘ ... or anticipating some greater misfortune than they are about to suffer, soldiers regard the future with greater fear.’) Finally, a decisive victory would free soldiers from the rigors of campaigning. Presumably, the longer the campaign, and the more important other factors such the need of a soldier to return to farmland, (de Ligt, 2012, 172), the availability of otherwise of booty (De Ligt, 124) or attend other private affairs (cf. Japanese bushi: Friday, 2004, 119), the more intense this desire to 'get it over with' might have been. Probably for such reasons, it was not unheard of for Roman soldiers to pressure, or attempt to pressure, their commanders into fighting (Caes. BGall. 7.19, 7.47, 52, BCiv. 1.64, 71, 3.74; Livy 25.14, 37, 33.8, 44.36). 333 Bloodlust: Pol 18.26; Caesar BCiv. 31; Livy 25.37, 37.43). Sense of imminent victory: Xen. Hell. 6.2.20–3; Caes. BGall. 2.27, BHisp. 31; Livy 10.38, Livy 15.7, 28.15, 33.36, 35.5; App. Mith. 44). 334 Goldsworthy, 1996, 285–6. When Sabin (2007, 413, 6) states that an army in retreat was usually pursued, he may do so with some justification.


97

that the Gauls would have felt themselves cowardly if they did not attack the retreating

Romans. Others in their position could easily have felt the same.

To summarise, victorious armies not uncommonly decided against pursuit for fear of

becoming disordered against an enemy that was still dangerous, or possessed fresh troops in

reserve, and thereby becoming vulnerable to counterattack. The enemy’s proximity to high

ground or a strong defensive position could also cause potential pursuers to reconsider an

attack. For all of the tactical reasons against pursuit, however, the desire to attack will have

remained strong. It was thus probably the natural reaction of most soldiers, and armies, to

pursue a retreating or withdrawing enemy.

Conclusion

There is plenty of evidence to suggest that if there was an opportunity for decisive victory

against a retiring army, a pursuit would be mounted. Pursuit might have taken the form of a

victorious battleline simply pressing forward, or a cavalry attack. A cavalry pursuit severely

delayed retreating forces and could throw them into serious disorder (Livy 28.16.3–5), even

forcing them to abort a withdrawal and halt in an unfavourable position (Caes. BCiv. 1.79–

81). Cavalry pursuit was thus an ideal method of countering a retreat or withdrawal.

In the case of retreat from battle, the continuing advance of a victorious battleline could

easily cause a total rout if any shock occurred among the retreating troops (Livy 30.18.13).

The morale of a retreating army will probably have been tenuous, and the movement itself

could be physically difficult when combined with fighting off advancing enemies, so that an

orderly retreat from battle could, and often did, break down into flight. Even if an army found

refuge in a camp or a walled settlement, it could be forced out of this location and need to

withdraw further. In this situation (or whenever an army retired over a long distance) the

army will have needed to adopt a columnar marching formation and will thus have been

further exposed to cavalry pursuit, which could cause delay and disorder and even allow

following heavy infantry to catch up and make a full–scale attack.335 An army in retreat or

withdrawal while under pursuit will not have been able to rest physically or mentally, and this

enemy pressure and increasing psychological strain could easily stimulate despair or panic.

335 Livy 28.16–1–7; Sabin, 2007, 413.


98

Due to the difficulties in making a successful and orderly retreat from battle, Sabin argues

correctly that the general trend was for a defeated army to suffer disproportionate casualties

as it attempted to escape.336

The weight of examples suggests that more often than not, an army retreating directly from

defeat in battle will have been pursued in some way, but there are also numerous reasons that

victors may not have pursued a retreating enemy.337 Foremost among these reasons were the

appearance of order and strength among the retreating men, who thereby showed little

physical or psychological weakness for pursuers to capitalise on, the presence of hostile

reserves, and proximity to high ground or a fortified position. The essential fear for a

victorious general must have been that his forces would fall into disorder during pursuit,

potentially exposing the victors to a reversal of fortunes if there was some counterattack.

To summarise, an army in retreat or withdrawal will have needed to take some action to

prevent or counter enemy pursuit if it wanted to maximise its chances of successful

disengagement or escape. In particular, it was important to take action that preyed on the

victors’ fears of becoming disordered and suffering a counterattack. The tactics that the

Romans could use to achieve these things will be the subject of the following two chapters.

336 Sabin, 2007, 413, 416. 337 See App. A.


99

Antonius covered not only his rear, but also both his flanks, with numerous javelineers and

slingers, led his army in the form of a hollow square, and gave orders to his horsemen to rout

the enemy when they attacked, but after routing them not to pursue them further.

(Plut. Ant. 42.1)

Chapter Three

Roman Retreat Tactics, Part I: Ratio and Combined Arms


100

Introduction

In 36 BC, the imperator Marcus Antonius led an enormous army into Parthia to avenge the

crushing defeat of Marcus Licinius Crassus in 53 BC.338 The campaign went awry, and after

losing his baggage train and failing to force the Parthians into a decisive battle, Antonius was

forced to begin a long withdrawal back to Armenia (Plut. Ant. 38–40). During this withdrawal

the Romans were attacked by the Parthians, who harassed the Romans with massed archery

and threatened them with a defeat reminiscent of Crassus.339 Antonius made a number of

tactical responses, including organising his infantry into mobile squares, counterattacking

with cavalry, and supporting the soldiers with javelineers and slingers (Ant. 42.1). Were these

tactics typical of the Roman army in general, and were there other factors and techniques that

the Romans could rely upon to secure their retrograde manoeuvres?

The consequences of a failed retreat or withdrawal could be dire.340 An army would have

striven to avoid this scenario at all costs. However, I have argued that pursuit and attacks by

victorious armies often caused retrograde movements to fail, whether they were carried out

from battle, out of an encampment or city, or over a longer march. How then, did armies act

to prevent their retreats from failing in the face of an active enemy pursuit? In this chapter,

the training, tactics and junior leadership practices of the Republican Roman army will be

examined in the context of the conduct of retreats and withdrawals. Although there will be

more in–depth discussion of retreats than of withdrawals, the latter will be examined. This is

for the reason that the evidence for the tactics associated with retreats from battle is more

detailed, while withdrawals often took the form of marches, which are well understood. Also,

the focus of this chapter is on the mechanics of Roman retreat tactics rather than generalship.

The importance of the general will be touched upon in the opening discussion below,

particularly in relation to the commitment of reserves to support a retreat, but this will not be

explored in detail until the following chapter.

338 Syme, 2002, 263. 339 Plut. Ant. 41, cf. Plut. Crass. 24. 340 Sabin, 2007, 413.


101

I will begin with a brief examination of two accounts of successful retreats, showing some of

their common factors. One major factor, the maintenance of order among the infantry, will

then be examined in detail, beginning with an explanation of what ‘order’ actually entailed.

Two other factors, the use of high ground and reserves, have been touched on already, but I

will expand further in the following discussion. Following the discussion of order, the

training and tactics of Roman armies, particularly of the milites that formed the bulk of the

legions, will be analysed in reference to the conduct of retreats and withdrawals, and the

maintenance of order. The manner in which legionary reserves were used to support a retreat

will be examined in detail here. Following this, the roles of subordinate officers in a retreat or

withdrawal will be analysed, beginning with the principales, centuriones, and then others

such as tribuni. Finally, the tactical roles of light infantry and cavalry will be considered to

give a full picture of the Roman army’s combined–arms approach to retrograde manoeuvres.

The roles of light infantry and cavalry as reserves in support of a retreat will also be

discussed here. The conclusion will summarise the characteristics of successful retreats and

withdrawals and give an outline of the methods and tactics that Roman forces could use to

retreat or withdraw successfully.

Factors of Success in Retreats and Withdrawals

When an army retreated from battle, pursuit could easily cause it to break down, but the

presence of nearby refuge, such as a camp, steep hill, or a town or city, could reduce the

chance of significant disorder. An enemy would be less likely to pursue retreating forces that

maintained order in their formations, and if the retreating army had maintained some reserve

force that could counter any pursuit.341 Were there other factors that generally contributed to

a successful retreat? I will consider two examples below in which retreating forces

successfully fought their way to safety in order to draw out these key factors.

First, an example from the Bellum Alexandrinum (40), which describes the Battle of

Nicopolis in 48 BC between the Caesarian army of Gnaeus Domitius Calvus and the Pontic

army of Pharnaces II. The majority of the Roman army was overwhelmed by the pike–armed

Pontic infantry and routed, leaving the more experienced Legio XXXVI (34) the only force

341 The importance of order has been touched upon previously, particularly in regard to the Roman retreats at Geronium, 217 BC (Polyb. 3.105.3–7) and Ilerda, 49 BC (Caes. BCiv. 1.46).


102

remaining on the field:

Consequently the king's forces, victorious on their own right wing and in the

centre of the line, now turned upon the Thirty–Sixth legion. The latter,

nevertheless, bore up bravely under the victors' attack and, though surrounded

by large enemy forces, yet with consummate presence of mind formed a circle

and so made a fighting withdrawal to the foothills, where Pharnaces was loth to

pursue it owing to the hilly nature of the ground. And so ... the Thirty–Sixth

legion retired to higher ground with losses not exceeding 250 men.342

As the author of the Bellum Alexandrinum reveals, the Legio XXXVI (unlike the rest of the

army), resisted the enemy attack (fortiter vincentium impetum sustinuit), changed formation

(in orbem), and conducted a fighting retreat to nearby high ground (pugnans ... se recepit ad

radices montium), which it reached safely.343 The Pontic king was reluctant to continue his

pursuit on ground where the Roman legion would have a tactical advantage (propter

iniquitatem loci), and so the battle ended.

Several features stand out in this account. Since the Roman legion fought in close order (in

orbem), the legion changed its formation while still fighting (impetum sustinuit), and then it

moved in a specific direction (ad radices montium),344 it was almost certainly working under

the coordination of an overall leader (probably Calvus) who (along with other leaders)

ensured that the legion moved gradually and in a controlled manner. As it was extremely

difficult for troops to manoeuvre in close formation on the battlefield, even if well–trained, to

be able to conduct this manoeuvre under enemy pressure the men of Legio XXXVI must have

had some relevant training.345 The use of terrain played a decisive role, as it was only when

the Romans reached higher ground that the victorious Pontic army was discouraged from

further pursuit.346 The maintenance of good order and clear direction toward a place of

342 Trans. A. G. Way, 1955. 343In orbem may also be taken to mean a 'square' formation (see Caes. BGall. 5.33). The phrase is used several times in the Caesarian corpus (Caes. BGall. 5.33; BAfr. 15), and it certainly refers to a formation in which the Romans were defending themselves on all sides. 344 This phrase also appears in other texts (eg. Livy 22.17.2, 37.20.3, 40.22.1), and in each instance highlights the central role such terrain played in infantry manoeuvre. 345 Goldsworthy, 1996, 176–77. 346 This was not timidity on the part of Pharnaces, but typical caution by ancient commanders (e.g. Livy 28.15.11; Caes. BGall. 7.51). Also see Gilliver, 2007, 138.


103

refuge, the former a product of the army’s discipline and the latter a function of good

leadership, were therefore key features of this successful retreat.

Turning to a second example, in the case of Caesar’s retreat from the slope of Ilerda in 49

BC, recall that some of his men rashly pursued their enemies onto the slope in front of the

town of Ilerda, where the latter had rallied and for five hours prevented Caesar’s men from

retreating by charging them whenever they attempted to do so (Caes. BCiv. 1.45.2–4). Finally,

Caesar’s men in turn charged the enemy and drove them back, eventually into the town

(BCiv. 1.46.1–2), thus cowing them (nonnulla parte propter terrorem in oppidum). Caesar’s

cavalry came forward and interposed itself between the retreating infantry and the

Pompeians, providing the opportunity for retreat (tutiorem nostris receptum dat: 1.46.3).

Clearly, the manner in which Caesar’s men moved was organised and orderly, for, so far from

moving quickly or rashly they had actually remained in the same position for five hours until

they were able to secure their retreat (1.46.1). Both sides received constant reinforcements

(1.45.7–8) but this did not break the deadlock. While the enemy held the high ground, and

Caesar’s cavalry was separated from the infantry (1.46.3), Caesar’s legionaries could not

retreat, but when they did the Pompeians declined to pursue them for two reasons: the

effectiveness of their counterattack and the presence of Caesar’s cavalry.

In this case, several factors are evident. Caesar’s troops moved in an orderly manner. His

infantry was evidently very well organised, as it continued fighting in a locus iniquus for a

long time while infantrymen were shifted in and out of the fight. This must be a reflection of

the discipline of the troops as well as of the organised leadership of Caesar and his officers.347

In addition, infantry reserves were required to be constantly fed into the battle by both sides

to prevent the men from becoming exhausted (Caes. BCiv. 1.45.7–8). The presence of cavalry

reserves facilitated the final retreat, and of course, Caesar had a camp nearby (BCiv. 1.43), to

which his army certainly withdrew. As such, key features of this successful retreat were once

again the discipline of the milites in maintaining order and Caesar’s use of infantry and

cavalry reserves to prevent a premature retreat and to stop enemy pursuit, not to mention the

proximity of a fortified camp.

347 One of Caesar’s legions that participated in the battle was the Ninth, a veteran formation (cf. Keppie, 1984, 208).


104

While pursuits commonly occurred in response to a retreat directly from the line of battle, the

observation to be drawn from most accounts of successful retreats is that the retiring armies

were eventually able to dissuade continued pursuits or attacks through the maintenance of

order among the infantry, and other tactics such as seeking refuge on higher ground or in a

camp, counterattacks, and the presence and use of reserves.348 Some of these aspects of

retreat tactics have already been discussed (eg. high ground, camps), and some will be

explored further below, while the role of the general, in tying these factors together to ensure

a safe retreat, will be explored in more detail in the next chapter. The definition and

importance of order in retreats will be addressed first, followed by an exploration of several

factors that contributed to the maintenance of order and prevention of enemy pursuit,

including training, signalling, infantry reserve lines, junior leadership, and the use of light

infantry and cavalry.

Physical Order in Retrograde Movements

Greek writers are thought to have had a generalised conception of battle tactics in which

disorder was fatal to infantry formations, and order was maintained primarily through well–

planned deployment and use of flat ground, and so this geometric and schematic view of

tactics, emphasising proper deployment and use of flat ground, was believed to be

appropriate to maintaining order in the phalanx, whether in advance or retreat.349 To elaborate

on this concept, we can turn briefly to the Macedonian–style phalanx, a formation that was

almost invulnerable to frontal attack as long as it was correctly arranged.350 This can be seen

both in the case of the retreat of the argyraspydes at Gabiene in 316 BC (Diod. Sic. 19.43.4–

5), who easily fended off pursuing cavalry, and the Seleucid phalanx’s attempt to retreat at

Magnesia (App. Syr. 35), during which large numbers of Roman legionaries were reluctant to

close with it. Appian (Syr. 6.35) states that the Romans feared the ‘discipline and solidity’

(ἐμπειρίαν τε καὶ πυκνότητα) of the enemy, and the Romans only made a successful

attack after the panic of a number of elephants trampled men and so disrupted the phalanx

(6.35). Diodorus Siculus (19.84.6–7) also notes that when Demetrius Poliorcetes retreated

after his defeat the battle of Gaza, 312 BC, the open ground allowed his troops to retreat in

348 Higher ground: BAlex. 40; Sall. Jug. 98. Counterattacks: Caes. BCiv. 1.79. Reserves: Polyb. 3.105; Caes. BGall. 1.25. 349 Lendon, 1999, 282–5. 350 Polyb. 18.30.11, 31.5.


105

good order, making the enemy reluctant to pursue.

A phalanx that fell into disorder was a different matter. Polybius (2.65–9) describes such an

event in his account of the Battle of Sellasia, 222 BC. A Spartan phalanx of perioikoi and

other allies had been posted on a hill known as Euas (2.65), and at the height of the battle

they were attacked from by a force of Illyrians and Macedonian phalangites that advanced

uphill to meet them (2.68). Polybius criticises Eucleidas, the Spartan commander on Euas, for

squandering his superior position on high ground by waiting to meet the enemy attack,

arguing that he should have charged and disordered the Macedonian phalanx from their

higher position and driven it from the slope.351 The result was that they met the Macedonians

near the summit of the hill, so that after being pushed back they were, before long, fighting

desperately to retain control of the summit itself (2.68.7–8). They were then thrust back from

the summit and continued to retreat, but now downhill on the reverse slope of Euas (2.68.9).

In this position they were unable to reform, nor to break away from the enemy who was

following them aggressively, nor to push back the now advantageously–positioned enemy. As

Polybius puts it: ‘The consequence was that very soon they had to turn and take to a flight

(φυγῇ) which proved disastrous (ὀλεθρίῳ), as, for a long distance, it was over difficult and

precipitous ground (κρημνώδη καὶ δύσβατον) (Polyb. 2.68.10). In this case, the retreat

failed as a result of a combination of heavy pursuit and unfavourable terrain, the latter factor

brought on by the former. Conversely at Pydna, the pursuing Macedonian phalanx became

disordered from advancing uphill, and quickly fell victim to a counterattack (Plut. Aem. 20.7–

10). Evidently, the phalanx was as difficult to attack frontally in retreat as it was in advance,

as long as it maintained its proper formation. If the phalanx fell to disorder, however, it was

extremely vulnerable (Polyb. 18.28-9). Thus, in the case of a retreating phalanx, the extent to

which it suffered pursuit will, in most cases, have indeed been closely linked with the

orderliness it maintained in doing so.

351 Polyb. 2.68.3–5. It should be noted that the Spartans’ tactics were not totally without merit, as forcing the Macedonian phalanx to ascend the entire slope will have tired them out before the two lines clashed.


106

Since the Romans did not utilise the phalanx, but looser and more mobile formations, what

was their view of order in retreat?352 Onasander (Strat. 27) gives the following advice on

order:

One should command both retreats (ὑποχωρήσεις) and pursuits (διώξεις) to be

made in formation (τάξει), so that, if defeated, the soldiers may suffer less injury,

when in their flight (φυγαῖς) they encounter the enemy, by not being scattered, man

by man, and, if successful, they may inflict greater injury on the enemy by keeping

their ranks and companies unbroken (τάξεις καὶ λόχους ἰσχυρότεροι), appearing

stronger to the fugitives, and moreover being safer themselves. For often the enemy,

observing their opponents advancing without order, by a concerted plan about–face,

form ranks once more and reverse the pursuit. In a word, the general should say that

nothing is more advantageous to his men than remaining in rank (μένειν ἐν τάξει),

and nothing more dangerous than breaking ranks.353

While clearly influenced by the Greek emphasis on order, Onasander wrote for a Roman

audience.354 He does not speak here of the rigorous emphasis on deployment and orderly

arrangement that eutaxis implied.355 He discusses the advantages of maintaining order in

retreat (or even ‘flight’) in more general terms, stating that it was easier for ordered

formations to defend themselves against pursuers, which also made them appear more

intimidating.356 It is unusual that Onasander does not distinguish between an orderly retreat

(ὑποχώρησις) and a flight (φυγή), and if anything this demonstrates that his use of

terminology was not precisely technical. However, he argues clearly that retreating soldiers

that retained their close formation were not only more able to fight off pursuers, but were less

likely to be pursued in the first place.

352 Phang, 2008, 50–1. 353 Trans. from the 1928 Loeb Edition transcribed on lacuscurtius.org by the ‘Illinois Greek Club’. 354 Campbell, 1987, 13. Also see Formisano, 2017, 145. 355 Cf. Polyb. 18.31–2. In Greek, maintaining physical order within heavy infantry formations was known as taxis (‘proper formation’) or eutaxis (‘good order’). See Lendon, 1999, 282. 356 Close formation was also comforting, and for this reason soldiers will naturally crowd together when afraid (Rush, 2001, 320). This was doubly important in the retreat, when the enemy might be emboldened by the possibility of imminent victory (Xen. Anab. 6.5.17). However, it is notable that for swordsmen this could become counterproductive if the ranks became so compressed for men to wield their weapons properly (Caes. BGall. 2.25).


107

Latin sources give a great deal of further evidence on the Roman concept of order or ratio.357

The common use of the term servare ordines, to describe soldiers remaining in their proper

place in the ranks, demonstrates that the idea of staying in formation was straightforward for

Roman readers.358 The phrases conferti ordines (‘compact ranks’), in orbem (‘in a

circle/square’) and agmine quadrato (‘having formed a square’), also occur repeatedly,

especially in Livy, showing that legionaries regularly kept compact formations in battle.359 In

particular, the Roman concept of order emphasised keeping legionaries close to the manipular

standards, the signa, under any circumstances.360 The Roman general and writer Frontinus for

instance advocated decimation for men who abandoned their places in the line.361 Roman

soldiers also appeared to have remained facing the enemy while retreating (Caes. BCiv.

1.80.4), thereby showing they were ready to fight. In general then the Romans, whose

formations were more widely spaced than in the Macedonian phalanx, still considered

maintaining compact formations to be essential in battle.362 For retreating milites, being

combat–ready by remaining in somewhat close proximity to one another and their standards,

still facing the enemy, was evidently considered to be important and conventional.

So was ‘order’ simply the maintenance of a relatively close formation? Marcus Junkelmann

has provided a brief analysis of fighting retreats, and argued that retreating too quickly under

enemy pressure was basically impossible as deep formations would quickly become confused

and disintegrate.363 Yet he argues that as long as infantry maintained their formation and

357 For convenience, I will utilise the Latin term ratio as a synonym for military order throughout the thesis (cf. Caes. BGall. 2.19.1, 2.22.1, 3.14.4, 4.23.5; Livy 24.17.5.3, 24.28.9.2, 24.48.4.2, 27.15.12.1, etc) 358 Eg. Caes. BGall. 4.26.1.2, 7.23.5.3, BCiv. 1.71.3.5, 1.71.4.1, 2.41.6.5; Sall. Jug. 51.1, 80.2–3; Livy 8.34.10.4, 9.19.8.1, 24.48.12.1, 28.15.6.2, 30.18.12.4, 13.1, 42.64.10.3. Another term for order is ratio (‘tactics’, ‘calculation’, ‘method’), commonly used by Caesar (eg. BGall. 2.10.5, 2.19.1, 2.22.1, 3.14.4, 4.1.7, 7.1.6, 7.4.1, 7.6.3, BCiv. 1.44.3, 1.65.5, 2.18.6, 3.47.1, 3.74.2, 3.75.1). Ordinare is less common (cf. Livy 2.21.4, 34.14.6, 42.59.6; Front. Strat. 1.6.3). Of interest, according to Phang (2008, 51), disciplina militaris did not emphasise spatial order as did eutaxia. 359 Conferti ordines: Caes. BCiv. 1.71; Livy 31.43.3, 32.17.8, 42.59.5, 42.65.8. In orbem: Caes. BGall. 4.33.1, BAlex. 40.3, BAfr. 15.3; Livy 1.17.6, 4.39.4, 21.56.2, 23.27.7, 28.22.15, 28.33.15, 42.65.7. Agmine quadrato: Livy 2.6.6, 7.29.6, 10.14.7, 21.5.16, 21.32.1, 31.37.1, 35.3.2, 36.10.4, 39.30.9, 44.9.6. 360 Cf. Caes. BCiv. 1.44. The Romans went to great lengths to ensure men followed their signa (eg. Caes. BGall. 5.37.5; Phang, 2008, 65–6), even using them as a form of bait by throwing them into the enemy so that legionaries would attack more ferociously in order to recover them (Caes. BGall. 4.25–6; Livy 34.46.12; Plut. Aem. 20). 361 Front. Strat. 4.1.34. Also note the emphasis on order and holding ranks in the drill commands recorded in Maurice's Strategikon (12.14). 362 Polyb. 18.29.1–5, 18.29–31; Phang, 2008, 50–1. Also see Lendon, 1999, 285–287. 363 Junkelmann, 1986, 241.


108

moved slowly, they could potentially withdraw safely even while under attack by the

enemy.364 Junkelmann thus echoes what has been discussed above, that retreating infantry

should maintain relatively close and ordered ranks, but he adds that a steady, gradual rate of

movement was also important. For example, consider the following part of Livy’s (42.59.4–

5) account of the Battle of Callinicus, 171 BC, which describes how a division of Thessalian

auxiliary cavalry retreated:

For, retreating gradually (cedentes enim sensim) with formation intact

(integris ordinibus), after they had joined the auxiliaries of Eumenes, with him they

both provided a safe retreat (receptum) between their units (inter ordines) for their

allies scattered in flight, and when the enemy pressed on in less close order (minus

conferti), they even risked advancing and met many of the fugitives half–way. Nor did

the king's men, now themselves scattered in a general pursuit (passim sequendo), dare

to close in battle with an enemy in formation and advancing steadily

(ordinatis et certo incedentibus gradu).365

It is true that this account describes a retreat by cavalry and not by infantry, but it still

demonstrates the importance of close order (integris ordinibus ... conferti) and gradual

movement (cedentes sensim ... gradu), specifically as pursuers were reluctant to engage with

troops that retained such order. There are several other accounts of infantry retreats, mostly

successful, in which a gradual ‘step by step’ rate of movement is also described.366 As such, it

appears that maintaining order in retreating infantry formations consisted of at least two

components: close formation and gradual movement.

What links these two factors? Since a fighting retreat was so taxing upon its participants, one

could wonder if it would not have been better for a losing army to get away as quickly as

possible, and so reduce the time that retiring troops were exposed to enemy pressure. In fact,

just as a desire to ‘get it over with’ might compel soldiers to attack impetuously, it is not

difficult to imagine that soldiers in retreat would want to increase their pace as their panic and

enemy pressure mounted.367 As Goldsworthy has pointed out, it was challenging for large

364 Ibid, 241. 365 Trans. Evan T. Sage and Alfred C. Schlesinger, 1938. 366 Xen. Anab. 5.2.32; Polyb. 11.24; App. Syr. 35, BCiv. 2.80, 3.69, 4.128. Also note Livy 42.59.4–6. 367 Desire to get it over with: Goldsworthy, 1996, 191–2. Soldiers increasing their pace under pressure: see


109

bodies of troops to manoeuvre on the battlefield, and orderly movement could easily degrade

into confusion, but he has also noted that a rapid pace in an advance could be a symptom of

fear, and that such haste could lead to great disorder in a formation.368 Junkelmann echoes

this point, but applies it specifically to the fighting retreat.369 So, it is specifically to manage

these difficulties that close formation and gradual movement went hand–in–hand: infantry

needed to maintain close formation to resist enemy attacks and dissuade pursuers by showing

a solid line of armed men, and they also needed to move at a gradual pace so that they could

carefully maintain this close formation. Close formation and gradual rate of movement were

therefore both necessary to preserve order, servare ordines, during a retreat.370

Legionary Training, Tactics and Reserve Lines

It is clear that maniples and cohorts, and even cavalry units, had to maintain close formation

and a gradual pace in order to retreat or withdraw in an orderly manner, but this would have

been difficult under enemy pressure. The management of order and pace cannot always have

been simple in the relatively complex Roman triplex acies, consisting as it did of many

separate manipuli and cohortes divided into separate lines. Were the soldiers specifically

trained to retire in an orderly manner? And how did the lines of the triplex acies, and sub–

units of the legion (maniples and cohorts) interact in retreat? The answers to these questions

are limited by a lack of evidence and are also partly interlinked, but I will attempt to show

how Roman heavy infantry maintained order and countered pursuit during retreat or

withdrawal.371

Appian's account (BCiv. 1.128) of the retreat of Brutus' infantry at Philippi, in 42 BC. 368 Goldsworthy, 1996, 176–7, 194, 197. Even in a victorious pursuit, maintaining a gradual rate of movement remained important for excited troops (Xen. Anab. 8.19). Also see Xenophon (Hell. 6.2.20–3) for the difficulty of changing formation with a phalanx during combat. 369 Junkelmann, 1986, 241. 370 Despite withdrawals being carried out in different marching formations, these same factors will have applied to some extent, as marching infantry could be highly exposed to enemy marauders due to the increased distance and duration of the movement (cf. Caes. BCiv. 1.78–80). 371 These questions are not simple to answer, as the sources usually do not systematically discuss the specifics of training practices, manipular tactics (Goldsworthy, 1996, 171–2) or signalling (Southern, 2006, 199–200, 203). Also see Sabin, 2007, 399–400.


110

In terms of training practices, the Roman army always stood out from most of its

contemporaries for its disciplined attitude.372 However, due to the fact that Middle

Republican armies were raised yearly, they would not necessarily be trained over the long

term, with the exception of armies that remained in service for longer periods and so

accumulated more training and experience.373 Training outside the yearly cycle of active

campaigning may have taken the form of military exercises performed by young men on the

campus Martius, or instruction at home by family members.374 As units became longer

serving in the Late Republic and Empire, training thus became more systematic.375 Polybius

(10.20) provides an account of Scipio training his soldiers c. 209 BC, and this is the earliest

extant reference to an organised Roman training regime. Scipio exercised his troops for an

extended period in four day cycles, with a forced march on the first day, the second day being

dedicated to equipment maintenance, the third to rest, then sword and javelin practice on the

fourth day. A different training regime may be found in accounts of Scipio Aemilianus’

Iberian campaign (Livy 26.51.4). Livy (34.13) also describes another training regime that was

carried out by Cato the Elder in 195 BC, in which Cato would combine practice marches with

active campaigning and fighting to harden his men.

There are few other direct references to Middle Republican training methods. Among the

handful of Middle Republican references, given the prominence of the Scipiones originating

from Polybius, one could assume that the Scipiones were unusual in applying such systematic

training methods, although it is also possible that these training methods were not so unusual

and that Polybius simply emphasised their application to glorify his patrons.376 Evidently, an

organised system of training was considered good practice. However, these references to

training do not mention practice of battlefield manoeuvres. A partial exception was the

apparent prominence of forced marches in which men might have walked in columns, but

more emphasis was evidently placed on individual fighting skills.377 It should be noted that

the Mid–Republican army had minimal need for precise formation drill as the Macedonian

372 Keppie, 1984, 55. 373 Southern 2006, 133, 135. How Roman armies became more efficient over time: Sage, 2008, 160; Goldsworthy, 2011, 48. 374 Cic. Cael. 11; McDonnell, 2006, 181–4. 375 By the mid–first century BC the Roman army, particularly that of Caesar and his successors, had become more regularised in its organisation and training practices (Keppie, 1984, 61–7, 76–7, 132–3; Goldsworthy, Fall of Carthage, 361–2; Potter, 2014, 69–70). 376 Southern, 2006, 135. 377 Phang, 2006, 50, 57–61; Southern 2006, 135–6.


111

phalanx did, due to the more open ranks of sword–armed legionaries.378 Nevertheless, Roman

soldiers will surely have understood the importance of moving in an orderly manner during a

retreat.379 It is striking that in one of the oaths sworn by legionaries, the coniuratio, a soldier

stated that they would ‘not abandon their ranks for flight or fear, but only to take up or seek a

weapon, either to smite an enemy or to save a fellow citizen’ (Livy 22.38.2–5). This is

evidence that Middle Republican legionaries were formally indoctrinated into staying within

their ranks, as Onasander later recommended. Although training methods were still left to the

inclination of individual generals, it is at least clear that the Romans did appreciate the

importance of an organised training regime, and this surely included some practice in

battlefield movement (such as retreat) even if the Middle Republican evidence does not

specifically refer to it.380

The evidence is better for the Late Republic. In 105 BC, Rutilius Rufus instituted a legionary

training system based on that of gladiators, which was replicated by Gaius Marius and

became standard across the legions, incorporating individual weapon’s practice, forced

marches, running and jumping.381 In 46 BC, Caesar re–trained his men in response to new

enemy tactics non ut imperator exercitum veteranum victorem ... sed ut lanista tirones

gladiatores condocefacere, by instructing them exact distances to retreat and manoeuvre

when ordered as well as practising feigned retreats and counterattacks (BAfr. 71). Caesar

implies that the gladiatorial methods more commonly adopted by the Romans in the first

century BC included specific instruction on how milites should retreat in formation.

378 More open ranks: Lendon, 2005, 179–82; Rawlings, 2007, 58; Phang, 2008, 50, 57–61. 379 One can recall the Imperial–era advice of Onasander (Strat. 27), in which he recommends that soldiers must remain in rank during retreat. 380 Individual inclination: Southern, 2006, 135. Scipio: App. Hisp. 86. Cato: Livy 34.13. Also see Goldsworthy, 1996, 180–1. 381 Val. Max. 2.3.2; Plut. Mar. 13.1; Southern, 2006, 135.


112

Another way of looking at the problem is by considering Roman field signals. A general or

other officer could give simple orders by flag or more likely by bucina, cornu or tuba.382 All

legions possessed some number of bucinatores, tubicines and cornicines.383 As shown at the

Battle of Thapsus in 46 BC, in which an over–eager musician signalled a general advance

without Caesar’s approval (BAfr. 82), legionaries were in the habit of following musical

orders. The key question is whether or not there was a generally recognised signal for a

retreat. According to Southern, the tuba was typically used to signal advance and also

retreat.384 At Gergovia in 52 BC, Caesar ordered a retreat to be sounded (receptui cani iussit)

by tubae (Caes. BGall. 7.47.1). In his account of the indecisive fight between Hannibal and

the Roman consuls of 212 BC, Livy (29.19.5) states utrimque signum receptui datum,

although he does not specify the method by which this was done. Livy gives several other

examples of signals for retreats.385 The two phrases given above, receptui cani and signum

receptui, appear elsewhere.386 Further, during both the battles of Gergovia, 52 BC (Caes.

BGall. 7.51.3) and Ruspina, 46 BC (BAfr. 18.3), the signal for retreating legionaries to stop

and fight was turning the standards towards the enemy, implying that in retreat, they were

turned away from the enemy. As such, it is clear that there was at least one regular signal for

retreat used among the Roman legions, certainly in the Late Republic and probably earlier,

which in turn implies that retreats were a regular manoeuvre in Roman tactics. It stands to

reason also that if soldiers habitually retreated in response to a signal, whether it was a horn

blast or the turning of the standards, that they knew what to do, and so must also have

practised retreating.

382 Southern, 2006, 204. Flag: eg. Caes. BCiv. 3.89. Bucina: eg. Livy 3.22.6. A variety of signals could have been used, including hand signals (Lee, 1996, 201, ref. Amm. Marc. 24.6.13) and written orders (Southern, 2006, 203–4), but the latter were less suitable for a simple, but time sensitive order for a retreat, and the former would have taken longer to transmit across the army than the blow of a horn. Of course they were not a perfect method of signalling, as in the din of battle, soldiers might not hear the horn (Caes. BGall. 7.47.1). 383 Southern, 2006, 159; Sage, 2008, 19. 384 Southern, 2006, 159; cf. BAfr. 82. 385 Livy 2.62.2, 2.64.8, 4.31.3. 386 Receptui cani: Caes. BGall. 7.47.1; Livy 26.6.7, 31.42.1, 35.29.7, 37.32.6, 44.35.20. Signum receptum datum: Livy 9.32.10, 26.45.4, 29.19.5.


113

Even though there is no description of such training, the Romans were well known for

advancing on the enemy with impressive orderliness, for which they must have practised, and

given that retreat was an advance in reverse, they surely practised such movements also.387

Further, as a withdrawal often took the form of a regular march in column, the Romans could

apply the same training that allowed them to march forwards in an orderly manner. Such

training will have been simple and was perhaps too mundane to be described by the sources,

but if it was done on varying kinds of terrain and practised under some pressure, this skill will

surely have been very useful to retreating soldiers.

Perhaps even more distinctive than Roman training practices was their unique triple–line

deployment, the triplex acies, which was used regularly throughout the Middle Republic.388

Since a Roman army was generally deployed in three lines, how did these lines move during

a retreat? Note that Polybius (2.33.7–8) criticises the battle deployment of the consul Gaius

Flaminius (c. 223 BC) because of the general’s failure to prepare for retrograde movement:

... the Consul ... mismanaged the battle by deploying his force at the very edge

of the river–bank and thus rendering impossible a tactical movement peculiar to

the Romans, as he left the lines (σπείραις) no room to fall back gradually. For

had the troops been even in the slightest degree pushed back from their ground

during the battle, they would have had to throw themselves into the river, all

owing to their general's blunder.389

According to Polybius, if a Roman army deployed correctly, it was capable of carrying out

short, temporary retreats as a regular feature of their battle tactics. However, it is not clear

exactly what Polybius means. At a glance, he emphasises that the movement in question is

‘peculiar’ to the Romans, one imagines that he is referring in some way to the functioning of

triplex acies, which was also unique to the Romans. In general, Roman manipular tactics

required room to manoeuvre in all directions, but Polybius particularly emphasises space

387 Goldsworthy, 1996, 193–7. Also see Plut. Ant. 39.4. 388 The triplex acies was in regular use by the third century BC (Keppie, 1984, 20–2). With the exception of Hannibal mimicking the Roman three–line deployment at Zama (Polyb. 15.11), nobody appears to have followed this Roman practice. 389 Trans. W. R. Paton, 1922–7. The anti–Flaminian tone of this extract is typical of Polybius (Walbank, 1957, 208).


114

towards the rear.390 The amount of space Polybius refers to will probably have been at least

half a kilometre of depth, as there could be 100–200 metres between each of the three lines,

and Polybius implies that more space was also needed behind the rear line.391

There are two possibilities to explain what manoeuvre Polybius refers to, and why space

behind the army was so important. He either describes a straightforward retrograde

movement (ie. a short or temporary retreat), or something more complicated, such as the

‘line–relief’ system. With regard to the former, Lendon has argued that Polybius means a

uniquely Roman fighting style premised on ‘successive, calm withdrawals under enemy

pressure,’ by which he means that a Roman battleline could flex and give ground on the

battlefield, then regain its solidity.392 The ability of Roman battlelines to flex under enemy

pressure is shown in the Second Battle of Philippi (App. B Civ. 4.128) when a Roman army

deployed in triplex acies was forced to retreat, and initially did so in good order. At length

this retreat did collapse into flight, because the lines of the triplex acies became intermingled,

which combined with enemy pressure to produce disorder and a rout. Such a disaster could

clearly occur even to experienced and properly–trained legionaries.393 It is thus clear that for

any retreat to be successfully carried out by a Roman army deployed in triplex acies, it was

necessary that space was maintained behind the lines, as Polybius has described above.394

The matter of the triplex acies raises a second possibility that Polybius refers to the

nebulously–understood Roman ‘line–relief’ system, by which units from the rear lines or

units relieved those in the forward lines or units, and thus allowed fresh troops to join the

battle.395 Although line–relief is not well described in the sources, understanding the system

390 Walbank, 1957, 209. 391 Caesar (BCiv. 1.82), in describing the deployment of his own army and that of his opponents near Ilerda in 49 BC, states that each army took up a space of over 600 paces (approx. 450 metres) in depth. On the other hand, in Polybius' (15.11) description of Hannibal's triple–line deployment at Zama, he states that Hannibal placed his third line ‘more than a stade’ (approx. 200 metres) from the front line. The distances between lines in the triplex acies probably also varied according to circumstances and terrain. 392 Lendon, 2005, 182. Lendon uses the term ‘withdrawal’ to describe a short retrograde movement, which I would describe as a retreat. 393 The army of the liberatores at Philippi had a large core of ‘first–rate legions’ (App. B Civ. 4.59). 394 See Daly, 2002, 62–3. 395 There is extensive debate over how this system worked (Junkelmann, 1986, 242–3; Daly, 2002, 60–3; Lendon, 2005, 181; Cagniart, 2007, 92–3; Sage, 2008, 69–76, etc). However, the relief of forward units by units from the reserve lines does appear to have been achievable during the intermittent lulls in fighting that are now believed to have occurred in extended battles (Goldsworthy, 1996, 208; Sabin, 2007, 428–9). This gave the Romans the unique advantage of allowing the replacement of tired units in the front line with fresh ones (Caes.


115

is one key to validating infantry–level models of Roman combat.396 The issue is thus

important to understanding not only how Roman armies moved in battle, but also the role of

retrograde movement in the use of triplex acies. It has been argued that line–relief did not

occur at the level of individual soldiers.397 but rather occurred among larger units or even

entire lines of the triplex acies, although it is not clear whether gaps were maintained between

units during battle to allow rearward units to come forward.398 It is clear from ancient

evidence that rear lines were able to replace the front line when necessary, as occurred for

instance at Pharsalus (Caes. BCiv. 3.94.1). Since short retreats in battle were scarcely

‘peculiar’ to the Romans, it is hard to believe that these were the unique Roman movement

Polybius referred to.399 In fact, his emphasis on rearward movement suggests that, rather than

the reserve units coming forward to replace the forward units, the forward units withdrew to

the reserve units behind them.400 Livy (8.8) describes the line–relief system just so. If this is

the case, then short retrograde manoeuvres were a distinctive feature of the line–relief

system, and a fundamental part of Roman utilisation of triplex acies.

Overall then, what is the ‘tactical movement peculiar to the Romans’ (Polyb. 2.33.7)?

Polybius says that the triplex acies and the line–relief system (however it functioned) gave

the Romans a systematic advantage in battle.401 In particular, Polybius (18.32.2–6) argues

that the Roman use of reserve infantry lines was unique and a major component of Roman

tactical success against the phalanx. In regard to retreats, recall that maintaining a ‘safety

distance’ reduced the pressure on retreating troops, but also as long as space could be created

between the Roman army and the enemy, line–relief could be carried out, which was another

BCiv. 1.45, 3.94; Livy 8.8, 10), while the enemy would usually possess no such luxury. Also see Sabin, 2000, 7, 16; 2007, 413; Gilliver, 2007, 129–30. 396 Sabin, 2000, 7; 2007, 431. 397 Goldsworthy, 1996, 208. 398 Koon (2010, 85) argues line–relief could be done without filling the gaps between units, but Gilliver (2007, 129–30) has argued that it is unlikely legions fought with gaps between units in the front line. Taylor has argued the gaps were maintained but may have been partly filled by light troops (Taylor, 2014, 319–20). 399 Pol 2.33.7, cf. Sabin, 2007, 412. 400 This is the re–construction shown in Anglim et al. (2002, 52). On the other hand, see Caes. BCiv. 1.45: at Ilerda in 49 BC, Caesar could not withdraw his forward cohorts and had to bring reserves forward to replace them. However, this was a different period and a different army, and the situation was a product of a specific tactical quandary. This may not have been typical, hence why Caesar describes it in detail. 401 Polybius (18.31–2) considered that the legion's multiple–line deployment gave the Romans a major advantage over the phalanx, as it allowed the Roman reserve lines to conduct flanking manoeuvres against the advancing phalanx. In the context of Hannibal’s simulacrum of the triplex acies at Zama, Polybius also approved of the way it forced men in the front–line to fight (Polyb. 15.11).


116

mechanism for reducing pressure through the replacement of tired troops with fresh ones.402

As has also been observed, reserves could provide a decisive physical and moral support to

infantry carrying out a fighting retreat in open ground, and so these could additionally

perform this function.403 Thus, two inferences can be drawn from Polybius’ statement that the

Roman army required space to the rear for its ‘peculiar’ manoeuvre. The entire army needed

space to move so that one or all of the three lines could advance or retreat as necessary in

response to enemy pressure without becoming too congested, and also, space was necessary

both to the front (ie. safety distance) and rear (for short retreats) so that sub–units or an entire

line could conduct line–relief. The success the Romans could achieve by performing a

tactical retreat followed by a well–timed advance was demonstrated at the Battle of Pydna in

168 BC (Plut. Aem. 20.4–8), and if necessary the Romans could also use the line–relief

system to prevent a retreat, as Caesar did at Ilerda (Caes. BCiv. 1.45). Thus, small retrograde

movements contributed to the adaptability of Roman manipular tactics.

One more characteristic of the Roman army was that it fielded regularly sized and

independent armies, legiones, which were made up of consistently sized and organised sub–

units, cohortes and manipuli.404 These sub–units were fundamental to the Roman army’s

flexibility in battle, as legions not only had multiple lines that could support the fighting line,

but individual maniples and cohorts could manoeuvre somewhat independently and conduct

flanking attacks.405 The importance of maintaining reserves in battle and during a retreat in

order to threaten pursuers has been discussed above. The triplex acies was advantageous for

this reason, as there were always reserve lines present to support the fighting line. Roman

sub–units could perform the same function, as Caesar demonstrated by using several cohorts

to protect the retreat of his forces at Gergovia in 52 BC (Caes. BGall. 7.51), and by rotating

cohorts in and out of the fighting at Ilerda in 49 BC to prevent his men from becoming

exhausted.406 The organised sub–division of Roman legions into multiple sub–units thus

402 Caes BCiv. 3.94.2; Livy 2.20.6, 8.10.4, 9.32.9. Also see Goldsworthy, 1996, 208; Sabin, 2007, 428–9. 403 This is in fitting with what Lendon (2005, 182) has argued, as mentioned above. 404 From the third century to the first century, the Romans phased out the use of maniples as the main sub–unit of the legion in favour of larger cohorts, although cohorts continued to consist of maniples (Keppie, 1984, 63–6; Goldsworthy, 1996, 34). The last recorded use of maniples as independent units in combat was in Africa in 109 BC (Sall. Jug. 49.6). However even into the Principate, signa were only possessed by maniples and not cohorts, suggesting that the former remained an essential focum for legionaries (Rankov, 2007, 32), even if not an independent combat unit. 405 Keppie, 1984, 43. 406 Caes. BCiv. 1.45. Also consider the example of Pydna, 168 BC (Plut. Aem. 20.4–8).


117

provided another layer of reserves that the men of the fighting line could rely upon to support

them in retreat, whether through their mere presence, by line–relief, or by counterattack.

To summarise, Roman infantry training and tactics were suited to the conduct of retreats and

withdrawals in several ways. Milites received training in close order manoeuvre, and at least

in the Late Republic, a good general drilled them specifically in retreats and counterattacks.

The Romans appear to have had dedicated signals for retreat with both signa and tubae.

Further, the triplex acies system was suited to the conduct of retrograde movements because

it provided a reserve line that could support or relieve the fighting line when necessary.

Retrograde movements may even have been fundamental to the implementation of line–relief

system, assuming that it functioned through the retirement of the fighting line to be replaced

by a posterior line. Finally, the regular organisation of Roman armies meant that sub–

divisions of the Roman army could manoeuvre to support fighting units, assisting them in

retreat. Retrograde movements were thus a significant part of conventional Roman battle

tactics, and multiple factors contributed to the ability of Roman armies to maintain order

during a retreat.

The Roles of Subordinate Officers

One more factor to consider in the maintenance of order during retreats is the roles of the

various leadership figures in a Roman army who fought alongside the milites. These include

the various junior officers known collectively as principales, centuriones, the veteran

legionaries (veterani), the aristocratic tribuni, and other officers such as quaestores. The

details of how these various kinds of officers contributed to preparing soldiers and managing

them during retreat will be considered below.

The Imperial-era term principales describes various grades of junior officers below the

centurions, such as optiones, and others of lower rank including tesserarii, beneficarii,

signiferi and aquiliferi.407 These officers had the closest proximity to the milites, and so will

most likely have had important roles in the maintenance of ratio.408 It has been suggested that

the optiones, second in seniority to centuriones, were stationed at the rear of a maniple or 407 Southern, 2006, 333. 408 Phang (2008, 65–6) points out that close supervision was needed by officers to keep formations organised during battle.


118

cohort and used their bladeless spear of office to coerce wavering soldiers.409 The more junior

tesserarii may also have stood to the rear and assisted the optiones in this task.410 Yet another

type of principalis, the beneficiarius, an assistant to a centurio, could again have performed a

similar function.411 Even so, there is no direct evidence for these practices.412 In sum, the

optiones and possibly other principales probably formed a coercive element in the rear

ranks.413 In a retreat, this function will have been essential to ensuring that sub–units moved

at a gradual pace and so maintained their ratio, as well as preventing waverers from fleeing

and triggering a greater flight.

The best understood and probably the most important junior officers were centuriones, front–

line leaders, whose duties included keeping soldiers with the standards and leading them by

example, and they were probably also key officers in charge of training legionaries.414 They

were the overall leaders of centuries and maniples and will have been key leaders in the

cohorts, and so the management of a retreat at the sub–unit level must have been primarily

their responsibility, though, as discussed above, they had various assistants in the form of

principales.415 The hands–on leadership role of centurions is demonstrated by the fact that

during routs and withdrawals, where numbers are given by the sources the centuriones

suffered disproportionate casualties relative to legionaries.416 Most importantly, the

centurions are specifically mentioned in the sources, along with tribunes, as key officers in

ordering and organising retreats (Caes. BGall. 3.95) or withdrawals (App. Hisp. 82). During a

409 Goldsworthy, 1996, 182, 2004, 185; Lendon, 2005, 185; Phang, 2008, 66. 410 On the tesserarius see Southern, 2006, 93, 100, 333; Goldsworthy, 2011, 68–9. For the possibility they assisted the optio see Junkelmann (1986, 242–3), who supposes they were deployed at the rear of their centuria along with their immediate superior, the optio, contra Southern (2006, 333), who describes the tesserarius as a ‘clerical post’ whose ‘main responsibilities concerned guard duties’ (100). This narrow definition of the duties of the tesserarii is owed to Polybius (Polyb. 6.35–6). However, Polybius simply did not describe the battlefield duties of the tesserarius, as he also did not for the optiones (see Polyb. 6.24–5, 36–7), so his restrictive description of their duties does not necessarily give the entire picture. 411 Goldsworthy, 1996, 124; Phang, 2008, 66. Also see Livy 34.15.4; Maur. Strat. 12B16. 412 It is striking that optiones were not mentioned by either Caesar or Livy. In Livy’s description (35.5.10) of a battle between Gauls and Romans c. 193 BC, he states that the Gauls placed officers at the rear of their line with spears to dissuade waverers. Was Livy implying that the Gauls were adopting a common Roman practice? 413 This was a practice as recently as World War Two, and Rush (2001, 316) suggests that any Roman soldier turning to flee may very well have come face to face with an officer such as an optio. 414 Sabin, 2000, 15–6; Cagniart, 2007, 93; Phang, 2008, 66; Training: Southern, 2006, 136. In the Imperial period, there were also several other types of officers known to have been responsible for training soldiers, including the campidoctor, the magister campi and exercitator (Southern, 2006, 136). 415 Southern, 2006, 332–3. Cf. Caes. BGall. 2.25, 6.39. 416 Caes. BGall. 2.25, 6.40, 7.51, BCiv. 3.71; BHisp. 23. Cf. Lee, 1996, 211; Kagan, 2006, 169. Presumably, centurions were also high–profile targets for the enemy.


119

retreat, many centuriones likely remained in the front lines, providing an example to their

men, restraining them from flight and managing their pace of movement.417 To summarise,

while principales posted in rear lines will have supported a retreat by managing the pace of

movement and dissuading stragglers from flight, centuriones played the most apparent role,

providing charismatic leadership by sharing the dangers of the front line and ensuring that the

soldiers maintained order.

Standard bearers, such as the signiferi and the aquiliferi, could have also played a role. As

mentioned, reversing the standards may have been a signal for retreat.418 The Romans were

strict in ensuring that soldiers remained relatively close to their signa, which is to say, their

centuries or maniples.419 The signiferi could play a crucial role in halting routs, as well as in

motivating men to advance.420 Even more so the legionary standards, the aquilae, were of

themselves an intense focus for the attention of Roman soldiers, who would follow them into

great danger and go to great lengths to prevent their capture.421 Through a combination of this

devotion towards the signa or aquilae and their own courage, these principales could thus

have a disproportionate influence on the men around them.422 The signiferi might also have

been held to higher standards of discipline than the regular milites. For instance, after a

Caesarian rout at Dyrrhachium in 48 BC, Caesar specifically targeted several signiferi for

punishment, and in AD 16 when the rebel Tacfarinas attacked a Roman cohort in Africa, the

milites fled and their commander Decrius berated the standard bearers (signiferos) for letting

the men flee.423 Finally, the signiferi also played a role in transmitting orders between the

417 Note the prominent role of the centuriones in leading a breakout of trapped legionaries during a skirmish in Gaul (Caes. BGall. 6.40). 418 Cf. Caes. BGall. 7.51.3; BAfr. 18.3. 419 Sage, 2008, 123–4. Also see Caes. BCiv. 1.44. 420 Halting routs (Caes. BGall. 7.51). Motivating men to advance: Sabin, 2000, 15–6. That said, the signiferi appear to have generally remained close to the fighting (Junkelmann, 1986, 242–3; also see Livy 9.39; contra Sage, 2008, 72). 421 Livy (3.70, 4.19, 6.8, 26.5, 34.46) provides numerous examples of officers throwing standards into the enemy ranks to stimulate soldiers into retrieving them. Several such examples can be found in other sources (Lee, 1996, 208–9). There are also multiple accounts of the use of standards to lead troops forward (Caes. BGall. 4.25; Livy 10.36 21.36, 25.14, 27.14, 26.5, 39.31; App. Mith. 58). Note that in many of these examples, another officer has taken the standard from the signifer: perhaps the signiferi themselves were reluctant to throw or otherwise unnecessarily endanger the standard, but the key observation is that the standard itself was the focus of the soldiers, not necessarily the bearer. 422 Caes. BGall. 4.25. The milites also appear to have had high expectations of the standard bearers. According to Livy’s (3.27.8) fictionalised account of a campaign c. 458 BC, highly motivated legionaries urged on their standard bearer with the cry adcelera, signifer! To this they received the reply sequere miles! 423 Caes. BCiv. 3.69, 74; Tac. Ann. 3.20. Also see Goldsworthy, 1996, 163–4. In Appian's account (BCiv. 2.63) of


120

general or his legati and milites.424 As such, the visibility and reputation of signiferi and

aquiliferi, in addition to the fact that Roman milites were trained to remain close to their

standards as a matter of course, meant that standard bearers must have been key figures in

retreats. They could signal a retreat and conversely, they could sometimes prevent retreat or

flight simply by standing their ground.

Veterani may have played a role in maintaining ratio. One relevant point in the use of the

maniple was the possible disposition of more experienced milites in the second and third

lines.425 In general it could be assumed that the bravest men, and those seeking recognition or

promotion, usually made their way to the front lines.426 It is tempting to speculate that, to

counter the possibility of a rout beginning in the rear, experienced men were also deliberately

posted in the rear to keep an eye on the less experienced men in front of them, as was

considered good practice by some Greeks.427 Since the triplex acies, by having more

experienced men in the second and third lines stimulated those in front to fight, the veterans

could have performed this same function during a retreat or withdrawal, compelling less

steady troops to maintain close order and a steady rate of movement.428 Conversely, if

veterans were posted in the front ranks (Plut. Marc. 26.1), they could prevent the front lines

from moving too quickly and personally resist enemy pursuers. They could thus be of use

regardless of whether they were in the front or rear, and would only have been of greater

service if they were posted to both positions.

the Battle of Dyrrhachium, soldiers blamed signiferi for a rout by claiming that the standards bearers fled first and the soldiers were bound to follow them. 424 For example, signa were used to give the signal for battle (Goldsworthy, 1996, 149–50). 425 Keppie, 1984, 64. Also see Polyb. 15.11. Unfortunately, no direct information exists on the placement of individual soldiers within Roman units (Gilliver, 2007, 130). Even the formation and spacing of Roman infantry within maniples and cohorts is an area of debate. Lendon's 'cloud theory' (Lendon, 2005, 179–82) is one of the more recent theories, but what this would mean for the idea that there was 'standard' spacing for legionaries and units is unclear (cf. Polyb. 18.30.5–8; Plut. Ant. 39.4; Goldsworthy, 1996, 179–81; Sabin, 2007, 426). 426 Gilliver, 2007, 130. 427 Routs beginning in the rear: Goldsworthy, 2011, 185. Experienced men at the rear: Goldsworthy, 1996, 178; Daly, 2002, 159, 200. According to Goldsworthy, the Greeks also thought it sensible practice to put brave men in the rear ranks to restrain or deter less determined men from fleeing (ref. Xen. Mem. 3.19; Ascl. Tac. 14.6). However, to extrapolate from Greek practice is no sure method of determining Roman practice. 428 Keppie, 1984, 64. Also see Lendon, 2005, 187–91. See Polyb. 15.11: Polybius mentions how Hannibal forced his less reliable troops to fight by deploying his veterans behind them.


121

Finally, the functions of a legion's aristocratic subalterns, the tribuni and quaestores, are more

difficult to determine, for the evidence regarding the tactical functions of military tribunes is

less clear than it is for centuriones.429 The functions of quaestores are better known, and

these were more administrative than tactical.430 That said, it is striking that at the Battle of

Vosges, 58 BC, Caesar placed gave command of part of his army to a quaestor (Caes. BGall.

1.52). Evidently, some quaestors could be given senior tactical appointments if necessary.

Returning to tribuni, most of the evidence regarding their duties also concerns administrative

tasks, but scattered references to other duties such as organising the battle–line, performing

reconnaissance, commanding detachments, and advising and making reports to the general,

suggest that they could be employed in all sorts of military roles.431 Suolahti notes that,

unlike centurions, the tribuni ‘never functioned as tactical sub–commanders with fixed duties

to attend to’, but they were attached the general's staff and served as assistants in various

capacities.432 They could certainly assist the general in planning operations and deciding

tactics, and Caesar implies that they were, alongside centuriones, key leadership figures for

the milites.433 In the Pompeian rout after Pharsalus, Caesar states (BCiv. 3.95) that the fleeing

Pompeian soldiers followed the orders of centurions and tribunes (ducibus usi centurionibus

tribunisque).434 It is also evident from the high death rate of military tribunes in the Mid–

Republic that these officers often exposed themselves in battle, and so they must have been

generally prominent officers in action.435 During a retreat (or withdrawal), it appears clear

that the tribunes will not uncommonly have been active in leading the milites, along with the

centuriones, but their exact duties will have varied according to the situation and the

assignments of their imperator.

429 The Caesarian corpus, for instance, mentions centurions more often than tribunes (eg. BAlex. 40), Cf. Welch, 1998, 102–3. On the roles of tribunes also see: Suolahti, 1955, 43–51; Goldsworthy, 1996, 124–5; Roth, 1999, 277. 430 Roth 1999, 277. 431 Sage, 2008, 104–6. Also see Goldsworthy, 1996, 124–5; Suolahti, 1955, 43–51. 432 Suolahti, 1995, 47–50. 433 Caes. BGall. 6.39.2: perterritos ... ad tribunum militum centurionesque ora convertunt . Various duties: Caes. BGall. 4.23.5; BAfr. 54.1; Livy 25.14.2, 34.13.4. 434 Also note Appian's account (Hisp. 82) of the withdrawal of a Roman army from Pallantia in 136 BC, mentioning that both centurions and tribunes were responsible for organising the army. 435 Eckstein, 2009, 219.


122

It was probably not easy to convince men who were keen to disengage from a victorious

enemy to move slowly.436 While Roman soldiers might be motivated to hold their ground by

discipline, coercion and comradeship, it was often the boldness and aggression of individual

leaders that motivated troops to assault and defeat the enemy.437 Charismatic leadership was

thus an essential component of motivating soldiers to act as desired by their commanders. It

seems clear that many figures in the army could have played such roles, and in general, the

potential flexibility of Roman command arrangements is striking.438 But as Caesar (BGall.

6.39.2) affirms, when soldiers were hard–pressed, it was generally the centuriones and tribuni

to whom they looked for leadership. These were, therefore, the most important officers for

transmitting instructions and maintaining order among the milites during a retrograde

movement.

Equites, Velites et Auxilia

The above discussion has focused on milites, but heavy infantry was not the only component

of a Roman army. Light infantry and cavalry were also major parts of Roman legions and,

later in the Republic, so too were auxiliary forces. In the Middle Republic, the average legion

consisted of 3000 legionaries, 300 equestrian cavalrymen, equites, and 1200 javelin–armed

light infantry, velites.439 By the third century BC, the legions were also typically accompanied

by equal or greater numbers of infantry and cavalry provided by the socii.440 From the time of

the Second Punic War, Roman armies began to be regularly supplemented by auxiliary

infantry, particularly by Greek archers, Greek and Balearic slingers, and Numidian, Iberian

436 Goldsworthy, 1996, 194. 437 Goldsworthy, 1996, 281. Caesar's Bellum Gallicum is full of examples of such actions by signiferi and centuriones (eg. 2.25, 4.25, 5.44, 7.47, 50). 438 Under certain circumstances, almost any officer could command significant forces. For instance, in the absence of a more senior commander, one or two tribunes could command a legion (Gilliver, 2007, 137; Rosenstein, 2011, 136), as could a quaestor (cf. Caes. BGall. 1.52) and in one exceptional case a lowly centurion led an entire army against Hannibal (Livy 25.19.9–10). 439 Polyb. 3.21; Keppie, 1984, 33–5. 440 Rawlings, 2011, 51. The alae fielded by the socii were probably organised differently to the legions in the fourth and third centuries BC, but the military units of the Romans, Latins and socii became more homogenous over time until they were relatively indistinguishable (Rawlings, 53). Cohorts of Latins and socii operated independently at times from the second century onward (Keppie, 1984, 64). The sources generally do not distinguish between legions and allies in battle accounts (Rawlings, 52), and so it may be assumed that in the tactical sense they were very similar. Drawing out and analyzing any tactical distinctions that did exist is outside the scope of this thesis, and so it will be assumed that the analysis of this thesis also applies to the infantry and cavalry of the Latins and socii.


123

and later Gallic and German cavalry.441 In the Late Republic Roman armies could be highly

multi–ethnic, combined arms forces: Caesar’s army in 57 BC consisted of Roman and

Cisalpine Gallic milites, Treveran and Numidian cavalry, Cretan archers and Balearic

slingers, while the grand army that Marcus Antonius took into Parthia in 36 BC fielded tens

of thousands of non–Roman javelineers, slingers and cavalry.442 As such, any examination of

Roman retreat tactics in the Middle to Late Republic must not only consider the roles of

legionaries, but also of cavalry and light infantry, many of whom were foreign to Roman

military practices.

As noted, Roman cavalry (armed with sword and shield) were particularly useful as reserve

forces, threatening or counterattacking potential pursuers if necessary.443 The example of

Ilerda has already been discussed, in which Caesar’s cavalry threatened would–be enemy

pursuers to allow his infantry to retire (1.46.3). During Scipio Aemilianus’ Iberian campaign

of 134 BC, Appian (Hisp. 88) describes a fighting retreat by a Roman cavalry detachment

that had found itself in trouble after pursuing the enemy uphill to a position from which the

cavalry were unable to safely retreat. Scipio brought more cavalry to the rescue and split

them into two groups, who attacked both sides of the enemy and then withdrew further away

each time. In this way they gradually retreated to safer terrain while fighting off enemy

pursuers. During Antonius’ Parthian campaign of 36 BC, his cavalry were invaluable in

driving off Parthian archers so that his infantry could advance.444 Just as cavalry were a

potent weapon in pursuit, Roman cavalry were a powerful tool for supporting a retreat or

withdrawal.

The Romans also used a variety of auxiliary cavalry, armed differently to the Roman style.445

At the Battle of Callinicus in 171 BC, the Romans fielded a large number of Greek auxiliary

infantry and cavalry in a skirmish against their Macedonian equivalents, but after the Romans

were beaten back and began to retreat, the Thessalian auxiliary cavalry came to their rescue

441 Keppie, 1984, 78–9; Sekunda, 2007, 356–7. 442 Caesar’s army: cf. Caes. BGall. 2.7, 24. Caesar later made use of German cavalry as well (Caes. BGall. 1.80). Antonius’ army: Plut. Ant. 37.3, 41.4, 42.1. 443 Caes. BCiv. 1.46; Sall. Jug. 98.1; Plut. Ant. 41.5–42.1; App. Hisp. 88. By ‘Roman’ cavalry, I refer to those trained and armed in the Roman manner, including not only Italians but, later, also Iberians and Gauls (cf. Plut. Ant. 37.3). 444 Plut. Ant. 39.4, 42.1. 445 Depending on time and place, these included javelin–armed Numidians (Sabin, 2007, 422; Sekunda, 2007, 353), Thessalians (Livy 42.59.4–5) and eastern horse archers or cataphracts (Plut. Ant. 37.3).


124

(Livy 42.59.4–5):

... the Thessalian cavalry, which had been in support (in subsidiis), away a short

distance from the left wing and beyond the melee, though at first spectators of the

conflict, were later, when the decision had been reached, of the greatest use. For,

retreating gradually with formation intact (cedentes enim sensim integris ordinibus),

after they had joined the auxiliaries of Eumenes, with him they both provided a safe

retreat (receptum) between their units (inter ordines) for their allies scattered in flight,

and when the enemy pressed on in less close order (minus conferti), they even risked

advancing and met many of the fugitives half–way. Nor did the king's men, now

themselves scattered in a general pursuit (passim sequendo), dare to close in battle

with an enemy in formation and advancing steadily (ordinatis et certo incedentibus

gradu).446

The Thessalian cavalry was effective in securing the retreat of the rest of the light infantry

and cavalry. In this instance, they maintained close order and gradual movement. Note,

however, that the Thessalians also alternated retreat with advance to further delay the enemy

and to take the fight to the pursuers, who were reluctant to engage because they themselves

were disordered in pursuit. The rationale for Onasander’s (Strat. 27) argument that units must

maintain close order whether in advance or retreat is well–demonstrated in this example. This

case makes clear the value of proficient, light–armed cavalry in the event of a retreat.

Cavalry were even faster than light infantry, even if the cavalrymen themselves were

relatively heavily equipped, and so they could more safely advance and retreat repeatedly. In

this way a reliable cavalry unit could provide support for the slower withdrawal of infantry

formations and even of other cavalry. Cavalrymen equipped with missile weapons were

perhaps even better suited to the conduct of fighting retreats, as they had great speed and

manoeuvrability and could engage in fire–and–retreat tactics that not only threatened enemy

pursuers, but inflicted casualties. Some of the auxiliary cavalry fielded by the Romans, such

as Numidians, were particularly well–suited to this style of fighting.447 Both Roman and

auxiliary cavalry could therefore serve well in defending retrograde movements by milites.

446 Trans. Evan T. Sage and Alfred C. Schlesinger, 1938. 447 Sabin, 2007, 422; Sekunda, 2007, 353.


125

This is important to note, for the roles of Roman and auxiliary cavalry in supporting Roman

retreats has been overlooked in modern scholarship.448

The precise tactical roles of Roman light infantry (velites) are uncertain.449 The subject of

how the Romans used light infantry in battle has not traditionally been well studied, but it has

been suggested that light infantry skirmished at the start of a battle, sometimes for hours, and

then once they had exhausted their missiles they retired, playing no further part in battle.450 In

this schema of light infantry fighting, as they expended their missiles in the opening stages of

an engagement, they would have been of little assistance to the heavily–armed milites during

a retreat or withdrawal. However, significant work has been done recently on the importance

of missile combat by scholars such as Zhmodikov and Slavik.451 Essentially, Zhmodikov has

argued that fighting with missiles, in particular with pila, was heavy and ongoing throughout

Roman battles, and Sabin has incorporated this argument into his concept of ‘safety

distance’.452 Slavik has refined these arguments by considering the role of the velites, positing

that it was they and other skirmishers that did most of the ranged combat and not the

legionaries, who did hurl their pila (which were limited in supply) but then fought mostly

with swords, as had previously been generally recognised.453 Also, it has been noted that light

infantry frequently operated alongside cavalry in battle.454 Given this link, since cavalry was

used to support retrograde movements, the same logic should apply to light–armed infantry,

which surely also had a role in retreats and withdrawals.

448 For instance, this function of cavalry is not mentioned in McCall’s (2001, 53–77) detailed and considered analysis of Republican Roman cavalry tactics. Nor is it mentioned in the Cambridge History of Greek and Roman Warfare (cf. Sabin, 2007, 422–5; Sekunda, 2007, 352–4).Cavalry could also protect infantry during routs (eg. Xen. Hell. 6.5.13–4, 7.5.24–5; Sall. Jug. 103). 449 Slavik, 2018, 154–6, 164. 450 Sabin, 2007, 409. 451 Zhmodikov, 2000 and Slavik, 2018. 452 Zhmodikov, 2000, 67–70, 78. Zhmodikov argues that it was the milites who did much of this type of fighting, by exchanging pila for extended periods instead of clashing hand–to–hand. Sabin has combined this idea with his own concept of the ‘safety distance’ to argue that periods of missile combat (again with pila) gave heavy infantry space and time to recover between bouts of hand–to–hand fighting (Sabin, 2000, 12). 453 Slavik, 2018, 165–8. Whether Slavik is more correct that Zhmodikov or not regarding the preponderance of pila in missile exchanges, both scholars show that missile combat was a more significant element of Roman battles than has always been recognised. 454 Sekunda, 2007, 353.


126

It is possible that lightly equipped infantry could perform a similar function during a retreat

to the cavalry (Caes. BCiv. 1.79.1), but heavy infantrymen were probably too slow and

heavily burdened for such tactics. Indeed, enemy light infantry could pose great difficulties to

milites who found it difficult to catch them (BAfr. 71). Light infantry also had the freedom of

movement to carry more missile weapons and to move around the battlefield retrieving

weapons that had already been thrown, and so these velites could continue to engage in

missile combat for hours, if necessary.455 In terms of how light infantry worked alongside the

units of milites that formed the battleline, it has been suggested that if gaps were maintained

between Roman maniples or cohorts during battle, then light infantry could fill these gaps

and attack, retreat or support the milites as necessary.456 In particular, Slavik has argued that

light infantry remained in support throughout a battle and, as long as velites maintained their

supply of missiles, they could be used to hold the enemy in place to support manoeuvres by

the milites.457 Slavik points out that velites could thus be used to facilitate the conduct of

line–relief and, extending his logic, the same support could obviously be given to soldiers in

retreat.458 It should be noted that velites were gradually phased out of Roman armies from the

second to first centuries BC, but they were replaced with auxiliary light infantry of all kinds

that could skirmish in the same manner, who must have performed the same functions as the

velites had done in previous times.459 Thus, light infantry no doubt played an important role

in supporting retreats or withdrawals by Roman armies, throwing missiles and suppressing

potential pursuers as the milites disengaged, and perhaps cooperating with the cavalry to do

so.

Conclusion

In this chapter, I have argued that successful retrograde movements in the presence of the

enemy generally had two main features; they were orderly, and the retiring army fought off or

prevented close pursuit by the enemy. There are several components to these two features of

successful retrograde movements, and as I have discussed the use of camps and high ground

previously, in this chapter I have discussed other key factors, maintaining physical order and

455 Slavik, 2018, 168. 456 Taylor, 2007, 319–20, cf. Sall. Jug. 49.6. 457 Slavik, 2018, 168. Cf. Polyb. 3.110. 458 Slavik, 2018, 168. 459 Bell, 1965, 419–21. Cf. Caes. BGall. 2.10, 19, 24.


127

using reserves. When it came to maintaining order, retiring heavy infantry maintained ratio

by combining close order with gradual movement, and retiring while still facing the enemy.

In this way, retreating troops could still appear formidable and thereby dissuade enemies

from pursuing them. Discipline was necessary to facilitate this orderliness, and was produced

by training and the leadership of veterans and officers, particularly centuriones and tribuni,

who provided coercion, inspiration and direction. Roman infantry received enough training in

battlefield manoeuvre from the Middle Republic onward to perform retrograde manoeuvres

in battle, and by the time of Caesar their training in retrograde manoeuvres was even more

sophisticated. The Romans also had formal signals for retreat and were well organised into

formal units and sub–units, allowing some milites to retreat while others supported them by

line–relief, maintaining the integrity of the battleline. The Romans habitually maintained

reserve infantry lines in the triplex acies, but they also utilised separate cavalry and light

infantry units which were particularly adept at protecting retreating infantry and fighting

pursuers. Light infantry could use their missiles and freedom of movement to support

retreating milites, while cavalry were particularly capable at intimidating or actively

countering enemy pursuers.

Thus, factors of training, formal organisation, leadership and combined arms composition of

Roman armies all united to make Roman armies well–suited to conducting orderly retreats

and withdrawals. In addition to the discipline and organisation of milites and their junior

leaders, because Roman armies included strong contingents of soldiers who had variety of

armaments and roles, such as javelineers, archers and cavalry, and maintained reserve

infantry lines, they had several tactical options for dissuading or beating back enemy

pursuers. Most Roman armies, therefore, had all the tools they needed to maintain order in

retreat or withdrawal and also to counteract enemy pursuit, and so to retire successfully.


128

…when the vanguard was thrown into confusion the legion posted on that wing could not

stand its ground and withdrew to the nearest hill. Finding nearly the whole of his line panic–

stricken – an event as unusual as it was unexpected – Caesar exhorts his men and leads the

Ninth Legion to their support. He checks the foe who are pursuing our men with insolent

daring, and compels them again to turn and retreat to the town of Ilerda and halt beneath the

walls.

(Caes. BCiv. 1.44.4–45.1)

Chapter Four

Roman Retreat Tactics, Part II: Virtus, Consilium and

Scientia Rei Militaris


129

Introduction

Recall that at Ilerda in 49 BC, Caesar’s men had pursued their foes up a hill upon which sat

the eponymous town, where they found themselves unable to retreat safely (Caes. BCiv.

1.45). Prior to this impetuous advance, his men had actually been driven into retreat from

another hill close to Ilerda, but Caesar had personally intervened with reinforcements and

hurled back the enemy, thereby giving his excited troops the opportunity to pursue the enemy

to that difficult position on the slope of Ilerda itself (1.44–5). Caesar’s intervention appears to

have been decisive in preventing that part of his army from suffering a rout. Could a general

typically have had such impact, and what other actions could they take to secure a safe retreat

or withdrawal?

As pointed out by Sabin, good generalship was perhaps the most important factor in any

army’s victory or defeat.460 It has been demonstrated that most Roman armies had all of the

tools they needed to conduct orderly retrograde manoeuvres, but the importance of a

general’s decisions and leadership are yet to be examined in detail. This chapter will consist

of a more detailed discussion of the latter subject. First, a brief and general overview of the

Roman concept of a general, and the roles and responsibilities of the general prior to and

during battle, will be given. Following this, the general’s role will be examined in

circumstances more specific to retreats, including retreats from the battleline and into a

fortified position, retreats out of fortified positions, and marching withdrawals over longer

distances. The chapter will then conclude by summarising the discussion, and outlining the

relationship between generalship and the tactics of successful retreats and withdrawals that

have already been discussed.

460 Sabin, 2007, 430.


130

The Roman Imperator

The Romans had a variety of important officers, including legati, tribuni, quaestores and

centuriones. Under certain circumstances, any such officer could command significant

forces.461 However, a Roman general was usually a senior aristocrat, appointed by the senate

for a specific campaign, typically middle–aged and with some level of experience in military

matters.462 While the skill and experience of these men varied, the competitiveness of

command appointments meant that sooner or later, the Romans would oppose their enemies

with one or more capable commanders.463 Legati were of a similar background and held the

most responsibility next to the general himself, and so would most often lead the army in his

absence.464 For the purposes of this chapter then, the term ‘general’ will apply to these senior

men, whether imperator, legatus or otherwise, who commanded an entire Roman army.465

To begin with, how did the Romans themselves conceive of a good general? In his speech on

the Lex Manilia in 66 BC, Cicero listed a number of qualities that the ideal Roman general

should possess, namely ‘industry (labor) in business, fortitude (fortitudo) amid dangers,

energy (industria) in acting, rapidity (celeritas) in executing, wisdom (consilium) in

foreseeing’ (Cic. Leg. Man. 11.29). Several authors also use the phrase scientia rei militaris to

describe a general’s overall proficiency in military matters.466 Caesar, in his depiction of his

own actions at the Battle of Alesia, 52 BC, provides an exemplar for how such a general

might behave in action. Caesar, having trapped Vercingetorix and his army in Alesia and

circumvallated the town, predicted attacks from a Gallic relief army and so stockpiled large

amounts of supplies and built a second wall outside the existing wall (Caes. BGall. 7.74). At

461 Cf. Caes. BGall. 1.52; Livy 25.19.9–10; Gilliver, 2007, 137; Rosenstein, 2011, 136. 462 There were exceptions, particularly in periods of civil strife when the senate’s authority broke down. For instance, in 84 BC, at only 23 years of age, Pompey raised an own army from his own clients and went on campaign (Plut. Pomp. 6). Further, in the Late Republic onward, commanders were increasingly of non–aristocratic origin and the senate had less control over their appointment (Rosenstein, 2011, 144). Also see Campbell, 1987, 20–1. 463 Sabin, 2007, 433; Hoyos, 2011, 76. 464 For example, on occasions during his Gallic campaigns, Caesar left his legate Titus Labienus in command of several legions (Caes. BGall. 2.11, 4.38, 5.8). 465 Throughout the following discussion, I will utilise the term imperator as a synonym for a Roman general rather than as a specific descriptor for a general who had been acclaimed as such by his troops in the field (cf. Southern, 2007, 172). 466 Caes. BGall. 3.23.5, 7.57.3, BAlex. 31.1.9; Cic. Font. 43.6, Man. 28.3, 28.13, ad Fam. 13.57.1; Tac. Ann. 12.40.9, Front. Strat. 1.1. Cicero, again in his speech on the Lex Manilia (10.28), listed four key virtues of the good general as scientiam rei militaris, virtutem, auctoritatem, felicitatem.


131

the climax of multiple Gallic attempts to breach these defences, Caesar moved to high ground

and personally observed the fighting (BGall. 7.85), despatched timely reinforcements to

trouble spots (7.86–7), and finally intervened personally, bolstering the morale of his men by

the sight of his red cloak (7.87). In his actions at Alesia, Caesar displayed all of Cicero’s five

qualities, and scientia rei militaris. Such dramatic leadership may not have been the norm,

and it might be noted that Caesar remained detached from the fighting for as long as possible,

but it may be seen that a combination of energy and good decision–making characterised a

good general.

Roman generals who trained and disciplined their troops rigorously were praised by Roman

writers, for example the Scipiones Africanus and Aemilianus, the vir militaris Gaius Marius,

and the Julio–Claudian general Gnaeus Domitius Corbulo.467 Maintaining disciplina militaris

was certainly considered to have been an important part of a general’s responsibilities,

although excessive severitas could be frowned upon and a good general might also show

solidarity with the soldiers by participating in labor himself.468 Cicero also spoke favourably

of Fabius Maximus Cunctator, in that the latter was well–read in military history (Cic. Sen.

11). However, a man of less education but greater experience could also make an argument

for being a good general, as did Gaius Marius when he compared the leadership of the

traditional aristocracy to a novus homo such as himself: he said quae illi litteris, ea ego

militando didici (Sall. Jug. 85.13). Despite Marius’ criticism however, competition and the

leadership of men in the public space had always been an inherent part of an aristocrat’s

upbringing, and such experience will have lent itself well to military command.469 Cicero

certainly admitted the importance of experience in addition to learning from textbooks and

exempla.470 A good general thus kept control of his men, who were of a lower class, and

whether through education or experience, or both, he was knowledgeable in military affairs.

467 Africanus: Polyb. 10.20. Aemilianus: Livy 26.51.4. Marius: Sall. Jug. 63. Corbulo: Tac. Ann. 13.35. 468 Phang, 2008, 76, 112–3. Even Caesar, who could be mild with his soldiers even in the face of serious disobedience (Caes. BGall. 7.52; BAfr. 85), would still administer punishments when necessary or expedient (BCiv. 3.74). 469 Rosenstein, 2010, 367–371, 381. 470 Campbell, 1987, 21.


132

The Romans also had a variety of criticisms for generals whom they considered to be poor.

When Velleius Paterculus condemned the general Varus, who had been killed in the clades

Variana of AD 9, he said that Varus was a marcor ducis (Vell. Pat 2.119.2) who lacked spirit

for fighting (2.119.3) and judgement (120.5), while Suetonius said of the same general that he

had acted with temeritas and neglegentia (Suet. Tib. 18).471 In his account of the Battle of

Atuatuca in 54 BC, Caesar (BGall. 5.32–7) contrasts the behaviour of the two Roman

generals, Sabinus and Cotta. The former panicked and dashed around from cohort to cohort

aimlessly, spreading his own nervousness, while Cotta moved about encouraging men and

even led the milites in charging and fighting (5.33, 35). Evidently, a good general was

expected to be spirited, but also calm, decisive and effective in his actions.472

The Romans considered success in battle to be less a result of the general’s orderly

arrangement of forces in the Greek manner, and more due to his own display of virtus by

facilitating charges and showing personal charisma, and scientia rei militaris through

management of terrain.473 Good tactics had to be adapted to the situation, as even in a retreat

there was no strict doctrine that Roman generals would follow.474 Strict technical competence

was thus not a key Roman consideration for a good general. It is therefore clear that the

Romans thought of a good general’s military ability in terms of several key qualities,

particularly military knowledge (scientia rei militaris), courage (virtus) and wisdom

(consilium), which were products of his background, education and experience.475

471 On the fairness of these criticisms, cf. Ch. 6, 219–20. Temeritas was a criticism also levelled against Alexander the Great by Quintus Curtius (Alex. 5.11.10), but this was countered by Alexander’s exceptional fortuna (Baynham, 1998, 108). Varus was not so fortunate (cf. Tac. Ann. 1.61). Fortuna was thus also considered an important quality of a general that could counteract unfavourable qualities. 472 Indeed even a general who was defeated in battle was not necessarily looked down upon by his contemporaries, as long as he had displayed virtus. That is moral courage, in making and adhering to decisions, and also physical courage, by remaining near the front line to observe the fighting and encourage the milites. See Rosenstein, 1990, 151–2; Goldsworthy, 1996, 164–5, 169–70. 473 Lendon, 1999, 323–5. 474 Onasander’s discussion of the retreat simply states that a general should keep his troops in formation (Strat. 2.27). He does not prescribe how to do so. Frontinus (Strat. 1.5, 2.13) briefly lists a number of stratagems that a general might use to distract or deceive pursuers, but these will have been vague guidelines for a general in action. It was clearly up to the general to keep his troops in order, and perhaps deceive the enemy to allow a retreat, but the methods of doing so were very much left to the commander. 475 As mentioned, the Romans also considered luck important. Sulla, for example, made much of his fortuna, hence his agnomen Felix (Plut. Sull. 44).


133

The General’s Tactical Duties

While the Romans conceived of a good general in terms of his personal qualities, this belies

the fact that there were indeed many technical tasks he had to perform to do his job

effectively. The general had much to do even before battle was joined, as it was difficult for

him to influence events once his unwieldy battleline clashed with the enemy’s, and so he had

to do as much as possible to give his own army the advantage prior to this point.476 Logistics

were a key factor, and a good general would extensively plan for how his army would receive

sufficient supplies during a campaign.477 He would probably be assisted in this by other

officers such as the quaestor.478 Training the army and earning the respect and confidence of

the soldiers was a primary objective for a general.479 Prior to any fighting, the first thing a

general might have done was to gather and interpret intelligence in order to decide on his next

movements.480 However, throughout the Republic and even into the Principate, neither the

Romans nor their opponents had any significant institutional intelligence capabilities.481 The

'blindness' with which operations were conducted meant that commanders were constantly

reacting to the movements of their opponents and could easily blunder into their opposition,

forcing a battle in a location that perhaps neither general anticipated nor desired.482 Even

Caesar, who had an excellent reputation for reconnaissance, still found himself fighting in

unfavourable positions on multiple occasions.483 Although Roman generals had a consilium,

there was no formal staff structure and so the burden fell completely on the general to

organise the collection and interpretation of intelligence.484

476 Sabin, 2007, 406. 477 Roth, 1999, 154–5, 276–7, 327–8, 331. 478 Ibid., 277. 479 Sabin, 2007, 432; Phang, 2008, 76, 112–3. 480 Sabin, 2007, 406. 481 Austin and Rankov, 1995, 108, 246; Sheldon, 2001, 26, 30. Personal reconnaissance by the general and/or his subordinates was the main source of information that a general could utilise to plan for the location of a battle, and there were therefore great differences in the level of information one commander might have access to relative to another (Austin & Rankov, 1995, 86, 107–8, 244). Some generals were more lax than others (Sage, 2008, 135–6), but even the best commanders could find themselves fighting on unfavourable ground (Goldsworthy, 1996, 128). 482 Stand–out examples of how such reactive movement could have enormous consequences for a battle's location and outcome include the rushed deployment of Philip V's army at Cynoscephalae, 190 BC (Polyb. 18.22) and the decisive cornering of Afranius and Petreius by Caesar near Ilerda in 49 BC, after extensive manoeuvring (Caes. BCiv. 1.61–84). 483 Goldsworthy, 1996, 128. 484 Eg Caes. BGall 1.39–40. See Sabin, 2007, 406–7.


134

Having gathered some amount of intelligence and assuming the general’s army had not

blundered into the enemy, his next step was to decide whether a battle should be fought, and

to decide where and when to do so.485 A general might consult with his subordinates, and

perhaps even take into account the mood of the milites, in making such a decision.486 Ideally

a general would give this own army every possible advantage, fighting on terrain that suited

his army and disadvantaged the enemy’s.487 The general would then need to arrange his army

into an orderly and simple battleline that could be maintained during the chaos of battle, and

he might establish a specific reserve force to react to changing circumstances, although the

triplex acies itself meant that the Romans habitually deployed with reserve lines.488

A general would speak to his men, or as many as he could, to try and boost their morale.489

Generals could attempt to move units or men around to counter the enemy deployment and

pit strong units against weak ones, but this was laborious and probably not always

advisable.490 Scipio Africanus’ post–deployment manoeuvres at Ilipa in 206 BC constituted

an unusual case executed by a talented general with an unusually skilled army (Polyb. 11.23).

Following preparations, the two armies would likely engage one another.

Once battle was joined, the ways in which a general could influence matters were somewhat

limited due to the primitive methods of command and control in ancient armies.491 Roman

generals had a tendency to remain detached from front–line fighting so that they had a

broader view of events, unlike Alexander and his Hellenistic emulators.492 The general might

thus communicate by messengers or by musical and visual signals.493 His main duties during

the battle would be to inspire his men by his presence, and to commit reserve forces or his

485 Goldsworthy, 1996, 167. A general who had a strong enough advantage in intelligence could possibly even win a bloodless victory, as Caesar did at Ilerda, 49 BC, by manoeuvring the Pompeians into a position where they could not escape nor forage (Caes. BCiv. 1.73–84; Cagniart, 1995, 41–2). 486 Eg. Caes. BGall. 1.2, 1.52, 3.3; BCiv. 1.41, 1.72. 487 Sabin, 2007, 404, 406, 432. 488 Ibid., 404–5. Cf. Caes. BGall. 2.20; BAfr. 60. 489 Sabin, 2007, 406. 490 Ibid., 407. 491 Ibid., 403, 406. 492 Ibid., 408. Plutarch (Pyrr. 16) gives an idealized view of combining both the Alexandrian ideal of engaging in personal combat, while maintaining the presence of mind to manage the wider battle, but surely both fighting and managing the army was generally not possible. 493 Goldsworthy, 1996, 168.


135

own personal guard at decisive points in the struggle.494 For example, when Tiberius attacked

a Dalmatian stronghold in AD 9, as his plan was a straightforward one and he had little to do,

he sat to watch the fighting in full view of the army, so that his men would fight to impress

him, and also so that he could see clearly when and where to commit reserves (Dio 56.13.3–

4). Flank attacks were a simple but highly effective tactic, and generals might strive to effect

them while the heavy infantry clashed in the centre.495 If necessary, some generals would

become involved in the fighting itself to bolster the morale of their troops, and a very few

even fought in the front line.496 Overall, as a general’s influence in battle was limited, the

main functions of the general were to observe the fighting, encourage their men where

necessary, and direct reserves as the situation required.497

Finally, it must be noted that a battleline could be too long, and visibility on a battlefield too

limited, for a single man to oversee the fighting in every place.498 For this reason, a Roman

general would also assign subordinate officers to other parts of the line to perform a similar

role to his own, that is to oversee the fighting, encourage the men, and possibly manoeuvre or

commit reserves.499 Such officers might have been tribunes, praetors, quaestors, co–consuls,

or legati.500 The flexibility of the Roman command structure meant that subordinates could

have significant latitude in their own tactics, as for instance a tribune commanded the

decisive Roman flanking attack at Cynoscphalae in 197 BC (Polyb. 18.26), and at the Battle

of the Sabis in 57, Caesar’s left flank under Titus Labienus defeated the enemies opposed to

them and stormed the enemy camp while the rest of the battleline remained relatively

stationary (Caes. BGall. 2.23, 26). However, even if an army had multiple commanders or

sub–commanders of senior rank, there was typically one man who was the general.501 The

494 Sabin, 2007, 408. 495 Ibid., 411, 413. 496 Goldsworthy, 1996, 168–9. There are several examples of such behaviour among Roman generals, particularly those viewed favourably by the sources such as Marius (Sall. Jug. 98), Sulla (App. Mith. 49), and Caesar (Caes. BGall. 2.25). 497 Goldsworthy, 1996, 165, 169; Sabin, 2007, 432. 498 Sabin, 2007, 408. 499 Goldsworthy, 1996, 168. At Cynoscephalae in 197 BC, the decisive flanking attack was led by a tribune (Polyb. 18.26.1–3. When preparing for a major battle, Caesar appears to have placed his legions under the command a variety of senior officers (Caes. BGall. 1.52, 2.11). Polybius also gives several examples of split command (cf. 3.114, 14.4, 15.12). 500 On the active role of military tribunes, cf. Eckstein, 2006, 219. For examples of delegation or split command, cf. Polyb. 3.114, 14.4, 15.12; Caes. BGall. 1.52, 2.11. 501 For instance, in the campaign of Cannae, 216 BC, the consuls Terentius Varro and Aemilius Paulus alternated their commands from day to day (cf. Polyb. 3.110–2–4, 112–113.1), and when it came to battle, their battleline


136

difference between this man and the other senior commanders with similar roles is that he

could travel across the entire battlefield and intervene where he deemed necessary, as Caesar

did at the Battle of the Sabis, 57 BC (Caes. BGall. 2.25–6) and at Alesia in 52 (BGall. 7.86–

7), while subordinate commanders remained at their assigned part of the battleline.502 A

general also had the power to punish or reward his soldiers. For example, Caesar would

lecture his soldiers when they performed poorly, and punish certain individuals if necessary,

and a harsher disciplinarian, such as the Imperial general Gnaeus Domitius Corbulo, could

push his men to their physical limits and dispense brutal punishments, but also earn their

respect by displaying his own virtus.503 There was also occasionally the possibility of severe

exemplary punishment, such as decimation.504 As such, the general’s mere presence could be

a significant source of both motivation and coercion. Therefore, while in some cases there

were other senior officers who could encourage and lead the milites and even commit

reserves, the general’s presence had greater impact and the burden fell more heavily on him

to do these tasks in the correct time and place.

While the technical requirements of generalship in battle were rather simple, they required

very precise and calm judgement, which under the circumstances required significant

physical and moral bravery. As such, the Roman belief that a good general was primarily a

man of good character, possessing courage, knowledge and wisdom, accurately encapsulates

the personal qualities that a man will have needed to perform the various tasks and duties

tasks of a general well and under pressure.

Generalship in Retreats from the Battleline and Into a Fortified Position

In previous chapters, a variety of scenarios in which retreats were conducted have been

examined. The most prominent scenario concerns retreats directly from a battleline. In

addition, retreats into or out of fortified positions have been considered, as well as

withdrawals over longer distances. It has been shown that each situation was different and

required the use of different tactics and, therefore, different actions by a general. The roles

was split into a centre and two wings, with each consul commanding one wing and two consular legati commanding the centre (3.114), yet Terentius was clearly in overall command (3.113.1). 502 Goldsworthy, 1996, 168. 503 Caes. BGall. 7.52, BCiv. 3.73–4; Tac. Ann. 13.35–6. 504 Eg. Plut. Ant. 39.7. Decimation was uncommon in the Middle Republic, but came back into use during the civil strife of the Late Republic (Goldberg 2015, 158–9).


137

played by generals in these distinct circumstances will be discussed below. First, however, let

us consider how a general could prepare for retreat from battle and what he might do to

facilitate the process.

A battle would ideally be fought on ground of a general’s choosing, so how did his choice of

battlefield impact upon the possibility of retreat?505 Assuming the terrain was not so hostile or

unknown as to prevent the Romans from fighting a pitched battle, it was not uncommon for

the Romans to fight battles near high ground to which they could retreat if necessary, as at

Nicopolis in 48 BC, or (perhaps inadvertently) at the Romano-Numidian battle of 106 BC.506

It is striking that Polybius (2.32–3) castigates the consul Gaius Flaminius for deploying his

army against the Insubres in 223 BC with the river Clusius to its back, a calculated risk that

left the Romans with ‘no room to fall back gradually’.507 Evidently, a general was expected to

leave his army some line of retreat when choosing a place for battle. The use of rivers as

tactical devices or defensive barriers in battle was also fairly common.508 Rivers could also

be used to delay enemy pursuers, as Labienus did in 53 BC.509 That said, a river provided

inferior defence to that of high ground or a properly fortified camp (Livy 23.26.10). A general

will therefore have looked for terrain that allowed the placement of a strong camp, or at least

offered the army some prospect of safe retreat to a hill or town.

505 Sabin, 2007, 403–4. 506 Nicopolis: Caes. B. Alex. 40. Romano-Numidian battle of 106: Sall. Jug. 98. The Numidian battle was not a planned engagement for the Romans, so the presence of nearby hills was perhaps better luck than management. In hostile and unknown terrain, however, such as at Carrhae in 53 BC (Plut. Crass. 23–7) or Kalkriese in AD 9 (Dio 56.18–22.2), the Romans might be forced to fight with no safe line of retreat, a nightmare scenario for the general and his men. 507 Polyb. 2.32–3. Note that the consul Tiberius Sempronius Longus made a similar calculated risk at the Trebia in 218 BC (Polyb. 3.72), and the only part of his army to survive was the centre, which broke through the enemy and conducted a fighting withdrawal (Polyb. 3.74), while the rest of the army was destroyed with their backs to the river. 508 There are a few examples from our period of rivers being used as defensive obstacles, or otherwise interfering with the movement of combatants: Polyb. 2.33–4; Caes. BGall. 1.53, 2.23, 4.15, BCiv. 3.71; Livy 39.31; App. Pun. 102. Note also the predilection of the Achaemenid Persians for using rivers in such a way, such as at the Granicus in 334 BC (Arr. Anab. 1.13) and Issus in 333 (2.10), to reinforce their battlelines against Alexander III’s Macedonians. 509 Caes. BGall. 6.7–8. Also see Livy 37.38.1–4.


138

If the terrain was not favourable, the Romans could use other tactics to secure a retrograde

movement in open ground. In a retreat carried out at Dyrrhachium in 48 BC, Caesar utilised

several kinds of ad hoc obstacles, including ditches and hurdles, to delay pursuing troops

(Caes. BCiv. 3.46). In 211 BC, Gnaeus Scipio was withdrawing in an attempt to evade

Carthaginian forces that had recently destroyed his brother’s army, but was brought to battle

at Ilorca and forced to halt his army on a hill that was not steep enough to provide significant

defensive value (Livy 25.35–6). In this scenario, Scipio used sarcinae and other baggage to

create a makeshift rampart, somewhat hindering the enemy, although the Romans were still

overwhelmed (25.36.1–13). However, there are overall few examples of the use of obstacles

to assist fighting retreats in open ground. This is probably because there must have generally

been little time to organise or plan the use of obstacles to assist such a retreat.510 In general,

rather than rely on such tactics, the Romans fought in a location where the army had access to

a place of refuge in the event of defeat.511

Aside from preparing for the event, the personal actions a general could take to assist his

army during a retreat are revealed in a number of illustrative examples. Sallust's account of

the Romano-Numidian battle in 106 BC (Jug. 97–8) emphasises the role of Marius in

organising a safe retreat in an unfavourable situation. When his army was thrown into

confusion by a surprise attack, Marius intervened with his personal guard of cavalry, assisting

the hard–pressed groups of milites, charging the enemy, giving orders by hand gestures, and

generally rallying his forces so that he could withdraw them to a nearby hill (98.1–4).

Caesar’s accounts of his own battles demonstrate that, in the event his men had to retreat, he

would manoeuvre reserves to intimidate pursuers.512 At Dyrrhachium, 49 BC, he deployed

archers both to construct obstacles and to protect a retreating legion from pursuers (Caes.

BCiv. 3.46). Caesar clearly kept a close watch on what was happening at all times and would

personally intervene to encourage or reinforce his men if absolutely necessary.513 Caesar

implies, in his account of the Battle of Atuatuca, 53 BC, that a good general verbally

encouraged his men as they fought (Caes. BGall. 5.35.5). The fact that a general's death or

510 Much could be done with more time to prepare, such as Pompey’s elaborate preparations at Brundisium (Caes. BCiv. 1.27–8). 511 Higher ground: Polyb. 2.68, 3.86, 11.22; Caes. BCiv. 1.79, 2.41–2, 3.46, 3.95, BAlex. 40; Sall. Jug. 98; Livy 5.38, 32.12; App. Hann. 41, Pun. 103. The Romans usually had a camp near their chosen battlefields (Caes. BGall. 7.49–51; BCiv. 1.45–6; Sabin, 2007, 414). 512 Caes. BGall. 7.49; BCiv. 1.45. 513 Caes. BGall. 2.25–6, 7.87–8, BCiv. 3.65.


139

flight during battle could easily trigger a rout shows his importance in motivating the army,

even in retreat, as is shown by the reaction of the Carthaginian army when their general Mago

Barca fell during a retreat in 203 BC.514 Clearly, the importance of a general’s personal

presence during a retreat cannot be overstated, and a good imperator, such as Marius or

Caesar, was highly visible during a retreat.

Once the army was in retreat, it will commonly have withdrawn into its camp or a fortified

position, but such a manoeuvre was dangerous, perhaps impossible, if the enemy could not be

kept at arm’s length while the army moved into or out of a camp or town.515 Although

retreating armies were vulnerable when retreating into a fortified place, this did not mean

pursuers always had free rein. Consider Caesar’s account of the later stages of the Battle of

Pharsalus, 48 BC, after Caesar’s men had driven the Pompeians from the field towards their

camp, with Pompeius himself arriving first (Caes. BCiv. 3.94–6):

Pompeius ... left the field and straightway rode off to the camp. To the centurions

whom he had placed on duty at the praetorian gate he exclaimed in a loud voice that

the troops might hear: ‘Protect the camp and defend it carefully (castra et defendite

diligenter) if anything goes amiss. I am going round the other gates and encouraging

the guards of the camp ...’ Having said this, he betook himself to the general's

headquarters, mistrusting his fortunes and yet waiting to see the issue ... the camp was

being zealously defended (industrie defendebantur) by the cohorts which had been

left there on guard, and much more keenly still (multo etiam acrius) by the Thracians

and barbaric auxiliaries. For the soldiers who had fled from the battlefield (qui acie

refugerant milites) panic–stricken in spirit and exhausted by fatigue, many of them

having thrown away their arms and their military standards, were thinking more of

further flight than of the defence of the camp. Nor could those who had planted

themselves on the rampart stand up any longer against the multitude of javelins, but,

worn out by wounds, quitted their position (sed confecti vulneribus locum

reliquerunt), and forthwith all, following the guidance of centurions and military

tribunes, fled for refuge (confugerunt) to some very lofty hills that stretched up to the

514 Livy 30.18.12–4; Sabin, 2007, 432. 515 Cf. Pol 14.4–6; Livy 25.39; App. Mith. 44. See Ch 2, 84–6.


140

camp.516

As is clear, specific preparations had been made prior to the battle for the defence of the

camp in the event it was attacked or, presumably, if the Pompeians were forced to retreat into

the camp. There were centuriones on duty at the praetoria porta and evidently the other gates

were guarded too (... reliquas portas ... et castrorum praesidia confirmo). There were some

number of cohorts (ie. legionaries) left on guard–duty in the camp during the battle, in

addition to number of Thracians and other auxiliaries, who all were supposed to defend the

camp strongly (industrie defendebantur). In the event, the disorganised and demoralised men

who had fled from the battlefield were of little assistance, and Caesar’s army was able to

overwhelm the other defenders. However, the key point is that although Pompey’s army was

no longer fit for battle, his camp was well–defended enough by other, fresh troops that it

could make a strong effort at holding off an enemy attack. Posting soldiers specifically to

defend a camp during battle was probably a routine procedure by attentive generals.517 These

men thus provided a final line of defence against pursuers in the event that an army was

driven from the battlefield and was able to return to camp.518

Aside from having assigned these men to guard the camp in the first place, a general might

also do a circuit of the camp to ensure that everybody was in place to defend it, and give

encouragement to his men (Caes. BCiv. 3.94). However, Pompey himself withdrew to his

headquarters (se in praetorium contulit) instead of overseeing the defence (Caes. BCiv. 3.94),

and, when the camp was taken, he fled, stripping off his general’s insignia (3.96). In essence,

Caesar is accusing Pompey of not doing his duty as a general by not remaining to oversee the

actual fighting.519 According to Caesar then, a Roman general should have left all the gates

guarded and assigned reliable men to defend the camp in the event that the army was

516 Trans. A. G. Peskett, 1914. 517 At Magnesia (190 BC), Appian specifically points out (Syr. 6.36) that the Romans had not left any cavalry or light troops to defend their camp, as the general expected that a nearby river provided enough defence, and conversely that the Seleucids did indeed have men guarding their own camp. Appian thus implies that it was actually unusual for the Romans to have done this. 518 Livy (7.7.1–5) describes an engagement between the Romans and Hernici in 362 BC, in which the Romans sheltered in a camp, having previously been defeated, but when their enemy attempted to storm the camp the Romans defeated them and threw them back in confusion. 519 Note that Pompey had also previously left his men during the battle itself to withdraw to the camp (Caes. BCiv. 3.94). Whether or not Pompey truly behaved in a manner so uncharacteristic of such an experienced general, Caesar infers what his responsibilities should have been.


141

defeated. He should have encouraged the men and remained present during the fighting, and

as the camp fell, he should have withdrawn his soldiers to another defensible position, just as

Pompey’s subordinates did (3.95).

Other examples demonstrate better conduct by generals who found themselves in dire straits.

One might compare Pompey’s alleged behaviour at Pharsalus to the conduct of Aulus

Caecina Severus in Germany in AD 15. Caecina and his men, having been ambushed and

forced to encamp, and expecting a major enemy attack, were in low spirits (Tac. Ann. 1.65).

Confusion arose in the camp and in their panic the soldiers attempted to flee from the

fortifications, when Caecina intervened to stop his men and even threw himself down in the

gateway to prevent the men passing over him (Ann. 1.66). By a combination of the general’s

efforts and the entreaties of the tribunes and centurions, the men were brought back to order

and the army saved (1.66–8). A good general would thus personally intervene to prevent

premature retreat or flight into or out of a camp.520 If all attempts at retreat or withdrawal

failed, a general could always surrender to save his men. At Ilerda in 49 BC, after the

Pompeians failed to withdraw and were trapped by Caesar’s army, the Pompeian general

Afranius surrendered to Caesar, motivated by the plight of his suffering milites (Caes. BCiv.

1.84). Caesar, who was otherwise critical of Afranius (BCiv. 1.75) did not fault him in this

instance (1.85). And if neither retreat nor surrender were possible, a courageous general

would die with his men, as did Aemilius Paulus at Cannae (Polyb. 3.116), Aurunculeius Cotta

at Atuatuca in 54 BC (Caes. BGall. 5.37), and Quinctilius Varus at Kalkriese in AD 9 (Dio

56.21.5).

During a retreat from battle then, the Roman general needed to plan effectively by choosing a

battle site and camp location that offered refuge to retreating men, he had to deploy his army

in such a way that it could fall back if necessary, to boost the morale of the soldiers by

maintaining a presence and, if necessary, displaying his fighting qualities, and to make

decisions under pressure regarding matters such as when and where to retreat, and how to

utilise reserves.

520 Also note Livy 10.36.6–8.


142

Generalship During Retirement From a Fortified Position and On the March

In the event that an army had to depart from a fortified position near a watchful enemy, it will

have been a danger that the withdrawing army could be attacked while leaving its position, or

outmanoeuvred and entrapped.521 As such, the general had to devise tactics or stratagems to

avoid this scenario. The stratagems and tactics that a general could use to secure escape from

a fortified position, and to defend the army while it marched in the open, will be considered

below.

In AD 66, Cestius Gallus campaigned against Jerusalem in an attempt to put down a Jewish

rebellion, but was forced to raise his siege due to logistical troubles.522 The army made a

chaotic withdrawal to Gabao, where Cestius rested his army for two days and gave the enemy

time to strengthen their forces.523 The Romans then withdrew further, taking heavy casualties

along the way, and they just barely found refuge behind the walls of Beth–Horon.524 Here the

Romans became entrapped, but Cestius devised a stratagem wherein he left a small part of his

garrison behind to give the impression that the army had not moved, allowing most of the

men to withdraw unharmed (Joseph. BJ 2.551–3). Despite the errors that had led him to that

point, Cestius thus saved his army.525 Evidently, withdrawal from one defensible position to

another was hazardous enough, but even once an army reached refuge, if the enemy was

watchful then some stratagem was needed to secure escape.

There were a variety of such stratagems. At Brundisium in 49 BC, Pompeius Magnus was

trapped in the town and needed to organise a withdrawal, but the besieging forces of Julius

Caesar were well–prepared for this event (Caes. BCiv. 1.25). Pompey took various actions to

secure his escape, including dismantling some of Caesar’s obstacles, establishing obstacles of

his own to delay pursuers, and embarking a large part of his army silently during the day then

the rest (presumably the rearguard) hurriedly by night (BCiv. 1.27–8). In 45 BC, near Munda,

521 Sall Jug 76.1; Caes BGall. 7.11.6, BCiv. 1.25–8; Plut. Luc. 19. 522 Cestius had lost part of his baggage train to a Jewish attack (Joseph. BJ 2.517–22, cf. Gichon, 1981, 54–5). Josephus expresses confusion at Cestius’ decision to withdraw, while Gichon (56) argues that, despite Cestius’ sound tactical decision making during the siege itself, he had good reason to feel uneasy about continuing the siege with winter approaching and active enemy forces to his rear. 523 Joseph. BJ 2.544–5. Cf. Gichon, 1981, 57. 524 Joseph. BJ 2.546–50. Cf. Gichon, 1981, 57–9. 525 Gichon, 1981, 60.


143

Gnaeus Pompeius (son of the triumvir) set fire to the town of Ucubi to cover the retreat of a

garrison he had posted there (BHisp. 27.4). Rome’s enemies also used fire to secure escape

from towns, as did the Numidians at Thala in 108 BC, and Pontic soldiers at Amisus in 72

BC.526 Frontinus records numerous cases in which generals secured safe withdrawals by

various methods of deceiving or distracting the enemy, by setting fire to buildings or other

flammable objects to distract the enemy and mask movement, and even by feats of

engineering to create a new escape route.527 Finally, there are several examples of armies

using nightfall to cover withdrawals from camps or towns, including the Carthaginians at

Ilipa in 206 BC (Livy 28.16.3–6), the Gauls at Cenabum in 52 BC (Caes. BGall. 7.11), and

the Romans at Ilerda in 49 BC (Caes. BCiv. 1.67).

To summarise, there were a number of stratagems a general could use to give his army to

time to move out from a fortified position without harassment, including setting buildings

aflame, moving by night, and deceiving or distracting the enemy by various methods, such as

simulating movement or the lack thereof. However, in these circumstances the advantage

remained on the side of the pursuers, as they were in the physical position of advantage, and

no stratagem was necessarily proof against vigilant opposition (Caes. BCiv. 1.65–6). The

general will have had to use his creativity, decide what approach was best for the situation,

keep good control of his forces, and trust to fortune.

An army on the march (whether withdrawing or not) was vulnerable to attack, especially by

cavalry, or to ambush, and so a general also had to avoid or counter these dangers.528 To

begin with, the army’s ratio agminis will have been a primary consideration. There were of

course numerous ways of organising the army’s order of march. A general could assign

picked troops, such as the elite extraordinarii of the Italian socii, to form part of the column

that was most likely to be attacked (Polyb. 6.40). A general could assign various units of

infantry or cavalry to particular parts of a column to scout ahead or protect against enemy

attack, depending on his personal preferences and the circumstances.529 It was possible for an

army to march in a triple column (Caes. BCiv. 41.1), which would allow it to make a ninety–

526 Thala: Sall. Jug. 75–7. Amisus: Plut. Luc. 19. 527 Front. Strat. 1.5.2–5, 8–13, 16–9, 22. 528 Cavalry pursuit: Caes. BCiv. 1.79–81; Livy 28.16.3–5. Ambush: Polyb. 3.118; Vell. Pat. 2.117.1; Joseph. BJ 2.540–555. 529 Goldsworthy, 1996, 105–9.


144

degree turn and thus be standing ready to fight in the triplex acies.530 Fundamentally, an army

usually marched in a column, which made the army far more mobile, but more vulnerable to

being attacked and delayed.531

A key factor in determining an army’s marching formation was the presence of the baggage

train, the aptly–named impedimenta, which for an army of several legions and auxiliary units

might consist of thousands of pack animals, hundreds of wagons, and several thousand

military servants.532 The baggage train carried the soldiers’ food, armour, missiles and other

equipment, and may have consisted of around 175 metric tons of supplies for only a single

legion.533 This baggage was crucial to the army’s effective operation but the train was heavy

and its rate of movement was slow, rendering it vulnerable. Roman generals could assign

large forces to escort the baggage, as for instance Caesar did when he assigned two full

legions to this task prior to the Battle of the Sabis in 57 BC (Caes. BGall. 2.18). Marcus

Antonius too assigned at least two legions to guard his baggage at the outset of the Parthian

campaign in 36 BC (Plut. Ant. 38.2–3). Evidently, the defence of the baggage was held to be

extremely important. For this reason, when an army was advancing through hostile territory,

the Romans would often place the baggage inside a hollow square formation, although a

columnar formation could still flank the baggage train.534 However a general approached the

problem, in considering an army’s ratio agminis, whether in advance or withdrawal a general

needed to both defend the baggage and arrange the army in such a way that it could defend

itself if attacked.

What of the actual route of march? The crushing defeats of withdrawing Roman armies at

Carrhae in 53 BC (Plut. Crass. 29–31), Kalkriese in AD 9 (Dio 56.19–21), and Beth–Horon

in AD 66 (Joseph. BJ 2.546–552) demonstrate how marching through locus iniquus could

bring disaster. The army’s route will naturally have been the general’s responsibility (Caes.

BGall. 1.40.10–1), but will have depended on the intelligence available to him. As

mentioned, during this period the Roman army lacked professional intelligence gathering or a

530 Carter (1991, 190) states that this was a formation used for dangerous marches. Caesar used it repeatedly

(Caes. BGall. 1.49.1–2, 1.51.1–2, 4.14.1, etc). 531 Goldsworthy, 1996, 109. 532 Roth, 1999, 114–5. 533 Roth, 1999, 79–84. 534 Goldsworthy, 1996, 105, 109.


145

staff apparatus for this purpose.535 When operating in hostile territory then, generals were

often dependent on local information.536 Such information could be reliable, as it apparently

was for Caesar in Gaul and Marcus Antonius in Parthia, but treachery could also lead a

general into terrain where his army would be disadvantaged, as occurred in the cases of

Sabinus, Crassus and Varus.537 In essence, while the gathering and processing of intelligence

was almost entirely the general’s responsibility, and the effectiveness with which he used this

information to choose a path of withdrawal could be decisive, the limitations of Roman

intelligence gathering in this period meant that a general could not always have chosen the

perfect marching route, just as he could not have always chosen the perfect battlefield.

The baggage train significantly slowed an army, but what could a general do if his army

needed to withdraw more quickly? Consider the following extract from Dio’s account

(56.21.1) of the clades Variana of AD 9:

Accordingly they encamped on the spot, after securing a suitable place, so far

as that was possible on a wooded mountain; and afterwards they either burned

or abandoned most of their wagons and everything else that was not absolutely

necessary to them. The next day they advanced in a little better order.538

Evidently, Varus abandoned his baggage to prevent it from hindering his forces as they

continued their march (Dio 56.20.5). This is a description of a procedure for which numerous

parallels can be found. In 56 BC, in Gaul, Caesar’s legate Sulpicius Galba was advised that,

being trapped by the enemy in the Swiss Alps, he should consider abandoning the baggage

and march for safety (Caes. BGall. 3.3). In Gaul two years later, the legate Titurius Sabinus,

when ambushed in a defile near Atuatuca, ordered the abandonment of baggage to form a

better fighting formation.539 After Crassus’ defeat near Carrhae in 53 BC, it can be assumed

that, along with 4000 wounded, much of the baggage was abandoned in the retreat from the

535 Austin & Rankov, 1995, 29–30. 536 Pol 9.14.2–4. Also note Caes. BGall. 1.41.4; Plut. Crass. 21; Ant.46, 48.1–2. 537 Reliable: Caes. BGall. 1.41.4; Plut. Ant. 40.1–5, 46.2–48.1, 49.2.Treachery: Caes. BGall 5.33; Plut. Crass. 21; Dio 56.19. 538 Trans. Earnest Cary, 1924. 539 Caes. BGall. 5.33. Although Caesar criticises Sabinus’ tactics, he states that abandoning the baggage to form a square formation was appropriate in an emergency. Likewise, Dio evocatively describes (56.20.5) the impediment that Varus’ baggage allegedly created in the initial fighting, hence the need to retrench much of the baggage and depart in better order (56.21.1).


146

stricken battlefield.540 Livy mentions several instances of armies, Roman and otherwise,

abandoning their baggage temporarily or permanently to gain an advantage, as a result of

defeat.541 In his withdrawal from Jerusalem in AD 66, having already lost much of his

baggage train, Cestius Gallus entrenched his army and destroyed most of his beasts of

burden, except those required for carrying missiles and war machines, before setting out once

more.542 In AD 15, Caecina did not destroy his own baggage, but appears to have lost most of

it in the fighting (Tac. Ann. 1.65). Therefore, at least in the Late Republic and Early Empire,

there appears to be a Roman practice of abandoning baggage–temporarily or permanently–as

a standard procedure when withdrawing.

The two possible justifications for the abandonment of baggage were that the discarded

baggage might offer a tempting target for enemy plunderers and facilitate the army’s escape

or the army would be rendered lighter and faster.543 If Varus abandoned the baggage, it was to

render his army more mobile and effective in battle, considering the trouble caused by the

impedimenta during the initial attack (Dio 56.20.5). If the enemy were distracted by

plundering this baggage, so much the better. Thus, the abandonment of baggage was a

sensible and regular response to the need for rapid movement and increased short–term

fighting capability, such as in the case of retreating or withdrawing out of an ambush, but the

obvious supply problems this caused for the army meant it was an option of last resort.

How did a general protect his marching column from enemy pursuers and marauders? A

square formation allowed the army to defend itself from all directions and protect the

baggage inside the square, but this did not render the army proof from attack. This problem

and various solutions to it are demonstrated in Plutarch’s account of Marcus Antonius’ 540 Plut. Crass. 28.1. Surely thousands of wounded legionaries would not have been left behind if the army had retained the capacity to carry them, as the army appears to have carried those it could (cf. Crass. 27.6). 541 Temporarily: Livy 3.28.1. Permanently: Livy 4.39.6; 8.1.5–6. Although Livy’s descriptions of the tactics in these early Republican battles are probably invented, he can only have drawn upon his knowledge of contemporary military practices to fill in the blanks (cf. Oakley, 1997, 351). As such, these were probably genuine Late Republican practices. 542 Joseph. BJ 2.544–546. As Dio implies (56.21.1), Varus must have kept some of his baggage. For one reason: he needed to carry his wounded (Rost, 2016, 561). Recent archaeological finds imply the presence of a ballista in the Senke, which suggests that Varus might also have kept some of his war machines (Stark, 2016). 543 Cestius Gallus left his baggage behind in AD 66, although unwillingly, and as a result his enemies were distracted from their pursuit (Joseph. BJ 2.553–554). The ambush of the Nervii and their allies on Caesar’s army in 57 BC appeared to have been targeted towards the baggage (Caes. BGall. 2.17), as did the attack of the Thracians on Manlius Vulso in 188 BC (Livy 38.40.9–15). Also see Tac. Ann. 1.65. More mobile: Engels, 1978, 131–132, 153–156.


147

laborious withdrawal from Parthia in 36 BC (Ant. 40–50). Although the imperator formed his

army into a hollow square, he and the men were still regularly attacked by the following

Parthians (42.1). To fend off these attacks, he utilised large numbers of missile troops to

counter the withering fire of Parthian archers.544 He also used both cavalry and infantry to

counterattack and drive back aggressive pursuers, giving space for his army to continue

retiring under less pressure (41.5–42.4). He avoided the open plains where his slow–moving

force would be most exposed to the enemy’s massed cavalry and instead adhered to higher

ground on his march (46–7). He also encamped to recover the order of his forces when

necessary (47.4, 48.5).

There are parallels to all of these mitigating actions. As mentioned, in 48 BC Caesar also used

missile troops to cover the retreat of a legion at Dyrrhachium (Caes. BCiv. 3.46.1). The

Pompeian army, when withdrawing during the Ilerda campaign of 49 BC, used lightly armed

legionaries (expedites cohortes) to fight and counterattack pursuing cavalry (BCiv. 1.79). The

Pompeians also attempted to control the high ground along their path of withdrawal, using it

to fight pursuers from a position of strength and also denying control of these positions to the

enemy (1.79–80). Crassus also attempted to use high ground to his advantage in his

withdrawal from Carrhae, 53 BC (Plut. Crass. 29.5–6). Finally, halting a march and

entrenching the army in a camp or behind city walls, to recover the army’s order or as a last

resort if the army was unable to move successfully, was a technique used by various Roman

generals in comparable situations.545

To summarise, a general who had to withdraw his army out of a fortified position, or

withdraw over open ground, had first to deceive the enemy or manoeuvre in such a way that

he could bring his army safely into the open. He then had to organise his marching column so

that the army could move quickly but also fight off pursuers and protect the vulnerable

baggage train. He had to select an appropriate route for the army, and use terrain to his

advantage or change tactics if the terrain or tactical situation were unfavourable, for instance

by changing the army’s formation, abandoning the baggage train, or making counterattacks,

in order to throw off pursuers and bring his army safely to another place of refuge.

544 Plut. Ant. 42.1, cf. 37.3 on Antonius’ numbers of troops. 545 Caes. BCiv. 1.80–1; Joseph. BJ 2.550–2; Tac. Ann. 1.63; Dio 56.21.1.


148

Conclusion

The general was central in ensuring that a retreat or withdrawal was carried out with

appropriate methods and tactics. Before a battle, he gathered and interpreted intelligence,

conferred with his consilium, and chose a location for battle that ideally had nearby high

ground, a camp or a fortified position to which a retreat could be made. He then deployed his

battleline, incorporating reserve units that could protect retreating infantry. A general had

two main functions during a battle, and also during a retreat if he saw fit to order one (Caes.

BGall. 7.47.1), particularly if the retreat was challenged by pursuers. The first was to

demonstrate virtus, by keeping a close eye on the fighting, exhorting and inspiring his

soldiers. The second was utilising consilium by dispatching reserves or even his personal

guards where necessary to bolster wavering men and threaten enemy attackers or pursuers. If

an army had to retire into or out of a fortified position, whether forced to or volunteering to

do so, the general had firstly to ensure the ramparts or walls were adequately manned, and if

necessary exhort and inspire the defenders, then plan and execute the army’s march out from

the position in such a way that, preferably, they were not attacked while doing so. Deceptive

stratagems or withdrawing by night were two ways to achieve this. During the march, the

general had to ensure there was an adequate rearguard and forces suitable for

counterattacking any pursuers, establish ratio agminis that protected the army’s baggage and

flanks, defend the baggage during the army’s movement, order counterattacks if necessary,

maintain the army’s direction and use high ground to advantage if possible, and finally halt

the army or encamp if necessary. Without the general performing all of these functions

adequately, orderliness among the heavy infantry would not necessarily ensure a successful

retrograde movement, particularly in more complex situations and over longer distances. In

essence, a general had to be both visible and mindful of the situation, displaying himself to

the soldiers where necessary but also making broader decisions when necessary. These tasks

were simple but will have required tactical knowledge and fine judgement, scientia rei

militaris and consilium, as well as physical and moral bravery, virtus, to apply effectively.

As can be seen the two main features of successful retreats and withdrawals, the maintenance

of order and the management or prevention of enemy pursuit, were the responsibility of all

parts of the Roman army. Soldiers and their immediate leaders, particularly the centuriones

and tribuni, were mostly responsible for maintaining the correct infantry formations and


149

ensuring that they moved at a gradual pace while fighting off enemy pursuers. The general

assisted the maintenance of order through his selection of the battlefield and drawing up the

battleline, by posting and committing reserves to reduce the pressure on retreating troops, and

by exhorting and inspiring his men. The soldiers and junior officers could contribute to the

task of managing enemy pursuit by maintaining their order, as an orderly body of soldiers

was itself intimidating to pursuers. However, a number of the factors that contributed to

ensuring a safe retreat, including nearby high ground or fortified positions, the presence and

commitment of reserves and the allocation of men to guard camps, baggage or marching

columns were all the down to the general’s decisions. Even the effectiveness of the milites

and officers were to some extent the general’s responsibility, for he was oversaw the army’s

training and had some influence over the assignment and competence of his subordinates.

Defects in the ability of milites, centuriones and tribuni in maintaining order during a

retirement could potentially be repaired over time by an attentive and disciplinarian general.

But the army as a whole could be in the greatest danger if a general put them in a

disadvantageous position and had made insufficient preparations for a safe retirement, as

occurred at Geronium in 217 BC, or at Beth–Horon in AD 66. A successful retreat, therefore,

was due to the scientia rei militaris, consilium and virtus of the general, qualities developed

through a lifetime of education and experience, and which ideally he drew upon to calmly

and effectively lead his men and adapt his tactics to the situation.


150

When [the Gauls] unexpectedly saw men whom they believed to be in retreat advancing

against them in attack formation, they could not even sustain the assault, and at the first

charge they were put to rout and sought the nearest woods.

(Caes. BGall. 6.8)

Chapter Five

The Development of Roman Retreat Tactics from the

Middle to Late Republic (218–46 BC)


151

Introduction

In 53 BC, Titus Labienus was encamped in Gaul, when he was warned of the advance of a

formidable enemy army that he feared would overrun his small force (Caes. BGall. 6.7).

Labienus decided to withdraw from his position, with the hope that the enemy would advance

recklessly and give him some opportunity for a successful action (6.7). This is precisely what

happened, and as the enemy pressed forward in a disorderly manner, Labienus quickly drew

up a line of battle and counter–attacked, throwing the surprised Gauls into chaos and routing

them (6.8). In this manner, a single legion used retreat tactics to defeat a much larger enemy

force. Such was the efficiency with which these Romans could retreat, reform a battleline,

and counterattack. Was this efficiency typical for that period, and if so, how did the Romans

develop to that point?

In the preceding chapters, it has been established that the Roman army was well–suited to the

conduct of retrograde movements due to the ability of milites to maintain ratio during retreat

or withdrawal through training, efficient organisation and active junior leadership. In this way

they intimidated potential pursuers and retained their fighting ability even in defeat. Also, the

habitual Roman use of reserve infantry lines and the presence of light infantry and cavalry

could also be used to dissuade or even fight off enemy pursuers, thus preventing enemy

pursuers from applying too much pressure to retreating troops. The most important figure in

the successful conduct of a retreat or withdrawal was, however, the general, who trained the

army, chose many of his officers, chose the battlefield, drew up his army with a safe line of

retreat, managed the use of reserves, exhorted the soldiers, and generally reacted to the

developing situation.

However, it must be recognised that the Roman army was not a monolithic institution, but

one that changed significantly over time. There were several catalysts of change to the

Roman army during the Middle to Late Republic. First, the Roman army suffered some

significant challenges in foreign and civil wars that necessitated changes to the organisation

and tactics that had proven so successful in the Second Punic War.546 The poor performance

546 Although Rome had a period of spectacular success in its rapid conquest of much of the Mediterranean by the mid–second century BC (Polyb. 1.1.5), it thereafter had difficulties in various theatres over a broad space of time, including the surrender of Hostilius Mancinus at Numantia in 136 BC, reverses against Jugurtha in Numidia, the epic defeat at Arausio in 105 BC, the Mithridatic, Servile and Social Wars in the early to mid first


152

of the Roman army during the mid to late second century was a stimulus to military reform,

with the challenge taken up by generals such as Scipio Aemilianus and Gaius Marius.547

Second, the military, political and fiscal crises that contributed to the fall of the Republic also

stimulated change in the Roman army and culminated in the permanent establishment by

Augustus of a new, manageably sized, and more structured army.548 Overall, the Roman

army’s organisational changes (eg. use of cohorts, regularisation of training practices and

command appointments) appear to have been gradual. Changes to its tactics were surely

gradual also. Fortunately, because most sources for this period concern Roman military

matters, and the Roman army was the dominant Mediterranean military force for so long, the

development of Roman retreat tactics over time may be examined in detail.

In this chapter, an examination will be made of the development of Roman retreat tactics

from the late third century BC, when the Roman army first emerges into clear view thanks to

Polybius, to the mid–first century BC, when Caesar provides highly detailed accounts of

several contemporary wars. The purpose of the chapter is to examine the links between the

tactical and organisational development of the Roman army itself, the differing tactics of

certain generals, and observable changes in how the Romans conducted retreats over time.

The chapter will be split into three sections. The first will examine retreat tactics in the

Middle Republic (218–168 BC), with a particular emphasis on the Battle of Pydna (168 BC).

This will provide an example of the sort of fighting retreat that the Mid–Republican Roman

army was capable of, and demonstrate how retreat tactics shaped the course of that battle. The

second part will focus on the period between the Middle and the Late Republic (149–102

BC). Here, the Battle of Nepheris (149 BC) and the Romano-Numidian battle of Sallust (106

BC) will be the foci for much of the analysis, as they display the potential flexibility of

Roman infantry and cavalry tactics. Finally, the third section will examine the tactics of the

Late Republic, particularly in the campaigns and various major battles detailed by Caesar, to

show changes in retreat tactics. The conclusion will identify the key factors and explain the

most significant changes in the Roman army’s retreat tactics over time.

century, and defeats against the Parthians in 53 and 36 BC (cf. Hoyos, 2007, 76). 547 Cagniart, 2007, 81–2; Hoyos, 2007, 76. 548 Keppie, 1984, 145–9; Eck, 2007, 114–22.


153

Retreats in The Punic and Macedonian Wars (218–168 BC)

There were many instances in which the Romans were hard–pressed or defeated during the

Second Punic War and the Macedonian Wars, and as such, there are half a dozen clear

examples of the Romans utilising retreat tactics in battle during this period. To begin with, in

218 BC in Cisalpine Gaul a Roman force under the praetor Lucius Manlius, after suffering a

heavy ambush at the hands of the Boii, fled from the carnage but rallied on high ground and

retreated in a more orderly manner from there (Polyb. 3.40.11–3). During Hannibal’s

overwhelming victory at the river Trebia the same year, the vanguard of the Roman army was

able to push through the centre of the Carthaginian battleline and withdraw in close order to a

nearby colony, where they were later joined by other survivors (Polyb. 3.74.3–6). In 217,

when the Romans were heavily defeated again at Lake Trasimene, a large body of soldiers

was able to force its way through to high ground and make an orderly retreat from that

position (Polyb. 3.84.11–3). Also in 217, when the rash attack of the magister equitium

Minucius was stopped by Hannibal and the Romans found themselves nearly surrounded, a

counterattack by the dictator Fabius Maximus was able to check the Carthaginian advance

and allow the Romans to retreat (Polyb. 3.104–5). In 211 BC, when Gnaeus Scipio was

confronted by the combined Carthaginian armies of Spain, after they had defeated his brother,

he attempted a fighting retreat under heavy harassment by Numidian cavalry until he was

forced to stand and fight, taking possession of higher ground for this purpose (Livy 25.35.7–

36.3). Finally, Polybius (18.25.4) notes that, early in the Battle of Cynoscephalae in 197 BC,

when the Roman left wing struggled to sustain the Macedonian phalanx as it advanced

downhill, the Romans retreated slowly, resisting the urge to flee.

As can be seen, the Romans applied retreat tactics on a number of recorded occasions, and

they generally used high ground to secure their retreats. The Romans routinely deployed their

forces in the triplex acies, and in battle the three lines advanced and retreated short distances,

individually or together as necessary.549 That said, in only two examples of retreats from this

period (Cynoscephalae and Pydna) did the Romans successfully carry out fighting retreats

under pressure from enemy heavy infantry. It was not always easy for the Romans to do so.550

549 See Ch. 3, 113–6. 550 The chaos at Geronium in 217 BC comes to mind, in which the Roman army of Minucius (cos. 221) was thrown into disorderly retreat, and might have been destroyed if not for the intervention of Fabius Maximus’ fresh legions (Polyb. 3.105).


154

Therefore, while there are numerous examples of Roman armies or detachments retreating in

battle during the Middle Republic, they could not do so easily if the enemy responded

aggressively. How effective were Roman retreat tactics in this period then, and how did they

develop?

Roman retreat tactics of this period are most clearly demonstrated in the case of the Battle of

Pydna, 168 BC. This was one of the last recorded major Mid–Republican battles fought by

Rome against an adversary that could offer her a serious challenge in a pitched battle. The

narrative of the battle is engaging drama. That year, on a plain south of the Greek city of

Pydna, a Roman army under the consul Aemilius Paulus met the royal host of Perseus, king

of Macedon, in the climactic battle of the Third Macedonian War. Perseus had scored a

victory against the Romans in 171 BC at Callinicus, and the first onset of the phalanx at

Pydna gave Paulus little reason for confidence (Plut. Aem. 19.2), but he bolstered the courage

of his men (19.3) who attacked with great élan, both sides suffering heavy losses (20.1–4).

The Roman front line was, nonetheless, beaten and gave way before the advance of the

phalanx, as did the rest of the Roman army, which could find no way to successfully

penetrate the wall of sarissae (20.4–6). However, as the phalanx advanced uphill against the

retreating Roman army, it encountered difficult ground that rendered its advance uneven,

breaking up its impenetrable line (20.7). Aemilius saw this opportunity and ordered his

maniples to attack gaps in the phalanx (20.7–8), because of which the Romans were able to

penetrate the phalanx and surround it in multiple places (20.9–10). When the sarissae had

been bypassed and the phalanx disordered by the terrain and the targeted Roman attacks, the

phalangites were easily overwhelmed (20.9–10) and most were killed in flight (21.6).


155

This battle has been presented as a straightforward example of one that was decided primarily

by superiority in Roman equipment and manipular tactics, but the Roman retreat has not

received significant attention as a turning point in the battle.551 Since Plutarch’s account

raises a number of difficult questions,552 which make it particularly difficult to reconstruct the

Macedonian battle plan,553 and other accounts are of no more help,554 it has been stated that

the battle cannot be reconstructed.555 Nonetheless, Plutarch’s account is not devoid of value.

For a start, it is one of the only major battle accounts from the period, and there is no decisive

551 Cf. Plut. Aem. 20.4–6. On the traditional presentation of Pydna as a victory showcasing the superiority of Roman equipment and formations over the Macedonian–style phalanx, as argued by Polybius (18.28–32), see, for instance: Fuller, 1965, 89; Hammond, 1984, 46; Goldsworthy, 2004, 104; Waterfield, 2014, 190. It is argued in particular that the more close–ranged equipment and fighting style of the Romans gave them a decisive advantage once the phalanx was penetrated (Lendon, 1999, 284; Goldsworthy, 2004, 103; Quesada Sanz, 2006, 21). The retreat during the battle is mentioned in passing by some experts. Hammond (1984, 46) states: ‘... it is apparent (though the sources do not say so) that the legionaries gave way and withdrew’. Goldsworthy (2004, 101) says that the Italians, when repulsed by the phalanx, ‘drew back and withdrew up slope towards their camp.’ However, Goldsworthy says nothing of whether or not the citizen legions retreated along with the socii,

as Polybius implies with the phrase ‘turning away’ (διατρεπομένων), but rather says that the attack of the

First Legion halted the phalanx’s advance. Nonetheless, these brief remarks are all that passes for an examination of the Roman army’s use of retreat tactics at Pydna. 552 Issues arise over the depiction of the alleged fragility of the phalanx and the lopsided casualty figures. For a start, almost any army that was disordered and surrounded would collapse: this was not a unique flaw of the phalanx (Sabin 2007, 412, 432). Plutarch’s account of the battle lionises the Roman army and its fighting style, and the ostentatiously low Roman casualties (Aem. 21.7) are suspicious (Sabin, 2007, 416). 553 The absence of Perseus from the account is strange, leaving open the question of who actually commanded the phalanx, and what happened to the Macedonian heavy cavalry. Perseus and his heavy cavalry had played a decisive role in the Battle of Callinicus, 171 BC (Livy 42.59.3), and apparently this impressive force withdrew intact from the battlefield at Pydna, accompanying Perseus (Plut. Aem. 23.1–2). It seems hard to believe that a man who had led his cavalry with great élan at Callinicus should have fled the battle as soon as it began, as Polybius stated (Plut. Aem. 19.3). Livy (44.42) claims, equally strangely, that Perseus fled with his companions, and that the remaining heavy cavalry simply retreated without fighting when the phalanx was beaten. Plutarch (Aem. 7–10) transmits an alternative explanation for this, attributed to Poseidonius, in which Perseus withdrew from the battle because he had been wounded fighting with the phalanx. This is also problematic, for why should Perseus have fought with the phalanx in the first place, and why take all his cavalry with him when he left the field, severely weakening the phalanx? It is also possible that Perseus simply bungled the battle, but the truth will remain unknown (Hammond, 1984, 47). 554 Cf. Hammond, 1984, 33, 38. Polybius evidently described the battle (Plut. Aem. 19.4), but his narrative is lost. Regarding Livy, his account unfortunately suffers from a lacuna of two pages, omitting the retreat (Hammond, 1984, 39). It is thus of little use. 555 Wheeler, 2006, 306. Wheeler does not explain his position here, but could have had several issues in mind, not only those mentioned above but also the non–presence of Roman elephants in Plutarch's account (cf. Livy 44.41.3–4) and the differing sequences of events leading to the Macedonian rout in Livy and Plutarch (cf. Livy 44.41.3–5 and Plut. Aem. 20.9–10). The similarities in the accounts of Livy and Plutarch are product of common source material, likely including Polybius (Hammond, 1984, 33, ref. Plut. Aem. 19.4; for Livy, cf. Erdkamp, 2006, 539–40). At best, the differences in the accounts probably reflect selective approaches to Polybius' account, whose own account, as the above discussion of casualties shows, would raise its own questions, even if it had come down to the present day.


156

reason not to trust the broad strokes of his narrative. The major movements by both armies

and, most importantly, the initial Roman retreat when forced back by the phalanx on flat

ground, are credible. The following analysis will, therefore, utilise Plutarch’s account of the

Battle of Pydna and focus on two particular aspects of the battle’s course: the retreat of the

Romans as they were initially repulsed by the phalanx (Plut. Aem. 20.4–6), and the disruption

of the Macedonian phalanx as it followed the Romans over the uneven ground (Plut. Aem.

20.7). This retreat, coupled with the adverse impact of the terrain on the phalanx as it

advanced, was a turning point in the fight. Focusing on this will allow a better understanding

of how retreat tactics shaped the course of the battle and demonstrate what the Roman army

was capable of in this period.

Plutarch (Aem. 20.5–6) thus describes the retreat of the Romans who first confronted the

phalanx:

When the first line had thus been cut to pieces, those arrayed behind them were

beaten back; and though there was no flight (φυγὴ ... οὐκ ἦν), still they retired

(ἀναχώρησις) towards the mountain called Olocrus so that even Aemilius, as

Poseidonius tells us, when he saw it, rent his garments. For this part of his army

was retreating, and the rest of the Romans were turning aside from the phalanx,

which gave them no access to it, but confronted them as it were with a dense

barricade of long spears, and was everywhere unassailable.556

It is not easy to visualise exactly what Plutarch describes. He states (Aem. 20.5) that there

was a retrograde movement (ἀναχώρησις) but emphasises that ‘there was no flight’ by the

Romans. It was only ‘part of the army’ (20.6) that was retreating at this point, namely the

forward units of Italian allies that had first engaged the phalanx (19.1–2, 4), but Plutarch also

makes it clear that the rest of the army also retreated (20.5). The Italians evidently formed the

first of a double or triple line, which would imply that ‘the rest of the Romans’ comprised the

principes of the second line and (perhaps) the third line, and that they were all retreating in

the approximate direction of the Roman camp (20.5).

556 Trans. Bernadotte Perrin, 1918. Note that I am following the LCL numeration from lacuscurtius.org.


157

There are few details with which to analyse the character of this retreat, but there are enough

to make several observations. First, it appears that only the socii, specifically the Paeligni and

Marrucini, were engaged in the initial clash with the phalanx, while the citizen legions came

up behind them later.557 Secondly, there is a suggestion from Frontinus (Strat. 2.3.20) that,

once Paulus had arrived with the Roman legions, the entire Roman army carried out a feigned

retreat in order to draw the phalanx onto broken ground, but Plutarch is clear that the Roman

retreat was genuine.558 One way or the other, the Roman general clearly maintained some

control of the troops, for he was able to reorganise and direct the cohorts for a counterattack

as soon as he saw the opportunity.559 Thus, Plutarch (Aem. 20.5) should be believed when he

says that there was no chaotic flight, but an orderly retreat. Further, as Hammond's analysis of

the terrain shows, the Macedonians had to advance around 100–200 metres before the

phalanx reached the most uneven ground, and so this is probably the approximate distance of

the Roman retreat, presumably towards the Roman camp that was only 400 metres from the

initial clash.560 The retreat probably took place at a steady pace, as the Macedonian–style

phalanx, in very close order and many ranks deep, could not in any case have advanced

quickly. The Roman army thus maintained its ratio even in this involuntary retreat.

Although in terms of discipline and training this Roman army was experienced, it does not

appear to have been an exceptionally efficient force,561 and Paulus did not appear to have

complete control of the army (Plut. Aem. 20.5–6), so one may ask why this retreat was so

successful. Plutarch does explicitly attribute the Roman counterattack to Paulus (Aem. 20.8),

and so he must have taken rapid control of the situation. Perhaps most importantly, Paulus’

instructions were transmitted quickly and efficiently to his subordinate officers, and in turn

again to the milites (20.8–9). The Roman army at Pydna was a sum of its parts, with some 557 Plut. Aem. 20.4, 6, 8. Also see Goldsworthy, 2004, 101. 558 Given Aemilius’ fear at the sight of the phalanx (Plut. Aem. 19.2) and his dramatic reaction when the socii were forced back (Aem. 20.6). 559 Plut. Aem. 20.8. Whether one adopts Plutarch’s account (Aem. 20.5–6) of an involuntary retreat, or Frontinus’ account (Strat. 2.3.20) of a deliberately feigned retreat, the effect was the same. The Roman army was able to effectively disengage from a disadvantageous fight and quickly re–engage when the tactical circumstances were more favourable. 560 Plut. Aem. 18.9; Hammond, 1984, 40, 46. 561 As the war had been ongoing for several years, and the Romans possessed a significant body of veterans (Livy 42.33–5, cf. Goldsworthy, 2004, 85), the army was reasonably experienced. Even so, Paulus had to whip the army into shape. The army’s discipline and effectiveness prior to Paulus’ arrival is criticised by both Livy (44.32.4–11) and Plutarch (Aem. 13.6). Paulus spent a short time training and readjusting his army on arrival, though it is possible his impact on the army has been exaggerated by favourable sources (Goldsworthy, 2004, 93).


158

highly motivated troops (20.2–4), efficient officers (20.9), and a reasonably decisive general

(20.8). It was a combination of these factors that allowed the army to retreat and then

counterattack successfully. However, as the retreat occurred before Paulus’ intervention, the

discipline and skill of the milites must surely have been the most important factor on this

occasion, which suggests that Roman troops were already well–trained and experienced in

carrying out simple retrograde manoeuvres in the Middle Republic.

Another important reason for the smooth conduct of the retreat was the terrain. The Roman

camp was situated on a hill, and the army drawn up on the flat ground at the base of this

hill.562 A standard explanation for the result of the battle (by both ancient and modern

commentators) was the break–up of the phalanx as it pursued the Romans over uneven

ground leading up to the Roman camp, after which the disordered phalanx quickly fell victim

to Roman sword–work.563 However, it should also be noted that the terrain assisted the

Roman retreat by slowing the advance of the phalanx and giving the retreating troops the

advantage of fighting from high ground, with the added confidence of having the camp

nearby. This was particularly helpful because, after the Romans’ initial repulse by the

Macedonians, the short distance between the Roman battleline and their camp meant that they

had little time and space in which to counterattack before being forced back into their

camp.564 However, the phalanx could not advance far before the terrain began to impact upon

it.565 The uneven ground rendered the advance of the phalanx uneven, making it less of a

threat to the retreating Romans, and providing Aemilius with his chance to counterattack.566

There were thus several ways in which the terrain assisted the Romans.

562 Hammond, 1984, 40. 563 Plut. Aem. 20.9–10; Lendon, 1999, 284; Goldsworthy, 2004, 103; Quesada Sanz, 2006, 21. Also see Polyb. 18.28–32. 564 At the first clash between the socii and the phalanx, the Macedonian army was only about 400 metres from the Roman camp (Plut. Aem. 18.9). 565 The contour lines of Hammond's map show that the ground became relatively uneven within approx. 100–200m of the phalanx's initial position, particularly in the centre–right of the phalanx's line, while to the left of the phalanx was a riverbed, roughly perpendicular to the Macedonian axis of advance, to contend with (Hammond, 1984, 40). 566 Though elements of the phalanx evidently moved too quickly to maintain their dressing (Plut. Aem. 20.7). Counterattack: Plut. Aem. 20.7–8.


159

How did ratio and generalship factor in? The Romans’ organisational advantages in

conducting retreats (the systematic use of reserve infantry lines, flexible sub–units and semi–

independent junior officers) were on full display in this battle.567 It is also true, as has often

been observed, that the superiority of the legion over the phalanx in manoeuvring on uneven

ground gave them a key advantage on the right terrain, but the battle had not begun on such

terrain, so when the Romans had been beaten on the flat ground, the conduct of an orderly

retreat allowed them to begin the battle anew on terrain that suited them. As such, while the

oft–observed (relative) inflexibility of the phalanx certainly was an important factor in the

battle’s result, it is particularly striking that the Roman formations could so much more easily

maintain their order while retreating over uneven ground.568 Aemilius also made an important

contribution by positioning the camp in a strong defensive position, to which a retreat could

readily be made. Conversely, it is also possible that the disastrous Macedonian uphill attack

was intentional, an attempt to press the Romans and storm their camp and thereby to win a

decisive victory. The quick advance of the phalanx and the relatively short distance between

the Roman line and their camp support this idea (Plut. Aem. 18.9). Perhaps the idea was not

without merit, but in practice it was a catastrophe, and Aemilius’ conventional, defensive

tactics secured his army’s retreat and set the stage for the decisive counterattack. Thus, the

successful retreat at Pydna was due to the superior ability of Roman formations to maintain

their order while moving over uneven ground, and the active and flexible, but conservative,

generalship of Aemilius Paulus.

In this period, the Romans were evidently capable of effectively using retreat tactics under

pressure from enemy infantry, but the retreat tactics at Pydna were simple: the beaten socii

and the legions behind them retired gradually up the slope of the hill to their rear, where the

general had secured a path of retreat by placing a fortified camp on the hill. The ensuing

counterattack required more flexibility, which the army inherently possessed in its ability to

subdivide into maniples, but the conditions for the counterattack were still created by the

retreat. No examples of Roman retreats in this period show any more sophistication than was

shown at Pydna: bodies of troops moved directly away from the enemy, towards the closest

defensible position, supported by reserve infantry lines and perhaps by cavalry and velites or

567 The legions come up to support the first line of socii: Plut. Aem. 20.6. Good conduct by officers: Aem. 20.1, 20.9. 568 Cf. Bell, 1965, 410.


160

auxiliary light infantry.569 The counterattack at Pydna was a product of the unique manipular

system, but its overwhelming success was due at least as much to the disorder of the enemy

as to the effectiveness of the Romans’ tactics. The Roman retreat tactics at Pydna were

effective under the circumstances, but not overly complex.

Roman tactics were relatively conventional in this period. The Roman army at this time was

still organised into the classic divisions of hastati, principes, triarii, velites, equites and socii,

as described by Polybius (6.21–6). As mentioned, new kinds of light infantry and cavalry

auxiliaries, such as Greek archers and Numidian horsemen, began to be incorporated into

Roman armies from the time of the Second Punic War onward,570 but there is no particular

reason that the gradual introduction of new types of auxiliaries carrying different missiles

should have brought any sudden changes to Roman tactics.571 The introduction of new kinds

of cavalry, however, might be a slightly different matter. For instance, Thessalian cavalry at

the Battle of Callinicus, 171 BC, did an exceptional job of supporting retreating Roman

troops (Livy 42.59.4–5), and the Numidians whom Romans began to utilise were particularly

adept in flexible advance–and–retreat tactics (Polyb. 3.116). This is not to say that the Italian

cavalry was no longer effective,572 but expert light cavalry, such as Thessalians and

Numidians, may have added a new flexibility to Roman retreat tactics, as these highly

manoeuvrable forces were better equipped for skirmishing with and thereby delaying

pursuers, and so assisting the retreat of infantry. That said, the examples of major retreats

examined above do not show any obvious evolution in retreat tactics from the Second Punic

War to the Battle of Pydna. One might note that the retreat tactics used by the Carthaginians

in the Second Punic War were in fact fairly similar to those of the Romans.573 Perhaps the

Romans had adapted their techniques from the Carthaginians. Whether they did or not, as the

success of the Roman retreat at Pydna shows, the battles Rome fought in the east following

the defeat of Hannibal presented no new challenges that necessitated further changes to the

Roman army’s retreat tactics.574

569 Cf. Polyb. 3.84.11–3, 18.25.4; Livy 25.35.7–36.3. 570 Keppie, 1984, 78–9; Sekunda, 2007, 356–7. 571 Cf. Slavik, 2018, 168. Also note Bell, 1965, 419–21; Quesada Sanz, 2006, 2–3. 572 McCall, 2002, 99. 573 Polyb. 11.21, 24; Livy 25.37.15, 28.13–5, 30.18, possibly also Hannibal's front–line at Zama (Polyb. 15.13). 574 Cf. Hoyos, 2007, 76.


161

Retreat Tactics from 149 BC to 102 BC

The period between the Battle of Pydna and the beginning of Caesar’s campaigns in Gaul

saw a great deal of warfare, and so it stands to reason that the Roman army developed its

practices in this period, but the details and timing are not clear. Fundamentally, the problem is

that there is no surviving contemporary source that describes military matters in significant

detail after the time of Polybius (c. 146 BC) until Caesar (58 BC).575 It is, therefore,

necessary to rely entirely upon non–contemporary sources for this period. There were, at any

rate, few large–scale battles in this period.576 These problems somewhat limit any analysis of

Roman tactical development in this period,577 but I will examine the few outstanding battle

accounts and look for features that separate Roman retreats in this period from those of the

preceding century.

One detailed example of a retreat from this period comes from Appian (Pun. 102–3), who

describes a fighting retreat by the Roman army at the Battle of Nepheris (149 BC) in the

Third Punic War against a formidable Carthaginian field army. According to Appian, the

Roman army led by the consul Manius Manilius advanced across a river to attack the

Carthaginians against the advice of a particular military tribune, Scipio Aemilianus, and after

a sharp fight the Carthaginians retreated to the safety of a fortified camp (Pun. 102). The

Romans became trapped between the unassailable Carthaginian position and the river they

had just crossed, and so were forced to retreat, which they commenced in an orderly fashion.

However, upon reaching the river, the army was constrained by the narrowness of the few

crossings and was unable to maintain its close order, stimulating the observant Carthaginian

general to attack while the Romans were disorganised, who then suffered heavy casualties

while attempting to flee the Carthaginian attack (102). In response to this situation Scipio

took 300 javelin–armed cavalry that he had in his immediate presence and collected as many

more as he could, and he made repeated hit–and–run attacks on the enemy to relieve the

575 While Polybius’ work covered events until 146 BC, it is highly fragmentary, and Livy’s books are totally lost after 167 BC, their contents known only through the Periochae. 576 From 146 BC until the Cimbrian Wars (113–101 BC), including the Jugurthine War (112–106 BC), most of Rome’s wars were against Gallic and Iberian tribes and mini–states that struggled to challenge large Roman armies in the field (Goldsworthy, 1996, 45–7). 577 There are two reasons for this. First, the limited sources do not allow the minutiae of Roman army’s development to be precisely tracked, and there are few detailed accounts of retreats or withdrawals that could be examined that might shed light more specifically on the development of retreat tactics in this period.


162

pressure on the infantry, allowing most of the army to cross the river successfully (103).

Meanwhile, a detachment of four cohorts had become separated from the rest of the Roman

army and still remained on the Carthaginian side of the river, but they were able to defend

themselves on high ground (103). To rescue these men, Scipio again selected a number of

cavalry, and took control of a hill that threatened the enemy, who fled to avoid being trapped

by two Roman forces on high ground (103). Similar themes can be seen in this account as in

Sallust’s Romano-Numidian battle account.578 Decisive intervention by an officer

accompanied by cavalry was able to relieve the pressure on the infantry and allow a

successful retreat, while some cohorts, even separated from the rest of the army, were able to

defend themselves against significant numbers of the enemy, essentially acting independently.

As usual, in the absence of a fortified position, high ground was the natural place for

beleaguered troops to withdraw to.

Another retreat recounted by Appian (Hisp. 88), during Scipio Aemilianus’ Iberian campaign

of 134 BC, describes a fighting retreat by a Roman cavalry detachment that had got itself into

trouble by pursuing the enemy uphill, from where it was unable to safely retreat due to the

presence of the enemy on higher ground. Scipio brought more cavalry to the rescue and split

them into two groups. These alternately withdrew and threw javelins, but they gradually

made their way back to safer terrain while still fighting off enemy pursuers. It is not clear

whether the cavalry accompanied auxiliaries or not, but the tactics are reminiscent of the

cavalry Scipio had earlier used in Africa (App. Pun. 102–3). Since the Romans occasionally

utilised javelin–armed Numidian auxiliaries that were highly adept in fire–and–retreat tactics

in the second century,579 it was probably javelin–armed auxiliaries (such as Numidians) that

Scipio led on both occasions. What becomes clear is that the increasing use of auxiliary

cavalry gave the Romans a broader array of options, not only in carrying out cavalry retreats,

but in using cavalry to support retreating or besieged infantry.

578 Sall. Jug. 98. Cf. Ch. 1, 52–3 and Ch. 5, 159–60. 579 Sabin, 2007, 422; Sekunda, 2007, 353.


163

An even more detailed account of a retreat is to be found in Sallust’s description of the

Romano-Numidian battle in 106 BC, when the army of Gaius Marius was ambushed by

significant enemy forces on the verge of nightfall, just as the Romans were going in hiberna

(Sall. Jug. 97.3). The Roman army was thrown into confusion and the battle broke up into

numerous scattered fights and skirmishes, where differing groups of milites retreated, stood

their ground or charged as necessary, until eventually they formed into a defensive circle

(orbis fecere) and fought the enemy in that formation.580 Marius himself personally

intervened with a hand–picked cavalry squadron (turma), attacking the enemy and bringing

relief to certain groups of infantry, and he was at length able to draw the army back together

and organise a retreat to a nearby hill (98.1–5).

Sallust’s account (Jug. 97.5) of the performance of the milites in this confusing situation is

fairly glowing, and the intervention of Marius and his handpicked cavalry is emphasised as

the key factor in the Roman recovery (98.1). As discussed, this battle presented a highly

chaotic situation and a formidable tactical challenge.581 The Numidians had judged that

nightfall would hinder the Romans (97.3), particularly given the Romans’ unfamiliarity with

the terrain and their flank–and–rear attacks caused much disruption (97.3–5). Perhaps the

Africans expected that an orderly Roman retreat would not be possible. They might thus have

expected a decisive victory, but the Romans performed coolly under these pressing

circumstances. Two pertinent observations reveal, first, the ability of Roman infantrymen to

reorganise themselves into defensive formations, much like a phalanx (Sall. Jug. 97.5), and,

secondly, the decisive reserve role played by Marius’ handpicked cavalry in the absence of

the orderly reserve lines of the triplex acies (98.1). The milites manoeuvred themselves into

defensive formations on their own initiative and so were evidently well–trained and

experienced.582 The cavalry, which almost single–handedly restored the battle, must also have

been well–trained. It seems that these infantry and cavalry tactics were advances from earlier

practices.

580 Sall. Jug. 97.4–5. The orbis was probably a tight circle of men rather than the sort of hollow square that Roman armies later used on the march (Paul, 1984, 240). Cf. Plut. Ant. 42.1. 581 Cf. Ch. 1, 50–2 582 Sallust (Jug. 97.5) calls a number of the men novi, but as Paul (1984, 239–40) explains, these infantry were at the end of their second campaign. Novi might refer to the cavalry who recently arrived with Sallust, or simply be a scribal interpolation (Ibid., 240).


164

There are a handful of other less informative accounts of retreats. Appian (Hisp. 81–2)

describes the withdrawal of a Roman army, commanded by Aemilius Lepidus Porcina and

Junius Brutus Callaicus, from Pallantia in northern Spain, 136 BC. At the siege of Pallantia,

due to supply shortages, the Roman army was suddenly ordered to depart in the last watch of

the night, but the tribunes and centurions did not have time to organise the army properly,

with the result that the sick and wounded (and presumably supplies) were left behind (82). As

the army departed at dawn amidst this chaos, the Pallantines harassed the Romans and

inflicted significant casualties on them until night, when the Roman army became inert from

exhaustion and, counter–intuitively, the Pallantines withdrew (82).

The causes for this chaotic withdrawal are clearly explained by Appian: the generals gave the

orders at last watch, leaving only a couple of hours for the orders to be implemented before

daylight, and so the army was not properly organised when it retired.583 Presumably, the only

reason the army was able to continue withdrawing while the Pallantines attacked is because

the latter were vastly inferior in numbers. Appian’s description of the chaos of the

withdrawal, particularly the abandonment of the wounded and the disorder caused by the

latter begging the army not the leave them, is reminiscent of Plutarch’s description of the

disastrous Roman withdrawal after the Battle of Carrhae in 53 BC.584 This may be a trope,

but the abandonment of sick and wounded was clearly conceived by some Roman writers to

be the antithesis of an orderly retreat or withdrawal. It can be assumed then that a well–

devised retreat would include preparations for the sick and wounded, transport for the army’s

baggage, and much more extensive logistical preparations than occurred at Pallantia in 136

BC, though these precise details elude us. In this case it was only the vast tactical superiority

of the Roman army that preserved it from the ultimate fate that poor leadership nearly

brought.585 The main lesson here is that a withdrawal took more than a couple of hours to

organise, and that it was the responsibility of the general to plan for this and to allow the

tribunes and centurions time to make preparations.

583 Orosius (Adv. Pag. 5.5.13) has a less detailed but slightly different account, in which due to the flight the Romans lost six thousand men, their camp and their arms, but he also blames Aemilius for the disaster. 584 Plut. Crass. 27.5. As the two descriptions are so similar, they could be tropes regarding desertion of wounded soldiers. That said, there is nothing unbelievable about them in the circumstances. 585 The Roman army obviously maintained a reasonable level of ratio, for despite the casualties suffered, the army suffered no major collapse.


165

In 102 BC a Roman army defending the river Atiso against the Cimbri, under the consul

Lutatius Catulus, panicked from fear of the aggressive enemy and, against their commander’s

orders, it withdrew in disorder (Plut. Mar. 23.2–5). Plutarch claims that, when Catulus could

not restrain the milites from abandoning their positions, he brought himself with his standard

to the head of the withdrawing troops, to give the appearance that the army was not fleeing,

but following its general in retreat (23.5). It could be observed from this incident that the

position of a general and his standard were important in determining the character of a

retrograde movement in the eyes of his men (and perhaps the enemy), and that even in this

later period of Republican history, it was important for a general or his standard to be visible

and prominent in a retreat or withdrawal. Catulus’ active leadership in this example stands in

implicit contrast to that of the consul Aemilius at Pallantia, as the latter was punished on his

return to Rome (App. Hisp. 83). From this brief example, we can determine that effective

management of a retirement by senior officers was crucial to its orderliness and thereby its

success.

Returning to the more detailed examples of Nepheris and the Numidian battle of 106, both

old and new observations can be made. Firstly, the importance of terrain to the success or

otherwise of a retrograde movement, as has repeatedly been raised throughout this thesis,

continues to be displayed in the preceding battle accounts. What appears to be new, however,

is the use of cavalry reserves, as evident in Scipio’s tactics at Nepheris (App. Pun. 103), as

well as Marius’ tactics in Numidia (Sall. Jug. 98.1).

At both Nepheris and in the battle of 106, relatively small numbers of cavalry under the

general’s direct command were used as a rapid reaction force.586 This is not the same as the

straightforward retreat tactics used at Pydna, as repeated manoeuvres and attacks by cavalry

along with the management of scattered infantry units in order to sustain a retreat, are more

complex tactics. It is true that, in the Mid–Republican period, both Romans and non–Romans

utilised cavalry to support infantry battlelines.587 And yet, examples of cavalry supporting the

retirement of infantry are few and far between, and there is no example of cavalry being used

586 Sall. Jug. 98.1; App. Pun. 10. 587 See Ch. 3, 122–5. Numidian cavalry in Carthaginian service were particularly efficient in the Second Punic War (eg. Polyb. 3.117), but they were not necessarily used to support retreating infantry (cf. Polyb. 11.21, 24; Livy 28.13–5, 30.18).


166

with such flexibility as at Nepheris or the Numidian battle.588 Even in the Second Punic War,

cavalry were still mainly deployed as skirmishers or for major flanking manoeuvres.589 On

the other hand, the use of cavalry in counterattacks to support struggling infantry is more

common in battles of the later second century, but these tactics sometimes required a hands–

on leadership style: Scipio utilised his personal cavalry detachment at Nepheris, and Sallust

emphasises that Marius utilised handpicked cavalry as his reserve force at the battle of

106.590 The personal interventions of these generals with their bodyguards were decisive to

the Roman recovery in those battles. It may be that this more flexible and situational use of

cavalry reserves, under the direct control of the general or a senior officer, and particularly to

support retiring or fleeing infantry, was a development that began in earnest throughout the

mid to late second century.

It is worthy of note that in both battles, isolated infantry units were able to organise

themselves into defensive formations, such as a closed circle, on their own initiative and to

fight effectively.591 There is nothing complex about these actions, since they were merely

groups of men conglomerating as might be expected when they were surrounded by enemies,

but it was impressive that the Romans could fight effectively in such a situation, for

commonly soldiers who were isolated and surrounded in this way were defeated.592 Such

discipline could only be a product of experience in battle, as Roman training methods could

not prepare soldiers for such ad hoc manoeuvres.593 Throughout the second century, Roman

soldiers had a tendency to serve for longer periods and therefore accumulate more military

experience, but not necessarily in major battles.594 It may be that the low intensity, long

588 Cavalry could be used simply to threaten enemy pursuers by their presence (Polyb. 18.22; Caes. BCiv. 1.46; Livy 42.59), but the cavalry support organised by Scipio and Marius appears to have relied more on speed and moving from one specific point of the battlefield to another. The cavalry were therefore committed to specific parts of the battle with greater precision, whereas early examples tended to see the cavalry remaining in a large mass just as they were originally deployed. 589 Eg. Polyb. 3.72, 113, 115, 11.21, 14.8, 15.11–2. 590 Sall. Jug. 98.1; App. Pun. 103. Handpicked cavalry were also used by Caesar (Caes. BGall. 1.42). Caesar later utilised German mercenaries, finding them the most effective cavalry against the Gauls (BGall. 7.67, 70, 80; Keppie, 1984, 100). Both of these generals sought to utilise the most effective cavalry, regardless of their origin. 591 Sall. Jug. 97.5; App. Pun. 103. 592 Men fighting in a circle could be vulnerable if they became too compressed (Polyb. 3.116; Caes. BGall. 2.25; Livy 23.27.1). Livy usually describes such a formation as a product of desperation, foreshadowing the deaths of the beleaguered men (4.39.4–5, 6.4.10, 21.56.2, 23.27.6–7, 28.22.15, 28.33.15, 42.65.6–7). 593 Cf. Ch. 3, 110–113. 594 Serrati, 2007, 495. The armies of Scipio and Marius were experienced, though their discipline required some


167

duration guerrilla warfare in Spain during this period taught Roman soldiers how to fight

better in isolated or ad hoc formations under more chaotic circumstances.595

There were several changes to the Roman army’s structure that also took place in this period.

In the last decade of the second century, the legion of 30 maniples was formally replaced by a

legion of 10 cohorts, and skirmishers were integrated into the cohorts.596 These skirmishers

were no longer javelin–armed velites, which were formally removed from the legion, but

foreign auxiliaries, commonly armed with slings and bows.597 The evolution towards this

closer cooperation between light and heavy infantry is implied in one of the last literary

mentions of maniples being used in battle, at the Battle of the Muthul, 108 BC (Sall. Jug.

49.6). Metellus Numidicus reportedly deployed several lines of reserves, disposed slingers

and archers in the gaps between his maniples, and placed the cavalry on the wings as usual,

suggesting that the light infantry would fight in the battleline and cooperate closely with the

milites (Jug. 49.6). The recent work of Slavik has suggested that perhaps the velites could

have performed a similar function, but the formal integration of skirmishers into cohorts

suggests an even closer cooperation between light and heavy infantry.598 Also, Gaius Marius

reduced the Roman army’s impedimenta and enhanced the legionary training regime, a major

step towards standardising the quality and manoeuvrability of Roman armies.599 The

increasing use of auxiliary cavalry, and the corresponding increase in advance–and–retreat

tactics used by these cavalry, has been described.600 The changes in light and heavy infantry

equipment, training and organisation, however, surely meant that the legionary sub–units

were more capable of fighting and retreating independently, as they were larger and more

closely supported by light infantry and cavalry. Their increased capability was certainly

demonstrated at the Romano-Numidian battle of 106, and even earlier at Nepheris.

sharpening (Goldsworthy, 2004, 120–1, 139). 595 Cohorts: Bell, 1965, 416–8; Sekunda, 2007, 356. On the difficulties of the extended fighting in Spain, cf. Bell (1965, 410–2), Cagniart (2007, 81) and Hoyos (2007, 76). Roman legionaries evidently could adapt their fighting style to suit this kind of warfare (Caes. BCiv, 1.44; Potter, 2010, 328). It must be noted that if the Romans did learn from these experiences, most of Marius’ men at the battle of 106 BC can only have absorbed such lessons from institutional knowledge rather than personal experience. 596 Cagniart, 2007, 86. 597 Keppie, 1984, 78–9; Rankov, 2007, 32; Sekunda, 2007, 356–7. 598 Slavik, 2018, 168. 599 Cagniart, 2007, 26–7. 600 Cf. App. Pun. 102–3; Hisp. 88.


168

That said, not all of the retreats and withdrawals mentioned above were necessarily

challenged by comparable or even disciplined bodies of troops, since the Roman army

usually displayed overwhelming superiority in open battle against its opponents in this

period.601 As such, effective tactics were not always necessary to save retreating Roman

armies. For example, the survival of the army at Pallantia (App. Hisp. 82) may be attributed

to overwhelming battlefield superiority, and probably also to superior numbers, while the

disordered withdrawal at the river Atiso in 102 BC was not effectively followed up by the

Cimbri, who were separated from the Romans by the river. The occasionally cavalier

performance of Roman forces in the same period also resulted in a number of avoidable

reverses.602 If the performance of the Roman army was uneven during the second century,

then so may have been its tactical development. Although some milites may have learned to

fight in isolated formations during this period, the intervention of commanders with cavalry

forces appeared to be the more important factor when facing defeat, in coming to the

infantry’s aid. Perhaps only particularly capable and active generals, such as Scipio and

Marius, utilised such tactics. It is significant that both of these generals had reputations for

military reform and thorough training practices.603 Conversely, poor leadership was the

reason the Romans suffered so much at Pallantia in 136, while good leadership preserved

some semblance of order at Atiso in 102. Therefore, the quality of an army’s leadership

remained a deciding factor in whether or not it utilised successful retreat tactics in this period.

Even considering the army’s gradual organisational change, most of the inherited institutional

knowledge that contributed to the successful retreats at Nepheris and in Numidia probably

occurred at the level of commanders and their officers, including centurions.604

601 Eg. Numidian indiscipline in 106 BC (Sall. Jug. 98.5–99) and the panic of the Carthaginians at Nepheris (App. Pun. 103). Cf. Gilliver, 2007, 156–7. 602 There are several examples of Roman indiscipline in this period, such as against the Germans (Plut. Mar. 23.2–5), Numidians (Sall. Jug. 44) and Carthaginians (App. Pun. 115–6). 603 Plut. Mar. 13.1–2, 25.1–2; App. Hisp. 86. Also see Bell, 1965, 415–7. 604 Note that Marius served with Scipio at Numantia, and so had the opportunity to learn his elder’s tactics (Plut. Mar. 3–4).


169

Retreat Tactics in the Late Republic

There are many more examples of complex fighting retreats and withdrawals in the Late

Republic than in earlier periods. To some extent, this is a product of the plentiful sources that

discuss this period, including Caesar, Plutarch and Appian.605 However, it is also the case that

Rome fought an increasing number of major wars and battles in the final century of the

Republic and Roman armies were defeated with some regularity, particularly in the Civil

Wars. Roman armies were thus commonly forced to carry out retreats and withdrawals. The

commentarii of Caesar are the main contemporary source for the mid–first century BC, and

provide highly detailed accounts of Roman tactics.606 The following analysis will, therefore,

focus on various retreats and withdrawals described in the Caesarian corpus.607 In total, I will

examine five battles, those at Gergovia (52 BC), Ilerda (49 BC), Dyrrhachium (48 BC),

Nicopolis (48 BC) and Ruspina (46 BC).

Caesar’s army, by the time it had finished campaigning in Gaul and was engaged in the Civil

Wars, was exceptionally skilled and well–led. Even so, in two of the following case studies

(at Ilerda and Dyrrhachium), Caesar’s men were facing other experienced Roman armies, and

so the former did not hold the sort of decisive tactical advantage that, for instance, saved the

Roman forces at Pallantia in 136 BC.608 Further, in several of the following instances,

Caesar’s men had to retreat downhill, which was one of the worst situations a body of

infantry could find itself in.609 In particular, when Pompeius saw that Caesar’s men had to

retreat per declive at Dyrrhachium in 49 BC, he was allegedly confident that they would

suffer heavy casualties.610 Therefore, even if Caesar’s army was of superior quality, it still

605 Cf. Caes. BGall. 1.25, 1.79, 2.23, 2.41–2, 3.46, 3.95, 5.37, 6.40; BCiv. 1.45–6, 1.79, 2.41–2, 3.46; BAlex 40; Plut. Crass. 27.3–29.7; Ant. 40–50; App. B Civ. 2.80, 3.69, 4.128. 606 Southern, 2006, 21–3. 607 This includes the Bellum Alexandrinum and Bellum Africum. The composer of these books was probably Aulus Hirtius, who drew upon the writings of Caesar and other participants (Gaertner 2017, 275–6). 608 Ilerda: BCiv. 1.43–4. Also see Carter, 1993, 178, 181. Caesar describes Afranius and Petreius as highly experienced and skilled (BCiv. 3.73). Pompeius’s troops were not of the same quality as Caesar’s (Carter, 1993, 177), but Pompey’s own extensive experience and the ability of his army to outpace Caesar’s in the construction of fortifications at Dyrrhachium, for which Caesar makes excuses (BCiv. 3.44–5), show that the latter’s army was of respectable quality. 609 Caesar faced this problem at Gergovia (Caes. BGall. 7.51), Ilerda (BCiv. 1.45) and Dyrrhachium (BCiv. 3.45–6. A downhill pursuit was a great threat to a body of infantry: at the Battle of Sellasia, 222 BC, when a Spartan phalanx was forced from the summit of a hill on which it had taken position, the phalanx was forced to retreat downhill and so was broken by their pursuers with heavy casualties (Pol. 2.67.3-10). 610 Caes. BCiv. 3.45. As will be seen, if Caesar had not made significant preparations before withdrawing his


170

faced great challenges, and so by examining Caesar’s campaigns it will be possible to

examine the tactics that a Late–Republican Roman army could use to retreat under the most

difficult circumstances, and to compare these tactics to those of other Roman armies in the

same period. It will then be possible to outline the differences between the most advanced

retreat tactics utilised by Republican Roman forces and those of earlier periods.

One of Caesar’s first reported fighting retreats occurred at the Gallic oppidum Gergovia in 52

BC. Caesar had ordered an attack against the Gauls encamped on the slope leading up to the

oppidum, and his troops carried this out successfully, but they then, allegedly against Caesar’s

wishes, pursued the Gauls all the way to the walls of Gergovia itself and made a botched

attempt to storm the town.611 This left the Romans exposed on the slope. The Gauls, with the

advantage of higher ground and the fortifications of Gergovia, checked the Roman advance,

and when the Romans allegedly mistook auxiliary cavalry riding around their flanks for the

enemy, the milites were forced back down the hill (deiecti sunt loco), and hundreds of

Romans were lost (Caes. BGall.7.51). It is striking that Caesar does not describe this as a

flight (fuga), but the fact that his soldiers appear to have turned away from the enemy to

escape suggests that it was a flight,612 albeit a limited one in which the soldiers were most

keen to escape the locus iniquus rather than the entire battle.

Although a downhill flight against a pursuing enemy was a worst–case scenario for an

ancient army, the Romans did not suffer catastrophic casualties because Caesar had taken

precautions and deployed two legions near the bottom of the slope, preventing the Gauls from

safely pursuing (Caes. BGall. 7.51). Thanks to this support, as soon as Caesar’s fleeing troops

reached level ground, they reformed their battleline, a display of order that itself would also

have discouraged pursuit (BCiv. 7.51). While cavalry could also have been useful in

threatening the pursuing Gauls, in this case, they had inadvertently triggered the rout due to

their barbarian appearance (7.50). Caesar was fortunate to have had large numbers of infantry

men (BCiv. 3.46), their fate would likely have been comparable to that of the Spartan phalanx at Sellasia (Polyb. 2.68.9–10). 611 Caes. BGall. 7.45–8. Caesar blames his men for not retreating when he ordered them to do so (BGall. 7.45, 47). It is also possible that Caesar's description of his battle plan intentionally concealed his own role in the defeat. Kagan (2006, 169–180) explores this event in detail, and finds it believable that Caesar’s plan was well–conceived, but that he lost control of it due to the actions of a small number of overly–aggressive soldiers. 612 Caes. BGall. 7.51: nb. infestis contra hostes signis constiterunt, suggesting that the soldiers had turned their backs. Cf. Kagan, 2006, 169.


171

in reserve, as otherwise his retreating force may very well not have turned to face the enemy

once it reached flat ground. If the Gauls had continued their pursuit unimpeded, Roman

casualties could have been much higher, such was the danger of fighting on a slope. Caesar

did well to have kept significant reserve forces he could in reserve, which had the effect of

limiting the enemy pursuit, but the Roman action had been a far cry from an orderly retreat.

One can, therefore, see from the example of Gergovia how even an experienced army and

commander could suffer heavy casualties if they were beaten and unable to retreat in an

orderly manner. Fighting on a slope only worsened the situation.

Caesar faced a similar situation two years later at Ilerda. At the onset of the civil war in 49

BC, after Pompeius Magnus had withdrawn from Italy, Caesar went west to Spain in order to

deal with an experienced Pompeian army led by the viri militares Lucius Afranius and

Marcus Petreius.613 Caesar confronted the two generals and their five legions at Ilerda, in

north–eastern Spain. A back–and–forth series of clashes over strategic ground near the town

culminated in a pursuit of the Pompeians by a large part of Caesar’s infantry up to the walls

of Ilerda itself.614 The approach to Ilerda was a long and narrow incline, leaving no space for

manoeuvre and heavily exposing Caesar’s men to missiles.615 Caesar reports that only three

cohorts standing side–by–side could fit into this space (BCiv. 1.45.4). This prevented Caesar

from sending either cavalry to assist the trapped men or greater numbers of infantry, while

any attempt to retreat by Caesar’s men was punished by the enemy who were holding the

higher ground.616 Caesar’s men were thus obliged to stand and fight in this locus iniquus.

613 These generals were formidable opponents, experienced in military affairs and loyal to Pompeius (Syme, 2002, 31, 396). Caesar himself respected their scientia rei militaris (BCiv. 3.73), and their men were the most loyal and experienced Pompeian troops (Carter, 1991, 207). Petreius was the more experienced of the two, but it is unclear who was the senior commander (Goldsworthy, Caesar, 482–3). 614 Both sides had encamped in close proximity to the oppidum itself (Caes. BCiv. 1.41–2), which dominated the area from the top of a slope that extended for several hundred feet down to level ground (BCiv. 1.45.4–5). The Pompeians chose a hill near Ilerda for their camp, and Caesar entrenched himself on the low ground, close to the Pompeian camp (1.41). However, there was another, lesser hill between the Pompeian camp and Ilerda, which Caesar decided to seize in order to cut off the enemy's line of retreat, as well as their access to the supplies they held in the town (1.43). A battle developed around this hill as the Pompeians rushed to prevent Caesar from seizing the ground: Caesar was initially repulsed, but eventually having committed four legions to the action, he drove the enemy back, with the Ninth Legion recklessly pursuing the enemy to Ilerda itself, where they found themselves fighting on the uphill slope, unable to retreat safely (1.43–5). Not all of the action is clearly described by Caesar (Carter, 1991, 191), but the battle was evidently a fluid one. 615 Caes. BCiv. 1.45.4–6. Steep rock ridges flanked both sides of the slope up to the town, trapping Caesar's men between them (Carter, 1991, 192). 616 Of the first, failed, attempt at withdrawal, Caesar says only: Hinc se recipere cum vellent, rursus illi ex loco superiore nostros premebant (BCiv. 1.45). The space was narrow, perhaps only around 440m wide


172

The Pompeians continually poured fresh milites into the battle, and, to prevent a rout Caesar

was forced to do the same, replacing exhausted troops with fresh men. Much of the fighting

was probably conducted at close range but not close quarters, as the Pompeians threw many

missiles but, after five hours had elapsed, Caesar’s troops finally ran out of missiles and drew

their swords.617 The Caesarians then made a charge, driving their enemies back to the walls

of Ilerda, where the ground flattened out, and cowing the Pompeians sufficiently that a

chance for retreat finally presented itself (Caes. BCiv. 1.46). At this time the cavalry was also

finally able to make its way forward, and by placing itself between the two armies (inter duas

acies), it allowed the milites to withdraw safely (BCiv. 1.46.3) The battle thus terminated

indecisively, but both sides claimed victory (1.47).

Several characteristics of a successful retrograde movement were displayed on the slope

beneath Ilerda. Both sides maintained ratio by fighting in close order, as they were boxed–in

by the terrain leading up to the oppidum.618 It is unclear whether or not Caesar’s soldiers,

when they eventually were able to retreat, moved at a gradual pace and thereby maintained

their formation, but it is fair to assume as much.619 The discipline of Caesar's highly–

experienced troops was unquestionably first–rate.620 The skill of the Pompeian troops is also

not to be denied.621 The Caesarians were well–led, demonstrated not least by Caesar's active

role in reinforcing wavering units and managing the movement of reserves (1.45), but Caesar

also sees fit to note the deaths of Caecilius, a courageous former centurion and a Pompeian

primipilus, hinting at the prominent role played by the opposing centurions in battle and

retreat.622 The casualties among senior centurions demonstrate that they managed the bulk of

(Goldsworthy, 1996, 138; cf. BCiv. 1.45.4). 617 Large numbers of missiles thrown by both sides: Caes. BCiv. 1.45–6. This may help to explain Caesar's comment on superior enemy numbers (BCiv. 1.46), as the slope of the ground allowed the men higher up to throw missiles into the action more effectively than those further down the slope could return fire (1.45), allowing more of the Pompeians to contribute to the action. 618 … tantum in latitudinem patebat, ut tres instructae cohortes eum locum explerent (BCiv. 1.45). An overly constricted fighting space could be a hindrance to swordsmen (Caes. BGall. 2.25), hence why no attempt was apparently made to squeeze more than three cohorts into the available space (BCiv. 1.45). 619 Caes. BCiv. 1.46.2: …facilis est nostris receptus datus. If the men were given an easy retreat, there was no need to rush, and Caesar is clear about the importance of maintaining formation (cf. BCiv. 1.44.3). 620 Note Caesar's emphasis on his men keeping to their ranks under pressure (BCiv. 1.44), and when his army began to waver as a result of the vanguard's initial repulse between the two camps, Caesar sees fit to describe this situation as ‘unusual’ (1.45). The stoicism with which Caesar's men fought from disadvantageous ground for hours (1.45–6) and the smooth operation of reinforcing the forward troops (1.45), are also notable. 621 Caes. BCiv. 1.43–4; Cagniart, 1995, 34. 622 Caes. BCiv. 1.46.4–5. According to Carter (1991, 192), the Caesarian ex–centurion (Q. Fulginius ex primo


173

the hand–to–hand fighting, and Caesar maintained the traditional commander’s role of

organising reserves and perhaps exhorting his troops. These tactics and techniques were not

necessarily unique to this period, but the efficiency with which they were performed and the

retreat was managed, particularly the transfer of reserves, is striking.

Finally, terrain was crucial to the manner in which the battle unfolded. The spectre of the

Battle of Gergovia hung over the engagement at Ilerda.623 At Gergovia, Caesar had had the

opportunity to plan ahead, and so prevented a major disaster by his judicious placement of

reserve units to deter Gallic pursuit (Caes. BGall. 7.49–51). This was impossible at Ilerda, so

the fight was determinedly maintained by Caesar’s men on poor ground for hours with

reinforcements constantly being sent uphill and exhausted men being drawn back (BCiv.

1.45–6). The disadvantageous terrain meant that Caesar was forced to utilise reserves, in this

case cavalry, to support the retreat, but it was the final courageous charge by his infantry that

made it possible for the reserves to come forward in the enclosed space.

The comparison between Gergovia and Ilerda shows that when a body of troops found itself

fighting on disadvantageous ground, it was better for them to maintain the fight in situ until a

safe retreat could be organised, than to flee downhill, even toward a better position. It would

appear that Caesar and his men had learned this lesson from Gergovia, and had thus

sharpened their ability to retreat in difficult circumstances.624 One might note that Caesar

once again accuses his men of being overeager in advancing into such an unfavourable

position in the first place, just as he had at Gergovia.625 Perhaps he was simply absolving

himself of blame, or perhaps the men really were overly aggressive.626 Even if the latter, their

coolness once in a potentially disastrous situation was certainly much greater than it had been

at Gergovia. This was probably a reflection of their own experience in retreats, Caesar’s

hastato legionis xiiii) was a re–enlisted veteran, who was obviously well–known, rather than a currently serving centurio. Potter (2010, 321–5) discusses the increasing importance of the centurionate in tactics and advising commanders in this period. The Pompeians also lost four other centurions among over two hundred casualties (Caes. BCiv. 1.46), a rate of attrition that demonstrates their front–line leadership roles. 623 Lendon, 2005, 222–4. 624 Many of Caesar's legions, such as the Ninth that was first to push onto unfavourable ground at Ilerda, had been campaigning continuously with Caesar for years (Keppie, 1984, 132, 135, 142). These legionaries were thus in a position to recall and apply the lessons of past failures. 625 Cf. Caes. BGall. 7.45, 47 and BCiv. 1.45.2. 626 Cagniart (1995, 34) suggests that Caesar’s original plan to seize a hill near Ilerda (Caes. BCiv. 1. 43) was well conceived but poorly executed by the soldiers themselves, who were overly aggressive.


174

careful management of reserves, and the usefulness of the Caesarian cavalry (Caes. BCiv.

1.46), in contrast to the mounted auxiliaries that had frightened the milites at Gergovia

(BGall. 7.50).

Caesar’s men were not the only ones who utilised retrograde manoeuvres during the Ilerda

campaign. Following the fighting around Ilerda itself, Afranius and Petreius decided to

withdraw (locis excedere) to another part of Spain (Caes. BCiv. 1.61.2). In the process they

found themselves in a race against Caesar to reach strategic ground with which they could

secure their escape (BCiv. 1.66, 1.69). While they marched, the Pompeian soldiers were

harassed by Caesar’s cavalry, and, being so delayed, they eventually found themselves cut off

from their intended avenue of escape, whereupon they decided to return to Ilerda.627 During

this second movement, Caesar’s cavalry again attacked the Pompeian army as it moved

(1.78.4), creating such difficulties that the latter were eventually forced to encamp on nearby

high ground (1.78–80). Caesar described the fighting withdrawal of the Pompeian rear–guard

in detail (1.79.5–80.1). He states that the Pompeians utilised lightly equipped legionaries

(expeditae cohortes) to form the rear–guard and that these routinely halted to protect the army

as it passed on flat ground (1.79.1), or if their parent army had to climb uphill, the rear–

guards would ascend high ground first and fight from there to protect their comrades as the

latter ascended (1.79.2). However, if the Pompeian army was forced to descend, and its

pursuers thereby were able to attack the rear–guard from higher ground, then the retiring

troops had great difficulty (1.79.3). Caesar (1.79.5–80.1) described the Pompeian tactics

under these difficult circumstances:

The only course left for them was, whenever they approached such places, to

order a halt of the legions and to repel the cavalry by a vigorous charge, and

when they had dislodged it, starting forward immediately at a run, to descend in

a body into the valleys, and so, after crossing them, again to halt on the higher

ground. For they were so far from being aided by their cavalry, of whom they

had a considerable number, that they actually received them for protection,

demoralized as they were by the previous battles, into the centre of their

column, and none of them could stray from the route without being caught by

Caesar's horse. Fighting in this way, men advance slowly and tentatively,

627 Caes. BCiv. 1.78. The cavalry harassment allowed Caesar's infantry to catch up (Carter, 1991, 202).


175

frequently halting to support their comrades, and so it happened on this

occasion.628

There are several observations to be drawn. The first is that the Pompeians had an orderly

system for protecting the army as it withdrew, with specifically appointed and equipped

legionaries standing guard as the army passed, adapting their tactics according to the terrain.

The relationship between tactics and terrain was a close one. Unsurprisingly, whoever held

the high ground had the advantage. The Pompeians had a regularised response even where the

terrain did not allow them to continue withdrawing with their usual methods: they halted the

army, dispersed their pursuers, then moved forward and crossed the difficult ground as

quickly as possible. According to Fuller, the army probably marched in a square formation,

protecting it on all sides.629 Caesar implies that he has some experience seeing Roman troops

fight in this manner.630 It is also evident that, as the Pompeian cavalry was of no use, the

infantry had greater trouble fending off Caesar’s cavalry. In this scenario, a number of

centurions were again killed (Caes. BCiv. 1.80.5), demonstrating again the key leadership

roles of centurions during a withdrawal. To summarise, the above account shows once again

that terrain heavily shaped the chosen tactics in the course of a retreat, that the Romans relied

upon cavalry to most effectively carry out fighting retreats or withdrawals but could replace

them with legionaries if necessary.

In the year following the campaign at Ilerda, a Caesarian army was defeated by Pharnaces of

Pontus in the Battle of Nicopolis, 48 BC. This engagement, however, saw the fighting retreat

of a single legion, after the rest of the Roman army had been defeated (BAlex. 40). The Legio

XXXVI, alone on the field after the defeat of the other Roman forces, resisted enemy attacks

despite being surrounded (circumdata), formed a circle or square (in orbem) and carried out a

fighting retreat to the nearest high ground (pugnans ... se recepit ad radices montium),

thereby terminating the battle (BAlex. 40). There is no evidence of the use of reserves here,

and so these Romans were fortunate that high ground was located close by. However, they

were also well–served by their use of a closed, defensive formation while retreating or

628 Trans. A. G. Peskett, 1914. 629 Fuller, 1965, 204. This is implied by the fact that the cavalry could be hidden in the centre of the army (Caes. BCiv. 1.79.5: medium reciperent agmen). Although a square formation could be though too unwieldy for an advance of any speed, Antonius successfully moved in such a formation in the open spaces of Parthia (Plut. Ant. 42.1). 630 Caes. BCiv.1.80.1: Tali dum pugnatur modo ... ut tum accidit.


176

withdrawing, just as was utilised at Ilerda by Afranius and Petreius (Caes. BCiv. 1.79.5) and

would be by Marcus Antonius in Parthia (Plut. Ant. 42.1). This is as such a standardised

Roman tactic in this period.

Returning to Caesar’s own campaigns, again in the year after Ilerda, Caesar fought an even

more complex battle near Dyrrhachium, this time against Pompeius himself. After having

withdrawn from Italy, Pompeius had stored his army's supplies and equipment in the town of

Dyrrhachium, which rendered it an ideal target for Caesar (Caes. BCiv. 3.41). After Caesar’s

arrival in the region, a variety of complex manoeuvres and counter–manoeuvres by the

opposing armies took place, culminating in the establishment of a convoluted series of

fortifications from which the two armies would frequently skirmish.631 One such skirmish

saw Pompeius press upon a hill recently occupied by Caesar’s men that had not yet been

properly fortified (BCiv. 3.45). This hill was defended by the Legio IX, which was one of the

legions that had fought on the slope of Ilerda the previous year.632 Pompeius sent forward a

detachment sufficient to harass this legion so thoroughly with missiles that Caesar, observing

the mounting casualties, ordered his legion to withdraw.633 However, as at Ilerda, the line of

retreat was per declive, and when Caesar's men attempted to withdraw, the Pompeians

attacked and prevented them from moving (3.45). At this juncture, Pompeius allegedly made

the remark that he would be a worthless general if he allowed Caesar’s men to retreat without

heavy loss.634 Evidently, Pompeius saw the Ninth's situation as all but irretrievable.

631 On his arrival, Caesar intended to cut Pompeius' army off from its supplies, just as he had attempted at Ilerda (Caes. BCiv. 1.43), and he managed to outmanoeuvre his opponents and encamp between the Pompeians and Dyrrhachium, which forced Pompeius to encamp on the coast of the Adriatic and keep himself supplied by sea (BCiv. 3.42). Caesar then began taking control of high ground around Pompeius's camp and occupying these hills with forts and garrisons, then constructing fortifications between these points, in order to entrap Pompeius (3.43). In response, Pompeius likewise fortified high points and constructed counter–fortifications to extend and secure his own perimeter (3.43–5). The result was a complex series of fortifications that loosely took the form of two (uneven) concentric lines, drawn along the hills possessed by one side or the other (see Keppie, 1984, 107). Pompeius decided to avoid a general engagement, and so he adopted a policy of attacking Caesar's men with light troops to harass and wound the Caesarians attempting to fortify certain hills (3.44). 632 Cf. Caes. BCiv. 3.45. The legion was commanded by the future triumvir Marcus Antonius. The courage and aggression shown by this legion, combined with the difficult situations they could find themselves in, bears all the hallmarks of Antonius’ skill and valour, as well as his impetuosity (Syme, 2002, 104). 633 Caesar, cum suos ex omnibus partibus vulnerari videret, recipere se iussit et loco excedere (Caes. BCiv. 3.45.4–5). Note that Caesar orders two different manoeuvres, recipere se and loco excedere. The implication is probably that they should give ground to the enemy (se recipere), and also remove themselves from the location entirely (loco excedere). 634 Caes. BCiv. 3.45.6: non recusare se, quin nullius usus imperator existimaretur, si sine maximo detrimento legiones Caesaris sese recepissent inde, quo temere essent progressae.


177

Due to the highly diffused dispositions of his forces at Dyrrhachium, Caesar was unable to

deploy large numbers of reserves, as he had done at Gergovia or (with some difficulty) at

Ilerda.635 He now resorted to placing obstacles on the downhill slope that formed the path of

the Ninth Legion’s retreat, with a rapidly formed trench placed between these obstacles, so

that his troops would have defensive fortifications to assist them in their retreat and that the

pursuers would be disrupted (Caes. BCiv. 3.46). He also deployed slingers to support his

retreating men, and, having made all these preparations, he ordered the withdrawal (BCiv.

3.46). As expected, the Pompeians pursued Caesar’s retreating men closely and crossed the

makeshift fortifications Caesar had thrown in their path, so that Caesar, fearing that the

appearance of flight would invite a more serious attack (veritus, ne non reducti, sed reiecti

viderentur, maiusque detrimentum caperetur), ordered Marcus Antonius and the Ninth to

cease the retreat and counterattack their pursuers (3.46). Throwing their pila and charging

uphill, they put the surprised Pompeians to flight, after which the Caesarians quietissime se

receperunt (3.46).

This disengagement, while it was of a smaller scale than those at Gergovia and Ilerda,

occurred under particularly difficult circumstances. Pompeius's alleged boast makes it clear

that the expected result of a downhill retreat against aggressive pursuers was widespread

massacre of those retreating.636 Caesar and his army had experienced the challenges of a

downhill retreat in the past, in the debacle at Gergovia and more recently at Ilerda. The

difference between this incident and those previous retreats is that, in the former case, Caesar

could not prevent the enemy pursuit in any way. At Gergovia, Caesar had positioned the

Tenth Legion and cohorts from the Thirteenth in support of his retreating men (Caes. BGall.

7.51). At Ilerda, Caesar was able to maintain the battle in place for hours until circumstances

were more favourable for a withdrawal. Neither option was possible at Dyrrhachium, where

Caesar's forces were more thinly spread and the Ninth's accumulating casualties meant that he

could not delay.637 At both Gergovia and Ilerda, Caesar had not only extensive infantry

635 Pompeius's forces controlled a line of fifteen miles, and, as this was the interior line, Caesar's surrounding forces can only have been spread over an even greater distance (Caes. BCiv. 3.44). Gergovia: BGall. 7.49–51. Ilerda: BCiv. 1.46. 636 This insinuation seems clear, but it is also worth noting that Caesar included his offhand insult towards Pompeius’s generalship as an apology for what was actually a significant reverse for his own forces (Fuller, 1965, 221). Even so, Caesar (BCiv. 1.46.4) had apparently expected a worse outcome from this incident. 637 Caes. BCiv. 3.44–5. While Caesar only lost 5 men in the final clash (BCiv. 3.46.5), he emphasises that many had already been wounded (3.44.6: multi… ex nostris vulnerabantur).


178

reserves at hand, but well–guarded camps.638 At Dyrrhachium, all Caesar had to support his

retreating men were some slingers and makeshift obstacles. The one saving grace was that the

men who pursued the legionaries were not themselves heavily armed, but light infantry,

hence their vulnerability to a stiff counterattack.639 Even so, this varied utilisation of light

infantry by both sides at Dyrrhachium has few precedents. At Dyrrhachium, Pompey used

light infantry to assault a position held by legionaries, and Caesar used his own light infantry

to support their retreat. Thus, the use of light infantry at Dyrrhachium shows that such

auxiliaries provided important defensive and offensive capabilities to a Roman general, and

could be assigned to specific tasks like a regular military unit.640 The velites and auxiliaries in

the Middle Republic do not appear to have been used with such flexibility. The case of

Dyrrhachium thereby shows the increasing important of light infantry in battle and

particularly in retreat, demonstrates the use of makeshift obstacles to support a retreat, and

emphasises the importance of counterattack in a difficult retreat or withdrawal.

In 46 BC, at Ruspina in Africa, the entirety of Caesar’s army again found itself forced to

retreat from an unfavourable action. While Caesar was fighting against the veteran Petreius

and his own erstwhile legate Titus Labienus, his army was surrounded and found itself

shaken by the novel cavalry tactics of the Pompeians (BAfr. 14–5). In response, Caesar

reorganised his line in an unconventional manner, in a battleline with alternating cohorts

facing forwards and backwards, from which formation his infantry counterattacked and

withdrew as necessary, while the cavalry cut off enemy forces from assisting each other (17).

After driving the enemy back in this manner, Caesar began to retreat to his camp in battle

formation.641 However, the enemy pursued heavily with large numbers of cavalry, forcing

Caesar to stop and fight in open ground (18). Harassed by the enemy’s skirmishing tactics

and unable to complete his retreat, Caesar launched a major counterattack and drove off the

enemy to a distance, then he took possession of some nearby high ground and gradually

completed his retirement while still maintaining battle formation.642

638 The camp at Gergovia: Caes. BGall. 7.35–6, 41. At Ilerda: BCiv. 1.41–2. 639 Carter, 1993, 180, cf. Caes. BCiv. 3.46. 640 Note also the light–armed legionaries that Afranius and Petreius used to support their withdrawal at Ilerda (Caes. BCiv. 1.79.1), and Antonius’ reliance on his missile–armed auxilia during his withdrawal from Parthia in 36 BC (Plut. Ant. 41.5–42.1). 641 BAfr. 17.2: ad sua praesidia sese, sicut erat instructus, recipere coepit. 642 BAfr. 18: ita uti erant instructi leniter se ad suas recipiunt munitiones. Appian (BCiv. 2.95) provides a different account of the battle’s end, stating that Petreius simply let Caesar go in order not to deprive his superior Scipio of the glory. Fuller (1965, 270) provides a rationalisation of this seemingly unlikely version of


179

Caesar’s tactical creativity and the coolness of his men under pressure are again in evidence

here. In the first attempt at retreat, Caesar created the opportunity through the novel

dispositions of his battleline and the use of cavalry manoeuvres and infantry attacks. It is then

shown once more that cavalry superiority could be a decisive factor in a fighting retreat over

flat ground, for the relative weakness of Caesar’s cavalry caused this retreat to stall, and so

the infantry assisted the cavalry by also counterattacking.643 This counterattack created the

space Caesar needed to complete his retreat; during which he briefly took possession of

nearby high ground, but it was the enemy’s defeat rather than the terrain that proved most

beneficial.644 Without the disturbance of enemy pursuit, the retreat was performed in an

orderly manner, at a gradual pace and in battle order.645 As usual, terrain remained a factor,

and the effectiveness of Caesar’s cavalry was key. But more than these things, it is evident

that just as in numerous other cases the best way to guarantee a retreat in the face of heavy

pursuit was to counterattack, and Roman infantry and cavalry could support each other in

doing so with great effectiveness.646

Caesar’s retreat tactics at Ruspina were possibly unique, as there is no other extant example

of an army being arranged so that alternating cohorts in a battleline faced to the front and

rear. Caesar’s contemporaries usually used a circular or square formation, as did Afranius at

Ilerda (Caes. BCiv. 1.79.5), the Legio XXXVI at Nicopolis (BAlex. 40), and Marcus Antonius

nearly two decades later in Parthia (Plut. Ant. 42.1).647 Caesar actually found the compression

of his army in orbem at Ruspina to be a hindrance (BAfr. 15.3–5), and so he extended his line

to fight more easily.648 He had done a similar thing at the Battle of the Sabis in 57 BC (Caes.

BGall. 2.21). Evidently, it was important that an army and even individual soldiers

maintained a certain amount of space to move and fight properly. Even so, the use of square

events, but the Caesarian version appears more plausible. How could Petreius, who had been defeated by Caesar before and knew his quality, let the latter escape unharmed if the opportunity for a more decisive victory was present? Caesar certainly would have taken advantage of such an error. 643 BAfr. 18: ad insequendum hostem perseverandumque cursum tardiora ... Note that the Pompeians struggled to maintain their retreat at Ilerda for similar reasons (Caes. BCiv. 1.79.5). 644 BAfr. 18: adversarii male accepti tum demum se ad sua praesidia contulerunt. 645 BAfr. 18: ita uti erant instructi, leniter se ad suas recipiunt munitiones. 646 Dyrrhachium: Caes. BCiv. 3.46. Cf. Antonius’ tactics against Parthian pursuers in 36 BC (Plut. Ant. 41.4–5, 42.2–4, 41.5, 42.1. 647 Caesar: BAfr. 15 (in orbem), 18 (... circumdatis). Afranius: Caes. BCiv. 1.79.5. Legio XXXVI: BAlex. 40. Marcus Antonius: Plut. Ant. 42.1. 648 Perhaps the fact that Caesar had a relatively small force at Ruspina, and many raw recruits (BAfr. 16.4), rendered his usual tactics less effective.


180

and circular formations by entire armies or legions, as well as the more creative arrangements

such as Caesar used at Ruspina, are Late Republican phenomena: no descriptions can be

found of such tactics attributed to earlier periods, at least with such large forces.649 Just as the

phalanx had difficulty manoeuvring on uneven ground, so Roman square formations could

have the same difficulty (Dio 56.13.5), but on open ground such a formation could give a

retreating or withdrawing army excellent protection (Plut. Ant. 42.1). By the mid–first

century BC, the Romans had absorbed this technique into their general military repertoire and

thus improved their ability to retire even while under attack. An army as experienced as

Caesar’s, with a commander such as he, was evidently capable of even more complex

defensive manoeuvres.

As usual, the terrain and the use of reserves were key factors in retreats. At Gergovia, Ilerda,

and Dyrrhachium, the slope in fact determined the course of the battle. At Ilerda it was an

effective system of reserves, command and control that allowed Caesar to replace exhausted

troops with fresh troops and helped his men to stand their ground, avoiding a repeat of

Gergovia. The cavalry was far better utilised at Ilerda. As soon as it could, it came forward to

cover the retreat (Caes. BGall. 1.46). Conversely, at Gergovia the sudden appearance of

Caesar’s auxiliary cavalry had frightened the legionaries and actually contributed to a partial

rout (BGall.7.50). Weak cavalry forces were a major disadvantage. This caused significant

problems for Afranius and Petreius at Ilerda, and for Caesar at Ruspina.650 However, good

infantry could pick up the slack. At Dyrrhachium, with no immediate cavalry reserves Caesar

utilised missile troops to support his retreating men, and manipulated the terrain through the

manufacture of obstacles. At Ilerda, the Pompeians utilised lightly equipped legionaries to

cover the heavy infantry, and at Ruspina Caesar used his regular cohorts in an unusual

formation to support the cavalry. By carrying out sharp counterattacks where enemy pursuit

could not be stifled by the presence of reserves, the Romans could even retire downhill,

something that would normally be disastrous for a hard–pressed body of infantry.651 Not only

could Caesar’s men do this, but so did Pompeian troops at Ilerda (Caes. BCiv. 1.79.5–80.1).

Caesar’s use of terrain and reserves to support his retreats was therefore often successful and

649 Livy repeatedly describes men fighting in orbem (4.39.4–5, 6.4.10, 21.56.2, 23.27.6–7, 28.22.15, 28.33.15, 42.65.6–7), but Livy usually describes such a formation as a reaction to desperate circumstances rather than an effective tactic. In the Late Republic, however, forming a circle or square is often a deliberate, calculated and effective response to enemy attacks, whether initiated by veteran soldiers or by officers. 650 Caes. BCiv. 1.79.5; BAfr. 18. 651 Polyb. 2.68; Caes. BGall. 7.51.


181

well managed, but his practices were not out of keeping with the tactics of other

contemporary commanders.

There were several major changes in the Roman army’s structure and tactics that occurred

during the first century BC. The trend of utilising significant numbers of foreign light

infantry and cavalry continued apace, to the point that Marcus Antonius marched into Parthia

in 36 BC with nearly as many auxiliary light infantry and cavalry (including Iberians and

Gauls) as he had legionaries (Plut. Ant. 37.3). By this time Roman and Italian cavalry had

been totally phased out.652 The result of this total shift to use of foreign cavalry was that the

Romans could field large numbers of equestrians suited to a variety of tactics, such as the hit–

and–run style of Numidians and eastern light cavalry. More importantly, the close

cooperation of infantry and cavalry shows that the Roman ‘heavy’ cavalry (probably Iberians

and Gauls in most cases) lacked none of the fighting ability of Italian cavalry, but in line with

the army’s increasing professionalism, were even better trained and practised at cooperating

with infantry and supporting retrograde movements. The permanent establishment of

professional auxiliary units was a later development, but Caesar’s own cavalry lacked

nothing for its impact on the success of his retreats.653

Further, the experience and professionalism of both milites and imperatores continued to

evolve. Viri militares, such as Afranius and Petreius, Pompey, Caesar and Antonius had

military experience perhaps even greater than the most outstanding generals of previous

generations, such as Scipio Africanus, Gaius Marius, Sulla and Lucullus. The soldiers

themselves had, by the time of Caesar, evolved into professionals who fought for a living,

and for their general, rather than for the state.654 The incredible disciplina of this breed of

milites is evident in their conduct of complex fighting retreats at Ilerda, Ruspina, and other

instances detailed above.655 The tactical flexibility of these legionaries is also evident in their

ability to adopt tactics from their enemies (Caes. BCiv. 1.44). Indeed, legionaries could

evidently perform the functions of light infantry in this period, whether they were using

652 Cagniart, 2007, 87. On the reasons for this, see McCall, 2002, 110–23, 141. 653 Cagniart, 2007, 88. 654 Ibid., 93. 655 Cf. Caes. BCiv. 1.43–7; BAfr. 14–8. In the Battle of the Sabis, 57 BC, it is striking how when a Roman army was caught by surprise the milites quickly organised themselves into units without the involvement of their commander (Caes. BGall. 2.21). By this time, Roman soldiers were evidently highly skilled at organising themselves under pressure.


182

Iberian–style skirmishing tactics or fighting alongside cavalry or outside regular formations

as antesignani.656 Roman legionaries had always shared similarities with Iberian fighting

styles, so the continued incorporation (or re–incorporation) of these fighting styles into

Roman infantry tactics demonstrates their continued flexibility in response to

circumstances.657 Thus, the Roman army’s retreat tactics were not only developing in one

direction, but they were in some state of flux depending on circumstances, with the main

underpinning factor being that their tactics became more complex over time.

As such, while terrain or the nearby presence of a fortified position have been shown as

decisive factors in the success or otherwise of a fighting retreat or withdrawal, by the mid

first century BC, Roman retreat tactics had become so effective that retrograde manoeuvres

could potentially be carried out with success not only in the absence of high ground, but even

in the presence of a decline, and in the face of enemy pursuit. The availability of reserves or a

rear–guard, whether these were legionaries, cavalry, or light infantry, and the general’s

utilisation of these troops through redeployment and counterattack, were the decisive factors

in the success or otherwise of a retrograde movement. The flexible and close cooperation of

light infantry, heavy infantry and cavalry in retreat and counterattack, and the flexible

adaptation of Roman tactics in general according to terrain and enemy, are probably the most

outstanding features of retreats and withdrawals in the Late Republic.

Conclusion

Roman retreat tactics developed over time. In the Middle Republic, such tactics were fairly

straightforward: the reserve lines of the triplex acies would support the front line if it was

repulsed, and cavalry posted on the flanks could threaten pursuers and allow a Roman army

to retreat in an orderly manner, usually towards high ground and/or a fortified position. For a

Roman army to retreat successfully in this period, opposing heavy infantry had to be

dissuaded from closely following the retreating troops, and ideally either high ground or a

fortified position needed to be located nearby. If these conditions were not met and the

Romans could not reduce or remove enemy pressure, it was difficult for a successful retreat

or withdrawal to be undertaken. 656 On Caesar’s antesignani¸see Carter, 1991, 191. On Roman adaptations to the hit-and-run tactics encountered in Spain, see Bell (1965, 416–9). 657 Cf. Livy 7.10.5; Quesada–Sanz, 2006, 20–1.


183

Throughout the mid to late second century BC, Rome had few enemies powerful enough to

challenge her armies on the battlefield, although at times the Romans still had to carry out

retreats or withdrawals due to logistical or other tactical challenges. The most serious threats

that faced the Roman army in this period came from the engagements and sieges of the

Numantine war, the Third Punic War, the Jugurthan War and especially the Cimbrian War.

The weakness of sources in this period makes it difficult to conclude with certainty, but the

adaptive and charismatic leadership of Scipio Aemilianus at Nepheris and Gaius Marius in

Numidia, featuring the use of mobile cavalry reserves to support beleaguered and isolated

infantry units, hints at an emerging yet heterogeneous sophistication in retreat tactics

throughout this period. The presence of nearby high ground, or a camp, or some other

obstacle to enemy pursuit (such as the river at Nepheris), was still essential, but it seems that

from this period onward, particularly capable commanders could utilise an active leadership

style and highly mobile reserves to complete a successful fighting retreat even in the absence

of a fighting line, signifying some development from the more restrictive tactics of the triplex

acies. The reasons for this development can most likely be found in the dynamic conditions

of the skirmishes and small–scale battles that Roman soldiers mostly engaged in during this

period, the more specialised auxiliary cavalry that the Romans were fielding in increasing

numbers, as well as the emergence of a caste of semi–professional generals.

Regardless of the patchy and unclear evolution of retreat tactics in the second century, by the

mid–first century, these tactics emerge as clearly more sophisticated. The Romans’ use of

reserves was more complex, including not only the sort of cavalry support seen in the

previous century but also an increased use of light infantry, including archers and lightly

equipped legionaries. There is a clear difference in the Late Republic between the close

integration of missile–armed auxiliaries and legionary cohorts, and the separate divisions of

velites that marked the Polybian era. Legionaries themselves also showed increased

flexibility, adopting light infantry tactics where necessary. Cavalry certainly functioned in

closer support of retreating infantry than they had in the previous century and a half. The

increasingly long service periods and more regular battle experience of Roman soldiers in the

mid–first century meant that they became more experienced and skilled, showing their

discipline during the stress of a retreat. Experienced troops in this period could maintain

order and withdraw gradually or, if necessary, hold their ground and launch counterattacks


184

even under intense pressure or on unfavourable terrain.658 They also demonstrated the ability

to form and maintain circular or square defensive formations during combat, either as

haphazard amalgamations of units or as an entire army.659 By using these various and

complex tactics, the Romans could retire on very unfavourable ground or over very long

distances.

Structural changes to the Roman army were a significant factor in this increased capability.

Particularly from the first century onward, the introduction of enhanced training regimes, the

formalisation of cohorts, the integration of missile–armed auxiliaries with heavy infantry

units, and increasing use of auxiliary cavalry increased the strength of Roman infantry units

while making the army more flexible as a whole. A general therefore had more options for

using light infantry and cavalry to protect the milites during a retrograde movement, while the

milites themselves, formed in greater numbers and trained more efficiently, were more

resilient to enemy pressure.

Although Roman retreat tactics became more sophisticated, old tactics and considerations

remained relevant. High ground and camps continued to be utilised as intermediate or

ultimate objectives for retreating and withdrawing troops. The use of multiple battles lines in

retreat was still wise, as seen when Caesar was saved by a reserve battleline at Gergovia.

Counterattacks following retreats were also not totally new tactics, as the Romans had, for

instance, counterattacked at Pydna.660 In general the changes in both Roman army

organisation, and more specifically in retreat tactics, were gradual and not necessarily linear,

as different armies fluctuated in quality throughout the second century and certain armies re–

adopted Iberian fighting styles due to specific campaign experiences.661 But the variety and

quality of Roman retreat tactics in this period was generally superior. Terrain, order and the

prevention of pursuit no longer dictated the success of a retrograde movement. The Romans

could utilise mobile reserves or a rear–guard, and counterattacks, to secure a retreat under

658 Eg. Ilerda (Caes. BCiv. 1.45–6), Nicopolis (BAlex. 40). 659 Eg. Numidia in 106 BC (Sall. Jug. 97.5), Nicopolis (BAlex. 40) and Antonius’ retreat from Media (Plut. Ant. 42.1). 660 Plut. Aem. 20.7–8. However, the counterattack at Pydna was not intended to facilitate the retreat, but to re–initiate the battle. Counterattacks to facilitate retreats were, on the other hand, fairly common by the mid first century: examples include Ilerda (Caes. BCiv 1.79), Dyrrhachium (BCiv. 3.46), Ruspina (BAfr. 18), and Antonius’ retreat from Parthia (Plut. Ant. 41.4–5, 42.1–4). 661 On Roman challenges in Spain, see Cagniart (2007, 81) and Hoyos (2007, 76). Legionary adaptations: Caes. BCiv, 1.44. Also see Potter, 2010, 328–9.


185

almost any circumstances.

However, even the best soldiers could panic.662 As there were limits to what even highly

skilled men could achieve under unfavourable tactical circumstances, good decision–making

by generals to avoid such unfavourable circumstances was essential. The leadership of front–

line officers such as centurions was always important, but the personal intervention of a

commander who was trusted by his men could make a decisive impact.663 But more

importantly, generals chose the ground on which battles were fought, built camps in strategic

positions, and were in charge of the deployment and commitment of reserves.664 This was not

only the case in the Late Republic. At Pydna, Aemilus deployed his army on ground that

allowed a safe retreat, and in the second century Scipio and Marius, caught in difficult

situations, were able to use reserves and decisive action to manoeuvre their forces into more

favourable positions. But a general of the mid–first century BC had the army’s long term

organisational and tactical improvements detailed above to draw upon. A general of the mid–

first century could thus more reliably call upon effective reserves in the event of a retreat or

withdrawal, expect his troops to maintain ratio, and count on them to perform a counterattack

if safe retirement was otherwise difficult. On top of these factors, the generals of this period

were more experienced and therefore more skilled than most of their predecessors. The

Scipiones, Africanus and Aemilianus, and Gaius Marius, were early exemplars of a more

professional class of general. Yet they paled in experience next to Pompeius, Caesar, Afranius

and Antonius. Further, while certain generals or commanders, such as Marius and Caesar,

might appear to have marked milestones in the development of the army’s skill and

professionalism, they did not change the army so much as make the best use of the system

that already existed, and later generals drew upon the accumulated experience of their

predecessors.665 Thus, later generals such as Caesar and Antonius had more effective military

forces to draw upon in the first place. In effect Roman soldiers, officers such as centurions,

and generals all became institutionally more effective at maintaining order and managing

enemy pursuit during retreat or withdrawal.

662 Caes. BGall. 7.50; BCiv. 2.41–2. 663 Cf. Marcus Antonius’ interventions in Parthia, 36 BC (Plut. Ant. 42.1, 43.1–3, 47.4, 48.4). 664 Goldsworthy, 1996, 165–9. 665 Cagniart, 2007, 93; Potter, 2010, 328–9.


186

All of these factors explain the exceptional success of Caesar in conducting retreats under

difficult circumstances. It is a fact that Caesar’s men were of superior discipline.666 But it

was Caesar who ensured the presence of infantry reserves and a nearby camp that supported

the fleeing legionaries at Gergovia, and of cavalry reserves and a proximal camp that saved

his men at Ilerda. Caesar organised the use of obstacles and support troops that allowed his

men to withdraw down a steep hill at Dyrrhachium, and he oversaw the redeployment of the

battleline and the repeated infantry and cavalry counterattacks that saved his army at Ruspina.

A less effective general may not have made these same preparations and decisions, and, in

any of these cases, even an army of such quality as Caesar’s could thus have suffered a

serious disaster. Perhaps it is the case that a more cautious general might not have found his

army in such unenviable situations in the first place. But counterattack was the best way to

support a retreat, and Caesar’s aggressive approach and spirited troops were more than equal

to such situations. As a result, in the conduct of withdrawals, Caesar outshone even Afranius,

Petreius and Marcus Antonius, who were in their own rights highly decisive commanders

when carrying out retrograde manoeuvres.

666 Note for instance the exceptional discipline of Caesar’s men when hard pressed at the Sabis (Caes. BGall. 2.27) on the slopes of Ilerda (BCiv. 1.46) or in the advance at Pharsalus (BCiv. 3.93).


187

For so it is with human affairs; in time of victory the very cowards may brag, while defeat

discredits even the brave.

(Sall. Jug. 53.8)

Chapter Six

Disastrous Withdrawals in the Late Republic and Early

Empire


188

Introduction

By the middle of the first century BC, the increased skill of Roman commanders, the

accumulated experience of soldiers embroiled in constant warfare, and general improvements

in the use of cavalry and light infantry meant that Roman forces could carry out successful

retreats and withdrawals in the most difficult of circumstances. In this period, Roman armies

demonstrated their skill in retreats at engagements such as Ilerda, Dyrrhachium, Nicopolis

and Ruspina. However, in the same period there were other Roman armies that suffered

significant casualties or even annihilation when attempting retreats from unsuccessful battles.

The total defeat of Crassus’ proconsular army at Carrhae in 53 BC (Plut. Cras. 27–31) and

the enormous casualties suffered in Marcus Antonius’ epic withdrawal from Media in 36 BC

(Plut. Ant. 41–50) are only two such examples.667 Even into the Principate, as the Roman

army became more professional, disasters still occurred when they attempted to escape

unsuccessful engagements. Perhaps the most famous of these was the Roman catastrophe at

Kalkriese in AD 9, the clades Variana.

As the preceding chapters have mainly focused on the success and evolution of Roman retreat

tactics, the question is now raised as to why it was that the Romans, who had become so

proficient at retrograde movements, could still suffer utter defeat in such scenarios? In

previous chapters it has been seen that a retreat or withdrawal could easily fail as a result of

enemy pursuit, but that as long as an army maintained ratio, was well–led and utilised terrain

or a nearby camp as a safe haven, or otherwise utilised reserves or support troops to

intimidate or hinder enemy pursuers, then a fighting retreat could be carried out under very

difficult conditions. By the mid–first century BC, the Romans had mastered these tactics, and

yet they still did not always succeed. Under what circumstances, then, were the Roman

army’s highly developed tactics ineffective?

The focus of the following analysis will be on three major Roman defeats in which

withdrawals were central to the course of events. The first case study will be the Battle of

Carrhae in 53 BC. The second will be Antonius’ strategic retreat from Media Atropatene in 36

BC. Finally, a study will be carried out of the clades Variana in AD 9. In the conclusion, the

failures of the Roman withdrawals at Carrhae and Kalkriese will be compared and contrasted

667 Another is the destruction of Gaius Scribonius Curio’s forces at Bagradas in 49 BC (Caes. BCiv. 2.42–3).


189

to the relatively successful withdrawal from Media in 36 BC in order to demonstrate further

prerequisites for successful retreats and withdrawals that have not yet been identified.

The Battle of Carrhae, 53 BC

The Battle of Carrhae was probably the worst foreign defeat suffered by a Roman army in the

first century BC. The disaster began with a crushing battlefield defeat (Plut. Crass. 23–27.4)

and continued with the partial disintegration of the Roman army as it withdrew over the

following days (Crass. 27.5–29.4), until most of the remaining soldiers were trapped and

destroyed attempting to escape into Armenia (29.5–31.7). The main sources for the battle and

the surrounding events are Plutarch and Cassius Dio, but they give conflicting accounts.668 Of

the two, Plutarch is the more detailed and reliable source, and so the following analysis will

mostly utilise his account.669

The Roman general who presided over this defeat was Marcus Licinius Crassus, the triumvir,

by no means an inexperienced commander.670 Crassus had entered Syria in 54 BC as

proconsul in order to conduct a major campaign against the Parthians. For this purpose he

raised an army of seven legions, four thousand cavalry and a similar number of light

troops.671 Crassus led his army into the open desert to the east of his province and

encountered the Parthian army somewhere near Carrhae.672 Broadly, the Romans were caught

668 Plut. Crass. 18–31; Dio 40.16–30. Livy’s account of the battle is lost, and later accounts may have been based on his version (Sampson, 2015, 188). Brief references to the battle can be found in several sources, including Caesar (BCiv. 3.31.3), Velleius Paterculus (2.46.4) Appian (Syr. 9.1) and Orosius (Adv. Pag. 6.13) 669According to Tarn (1951, 50, cf. 51–3), an unnamed Mesopotamian Greek was probably Plutarch’s source for both Crassus and Antony. Tarn (51) considers the account of Carrhae to be better than those of most ancient battles. On the other hand, Dio’s account is much less contemporary and is very different to that of Plutarch (see Sampson, 2015, 122–3). Dio had a general tendency to recast battle narratives according to prevailing stereotypes (Pelling, 1988, 278). This suggests that Plutarch should be considered the more reliable account of the battle. Also see Sampson (2015, 122–3) on Dio’s weakness as a source for this battle. That said, the sources universally blamed Crassus for the defeat, rather than accept the limitations of Roman military power (Overtoom, 2018, 20–1). Plutarch is no exception to this (cf. Ward, 1977, 1–3), and this must be kept in mind when considering his account. 670 Plut. Crass. 6.6, 10.1–11.7; cf. Sampson, 2015, 79–80. 671 Crassus engineers his proconsular command: Plut. Crass. 14–5. Crassus' ambitions, perhaps overstated: Crass. 16.1–2, cf. Sampson, 2015, 98. Troop numbers: 20.1. Of the cavalry, many were Gauls sent by Caesar, and the light troops evidently included a significant number of archers (25.2). Also see Sampson, 2015, 114–5. The increased use of cavalry and light infantry in the Late Republic is evident in this army composition, though it would still prove insufficient for what was to come. 672 Modern Harran in south-east Turkey. Crassus allegedly neglected advice to march through Armenia and


190

flat–footed in open terrain against a Parthian army, consisting only of cavalry archers and

some cataphracts, which parried all Roman counterattacks and subjected the legions to an

arrow barrage interspersed with cavalry charges for many hours, leaving the Roman army

broken and bloodied (Plut. Crass. 23.3–27.2). Of particular interest during the battle itself is

that, when Roman missile–armed infantry attempted to attack the Parthians, they were beaten

back by the enemy archery and sought refuge inside the formation of the milites (24.4). From

this it can be noted that the light infantry and milites did cooperate, but not effectively.

Crassus' son was killed during the action (Crass. 25.11–26.4). As a result of this and the sheer

scale of the defeat, the general allegedly suffered a nervous breakdown.673

The Romans then carried out a chaotic night retreat to Carrhae.674 This was probably

necessary, but it went poorly. Thousands more soldiers were killed as a result of being

abandoned in the camp or lost (Plut. Crass. 28.1–2). The Romans were able to reorganise

themselves at Carrhae, but they were forced to retire again to avoid being besieged (Crass.

29.1–3). Crassus and his remaining forces attempted to make their way back to Syria but

were outmanoeuvred and trapped on some high ground en route, where they fended off an

initial Parthian attack (29.6–7). Crassus was apparently compelled by his desperate soldiers to

parley with the Parthians, but shortly after doing so he was killed during a scuffle (30–1).

With the murder of their commander and several senior officers the Roman army’s spirit

collapsed. Some men surrendered, and at night the rest scattered in disorderly flight, but few

escaped (31.7). The remnants of the army that had withdrawn with Crassus and his legate

Octavius were thus totally destroyed.675 The defeat was total.676

avoid the very open terrain east of Syria (Plut. Crass. 19.1–3). Crassus’ initial advance: Crass. 20. Crassus allegedly took bad advice to change his course: 21–2. Encountering the Parthian army: 23. Nearby cities, including Carrhae: 25.11. Crassus’ neglect of apparently more sage wisdom may be a trope, foreshadowing his failure: note that, for reasons of supply, Alexander the Great took a very similar route to Crassus when he campaigned in the region (see Engels, 1978, 67–9, 121–2). Taking this route was not a blunder per se, but it was unquestionably difficult. 673 If Crassus succumbed to despair as Plutarch describes (Crass. 27.4–5), he did not do so immediately (cf. 26.5–6). 674 Plut. Crass. 27.3, 28.1. 675 Crassus had four cohorts of infantry, a few cavalry, and five lictors with him (Plut. Crass. 29.6), while Octavius led a detachment of about 5000 men (Crass. 29.5) 676 In numerical terms the defeat was a catastrophe, with around 30,000 soldiers lost (Plut. Crass. 31.7). It was also a strategic and diplomatic calamity, as Rome was humiliated by the loss of several legionary aquilae, Syria was exposed to extensive raiding, and Rome’s reputation in the east was damaged (Sampson, 2015, 169–72).


191

While the Roman defeat on the battlefield near Carrhae had been the main event of this

disaster, the situation was significantly worsened in the withdrawals of the following days. If

20,000 men died in the entire debacle, nearly half of these (not to mention Crassus) may have

died during the manoeuvres following the battle.677 10,000 were also captured in the days

after the battle.678 It was, therefore, during the two withdrawals following the main battle that

the Roman army was truly destroyed, and so it is necessary to examine these manoeuvres and

what went wrong with them to fully explain the defeat.

Plutarch’s assessment of the army’s performance, beginning with the night march from the

battlefield to the army’s destruction, is far from glowing. A cavalry commander named

Ignatius abandoned the army, taking his horsemen with him (Plut. Crass. 27.6–7). During the

escape from Carrhae towards Syria, the quaestor (later liberator) Cassius made his way out

without supporting Crassus (Crass. 29.4). At the final confrontation, the soldiers themselves

coerced their general into making his fatal meeting with the enemy (30.5). However, there are

also several examples of spirited performance. When the legate Vargonteius was trapped on a

hill with four cohorts in the initial retirement, he and his men did not surrender but fought to

the death (28.2). The legionaries of Octavius who came to assist Crassus vigorously defended

him from attack (29.7). Even though these soldiers also forced Crassus to parley with the

Parthians, the army’s panic upon his death at least demonstrates that they had not planned to

sacrifice the general’s life for their own safety.679 When Crassus went to meet the enemy,

even though he was apparently expecting to be killed (30.5), he was accompanied by a

number of determined officers and soldiers, and he even had to dismiss his lictors (31.1). At

the last, when the Parthians attempted to kidnap Crassus, his comites intervened and a

number were killed, even though some of these were apparently unarmed (31.4–5). These

latter examples are promising reflections on the fides and virtus of at least part of the army.

However, the fact that Plutarch gives several examples of extreme indiscipline, particularly of

677 Twenty thousand: Plut. Crass. 31.7. Four thousand killed in the camp: Crass. 28.1. Four cohorts destroyed under Vargonteius: 28.2. Also see Sampson, 2015, 138. Up to five thousand may have died during the final collapse after Crassus' death (cf. Crass. 29.4–7). 678 Plutarch does not explain this figure, but he does mention that the Parthians took a number of lost legionaries prisoner in the morning after the initial withdrawal from the battlefield (Plut. Crass. 28.1), and a number of men surrendered when Crassus was killed (Crass. 31.7). Considering that the Parthians killed everyone they found in the abandoned Roman camp near the battlefield, it is likely that most of these prisoners were taken after the first night. 679 Plut. Crass. 30–1. The legionaries’ stalwart defence of their general when under open attack shortly beforehand also demonstrates their continued willingness to defend Crassus.


192

the flight of the cavalry and of the milites threatening their general, demonstrates that it was

difficult for Plutarch to heap unqualified praise on Crassus’ army. There was too much

evidence of licentia and ignavia.

It is worth looking at the army’s experience and training to help explain its apparent

indiscipline. Sampson claims that, based on Crassus' previous experience with disciplining

and hardening raw recruits in the revolt of Spartacus, and on the fact that, by the time of

Carrhae, Crassus' recruits had been under arms for at least a year and had accrued some battle

experience, Crassus' army was ‘up to the expected Roman standard.’680 However, a year of

such experience with no major battles paled relative to the experience and skill of veteran

armies, such as those of Caesar and Petreius.681 A year or so under arms and a small amount

of campaign experience could in no way have prepared the army for its harrowing experience

on the field at Carrhae and the chaos and terror of the ensuing withdrawal. The instances

already given of the flight of cavalry and insubordination of legionaries, not to mention the

army’s total breakdown upon Crassus’ death, are evidence that, even if the army had been of

a reasonable standard prior to the battle near Carrhae, it was on the verge of disintegration

following its defeat in the field.

However, the army itself cannot be blamed without considering the role of its leadership. The

key officers named by Plutarch (Crassus aside) are the quaestor Gaius Cassius Longinus and

the legati Octavius and Vargonteius.682 As the general, Crassus dominates the account, but

Plutarch is critical of his performance. He alleges that, following the battlefield defeat, an

exhausted Crassus (not only defeated but having lost his son) lay in his tent after the

Parthians withdrew and could not be roused to action by his officers.683 Even if Plutarch

exaggerated Crassus' despair, he gives little evidence of Crassus' leadership during the initial

retirement. Plutarch contrasts Crassus’ alleged lethargy with several examples of energetic

action by other Roman officers: with Octavius and Cassius who apparently organised the

680 See Sampson, 2015, 115. Hardening troops against Spartacus: Plut. Crass. 10. Battle experience prior to Cannae: Dio 40.12.2 681 Afranius and Petreius: Caes. BCiv. 3.73. The Pompeian troops at Ilerda were the most loyal and experienced (Carter, 1991, 207). Caesar's men were of exceptional discipline (Caes. BCiv. 3.93.1). 682 Cassius: Plut. Crass. 18.4–29.4, passim. This Cassius is the future liberator. Octavius: Crass. 27.5, 29.5–7, 30.5–31.5. Vargonteius: 28.2. The praenomina or cognomina of Octavius and Vargonteius are not given. A tribune by the name of Petronius is mentioned (30.5, 31.4–6), as well as a craven cavalry leader named Ignatius (27.6–7) and the commander of the Roman garrison at Carrhae is named as Coponius (27.7–8). 683 Plut. Crass. 27.4–5. Plutarch also criticised Crassus’ deployment for the battle in the first place (Crass. 23.3).


193

withdrawal to Carrhae themselves (27.5); the garrison commander at Carrhae, Coponius, who

left the safety of the town to locate the army and guide it back safely (Plut. Crass. 27.6–8);

and the officers Petronius, Vargonteius and Octavius who are generally singled out for their

selfless loyalty and courage.684 Crassus apparently recovered himself at Carrhae and once

again took command of the army, but he is depicted as having almost totally lost control of

events.685 If Plutarch’s account of Crassus leadership is accurate, it is the most obvious

explanation for the failure of the army’s withdrawal from Mesopotamia.

On the other hand, there are also hints of questionable conduct by some of Crassus’

subordinates. Of primary concern is the bias of Plutarch’s account towards Cassius, who is

mentioned repeatedly, invariably favourably, and often as an opposing voice to the tactics of

Crassus.686 The impression given is that Cassius was the army’s more competent co–

commander.687 It is implausible that Cassius should have been making important decisions

such as agreeing to parley with the Parthians, even in Crassus’ presence (Plut. Crass. 28.4–5),

but then took the cavalry with him to Syria and left Crassus to face the Parthians alone

(Crass. 29.4). Plutarch blames the trickery of local guides for this, but this is not

convincing.688 There was already tension between Cassius and the imperator before the

defeat at Carrhae, and this relationship can only have soured further under the

circumstances.689 Treachery is a central theme of Plutarch’s account, and yet he never accuses

Cassius of what appears to be desertion.690 At least one other officer, Ignatius, certainly

deserted with his troops (27.6). It is probably no coincidence that Ignatius and his detachment

comprised cavalry, as did Cassius’ force that fled to Syria (29.5), not to mention the fact that

Crassus’ cavalry numbers were at any rate much-reduced following the disastrous battle

(20.1). Mounted men were far more able to flee the Parthian advance than were the pedites,

who were probably also burdened with wounded men. One might also doubt that, in the final

684 Plut. Crass. 28.2, 29.5–7, 31.4–6. 685 Plut. Crass. 28.3–5, 29.2, 5–6, 30.2–31.5. 686 Cassius himself was probably one of Plutarch’s sources (Ward, 1977, 295). Cassius was seen by some writers as a better general than Crassus (cf. App. B Civ. 4.59), even though he was never in full command. 687 Plut. Crass. 20.2, 22.4, 23.3, 23.3–4, 27.5, 28.3–5, 29.1. Note that during the battle, Cassius commands one wing, while Crassus’ son Publius commands the other, suggesting that these two were of roughly equal seniority, but certainly subordinate to Crassus, who remained in the centre (Crass. 23.4). 688 Plut. Crass. 29.4. The traitorous guide or the turncoat dooming the army of the protagonists is a recurring theme in Greek literature (Hdt. 7.213.1, contra Plut. Ant. 41.1–4, 46.2 in which the guides were trustworthy). 689 Cf. Plut. Crass. 22.4, 23.3. 690 Treachery: Plut. Crass. 21.1–5, 7, 22.1–2, 27.6–7, 29.2–5, and possibly 31.3–4.


194

withdrawal from the town of Carrhae, Crassus commanded only a small detachment of troops

while his remaining legate, Octavius, had a much larger force (29.5–6), and Cassius had most

of the cavalry (29.4). It would seem unlikely too that the army was deliberately split into

several divisions in the withdrawal from Carrhae, but Plutarch again explains this

conveniently by emphasising the unreliability of the local guides (29.2–6).

In light of these issues, two possibilities must be considered. The first is that, as the account is

overwhelmingly pro–Cassius, Crassus is deliberately presented in such a way that he

shoulders all blame for the defeat while Cassius is vindicated for his involvement. The

second possibility is that Crassus was, due to injury or stress, unable to command the army

effectively, which is why his subordinates made important decisions and commanded major

divisions of the army in the withdrawal from Carrhae. It should be noted, however, that

physical wounds are mentioned by neither Plutarch nor Dio, and Dio does not even mention

the general’s mental breakdown immediately following the defeat.691 However, Plutarch’s

account of the withdrawal does not dwell on details other than to reveal more about the

character of Crassus, who is clearly depicted in this case as a victim of fortuna and the

treachery of easterners.692 In light of the pro–Cassius bias of Plutarch’s account, it is not

necessary to believe that Crassus was actually incapacitated. Even if he was, he quickly re–

assumed command in the withdrawal from the battlefield (Plut. Crass. 27.8). However, he

was probably shaken by events and so became more reliant upon the advice of his most

senior officers, Cassius and Octavius.693 It is impossible to know why Crassus and Octavius

agreed to leave Cassius in charge of most of the cavalry. Perhaps it was Cassius’ doing, or

perhaps the cavalry who were not suited to the terrain along Crassus and Octavius’ path.694

That Cassius did not do his duty is clear, and his sole influence over Crassus is probably also

exaggerated, considering that Octavius was himself entrusted with a large group of soldiers

(29.5). Crassus was, therefore, probably undermined by his most senior subordinate, with the

effect that he was deprived of cavalry that was desperately needed to secure his own

withdrawal. Therefore, Plutarch was unfair to have blamed the debacle almost solely on

691 Plut. Crass. 27.4–5; Dio 40.25.1. Orosius (Adv. Pag. 6.13) also does not mention the breakdown. 692 Plut. Crass. 29.5, 30.3; Geiger, 2014, 293. 693 We cannot really know who ordered the departure towards Carrhae while leaving behind thousands of wounded. Perhaps abandoning these men was not a conscious decision, if the Roman army abandoned baggage to move more quickly but then did not have enough wagons to carry the wounded. 694 Marshes and hills feature along Crassus’ route (Plut. Crass. 29.5–7).


195

Crassus’ generalship.695

The combination of the army’s fragile cohesion after its defeat, poor leadership, and the

separation of the infantry and the remaining cavalry, had a drastic impact on the quality of the

army’s retreat tactics. First, the fact that the army did not stay together in the initial retirement

from the battlefield to Carrhae had dire results.696 The army’s disorder and fear of attack

resulted in a halting, concertina–like movement away from the battlefield, creating

unnecessary delay and giving pursuers time to catch stragglers (Plut. Crass. 27.6). Perhaps if

the Romans had marched at a steadier pace, they might have reached Carrhae at an earlier

hour and with more order, allowing them to send out men to help recover some of the

stragglers (Crass. 28.1–2). Crassus' advance guard of cavalry, which deserted upon reaching

Carrhae, could have been of great assistance in this matter (27.6–7). In the second withdrawal

from Carrhae, the army became divided into at least three divisions, with even more dire

results: the infantry was overtaken by the mounted Parthian army and came to the fate

described above, while much of the cavalry fled to Syria with Cassius (29.5). Therefore, the

withdrawals were poorly executed, but the reasons why, and who was responsible, are not

immediately apparent.

As argued in previous chapters, it was important for a retiring army to move towards

defensible positions where necessary, to maintain ratio, and to manage enemy pursuit. It must

therefore be noted that the repeated Roman use of night marches to escape was perfectly

logical in light of the fact that the Parthians were unlikely to pursue by night, and the Romans

were unable to confront them on open ground.697 Marching by night was therefore one way

of managing enemy pursuit. The actual direction of withdrawal was also logical. Carrhae was

the nearest defensible position to the battlefield, as were the hills that Octavius and Crassus

reached in the second withdrawal. The separation of the army, particularly Cassius’ cavalry

from the bulk of the infantry was also, as mentioned, due to treachery (Plut. Crass. 29.2–6)

and not a deliberate decision by Crassus. The Romans’ use of terrain was also not without

merit. The route that Crassus took in the second withdrawal was probably the correct one, as

695 Plutarch inherited the aristocratic disdain for Crassus and his new wealth, and so held a bias against Crassus (Ward, 1977, 1–3) even aside from the fact that Cassius himself was probably one of his sources (Ward, 295). 696 There was the mass killing of 4000 wounded in the Roman camp, and a significant force under Vargonteius was cut off and destroyed (Plut. Crass. 28.1–2). 697 Night marches: Plut. Crass. 27.5–6, 29.2–3. Parthian dislike of night operations: Crass. 29.3.


196

there was high ground to the north that could protect him from the Parthian cavalry (Crass.

30.1). It is striking that Octavius actually reached a position of safety, but then returned to

protect Crassus (29.5). In 36 BC, Marcus Antonius was also able to escape Parthian pursuit

through the hills towards Armenia (Ant. 41, 49). According to Plutarch and Dio, this outcome

is exactly what the Parthian commander Surenas feared.698 When Crassus was trapped by the

Parthians, Octavius’ charge to his defence was able to sweep the Parthian cavalry from the

hill and somewhat demoralise the enemy (Crass. 29.7–30.1). The possibility for some

successful fighting thus existed as long as terrain was on the Romans’ side (30.1), and so the

decision of Crassus and Octavius to attempt a retirement to the north was tactically sound.

Indeed, much of Crassus’ decision making appears to have been reasonable, so what went

wrong?

Crassus’ position was highly precarious. The use of high ground did not guarantee an army’s

safety, as seen in the case of Vargonteius (Plut. Crass. 28.2), but when Crassus was caught by

the Parthian vanguard, with the assistance of Octavius his forces were able to resist Parthian

attack, just as Antonius’ men later did in 36 BC.699 At this stage, the Romans needed to

continue their march for another two kilometres to reach safety (Crass. 29.5–6), but for this

to be possible, a force that could act as a rear–guard and carry out counterattacks was

required.700 However, the loss of almost all of the army’s cavalry and light troops left the

Roman army devoid of forces that were suitable for this task.701 If the Roman army had kept

its forces together after leaving Carrhae, including Cassius’ 500 cavalry, it would have been

in a stronger position, and may have been able to attempt a fighting withdrawal against the

Parthian vanguard to reach safer ground. Without a sizeable force of cavalry, this was

difficult, but it must be noted that Crassus specifically argued for waiting on the hill until

night and then continuing the march into the foothills, when he was aggressively overruled by

the panicked milites (30.4). At this critical juncture, Crassus’ tactical judgement was sound,

but he was prevented from effective action by the indiscipline of his men.

698 Plut. Crass. 30.1; Dio 40.25.4–26.1. 699 Plut. Crass.29.6–7–30.1, cf. Plut. Ant. 45.1–3. 700 Cf. Plut. Ant. 42.1. 701 Many light troops and apparently most of the cavalry perished in the main battle (Plut. Crass. 20.1, 25.2–12), several hundred cavalry fled with Ignatius (Crass. 27.6–7), and 500 more fled with Cassius (29.4).


197

With all of this said, we must reckon with the central fact that the army had suffered

enormous casualties in battle. Even if Crassus were incapacitated in the initial night

withdrawal, which contributed to the chaos, the fact that the Roman army left thousands of

wounded behind in the camp had nothing to do with poor leadership or undisciplined troops

(Plut. Crass. 28.1). This was a result of the simple fact that the Roman army had suffered too

many casualties in battle, especially among the valuable light troops and cavalry that were so

important for fighting off the Parthian cavalry, and will have been carrying large numbers of

wounded.702 It is also true that the lack of cavalry could have had an impact on Crassus’ and

Octavius’ tactical options once they were trapped in the second withdrawal, but the reason

that men were deserting was because of the psychological impact of the battlefield defeat;

since they believed themselves beaten and so tried to save their own lives. Again, losses of

cavalry and light troops, vital to supporting heavy infantry in retrograde movements, were

greater during the main battle near Carrhae than the number lost to desertion during the

withdrawal. The Romans also lost enormous numbers of heavy infantry in the main battle.

Certainly by the time Crassus and Octavius were confronted by Surenas north of Carrhae, the

Roman army was something like a quarter of its former size.703

The performance of the Parthians cannot be ignored in explaining the Roman failure in 53

BC. Of course, Surena’s battle plan was an ingenious counter to Rome’s predominantly

infantry–based tactics (Plut. Crass. 24, 27). With their superiority in cavalry, it was natural

that the Parthians overtook the beaten Roman army in the open terrain of Mesopotamia, and it

was apparently Surena’s plan to trap the fleeing Romans before they reached the more

mountainous terrain to the north (Crass. 30.1). However, the fact that the Parthians did not

pursue at night gave the Romans a major opportunity to escape (29.3). Even with the delays

caused by the betrayal of Crassus’ guides (29.5–6), Crassus was very close to safety when he

was brought to a halt by the Parthians (29.5–6). If his men had listened to him and continued

the withdrawal, they might well have escaped. Thus, although the Parthians did pursue as

they ought to have done, they did not do so with exceptional efficiency. The performance of

the Parthians cannot be given as the main explanation for the failure of the Roman

withdrawals.

702 Cf. Plut. Ant. 42.1. The fate of the light troops is not mentioned, but many probably died during the battle (cf. Plut. Crass. 24.4, 25.2., 25.12). Regarding the cavalry, of the approximately 4,000 that Crassus had in the first place (Crass. 20.1), he only had a handful with him in his final confrontation with the Parthians (29.6). 703 Cf. Plut. Crass. 20.1, 29.5–6.


198

It would be unfair to place total blame on the tactics of Crassus (and on whoever was in

charge immediately following the main battle) for the failure of this withdrawal. Crassus was

a competent tactician.704 Escaping to Carrhae, which was the closest defensible position, was

the sensible decision. Withdrawing out of Mesopotamia using the high ground to the north

was also sensible. Moving at night also nullified much of the Parthian pursuit, and was the

only way the diminished Roman forces could have done so. However, the army did not

maintain ratio, but continuously became separated and was plagued by desertion and mutiny.

This confusion and indiscipline were inevitable products of the decisive defeat and enormous

casualties suffered in the main battle.705 Perhaps when Crassus’ army was finally trapped by

the Parthians, if they had lost fewer light infantry, cavalry and legionaries during the previous

marches, they could have continued a fighting withdrawal utilising the high ground along

Crassus’ chosen path. Waiting until night and then attempting a retreat into the Armenian

foothills was no doubt the only remaining possibility for a successful withdrawal. But the

corrosive effects of the army’s total defeat, the misfortunes that led to the army’s separation,

and the collapse of the army’s nerve in their final argument with Crassus, doomed them all.

Marcus Antonius’ Parthian Campaign, 36 BC

In 36 BC Marcus Antonius, the triumvir, made a major military expedition into the Parthian

Empire. It was his intention, among other aims, to avenge Crassus and his legions and re–

establish Roman dominance in the east.706 The campaign did not go as planned, and the

Roman army was forced to retire from Mesopotamia after several reverses, but despite heavy

pursuit the Romans were able to extricate themselves. Once again, Plutarch is the main

source for these events, and he provides enough detail to analyse the Roman army's tactical

performance during this campaign.707 A comparison of Antonius’ performance with the

704 Ward, 1977, 289, 295. 705 For a modern study of the corrosive effect of heavy casualties on the cohesion, tactical efficiency and morale of soldiers, particularly due to disproportionate casualties among veterans and junior officers (such as centuriones: cf. Caes. BGall. 7.51; BCiv. 3.71), see Rush (2001, 323–4, 333–5, 343). 706 Syme, 2002, 263. 707 As mentioned, Tarn argues that the same man, an unknown Mesopotamian Greek, was the main source for both Crassus and Antony. It is certainly evident that there are similarities in the two accounts, for instance the abandonment of the Armenian allies causing problems for both generals (Plut. Crass. 22.2–4; Ant. 39.1, 50.2), and the appearance of foreign guides, with very different results for each general (cf. Plut. Crass. 21.1–5, 7, 22.1–2 29.2–5; Ant. 41.1–4, 46.2). However, Pelling (1988, 221) argues that Quintus Dellius is Plutarch’s main source for Antony, and that some of the criticism of Antonius’ decision–making originates with him (cf. Plut. Ant. 37.5–38.1). On the other hand, the emphasis on Antonius’ ‘military brilliance is likely to be [Plutarch’s]


199

disastrous withdrawal from Carrhae will demonstrate what, if anything, Antonius and his

army had learned from their counterparts at Carrhae, or how the circumstances differed to

allow the Romans to withdraw successfully on the later occasion.

Antonius’ campaign began with an advance from Armenia into Media Atropatene, where he

eventually laid siege to the Parthian royal city of Phraata (Plut. Ant. 38.1–3). However, in the

meantime Antonius’ baggage train was destroyed and he was abandoned by his Armenian

allies and their six thousand cavalry.708 Antonius was able to trick the Parthians, led by their

king Phraates IV (Ant. 40.1) into fighting a pitched battle (39.3–4), and the Romans were

successful in this main engagement (39.4–5) but despite launching an extended pursuit,

Antonius was unable to inflict significant casualties on the enemy, and the Parthian army

escaped mostly unharmed (39.2–5). As the Romans retreated to their camp the Parthians

regrouped and attacked them, forcing the Romans to fight them off as they retired (39.6). The

Roman camp itself came under attack but it was successfully defended (39.7). Unable either

to conduct the siege of Phraata without sufficient supplies or heavy equipment, or to force the

fast–moving Parthians into a decisive battle, a dejected Antonius began to withdraw back

towards Armenia under a truce (40.1–4).

In spite of this, the Parthians attacked Antonius on the march (Plut. Ant. 41.3–5). Perhaps the

truce was a stratagem to force the Romans to march out into the open where they would be

vulnerable. However, in defending his marching army Antonius succeeded where Crassus had

failed, since he was able to utilise combined arms to good effect by inflicting casualties with

massed javelineers and slingers and then scattering the enemy with his Gallic cavalry (Ant.

41.4–5). Antonius continued his withdrawal in a square formation, repeating these tactics

when challenged, and after several minor reversals over a period of four days, the Parthians

were dissuaded from further attacks (42.1). However, on the fifth day of the withdrawal

Flavius Gallus, one of Antonius’ legates, was authorised to engage the Parthians in a more

decisive fight (42.2). He was outmanoeuvred and Antonius’ other legates only reinforced him

own’ (Pelling, 1988, 221). Cassius Dio (49.24–33) gives the only other detailed account of this campaign, and it is relatively consistent with Plutarch’s, but Plutarch gives far more detail on the withdrawal from Phraata (cf. Plut. Ant. 41–9 and Dio 49.28–31.2) and so he is the better source for our purposes. 708 The baggage train had consisted of two legions with the entire army’s impedimenta and was commanded by Oppius Statianus, but was totally destroyed (Plut. Ant. 38.3). The desertion of the Armenians: Ant. 39.1, cf. 37.3. Also see Syme, 1960, 264. The abandonment of the Armenians could also have contributed to the disastrous loss of the baggage train (Dio 49.25.5; Pelling, 1988, 226).


200

haphazardly, throwing the entire army into turmoil so that Antonius had to personally

intervene (42.3–4). The debacle resulted in thousands of casualties, but Antonius displayed

his leadership qualities and successfully encouraged further efforts (43). When the Parthians

attempted to overwhelm the Romans the next day, they were firmly resisted and beaten back,

and this pattern was repeated for several days (45.1–3), even as the Roman army began to

suffer from famine (45.4–6).

Plutarch mentions at this point that the Romans had abandoned some equipment, as many

mules or oxen had died and wagons were needed to carry wounded (45.4). This implies that

Antonius was also adopting the typical Roman practice of lightening the baggage as

necessary.709 Of particular note during this stage of the withdrawal were the tactics Antonius

used to fight off the repeated enemy attacks: the milites would enclose the lightly armoured

auxilia within their formations, defending them from arrows with their shields; and, when the

Parthians approached very close the legionaries would charge and, using their pila as spears,

rout the enemy (45.1–3). Intimate cooperation between light infantry and milites is evident

here.

Continuing his withdrawal, Antonius had to choose between two routes, one being on plains

where his army was vulnerable to attack, and the other along a safer, mountainous route,

which had little water.710 Antonius chose the latter, wisely using high ground to protect his

forces, but taking the calculated risk of withdrawing by night.711 During the march, a brawl

allegedly began in the rear–guard and escalated to the point where the army threatened to

disintegrate, forcing the general to encamp and restore order.712 In the morning, with Roman

discipline restored, the Parthians made a final attack (Plut. Ant. 49.1). Antonius confronted

this attack by having the milites continue to withdraw under the cover of their shields, while

the missile–armed infantry fought the enemy archers (Ant. 49.1). The Romans were thus able

to reach the river despite Parthian pursuit, at which point the latter finally gave up (49.2).

709 Cf. Ch. 4, 145–6. 710 The Parthians, just as they had at Carrhae (Plut. Crass. 28–30), had attempted to lull Antonius into complacency, but a friendly Parthian named Mithridates informed Antonius that his enemies would ambush him if possible (Plut. Ant. 46.2–4). The guide recommended that he should therefore avoid the plains (Ant. 47.1). 711 On the dangers of night withdrawals, cf. Ch. 1, 51–5. 712 Antonius chooses his route: Plut. Ant. 47.2. Disorder in the army: Ant. 48.1–3. The army encamps: 48.5. Plutarch alleges that Antonius, thinking his army had been attacked and defeated, had first considered suicide but was dissuaded (48.3–4). This may be a literary flourish foreshadowing his eventual fate after Actium (76–7).


201

After crossing into Armenia, Antonius found that he had lost around 20,000 infantry and

4,000 cavalry.713 Despite this, the campaign has been described by Syme as ‘a defeat, but not

a rout or a disaster’.714 Antonius had successfully withdrawn most of a very large army from

hostile territory, under intense pursuit, through a gruelling twenty–seven day march (Plut.

Ant. 50.2). His achievement compares favourably with the disintegration of Crassus’ army

during the failed withdrawals of 53 BC. At Carrhae, heavy casualties resulting from the

battlefield defeat had made it difficult for the army to escape successfully, but Antonius’ army

also suffered very heavy casualties. Why, then, was Antonius more successful than Crassus?

Antonius’ army fought numerous battles during the withdrawal, but was never decisively

defeated.715 The army moved slowly while under attack, and stopped entirely when

necessary.716 The Romans maintained close order and a gradual rate of movement when

moving in the presence of the enemy, and this orderliness helps to explain their general

success in fending off Parthian attacks.717 Antonius’ retreat tactics were also aggressive when

necessary. On several occasions his army stopped and fought, or even advanced towards the

enemy.718According to Plutarch (Ant. 42.1), Antonius organised the army into a hollow

square (πλαισίῳ) for much of the march, with his flanks and rear protected by looser

formations of light troops, and instructed that cavalry only pursue the enemy for short

distances.719 This was apparently highly effective (42.1). The Romans also formed a testudo

under missile fire, the milites defending the light infantry, and were in close enough

formation when charged by Parthian cavalry that they repelled the latter by wielding their

713 Plut. Ant. 50.1, cf. Syme, 2002, 264–5. Livy (Per. 130) provides a lower casualty estimate at two legions lost (with Statianus) and 8000 men lost on the retreat. Evidently, the casualties were significant. 714 Syme, 2002, 264. 715 Antonius’ army suffered only one defeat worthy of mention, but even here Plutarch gives no reason to assume the Romans became seriously disordered, as the legionaries’ casualties may be attributed to their becoming surrounded by horse archers whilst other Roman officers reacted indecisively in reinforcing them (Plut. Ant. 42.2–4). Antonius, nonetheless, successfully drove off the Parthians and extricated his army (Ant. 42.4). 716 Slow movement while under attack: Plut. Ant. 45.2, 49.1–2. Army stops to fight: Ant. 41.4–5, 42.2–4, 45.2–3. 717 Plut. Ant. 39.4, 41.4–5, 45.1–3, 49.1–2. This is marked contrast to the disorganisation of retreating Roman forces after the defeat at Carrhae, with the exception of Octavius’ counterattack in the final confrontation (Plut. Crass. 27.5–7, 29.3–6, 29.7–30.1). 718 Stopped and fought: Plut. Ant. 41.4–5, 45.3. Advanced towards enemy: Ant. 42.2–4. According to Plutarch (50.1), Antonius was victorious in eighteen separate engagements during the retreat from Atropatene. 719 During Tiberius’ campaign against the Dalmatians, AD 9, he also used a mobile square formation, but in an attack through more difficult terrain (Dio 56.13–4). This may be evidence of the further refinement of the tactics Antonius used at 36 BC.


202

pila as spears (45.2–3). The Romans also utilised a square formation and formed testudines at

Carrhae, but Parthian cataphracts were able to attack the legionaries with some success.720

Further, while Antonius was able to successfully counterattack and inflict casualties by using

light armed troops and cavalry (42.1), similar tactics at Carrhae were defeated by Parthian

archery (Crass. 24.4). Antonius’ battle tactics were therefore more effective than those of his

predecessor in 53 BC. What is striking is how varied Antonius’ tactics were. Whether the

Parthians attempted to batter the Romans with missiles, or charge them with spears, both of

which were effective modes of attack at Carrhae (Crass. 27.1–2), Antonius had an effective

tactical response.

Was there some difference between the Parthian army and tactics in 36 BC, and those of the

victors at Carrhae? The distinctive tactics of the Parthian general in 53 BC, Surena, were

marked in particular by the abnormally extensive stores of arrows carried by a large force of

camels (Plut. Crass. 25.1), and by the successful ambush of Publius Crassus when he pursued

too far from the main army (25.1–6). However, it is striking that Antonius never put his army

in a position to be bombarded by arrows for such a period of time, as he had very numerous

forces and always ordered charges or advances to drive the Parthians off when they

approached.721 He also never allowed his pursuing forces to chase the enemy too far away

from his main column (Ant. 42.1), thus avoiding the error of Publius Crassus. As such,

although Plutarch does not say if the unnamed Parthian commander of 36 BC had attempted

to replicate the effective tactics of Surena, Antonius at any rate never gave him the

opportunity to do so. That is not to say the Parthians were ineffective. There is certainly

plentiful evidence that the Parthians regularly utilised the mass archery that had been so

effective at Carrhae, as well as a variety of other tactics.722 That they were not successful,

however, must be explained by the Roman response, and the effectiveness with which

Antonius countered their tactics was certainly a leading factor.

720 Plut. Crass. 24.3, 27.1–2. The testudo does not appear until the mid–second century BC (eg. Livy 44.9.8), but by the mid–first century appears to have been in regular use (cf. Caes. BCiv. 1.97.5; BAlex. 40; Plut. Ant. 42.1). 721 Plut. Ant. 39.3–4, 41.5, 42.1, 42.4, 45.3. 722 Archery: Plut. Ant. 41.5, 42.4, 49.1. The Parthians also used mass charges (Ant. 45.3) and attempted to gain advantages through trickery and stratagem (40.3–4, 41.3–4, 46.1).


203

In comparing Roman tactics in 53 and 36 BC, the close cooperation between light and heavy

infantry in 36 BC stands out (Plut. Ant. 45.1–3). Although Crassus had also utilised light

infantry, milites and cavalry in combination, he did so in smaller groups rather than with the

entire army (Plut. Crass. 25.2). The quantity of Antonius’ auxilia is also striking. Whereas

Crassus had fielded about 4000 light troops and a similar number of cavalry, Antonius’ army

upon entering Armenia included at least double these numbers and perhaps significantly

more.723 Antonius’ heavy cavalry performed particularly well (Ant. 41.5), and it might be

noted that they were mostly Gauls and Iberians (37.3), just as Crassus’ cavalry in 53 BC

comprised mostly Gauls (Crass. 25.6–9). While the Gallic cavalry were defeated in 53 BC,

Plutarch explains that they chiefly struggled with the heat (25.9), and even so fought

extremely well. In contrast, since it was close to winter during Antonius’ campaign (Ant.

40.3), it is unsurprising that his much more numerous cavalry performed well.724 Plutarch

explains nothing about the composition of the auxiliary cavalry, and but it might be assumed

that among these were lighter cavalry types that would be suitable for skirmishing.

The Parthian archery was certainly not ineffective, as Plutarch (41.5) states that when

Antonius’ light troops sallied out they suffered many wounds, but in a pitched battle the

Roman firepower was actually superior.725 Antonius’ slingers actually shot further than the

Parthian archers and gave them significant trouble (Dio 49.26.2). As such, with a very strong

cavalry arm and an enormous number of troops the Roman army was more formidable than

had been Crassus’ in 53 BC, and this is undoubtedly a key reason as to why Antonius’ battle

tactics appear to have been generally successful.726 Antonius’ more complex and adaptive

tactics, and the significantly increased numbers of missile–armed infantry and cavalry in his

army, strongly suggest that he had thought seriously about how to counter the Parthian tactics

that were so successful at Carrhae, and from a tactical standpoint he was almost completely

723 Cf. Plut. Crass. 20.1 and Plut. Ant. 37.3. Plutarch states that Antonius had 10,000 Gallic and Iberian cavalry and 30,000 other auxiliaries, these being cavalry and light–armed troops, but Plutarch does not go into further detail. This enormous latter number probably includes the large Armenian force that abandoned Antonius prior to actually fighting the Parthian army (Plut. Ant. 37.3, cf. 39.1). 724 In the battle near Phraata, the Gallic cavalry were able to charge, fight an engagement, and pursue for 30 kilometres (Plut. Ant. 39.4–5). The heat was evidently not too much for them, although perhaps this pursuit occurred later in the day. 725 Plut. Ant. 41.5. Evidently, the loss of the Roman baggage train under Statianus (cf. Ant. 38.3) did not include all of the Roman supply of missiles. Perhaps the baggage that was lost consisted in large part of the siege train, but not all the infantry’s supplies (cf. Roth, 1999, 91). 726 Plut. Crass. 50.1. The failed Roman counterattack on the fifth day of the retreat is an exception to this (42.2–43.1).


204

successful.

Quantity was one thing, but the quality of the army must also be considered. When Antonius'

army first encountered the Parthians, the Romans demonstrated their discipline by marching

steadily and silently (ἀθορύβως καὶ σιωπῇ), in equidistant formations (εν διαστήμασιν

ἴσοις), all in the looming presence of the Parthian army (Plut. Ant. 39.4). Such orderliness

was to be expected of well–trained Late Republican legions. The army displayed the same

order and discipline while fighting and withdrawing on numerous occasions (Ant. 45.1–3).

On the other hand, after the blow to Roman morale caused by the loss of Statianus' legions

and the baggage, a depressingly indecisive victory outside Phraata, and a difficult retreat back

to the camp, some legionaries had fled from the ramparts of their own camp when attacked

by the Parthians, provoking Antonius to decimate them.727 This demonstrates the importance

of Antonius himself in maintaining the army’s discipline, as the incident of decimation (39.7)

demonstrated that he was both a stern and inspiring leader (43.1–2).

However, near the end of the extended march from Media, Plutarch records (Ant. 48.2–3) an

incident of great indiscipline during a night march, in which isolated thefts of gold and silver

escalated into attacks on Antonius’ personal baggage and brought the army, fearing it had

been attacked and defeated, into a critical state of confusion. Antonius allegedly reacted by

bringing the army into encampment so that he could re–establish discipline, and by the next

morning the army was steadier in spirit and cohesion (49.1), and performed correctly under

Parthian pressure (49.1–2). However it is difficult to believe that in hostile territory, closely

pursued by an enemy army, Roman troops should have chosen to fight each other and plunder

their own baggage, especially considering the same army’s respectable performance in battle

prior to and immediately following this alleged incident.728 It is perhaps more likely that

Plutarch has exaggerated a relatively minor crisis so that he could present Antonius as

727 Loss of baggage: Plut. Ant. 39.1. Indecisive victory: Ant. 39.5. Difficult retreat: 39.6. Flight and decimation: 39.7. The use of decimation had fallen out of use in the Middle Republic but had become more common again in the period of civil wars, as unmotivated men were forced to conduct seemingly endless campaigns (Goldberg 2015, 158–9). If Plutarch is to be believed on this point, and there is no good reason he should not be, this is evidence of a weakness in Roman morale at this stage. Also see Pelling, 1988, 227. 728 Pelling (1988, 237–8) notes that the inclusion of this crisis at the point of the narrative immediately preceding the Roman’s final salvation is a conventional storytelling technique, but does not doubt that this particular incident occurred.


205

considering suicide (Ant. 48.3), foreshadowing what would actually occur in 31 BC.729 As a

result, while there was an isolated incident of panic early in the campaign (39.7), and the

Armenian contingent was certainly unreliable, it was Plutarch’s overall assessment that

Antonius’ army in 36 BC was of excellent quality (43.2), and for the most part this is borne

out by its tactical performance.

The army’s leadership must also be considered. It has already been explained that Antonius

himself was a decisive commander who was well–prepared for the Parthian king’s distinctive

tactics.730 Plutarch gives a few details on Antonius’ key subordinates during the withdrawal,

naming Flavius Gallus (Ant. 42.2) and Publius Canidius Crassus (42.4) as senior legates,

Marcus Titius (42.3) as a quaestor, and Domitius Ahenobarbus as another senior officer

(40.5).731 The two legati are only featured in light of their alleged tactical errors, perhaps as

an attempt by Plutarch to absolve Antonius from blame for Gallus’ failed counterattack.732

The quaestor Titius appears as a foil to Gallus, briefly attempting to halt the ill–fated

counterattack, then he disappears from the narrative (42.3). As regards Gallus, even whilst

criticising his judgement, Plutarch emphasises his military experience, courage, and virtuous

death.733 For the most part, Plutarch’s narrative focuses on Antonius, as expected of a

biography, and Antonius’ charisma and tactical judgement are repeatedly praised.734 The

entire account is clearly crafted to praise Antonius for his leadership qualities in

difficultatibus, but there is no reason not to give the general some credit for the success of the

729 Plut. Ant. 76–7; Pelling, 1988, 238. 730 Antonius was also highly experienced by this stage of his career, having led an army successfully in Egypt in 58 BC (Plut. Ant. 3.1–6), distinguished himself in Caesar’s service (Ant. 7–8) and proved to be the most capable general of the civil wars that followed the Ides of March at Mutina (App. B Civ. 3.67–72) and Philippi (Plut. Ant. 22.1–4). 731 Canidius Crassus was consul in 40 BC. Marcus Titius was consul suffectus in 31. This Domitius Ahenobarbus was the son of the consul of the same name in 54 BC. Flavius Gallus is not otherwise known (Pelling, 1988, 231). 732 Plut. Ant. 42.2–4. Even so, it was Antonius that allocated extra troops to Gallus for the purposes of fighting a heavier engagement, and so he bears some share of responsibility for the affair, even if it was his intervention that put an end to the unfavourable fighting (Ant. 42.4). Also see Pelling, 1988, 231–2. 733 Plut. Ant. 42.2–3, 43.1. 734 Plutarch directly attributes the good order of the Roman army on the march to Antonius’ leadership (Ant. 42.1). Antonius intervened to rescue Gallus, assiduously restoring the spirits of his men after that reversal (43.1), and he appears to have been greatly respected by the milites for, among other things, sharing in the army’s labours (43.2–3). When the army was later passing through mountainous terrain, via a route bereft of potable water and suffering from Parthian harassment, Antonius’ encouragement again proved essential to maintaining the army’s animus (47.4). Finally, Antonius brought his army safely to the river Aras, and oversaw a careful and successful crossing in the presence of the enemy (49.1–3). See Pelling, 1988, 220–1.


206

withdrawal. Even so, it must not be ignored that Antonius was ultimately responsible for the

loss of his siege train and thus the failure of his campaign (Plut. Ant. 38.2–3). While the

judgement of the army’s leaders was imperfect at times, Antonius and his subordinates were

experienced and for the most part effective, and so they will have contributed to maintaining

the army’s order and discipline during the long and taxing withdrawal.

As imperator, Antonius was the ultimate arbiter of the Roman army’s direction of march and,

therefore, the terrain it encountered along its path. In general, the army adhered to undulating

terrain and avoided open plains wherever possible, and Antonius appears to have wisely

followed local advice in this regard.735 At one point the Parthians attempted to attack the

Romans whilst the latter were descending a hill, and on other occasions they made attempts

to divert the Romans into more open ground so as to utilise their cavalry and archery more

effectively.736 Such scenarios were grave dangers to the slower Roman army, which could not

escape quickly if it was heavily defeated. However, Antonius did manage to avoid these traps.

In the case of the one significant defeat the army suffered, Antonius was still able to drive the

enemy off and organise an orderly withdrawal.737 There was, therefore, no point at which the

Parthians were able to threaten the Roman army so greatly that the latter was forced to stand

and fight, as the detachments led by Crassus and Octavius in 53 BC were. As such, aside

from the fact that the Roman army was very formidable on its own account, Antonius’ use of

terrain was cautious and appropriate to circumstances, giving the Parthians little opportunity

to utilise their horse archery in ideal circumstances as they had at Carrhae.

If we were to compare the withdrawals of 53 and 36 BC, we might then conclude that

Antonius’ officers and soldiers were of superior quality and that Antonius was simply a better

general than Crassus. It must be noted that Antonius was a younger man, and did not have to

contend with the intense psychological blow that Crassus suffered when his son was killed.738

735 Plut. Ant. 41.1–2, 45.2, 47.1–2. Antonius’ sage use of local advice compares favourably with Plutarch’s depiction of Crassus, who allegedly rejected the sober analysis of Cassius in favour of a route chosen by a duplicitous Arab chieftain (Plut. Crass. 20–2). Antonius’ choice nonetheless turned out to be correct: Xenophon also profitably utilised a local guide during the March of the Ten Thousand (Xen. Anab. 4.19–27). This connection was not lost on Plutarch (Ant. 45.6). On the connection with Xenophon, also see Pelling, 1988, 221. 736 Attacking on a reverse slope: Plut. Ant. 45.2. Attempts to divert the Romans: Ant. 41.3, 46. 737 Plut. Ant. 42.4. The Roman army was evidently able to take care of its thousands of casualties following this reverse (Ant. 43.1), showing it retained its organisation and logistical apparatus. 738 Plut. Crass. 26.3, 27.4–5. Crassus had been approximately 60 when he died, while in 36 BC Antonius was 47, still in his prime as a commander. That said, this may not have been a hindrance at all, considering for


207

However, there were three important differences between Crassus and Antonius’ armies and

the circumstances under which they retired.

The first was Antonius’ army. It was at least twice the size of Crassus’, had more than double

the cavalry and light troops, and possibly ten times as many heavy cavalry.739 It thus had

superior firepower to combat the deadly Parthian archery (Plut. Ant. 41.5), and such numbers

of cavalry that it could not be isolated and destroyed as Publius’ 1300 horsemen had been at

Carrhae (Plut. Crass. 25.2–12). Secondly, as a result of the size and strength of Antonius’

army and his own cautious generalship, the Roman army in 36 BC had not been decisively

beaten in battle.740 The destruction of Antonius’ rear–guard and the defeat of Gallus’

counterattack were very costly in men, and by the end of Antonius’ campaign, 20,000

infantry and 4,000 cavalry had allegedly perished, but more than half by disease.741 However,

even if these casualty figures are to be accepted, Crassus’ army also suffered similar

casualties in the fighting at Carrhae and in the withdrawal immediately following, but this

was in an army half the size and that already lacked sufficient light troops and cavalry.742

Finally, the fatal weakening of the Roman army at Carrhae due to heavy casualties and

desertion meant that the Romans marched by night in order to avoid fighting the Parthians,

and due to the army’s battered state, unfamiliarity with the country and the inherent difficulty

of night withdrawals, this only increased the confusion and led to further casualties.743

Conversely, Antonius’ army, which was perfectly capable of facing the Parthians in open

confrontation, marched mostly during the day, and so Antonius could more easily maintain

instance that Antigonus Monopthalamus was 80 when he died fighting at the Battle of Ipsus (Plut. Dem. 29.4–5). 739 Cf. Plut. Crass. 20.1 and Plut. Ant. 37.3. Of Crassus’ cavalry, only 1000 were Gallic cavalry, and even these were not heavily armoured (Plut. Crass. 25.2, 7–9), whereas Antonius had 10,000 Iberian and Gallic cavalry (Plut. Ant. 37.3). 740 While the casualties during Gallus’ failed attack were very heavy (3000 dead and 5000 wounded), the Roman army was able to take control of the battlefield and take all of its wounded to safety, showing that the army was still completely functional (Plut. Ant. 42.4–43.1). The destruction of Antonius’ baggage under Statianus saw the loss of two entire legions (Plut. Ant. 38.3), but this was only a part of the army. 741 Cf. Plut. Ant. 50.1. Note that the source for this figure was probably Dellius (Pelling, 1984, 239), who later betrayed Antonius and so was not a wholly sympathetic source (cf. Pelling, 1984, 221) who might have been inclined to exaggerate the scale of Antonius’ failure. 742 Plut. Crass. 31.7, cf. 27.6–7, 28.1–2, 29.4–6. Also see Sampson, 2015, 146. 743 Crassus’ army retreats by night and becomes confused and separated: Plut. Crass. 27.5. 4000 men killed in the Roman camp: 28.1. Vargonteius gets lost and is killed with four cohorts: 28.2.


208

control of his forces.744 The one time that Antonius’ army did march by night, it fell into

confusion (Plut. Ant. 48.2–5). While one might think that marching by day was actually more

dangerous, as the Parthians did not tend to pursue at night (Plut. Crass.28.1) the fact that

winter was approaching (Plut. Ant. 40.3) probably meant that the temperature was more

bearable during the day, while the army’s hunger also meant that the Romans could not afford

the delay of moving slowly at night (45.4). Withdrawal by night would therefore have

probably been a greater hindrance than a help, and so Antonius’ decision to withdraw during

the day was correct, but it was only made possible by the ability of his army to fight off any

Parthian attack.

As such, the Roman withdrawal from Parthia in 36 BC must be considered a success, because

the majority of the army escaped over a very long distance despite close pursuit by a

formidable enemy. The key factor was that Antonius’ army was much larger and better

equipped to face the Parthians than Crassus’ had been. The strength of Antonius’ army meant

that the Parthians could not decisively defeat it in battle and could not inflict crippling

casualties, so that Antonius’ army retained its order and cohesion and was not forced to carry

out any desperate high–risk manoeuvres such as withdrawal by night. In addition to this, the

fortitude of the soldiers who endured nearly a month of skirmishes and forced marches was

praiseworthy. Their military skill in utilising varying formations, such as hollow squares and

the testudo, and counterattacking and retreating as ordered, is just another example of what

properly trained Late Republican legions were capable. The cooperation between missile–

armed light infantry and milites was much more effective in 36 BC (Plut. Ant. 45.1–3) than it

had been in 53 (Plut. Crass. 24.4), because Antonius’ army actually possessed firepower that

could rival its enemies.

The army’s leadership, particularly Antonius, deserves some praise. While Antonius must

ultimately shoulder the blame for the campaign’s failure in the first place, his leadership

during the withdrawal was cautious and orderly, utilising high ground as much as possible

and avoiding unnecessary engagement, but featuring aggressive counterattacks and

charismatic personal interventions when necessary.745 Antonius’ leadership during the

744 Plutarch (Ant. 47.1–2) emphasises the one occasion in which Antonius did retreat at night, pointing out that the Parthians usually did not pursue at night. This suggests that much of Antonius’ retreat, and the many battles fought with the pursuing Parthians (50.1), occurred during daylight hours. 745 Plutarch’s criticism (Ant. 37.5–38.1) of Antonius’ campaign timing is rejected by Pelling (1988, 221, 224),


209

withdrawal from Media was deliberately portrayed in glowing terms by Plutarch, but the

effectiveness of his tactics cannot be denied.746 The general fielded the correct balance of

missile–armed infantry, cavalry and legionaries, he used high ground and night marches to

avoid unfavourable confrontations, and a variety of combined arms–tactics, particularly

counterattacks, were utilised to fight off the enemy. The main way that this lengthy

withdrawal could have failed, considering the strength of Antonius’ forces, was for the

general to have blundered into an extended and very costly engagement. However, Antonius

had much better intelligence than Crassus had had in 53 BC.747 He took advantage of this,

and whatever the criticisms of Antonius’ rashness in other scenarios, his generalship was too

cautious to give the Parthians any real opportunity for a decisive victory.748

The Clades Variana, AD 9

The clades Variana was a major Roman defeat, occurring over three days in September, AD

9.749 In brief, Publius Quinctilius Varus was the governor of Germany and commanded a

significant army of three legions and several units of cavalry and professional auxiliaries

(Vell. Pat. 2.117.1). While en route to put down a rebellion, perhaps intending to encamp

closer to the Rhine over the winter, Varus and his army were ambushed by a force of

Germans, primarily Cherusci, led by their chieftain Arminius, who was also a Roman

auxiliary officer and perhaps a member of Varus’ consilium.750 The clades was not an

ordinary battle, but a series of ambushes, skirmishes and small battles that took place over an

and the main event leading to the campaign’s failure, the loss of Statianus and the baggage (Plut. Ant. 38.3) was not all Antonius’ fault: Antonius was close enough that Statianus could allegedly send for help (Dio 49.26.1), even if the former arrived too late. Dio (49.25.5) also claims that the Armenians could have assisted Statianus but neglected to do so. See Pelling, 1988, 225–6. It is possible that Antonius actually made very few mistakes after all: the formidable geographical challenges of campaigning in this region were probably the main factor in the campaign’s failure (see Engels, 1978, 67–9, 121–2). 746 Plut. Ant. 42.1; Pelling, 1988, 221, 231–3. The overall theme of Plutarch’s account of the Parthian campaign is to foreshadow Actium with several parallels, firstly a strategic blunder blamed on Antonius’ relationship with Cleopatra, followed by Antonius’ shame, then attempts (successful in Parthia) to repair the situation through inspiring leadership (Pelling, 220–1). While Plutarch criticised the influence of Cleopatra (221), we can gather that Plutarch genuinely considered Antonius to have been an excellent general. 747 Where Crassus was betrayed by his guide (Plut. Crass. 29.4), Antonius received valuable information on Parthian intentions (Plut. Ant. 46.2–4). 748 Plut. Ant. 38.1–3; Syme, 2002, 104. 749 Three days: Schlüter, 1999, 154. September: Schlüter, 1993, 45. 750 Varus’ intentions: Schlüter 1993, 45. The role of the Cherusci: Wolters, 2008, 119–121. Arminius: Vell. Pat. 2.118–119.


210

area of at least 50 square kilometres.751 The Roman army appears to have attempted to fight

its way free but, at some stage during the second and third days of fighting, was defeated

decisively, after which its remnants were scattered, destroyed or forced to surrender.752 The

defeat came as a great shock to Rome and has been viewed as a major strategic setback for

the empire.753 Meanwhile the Roman general, Varus, has received the lion’s share of blame

for the defeat.754

However, is it really fair to blame Varus’ decisions for what went wrong? What does this

defeat say, if anything, about the limitations of Roman retreat tactics in this period? The

argument will be made below that the clades, in large part, was a withdrawal that failed. The

following analysis will thus briefly analyse the complex evidence for the clades in order to

establish a general sequence of events, and then an argument will be made that Varus’ plan in

response to the initial attacks was to withdraw. The Romans’ tactics in this scenario, and the

factors leading to the Romans’ defeat, will then be assessed in order to provide an assessment

of Varus’ generalship and the Romans’ application of retreat tactics in this situation.

Among the literary sources on the clades, Cassius Dio (56.18–22.2) is the most detailed.755

Although Dio wrote over two centuries later, it is generally agreed that he had reliable

contemporary sources at his disposal and that most aspects of his version of events are

credible.756 He provides the most analytical account of the battle, focusing, not on the

character of Varus, but on factors that contributed to Roman defeat, such as terrain (56.20.1,

21.2), weather (56.20.3, 21.3) and disruption by the baggage train (20.2–5). He portrays the

751 Sheldon, 2001, 23–24; Wolters, 2012, 11. 752 See Dio 56.21. Some escaped to lead the Roman general, Germanicus, back to the site of the battle several years later (Tac. Ann. 1.61). According to Tacitus (Ann. 12.27), some Roman captives were liberated only decades later, but most of the men that surrendered were probably killed in the immediate aftermath of the battle (Vell. Pat. 2.119.4; Tac. Ann. 1.61). 753 Suet. Aug. 23: paene exitiabilem. Also Dio 56.23–24. In modern scholarship, it has been argued that it was a major setback (e.g. von Schnurbein, 2003, 105), although it has also been suggested that the effect of the clades on Roman policy has been exaggerated: cf. Heather, 2010, 5–6; Wolters, 2012, 13; Winkler, 2016, 50–51. 754 See Vell. Pat. 2.117–119; Florus 2.30.31; Dio 56.18–19: the worst criticisms began with Velleius Paterculus (Timpe, 1970, 123). Suetonius claimed that his defeat was ‘nearly fatal’ to the empire (Aug. 23) and sharply criticised Varus’ temeritas et neglegentia (Tib. 18). Tacitus (Ann. 1.61) was less critical, describing Varus as ‘ill–fated’, but naive (1.55). Aside from incompetence, the disaster was also attributed to divine anger (Winkler, 2016, 37). Modern critics: Benario, 1986, 402; Sheldon, 2001, 30; Baltrusch, 2012, 131. More sympathetic views: Syme, 2002, 511; Wolters, 2012, 10. 755 Manuwald, 2007, 431. 756 Wolters, 2012, 11; cf. Manuwald, 2007, 438.


211

Roman army as helpless.757 However, his emphasis on poor weather could be a literary

device.758 More critically, the discovery of the battle site at Kalkriese has invalidated Dio’s

characterisation of the terrain. In specific, Dio describes deep ravines and thick forests which

he alleges severely limited the manoeuvrability of Roman forces (Dio 56.20.1, 21.2).

However, the site at Kalkriese, particularly in the defile known as the Kalkrieser–Niewedder

Senke (Kalkriese–Niewedde depression) where the majority of Roman finds have been

unearthed, does not (and never did) have such landmarks.759 As such, Dio’s description of the

terrain at Kalkriese is wrong.760 This has ramifications for the accuracy of his description of

the fighting, as he relates the tactics used by the Romans to this incorrect description of the

terrain.761 As a result, the tactical details of the fighting in Dio’s narrative are not entirely

trustworthy.

That said, archaeological evidence provides some support for aspects of Dio’s narrative. The

size of the excavated battlefield (at least 30 square kilometres) is consistent with the duration

of the battle, which comprised a slow, halting advance by the Romans over several days.762

There is some support for Dio’s emphasis on the difficulties presented by Varus’ large

baggage train, which allegedly included a significant number of women, children and

servants and too many wagons and pack animals,763 as remnants of wagons have been found,

confirming their presence.764 According to Dio, the baggage train was a major factor in

757 Dio 56.20.5, 21.3–4. Manuwald, 2007, 440, 442. 758 See Dio 56.24.2–5; Manuwald, 2007, 446. Östenberg (2017, 253–9) has argued that Roman writers could perceive weather and terrain as intertwined with the divine in opposition to the Roman army. Conversely, such weather was fairly typical in that area and at that time of year (Murdoch, 2008, 107). 759 Schlüter, 1999, 128–131. It is possible that there was some earlier fighting further south and south–east, in which the ground was undulating and more heavily forested, which might partly explain Dio’s confused characterisation of the terrain. 760 Swan, 2004, 261. 761 See Dio 56.20–21. Dio does not mention the marshy, waterlogged ground. Perhaps he described the terrain from further south–east where the fighting could have started. Tac. Ann. 1.63–4 is much closer to the reality. 762 Schlüter, 1999, 154. The battlefield was at least 30 square kilometres, according to Rost/Wilbers–Rost, 2010, 118, contra Sheldon (2001, 23) who states it was approximately 50 square kilometres. However, Sheldon’s source for this is unclear. For comparison, note the case of Cestius Gallus in the defile of Beth–Horon in AD 66 (Joseph. BJ 2.546–551). On the third day of his retreat, Gallus' hard–pressed forces advanced at around half their normal speed (Gichon, 1981, 50, 59, 61–2). At Kalkriese, the Roman army was attacked over a distance of at least 15 kilometres, running east to west (Rost/Wilbers–Rost, 2010, 133). Tacitus’ mention (Ann. 1.60–61) of multiple camps discovered at the site of the clades also suggests multiple days of fighting, although squaring this temporally and spatially with Dio’s account is not straightforward (Swan, 2004, 264). 763 Dio 56.20.2. There is evidence for the presence of some women in the Roman train (Moosbauer/Wilbers–Rost, 2007, 33). 764 Their number cannot be known, or whether they were two or four–wheeled carts Rost, 2016, 563. The


212

Roman disorganisation during the initial ambush (56.20.5). Dio also states that the Romans

reorganised themselves into better order after an initial setback (56.21.1). Further, although

the presence of women was unusual for a military campaign, it is probable that Varus

intended to take up winter quarters after a brief campaign against the insurgents, which would

explain why his baggage train may have been larger than necessary.765 However, the presence

of thousands of military servants and a significant number of wagons was routine in Roman

military operations.766 Dio (56.21.5) also mentions that Varus was wounded during the

fighting and later committed suicide. Velleius (2.119.3) and Tacitus (Ann. 1.61) support his

account.767 As noted, Dio’s claims of Roman counterattacks being interrupted by thick

forests, but with no mention of the marshes or entrenchments that hindered them, conveys a

sense that, at some stage during the battle (perhaps at its climax) the Romans were unable to

utilise their formations and tactics effectively.768 Overall, one cannot disagree with Swan’s

claim that Dio’s account is reliable in some respects but ‘poor in details’.769

The archaeological evidence has not illuminated the matter of whether the Romans encamped

during the battle. Tacitus and Dio mention that the Romans encamped at least once during the

fighting, but no Roman camps have (yet) been definitively identified at Kalkriese.770 Dio

(56.21.1) mentions that the Romans made a camp to reorganise after the first day’s fighting

and they departed in the morning before the worst phase of the battle, which might be

identified with the fighting in the Senke. Dio explicitly mentions only one camp, whereas

Tacitus (Ann. 1.61) states that most Roman casualties were suffered between the first and

second Roman camps. The camp that Dio mentions (21.2) may be the first, well–laid camp

that Tacitus mentions (Ann. 1.61). The second camp of Tacitus (1.61) might then be identified

with the desperate fighting at the end of Dio’s account (21.3–5). While Tacitus does not make

the sequence of events clear, his second, ramshackle camp is not incompatible with Dio’s

archaeological finds from these carts are fragmentary (ibid., 561). On two and four–wheeled carts, see Roth (1999, 211–212): both two and four–wheeled variants were pulled by only two animals. The four–wheeled cart’s carrying capacity was not drastically higher than that of the two–wheeler, so it is not necessary to believe that Varus’ army had both types. 765 Dio 56.19.3–4; Schlüter, 1993, 45. The baggage train probably was a major hindrance during the fighting, but there was little else Varus could do about it aside from his reorganisation after the initial ambush (Dio 56.21.1). 766 Roth, 1999, 114, 211. 767 Also see Swan, 2004, 266. 768 Dio. 56.21.2; cf. Swan, 2004, 265. Also see Schlüter, 1999, 46–47; Großkopf/Rost/Wilbers–Rost, 2012, 102–103. 769 Swan, 2004, 250. 770 Wilbers–Rost, 2016, 575 and Crossland, 2017. Cf. Tac. Ann. 1.61; Dio 56.21.1.


213

account even if the latter does not explicitly mention both camps.771 It was, however, routine

for a Roman army to encamp if attacked on the march.772 Tacitus (1.63) depicts Caecina

doing this under similar circumstances in AD 15. Therefore, it is reasonable to accept Dio’s

description of the first camp, and to keep in mind that a second desperate fortification might

have been constructed during the most intense fighting.773

From the literary sources, it may be determined that the events and planning took place as

follows: the army of Varus with a significant baggage train, after departing the main Roman

camp near the Weser, moved northwest through the territory of the Cherusci towards a

rumoured rebellion, when it was attacked by Arminius in hitherto friendly territory.774 Varus’

original plan was most likely to intimidate the rebels into submission by his mere presence,

and perhaps to perform any actual fighting with auxiliaries, then to go into winter quarters

closer to the Rhine, hence the presence of so many civilians.775 The betrayal by Arminius

forced him to change his plans. Having survived the initial attacks and encamped, Varus

reorganised the baggage train and changed direction westward, in the direction of the Rhine

where greater Roman forces were concentrated.776 Varus likely expected that his experienced

army would stand up to further attacks and allow him to break out. However, the terrain and

the German attacks wore down his army so much that he was unable to escape the ambush.

Over a day or two of heavy fighting, including at least one major failed Roman counterattack,

the Roman army was reduced to a fragment of its former size (Dio 56.21.2–4). Varus

despaired and committed suicide, triggering a general sauve qui peut, and most of the army

was killed (56.21.5–22.2).

771 Cf. Goodyear, 1981, 96; Grethlein, 2013, 136. 772 Note, in particular, the reaction of Antonius when his army became disordered during a night march through the foothills of Armenia in 36 BC (Plut. Ant. 48.5). Also see Cestius Gallus’ retreat to a series of camps when attacked between Jerusalem and Beth–Horon in AD 66 (Joseph. BJ 2.541–252, 544–545). 773 Grethlein, 2013, 136. 774 Dio 56.19.4–5, 20.1–2. The place from which Varus departed is not known. Murdoch (2008, 99) suggests the camp was at Minden. Anreppen is another possibility. As for the location of the rumoured rebellion, the Angrivarii appear likely candidates considering Varus’ northerly direction of march (Murdoch, 2008, 103). 775 Schlüter, 1993, 45. On Roman tactics against rebellions, see Goldsworthy (1996, 90–95). The unsettled nature of Germania may have required the civilians to be brought into winter quarters with the army for their protection. On the fighting role of auxiliaries see Lendon, 2005, 242. 776 Dio 56.21.1. Varus was presumably heading towards the Rhine and the Lippe where most Roman bases were located (see Murdoch, 2008, xi).


214

So how do we characterise Varus’ manoeuvres? According to Dio (56.21.1), Varus abandoned

a part of baggage during the initial encampment so that the army could march with better

order. This was a standard Roman technique during withdrawals.777 If Varus abandoned the

baggage, it was to render his army more mobile and effective in battle, considering the

trouble caused by the impedimenta during the initial attack (Dio 56.20.5). Thus, Varus was

clearly not intending to stand and fight, but to keep moving, in an attempt to withdraw from

what he now realised was enemy territory. Velleius Paterculus, in his use of the term

egrediendi (escaping) implies that withdrawing was, or should have been, one of the army’s

intentions.778 For what other reason would Varus have destroyed most of his baggage

(presumably including food and other essentials), if not to move more quickly, disengage

from the enemy and reach safety as soon as possible?

The destruction of the baggage leads to other deductions. The closest Roman base was

approximately 100 kilometres away–back in the direction from which the army had come.779

Haltern was his more likely destination, as it was a large base close to other Roman bases and

the Rhine legions. However, it was over 100 kilometres away.780 As Varus had unexpectedly

found himself in enemy territory, any chance of obtaining provisions from local allies was

gone. Further, the sparse agricultural resources of this region rendered foraging for food and

fodder impractical.781 To reach Haltern from the army’s position at Kalkriese would take a

week at best, as the Romans were burdened with carts of wounded men and supplies.782 The

army was at the outermost limit of the distance it could operate from a base without grazing

for fodder, and its horses would probably collapse from exhaustion by the end of the week.783

777 Ch. 4, 145–6. 778 Vell. Pat. 2.119.2: Exercitus omnium fortissimus ... circumventus, cum ne pugnandi quidem egrediendive occasio nisi inique ... Also see Woodman, 1977, 199. 779 100 kilometres: von Schnurbein, 2003, 94. Also see Murdoch, 2008, xi. 780 Cf. Murdoch, 2008, xi. This distance is as the crow flies. The paths suitable for the passage of an army with wagons probably made the trip longer. Varus might not have known the precise distance, but he must have been cognisant that his destination was not particularly close by. 781 Timpe, 1995, 20. Water may have been easy to find in the wetlands in which the Romans were operating, especially if it was raining heavily. Food and foraging were more difficult. 782 Varus probably had wagons of wounded soldiers (Rost, 2016, 561, cf. BAfr. 21). In ideal circumstances, mule–driven carts could travel between 20 and 30 kilometres per day (Roth, 1999, 211). However, the situation in Germany will have enormously reduced this rate of movement. 783 An army could only operate further than 80 to 100 kilometres from a supply base by grazing their animals (Roth, 1999, 129, 198, 201). This was impractical in Varus’ situation. Horses, and other baggage animals, required a full rest day of grazing every five to seven days (Engels, 1978, 154–155). They could have been eaten, but probably not cooked, as the Germans will have prevented the Romans from collecting firewood.


215

Varus’ decision to destroy the baggage thus allowed the Romans only one week to survive.

Schlüter argues that this action rendered it impractical for the Roman army to wait out the

Germans in an encampment or to go back the way it came.784 Therefore, speed was of the

essence. Varus must have intended to march expediti, perhaps burdening his infantry and the

calones and other non–combatants with extra supplies to reduce the army’s diminishing

reliance on wagons and pack animals.785 The infantry could not be overburdened, as they

would be expected to fight, but reducing the army’s baggage too severely could cause food

shortages within only a few days.786 By the time Varus had dismissed his plan to confront the

northern rebellion, the only avenue of escape was through the Kalkriese–Niewedde

depression.787 Varus’ broader strategy must have been to retire from enemy territory by

marching westward through the Senke, to reach friendly territory near the Rhine.

Sheldon raised a key criticism of Varus’ decision–making process; that he did not reconnoitre

the defile before entering it.788 She assumes that reconnaissance was possible and that there

was an alternative route available. Regarding reconnaissance, by what means could Varus

have gained intelligence about what lay ahead? Of primary importance is the fact that

Arminius and his auxiliaries had probably been Varus’ main source of intelligence,

particularly as the army was passing through Cheruscian land.789 Polybius, no mean authority

on military matters, admits that a general could not always be personally acquainted with the

army’s path and sometimes had to rely on local guides.790 It is also essential to note that, in

this period, contra Sheldon, the Roman army had no professional tactical intelligence–

gathering apparatus.791 At any rate no scouting force could have survived, considering that

the vicinity of Kalkriese Hill was in possession of enemy troops. The defection of Arminius

and his auxiliaries deprived Varus of a vital source of local information and under the

circumstances, there was no way to rectify this deficiency.792

784 Schlüter, 1993, 46. 785 Note that expediti did not mean marching without baggage, but with minimal baggage (Roth, 1999, 81). 786 Ibid., 88. 787 Schlüter, 1993, 45. 788 Sheldon, 2001, 29–30. 789 See Murdoch, 2008, xi. 790 Polyb. 9.14.2–4. For example, Caesar relied upon Diviciacus in Gaul (BGall. 1.41.4). 791 Austin &Rankov, 1995, 108, 135, 168–169, 246; contra Sheldon, 2001, 29–30. Sheldon gives Varus too little credit, as Varus had local informants, chief among them Arminius. Varus’ other auxiliary troops did not have the local knowledge of Arminius’ auxiliaries (Timpe, 1970, 110). 792 Goldsworthy, 2007, 98.


216

Although the terrain in the Senke was unsuitable for an army, this does not mean a safer route

was available.793 Other Roman generals had found themselves in this type of position in the

past. In 188 BC, Gnaeus Manlius Vulso, while campaigning in Thracia, entered a narrow

defile and suffered heavy losses (Livy 38.40.6–15). He later successfully defended himself

against senatorial criticism on the grounds that he understood the danger but had no other

choice of route and could never have achieved anything without advancing aggressively

(Livy 38.49.5–13). If Varus had survived, he might have made a similar justification.794 In

108 BC Metellus Numidicus, while campaigning in Numidia, detected an enemy ambush

overlooking a defile through which he had to march (Sall. Jug. 49). Scarcity of supplies

forced him to enter the defile and suffer the ambush anyway (Jug. 50). Similarly, Varus must

have understood that he was taking a calculated risk.

For several reasons, primarily the logistical ones given above, Varus had no choice but to take

the most direct route to safety. He could not wait inside a camp or simply defend himself on

high ground. The route he needed can only have been through the Senke, where Varus surely

did not expect the enemy to commit themselves to a decisive battle, as the Germans were

usually worsted in such encounters.795 The fact that he kept some wagons shows his

confidence that they would be able to make their way through the defile safely. It is likely

that he did not expect to remain there for long. If he was not aware of the extent of the

preparations for the ambush in the corridor, he must have expected the Romans to force their

way through any resistance.796 The only way the Roman army could have been stopped from

passing through the Senke was if Germans were so numerous and well–prepared for intense

fighting that they could stop them.

793 Assuming Varus fully understood the perils of the Kalkriese path, perhaps he could have taken a westward path further south, avoiding the defile at Kalkriese (see Schlüter, 1992, 311). But even if he could have, this would simply have exposed his forces to the sort of hills and forests that Dio describes (56.20.1–4, 21.2) as being so debilitating to Varus’ army, by preventing it from deploying and manoeuvring its formations. 794 In AD 15, when the experienced general Aulus Caecina Severus was ambushed by Arminius in similar terrain, he also chose to push forward in search of better ground (Tac. Ann. 1.64–65). 795 It would have been reasonable for Varus to expect the Germans to launch a major attack on him and for his army would win such a direct encounter (Goldsworthy, 1996, 53). Again, note Tac. Ann. 1.63–64. 796 Schlüter, 1993, 36, 46.


217

As mentioned, the Roman army had a number of wagons accompanying them to the Senke,

many probably loaded with wounded soldiers. A usual order of march would have placed the

baggage in the centre of the Roman formations, but the narrow space would have made this

difficult as Roman army would have been divided in half by the baggage. The fatal chaos

caused by fighting troops entangled with wagons and other impedimenta is emphasised by

Tacitus and Dio.797 Although Dio’s statement (56.21.1) that Varus reduced his baggage and

proceeded from his first camp in a relatively orderly formation has been accepted, the

confined terrain near the rampart and the intensity of German attacks must have rendered the

remaining baggage an enormous liability, taking up valuable space that the army required to

fight and rally. Thus, the baggage was a serious impediment, and yet this was an unavoidable

complication of fighting on such difficult ground and was largely outside the general’s

control.

After the Roman army entered the Senke and encountered intense opposition, it was not

practical to turn back.798 If the Roman plan had been to force a passage west through the

defile, the lack of space and enemy attacks would have made it too difficult for the Roman

army to continue moving or properly defend themselves. There was no space to use tactics

that might have allowed the Romans to make a fighting retreat, such as forming a hollow

circle or square.799 Little could therefore be done but to attempt to fight in this

disadvantageous location.800 The only other option was to move north and utilise the pass on

the southern and south–western edge of the Great Marsh. There is some evidence that the

Romans attempted to do this, but Arminius would have predicted the directions in which the

Romans might flee and acted to prevent this.801

Failing to break through or take an alternative path through the Senke, the Roman army had

exhausted all options. At some point, casualties mounted to the point that the army’s cohesion

collapsed, logistical and medical capacities failed completely and groups of soldiers or

individuals attempted to escape. Of Varus’ involvement in this stage of the battle, it is only

797 Tac. Ann. 1.64–5; Dio 20.5. As another example, the Vitellians at Cremona in AD 69 were defeated after becoming entangled in their baggage (impedimenta) and wagons (vehicula) (Tac. Hist. 3.25). 798 Schlüter, 1993, 36; Sheldon, 2001, 30; Rost / Wilbers–Rost, 2012, 121. 799 Caes. BGall. 5.33, BCiv. 1.79.5; Plut. Ant. 42.1. Cf. Goldsworthy, 1996, 105. Cf. Ch. 5, 179–80. 800 If the rampart was German, it was the key obstacle that the Romans had to attack. There is evidence that they did so: Schlüter, 1993, 37; Moosbauer / Wilbers–Rost, 2007, 28; Wilbers–Rost, 2016, 573–575. 801 Schlüter, 1999, 131–133.


218

known that he was wounded fighting and, when the situation became hopeless, probably for

fear of capture, he and a number of senior officers killed themselves.802 It was a measure of

the difficulty of the situation that Varus was wounded, as a Roman commander of his stature

did not habitually expose himself to direct danger.803 Varus was no doubt engaging in active,

charismatic leadership in a desperate attempt to bolster the nerve of his crumbling formations.

Such behaviour was typical for a general attempting to support retreating troops.804 However,

at some point after being wounded, he killed himself to avoid capture, and his death no doubt

triggered a general rout, as Dio describes.805

To summarise, Varus’ tactics in response to the initial ambush were to encamp, minimise the

baggage train and attempt to force his way through the ambush. Although Varus plunged his

forces more deeply into the enemy ambush, his intention was to reach safety as quickly as

possible. He did not plan to confront the enemy in battle, but was carrying out a withdrawal

and utilised standard, aggressive, forward–moving tactics in doing so. Logistical constraints

forced the Romans to take the shortest feasible route westward to where Roman bases were

located. Although the Romans’ path through the Kalkrieser–Niewedder Senke comprised a

natural defile not suitable for an army, Varus likely expected that even on difficult ground the

Germans would be unable to defeat his men in a direct confrontation and that the Romans

would eventually emerge from the defile and continue their withdrawal westward. There was

no precedent in Roman experience for the intensity of the fighting they encountered in the

pass. It is thus difficult to blame Varus for his choice to move through the Senke when his

alternative options and his knowledge of the enemy forces were so limited. Once the Roman

army was trapped in the area of what is now called the Oberesch, it could not escape intact, as

the army was prevented by the terrain and enemy action from turning back, deploying and

fighting effectively, taking the route to the north along the edge of the Great Marsh or

breaking the enemy in a decisive fight. The intensity of enemy attacks stopped the Roman

army in its tracks, leaving it trapped in the ambush when it had originally intended to force its

way through. The army had to stop and fight under incredibly disadvantageous circumstances

and, while it attempted this, as shown by the evidence of heavy casualties, it failed. With the

802 Tac. Ann. 1.61; Dio 56.21.5. Also see Swan, 2004, 266; Rauh, 2015, 400. 803 See Gilliver, 2007, 136–137. 804 Sall. Jug. 98.1–4. Livy 30.18.12–4. Also note Caes. BCiv. 5.35.5. 805 Dio 56.21.5–22.1. The death of an army’s general was a severe psychological shock that could easily trigger a rout (Sabin, 2007, 431).


219

weakening of the army after its defeat near the rampart, a return to Varus’ original plan of

advancing through the ambush cannot have been possible.

Sheldon is probably correct in stating that the fate of the Roman army was sealed as soon as it

entered the Senke.806 As the Romans were prevented from escaping, breaking the enemy was

their best hope of survival, but there was probably no opportunity for a decisive clash. What

appears clear from the archaeological evidence is that the Romans attempted every

possibility: advance, stand and fight, change path and flee. Arminius seems to have planned

for every eventuality, and the numbers of German warriors only grew as the Roman situation

worsened (Dio 56.21.4). The Roman defeat was not due to any weakness or gap in their

tactical repertoire, but due to the sheer impossibility of their situation.

Just as Antonius succeeded where Crassus failed because the latter had been decisively

defeated while the former was not, Varus’ retreat tactics failed because in such terrain he

could not win a decisive battle and yet he was forced to fight one. There was no opportunity

on such ground to dissuade enemy pursuit, and movement under the circumstances was all

but impossible. Varus’ army was simply not strong enough, and the difficult terrain removed

any other advantages they might have had. The Romans could not form a square or circle, nor

manoeuvre their cavalry and auxiliaries in support of the milites (Dio 56.21.2). Perhaps a

final factor working against the Romans was the difficulty of night marches, which Dio

described as the Romans’ final attempt at escape.807 The last thing holding the army together

once it was trapped in the defile was probably Varus’ leadership. When he died a rout was

inevitable, but the fact of his suicide almost certainly demonstrates that there was nothing

else to be done.

Arminius must be credited for placing the Roman army into an impossible situation. In later

German campaigns, when Roman armies were placed in similar situations they did not

significantly change their tactics, but were simply fielding larger forces.808 Perhaps Augustus

806 Sheldon, 2001, 30. Also see Lienemann, 1993, 79. 807 Dio 56.21.3–4. 808 Caecina, finding himself in a comparable situation in AD 15, utilised similarly aggressive tactics, but with the crucial advantage of greater forces than had been possessed by Varus (Tac. Ann. 1.65). When Germanicus faced another comparable ambush in AD 16, he had a mammoth force of eight legions at his disposal (Ann. 2.17). Germanicus used artillery to escape another ambush by Arminius in AD 16 (Ann. 2.20), and archaeological evidence suggests that the Romans did likewise in another possibly comparable situation during


220

should never have left Varus to administer such a country with so few men.809 But it was the

betrayal by Arminius who held, for a German, a position of unparalleled trust and authority in

the Roman army that was the decisive factor.810 Any other commander in Varus’ position

would have been hamstrung by the betrayal of the auxiliaries depended upon for local

intelligence. He would have responded by abandoning much of his baggage, making a

fighting withdrawal along the shortest path towards the Rhine, and would almost certainly

have fallen into Arminius’ ambush. If Varus’ tactics displayed vulnerabilities that Arminius

could take advantage of, these vulnerabilities were systemic to the Roman army and did not

disappear with Varus’ death. When it came to Varus’ attempts to withdraw, his plan was

sound, but inside the defile at Kalkriese, he had neither the numbers nor the physical space to

utilise the army’s typical retreat tactics. Arminius had surely known that this would occur.

Conclusion

The subject of this chapter has been the limitations of Roman retreat tactics. It has been seen

that even the most advanced retreat tactics were not always successful. In general, as

highlighted through the examples of Crassus and Marcus Antonius, withdrawal in open

ground could be extremely challenging against an aggressive enemy. In these examples, the

presence of large numbers of enemy cavalry and horse archers made engagement in open

terrain difficult enough but, as examined previously, a fighting withdrawal could not be

carried out indefinitely under enemy pressure.811 With no high ground or camp to withdraw

to and no opportunity to decisively defeat the enemy, a long distance withdrawal could only

be successful if the marching army, like that of Antonius, was able to maintain order and

thereby utilise heavy infantry, missile–armed light infantry and cavalry in combination to

the Battle of the Harzhorn c. AD 235 (‘Auf den Spuren der Römerschlacht am Harzhorn‘, NDR (2015) accessed 2019. https://www.ndr.de/geschichte/schauplaetze/Die–Roemerschlacht–am–Harzhorn,roemerschlacht100.html). But as mentioned, the presence of at least one ballista at Kalkriese leaves open the possibility that Varus used the same tactics (Stark, 2016). Even Roman intelligence practices did not change significantly for a long time (Austin & Rankov, 1995, 245–6). 809 Wells, 1972, 238–9. 810 When Drusus had been trapped by Germans in a defile in 11 BC, he was simply allowed to escape through enemy laxity (Dio 54.33.3–4), but Arminius was a more disciplined ambusher (cf. Tac. Ann. 1.68). He was just as capable in a pitched battle (Tac. Ann. 2.17, 45–6). Varus had allegedly been warned about Arminius (Tac. Ann. 1.55, 67–8; cf. Swan, 2004, 258), but if this actually occurred, Varus probably chose to stay out of what will have seemed an internal German dispute (Sheldon, 2001, 15) and gave Arminius the benefit of the doubt by virtue of his professional position in the army (Timpe, 1970, 104). 811 Cf. Ch. 2, 89–92.


221

fight off or counterattack pursuers.

Marcus Antonius’ withdrawal from Media was in all ways better conducted than Crassus’

from Carrhae, but this was fundamentally due to the fact that Antonius’ army was strong

enough to avoid a decisive defeat and that it thus retained its discipline and organisation. It

was not enough for the army to contain experienced troops and be led by an experienced

general. For instance, the Caesarian army led by Curio, when it attempted to retreat at the

Battle of Bagradas, 49 BC, collapsed quickly under enemy pressure because it had taken

heavy casualties and the morale of the soldiers eventually failed (Caes. BCiv. 2.42). A similar

situation occurred at Carrhae, when the milites refused to listen to the sound advice of their

general and forced him to negotiate with the enemy, ruining their chances at safety (Plut.

Crass. 30.3–5). The size and strength of a retreating army, relative to its enemy, was thus

essential to the success or otherwise of a fighting retreat or withdrawal.

Withdrawal through very constricted terrain was even more difficult, because any army

passing through a defile was exposed to attack along its extended flanks and unable to

manoeuvre or counterattack. The failure of Varus’ withdrawal through the Senke in AD 9 was

due to the fact that he could not utilise his army’s superior battle tactics in such restricted

terrain, and so he could not throw off enemy attacks or dissuade pursuit. He was forced to

stand and fight a battle he could not win, and the heavy casualties suffered in this decisive

defeat made further movement while under pursuit totally impossible. If, as Dio describes

(56.21.3–4), the Romans made one last attempt at escape after the main defeat, the balance of

forces was by this point too far out of the Romans’ favour for such a fighting withdrawal to

be achievable, even if the terrain to the west and north–west was more favourable.

It is a generally applicable observation that an army of roughly equal strength to its enemy, or

greater strength, was better able to retreat or withdraw successfully than an army that had

suffered heavy casualties and a loss of cohesion. This, on the face of it, is unsurprising.

However, the underlying reason that this observation is true is because, as has been argued an

army required strong discipline, organisation and leadership to maintain a fighting retreat or

withdrawal. Heavy casualties eroded discipline, organisation and leadership in large part due

to the deaths of large numbers of veterans and junior leaders who were at the forefront of any


222

battle.812 This directly affected the quality of the army’s leadership and indirectly made it

more difficult to keep the army organised, while also affecting the discipline and motivation

of the troops through the loss of leadership figures who were intimate to the milites.813

To conclude, the Roman army’s retreat tactics were highly effective during the Late Republic

and Principate, but it was a partial condition of a successful fighting retreat or withdrawal that

the Roman army maintained ratio so that it could fight successfully if forced to do so, or at

least present a bold enough front that the enemy would not commit to a decisive fight. Any

army that did not have the strength to dissuade enemy pursuit through intimidation or actual

fighting could not sustain a fighting retirement over a long distance without taking refuge on

high ground or in a fortified camp. It might be said then that Roman retreat tactics themselves

had no real limitations, as long as the balance of forces was in the Romans’ favour. It was

possible that unfavourable terrain, such as that at Kalkriese, could put paid to a withdrawal

that might have been achievable if the action were fought in open ground. However, just as

logistical and other factors informed the fateful Roman plan at Kalkriese, some tactical

circumstances were simply not conducive either to fighting a successful battle or to

withdrawing safely. The Romans always needed space to deploy and adjust their infantry

formations and to utilise their light infantry and cavalry, the effective combination of which

was key to Roman retreat tactics. In these cases, even the most efficient army and commander

were fairly helpless if their enemy made no blunders. The success of Antonius’ withdrawal

through Media over a very long distance and under heavy pursuit shows that, even for the

well–organised and disciplined Romans, sometimes there was no substitute for numbers. If

Crassus or Varus had approached their final crisis with more men, they might have been

successful.

812 On the heavy casualties suffered by centurions, see: Caes. BGall. 2.25, 6.40, 7.51, BCiv. 3.71; BHisp. 23; Lee, 1996, 211; Kagan, 2006, 169. It is also likely that many veterans fought in the front ranks (Gilliver, 2007, 130) and would thus be first to die in any major reverse. 813 This phenomenon has been studied by Rush in American and German infantry units during the Battle of Hürtgen Forest, 1944 (Rush, 2001, 323–4, 333–5, 343).


223

Retreat? Hell, we just got here!

(U.S. Marine Captain Lloyd Williams)

Conclusion

Retreats and Withdrawals in Roman Warfare


224

Summary of the Thesis

This thesis has sought to answer the question of how retreats and withdrawals affected the

tactics and strategies used by the Romans during the Middle and Late Republic. This topic

has been under–appreciated, perhaps because so many ancient battle accounts end with routs.

However, a significant proportion of battles also ended with a losing army retreating mostly

intact, and armies also regularly made withdrawals over longer distances in order to improve

their tactical or strategic situations (Introduction, 19–20). These retrograde manoeuvres were

common enough that they must have been part of the normal skill–set of Roman soldiers, and

an important consideration in the tactics and strategies of Roman generals. Thus, by

examining why and how the Romans performed retrograde manoeuvres, this thesis has

offered a new perspective on Roman tactics and military decision making during battles and

campaigns, as well as on Roman military evolution, from the Middle to the Late Republic.

In Chapter One, an examination was made of the varying scenarios in which organised

retreats or withdrawals were necessary and could be feasibly conducted, and of the places

that retiring armies typically sought refuge. Defeat in battle was the most common such

scenario. This type of retreat could be performed deliberately or involuntarily, either to avoid

further fighting or to re–engage in a more favourable location. Armies would also generally

disengage if night fell during a battle, if an engagement was indecisive, or if the weather was

particularly poor, but these were not particularly common scenarios. Armies could also

simulate flight or a disorganised retreat to lure their enemies into an ambush or surprise

counterattack, but infantry–based Roman armies did not commonly do this. An army seeking

refuge following a retreat would usually make for the closest fortified position or high

ground, a fortified camp being the preferred destination, particularly if the intent was to avoid

further fighting. An army intending to re–initiate fighting in a better position might, however,

choose to withdraw onto a hill instead of into their camp. Finally, an army that could not

retreat to high ground or a nearby camp, or was forced to retreat further for any reason, could

make a withdrawal, a retrograde movement over a longer distance. A withdrawal had

different characteristics to a retreat due to the increased speed and distance that the army

moved. A withdrawal was generally carried out in a marching column or square formation,

whereas a retreat could be conducted over a shorter distance, with a formed battleline while

still facing the enemy. Both retreats and withdrawals appear to have occurred fairly


225

commonly and so were major features of Republican Roman warfare.

In Chapter Two, it was shown that retreating or withdrawing troops were not always able to

do so unmolested, as they were often pursued or attacked by the enemy. Pursuit most

commonly took the form of either a general advance by a victorious battleline or a cavalry

attack, the latter in particular being highly effective in open ground and presenting less risk to

the victorious army. It was also demonstrated that pursuit by the enemy had a significant

impact on the success or failure of the retrograde movement. The advance of a victorious

battleline could easily induce a total rout if any shock occurred among the retreating troops,

and an army in withdrawal could be delayed or seriously disordered by pursuing cavalry or

light infantry. Retreats and withdrawals were difficult and by their nature demoralising. They

could thus easily break down into mass flight under enemy pressure. Sometimes the victors in

a battle chose not to pursue a retreating enemy, particularly if the latter maintained good order

or possessed visible reserves. Many generals will have been cautious of unrestrained pursuit

of a retreating enemy, in case their forces fell into disorder and were victims of a

counterattack. However, the most common scenario for a defeated army was a disorderly

retreat, leading to flight and disproportionate casualties. To avoid such a scenario, a retreating

or withdrawing army needed to prevent or counter enemy pursuit.

Throughout Chapter Three, an examination was made of the factors that contributed to the

ability of Roman armies to carry out successful retrograde movements, including training,

leadership, reserves and combined–arms tactics. It was shown that an army would more

likely be successful in retreat or withdrawal if it maintained ratio, thereby intimidating

pursuers and making the army more capable of fighting off any attack, and also if it acted to

prevent or counter enemy pursuit, such as by utilising reserves to intimidate or to

counterattack. Roman infantry appear to have been trained to carry out retreats and

withdrawals, and maintained ratio in a retrograde movement by combining close order

formations with gradual movement. The disciplina required to maintain ratio during a retreat

or withdrawal was facilitated by this training and also the leadership of officers, who

provided coercion, inspiration and direction. Centuriones and tribuni appear to have been the

key figures that milites would turn to when under pressure, and so they were essential for

transmitting the general’s instructions and maintaining order among the infantry during a

retrograde movement. Key tactics used by the Romans to manage or prevent enemy pursuit


226

included deploying the reserve infantry lines of the triplex acies, or using separate arms such

as cavalry, javelineers (velites or later, auxilia), or archers to fight pursuers, while the milites

retreated in tightly–formed bodies. The Roman army thus had all of the necessary tools to

maintain order in retreat or withdrawal, and to counteract enemy pursuit.

Chapter Four explored the role played by the general prior to and during retrograde

movements. To begin with, a Roman general was usually an older and more experienced

man, with decades of political and military education and experience, which he drew upon to

form the essential military qualities of scientia rei militaris, virtus, and consilium. Before a

battle, it was the responsibility of a general to gather and interpret intelligence so that he

could decide where and when to deploy his army to fight, and Roman generals typically

chose a battlefield with a fortified camp or high ground close enough to their battleline that

an orderly retreat to that location should have been possible. It was also their responsibility to

deploy reserves that could assist the milites if a retreat became necessary. During a retreat,

just as during a battle, a general had limited influence over what was happening. Mainly, a

good general demonstrated virtus by staying close to the fighting and encouraging his men

verbally and with his presence, and utilised consilium by dispatching reserves or his personal

guard to support hard–pressed soldiers or to threaten enemy pursuers. If an army had to

retreat into or out of a fortified position, the general ensured the walls were adequately

defended, exhorted the defenders, and organised the army’s march so that it was not thrown

into disorder while entering or leaving the position. During a longer withdrawal, a general

firstly ensured there was an adequate rearguard or some force dedicated to fighting pursuers,

and established ratio agminis that suitably protected the army’s vulnerable elements, such as

the impedimenta. During the withdrawal the general ordered counterattacks when necessary,

and managed the army’s direction of march, taking the most direct practical route to their

destination but also using terrain to their advantage when possible, such as by adhering to

high ground or denying it to pursuing enemy forces. If something went wrong, the general

could abandon part of the baggage to make the army faster and more effective in fighting, or

halt the army and bring it into encampment. It was thus demonstrated that while all parts of

the Roman army were responsible for a successful retreat, as milites and their officers had to

maintain ratio and fight off pursuers, the decisions made by the general were key to the

outcome of affairs. A general trained the army, chose officers, selected the terrain and

position for a camp, deployed the army for battle and posted reserves. A successful retreat or


227

withdrawal depended on the maintenance of ratio and management of enemy pursuit, which

were heavily affected by the general’s decisions.

In Chapter Five, the development of Roman retreat tactics from the Middle to the Late

Republic was examined in closer detail. In the Mid–Republic the Romans could if necessary

utilise the reserve lines of the triplex acies, and cavalry deployed on the flanks, to support a

retreat by the fighting line. Throughout the mid–to late second century BC, however, hints

emerge of Roman infantry learning to fight more independently in semi–isolated units, while

some talented commanders, such as Scipio Aemilianus and Gaius Marius, utilised cavalry

more dynamically by manoeuvring and counterattacking where necessary to support the

infantry. The smaller–scale and extensive skirmishing of Roman campaigns in this period

may partially explain these changes, but what is indisputable is that by the middle of the first

century, Roman retreat tactics emerge as much more complex and effective than they had

been in the Mid–Republic. In the Late Republic, the Roman army’s organisational and

tactical development meant that generals more often had flexible cavalry and light infantry

units to rely on. Legionaries could themselves serve in the latter capacity, as expeditae

cohortes. Roman armies could utilise more complex formations, including hollow squares or

circles that could move while fighting in all directions, and these formations worked hand–

in–hand with mobile reserves to fight while moving, to stand their ground, or to counterattack

as necessary, in order to create the necessary space for a safe retreat. Structural changes to the

Roman army, such as enhanced training regimes, the formalisation of cohorts, the integration

of missile–armed auxiliaries with heavy infantry units, and increasing use of auxiliary cavalry

increased the strength of Roman infantry units while making the army more flexible as a

whole. Therefore, a general had more options for utilising light infantry and cavalry to keep

milites secure during a retrograde movement, while the soldiers themselves manoeuvred more

skilfully in stronger formations, thereby resisting enemy pressure more effectively. This

variety of tools meant that by the mid–first century BC, good quality Roman armies led by

skilled commanders, such as Marcus Petreius, Marcus Antonius, and particularly Julius

Caesar, showed the ability to carry out fighting retreats in the most difficult circumstances,

even over long distances, against heavy pursuit, and in unfavourable terrain.


228

In Chapter Six, the limitations of advanced retreat tactics were examined through the analysis

of three large–scale examples of failed Roman retrograde manoeuvres in the Late Republic

and Early Empire. The chaos of the retreat after the Battle of Carrhae in 53 BC was

contrasted with the far more orderly and successful retreat of Antonius’ army through Media

in 36 BC. This comparison showed that Antonius was successful where Crassus had failed

because he was able to prevent the Parthians from either decisively defeating him in battle, or

pressuring him enough that he could not continue his retreat. The various tactical and

leadership failures in the retreats at Carrhae ultimately stemmed from the enormous casualties

the Roman army had suffered in battle, which had shattered its tactical effectiveness.

Antonius never suffered from this disadvantage. In the third case study, that of Varus’ famous

defeat at Kalkriese in AD 9, there was an examination of the general’s decisions, particularly

regarding his choice of route through a defile. However, it was argued that Varus had little

choice in the matter, and had likely expected that his men would win any skirmishes in the

defile. Unfortunately, enemy resistance was greater than expected and his army was forced to

fight in extremely unfavourable terrain, resulting in the army’s decisive defeat. As a result of

this defeat and the heavy casualties suffered, the Romans could no longer threaten nor

counterattack their pursuers, or even move in an orderly manner, and so there was no further

opportunity for a successful retreat. The fact that the Romans lost heavily in the battle at

Kalkriese was the decisive factor in the failure of this retreat, suggesting that the army was

perhaps never as strong as it needed to be in order to escape its situation. At both Carrhae and

Kalkriese, casualties were decisive in preventing successful retreat. Very heavy casualties

weakened the army at all levels, particularly with disproportionate casualties among veterans

and centuriones who usually bore the brunt of the fighting.814 They also emboldened the

victors. A weakened army attempting to retreat against superior and confident forces was

unlikely to be successful no matter its chosen tactics, if it could not find immediate refuge. As

such any fighting retreat, even with the Roman army’s highly effective tactics, still required a

favourable balance of forces so as to prevent decisive engagement and allow successful

counterattacks. No matter how far the Roman army’s tactics advanced, the general rule still

applied that a retreating army under pressure needed at some point either to dissuade its

pursuers or to find secure refuge.

814 One modern study of soldiers’ morale in battle demonstrates that the loss of junior leaders, who were the soldiers’ most intimate leaders, had a critical impact on morale and discipline (Rush, 2001, 323–4, 333–5, 343).


229

The Relationship of Retreats and Withdrawals to Roman Tactics and Strategy

Two unique and well–observed features of Republican Roman warfare are salient in

determining the relationship of retrograde movements to Roman tactics and strategy. The first

is the Romans’ longstanding tendency to adopt tactics from their neighbours and enemies.815

In the Middle Republic, it is striking that the Carthaginians appear to have regularly carried

out orderly retreats and withdrawals, or attempted to do so, and that in this period their tactics

were not dissimilar to those of the Romans.816 In the second and first centuries BC, some

Roman legionaries also developed an advance–and–retreat fighting style in simulation of the

fluid fighting styles they encountered against Spanish tribesmen.817 The defeats and reverses

suffered by Roman arms contributed to this tendency of adopting successful tactics and

strategies, and will have affected the Romans’ development of more effective tactics and

strategies for conducting retreats and withdrawals, such as increased use of foreign light

infantry and cavalry, use of different formations such as defensive circles or squares, and

implementing deceptive tactics when retreating from camps. The development of Roman

retreat tactics from the Middle to the Late Republic was therefore a fluid and adaptive

process, shaped by Rome’s military experiences throughout this period.

The second relevant feature of Republican Roman warfare is two–fold, the organisation and

fighting style of Roman legions, incorporating the army’s professional core, the centurionate

and rank–and–file veterani, and the triplex acies. As observed in Chapter Three (113–6), the

Romans utilised a slow and steady advance–and–retreat fighting style with their multiple

battlelines, and from the Middle Republic were already capable of using these lines as reserve

forces, to advance and support the front line (Ch. 6, 157–60). The slow–and–steady fighting

style of a fully deployed Roman army was always well–adapted to the maintenance of ratio

in retreat, the systematic use of reserve lines gave the army an ideal tool for supporting the

fighting troops if they retreated, and the officers that formed the army’s institutional memory,

particularly the centuriones, developed more advanced tactical knowledge and so were better

able to provide leadership and tactically–sound directions during a retreat. The Romans

815 Goldsworthy, 2011, 24, 29–31. 816 Cf. Polyb. 11.21, 24; Livy 25.37.15, 28.13–5, 30.18. Also see Cagniart, 2007, 92. Hoyos (2007, 76) points out that the superior skill of Scipio Africanus’ army during the Second Punic War was ‘a counter, and a tribute to the professionalism of Carthage’s land forces in the Barcid era’. 817 Caes. BCiv. 1.44; Potter, 2010, 328.


230

diversified their tactics over time, incorporating more cavalry and various kinds of missile–

armed light infantry into their forces, and as legionaries and junior officers became more

experienced, the Romans were able to build upon the tools they already had to carry out more

complex retreats and withdrawals in more difficult situations. Eventually, as Roman battle

tactics became more varied, maintaining the triplex acies was no longer necessary to support

a retreat, as good generals diligently deployed specific units of infantry or cavalry in support

of retreating troops.

The Romans thus built upon their existing fighting style and military structure to become

more effective at retreats and withdrawals, just as they became more effective in general. The

Romans were capable of carrying out orderly retreats and withdrawals from at least the late

third century BC, but their methods of doing so were not initially unique to them. For several

reasons, including the adoption of tactics from their enemies, accumulated experience from

previous defeats and the army’s general increase in efficiency and combined–arms

cooperation, from the Middle to the Late Republic the Romans developed a more advanced

capacity, in both form and function, to conduct successful retreats and withdrawals. In other

words, the Romans not only became more effective, but their tactics and strategies for

carrying out retreats and withdrawals also became more and more distinctively theirs, in line

with the increasingly unique characteristics of the army and its tactics. The superiority of

Roman retreat tactics will have been a significant strategic advantage, for they were more

capable of retreating from battlefield defeats and withdrawing from unfavourable positions,

and could thus better preserve their forces for successful action elsewhere. And so by the first

century BC, Roman retreat tactics had a significant impact on the methods by which the

Romans fought battles and conducted campaigns, and their general success in doing so. As

demonstrated by Marcus Antonius in 36 BC (Ch. 6, 208–9), a strong and cohesive Roman

army with a competent general could successfully retreat and withdraw in dire circumstances

and against the most formidable enemies. The Romans were able to retain these advanced

tactics not only through the institutional continuity of the Roman army, provided by

professional officers such as centuriones, but also their developing reservoir of written

material on warfare.818

818 Most such manuals are lost, but Frontinus (Strat. 2.13) and Onasander (Strat. 27) both discuss retreats and withdrawals, demonstrating that these were later considered to be manoeuvres that a good general should understand and perform correctly.


231

Retreats and Empire

No argument is necessary to state that the Roman army was the dominant military force in the

Mediterranean for hundreds of years, perhaps from the fall of the Successor kingdoms in the

second century BC until after Imperial field armies began suffering major defeats to the

Goths in the in the third and particularly, the fourth centuries AD. In this age of dominance,

the Romans were capable of fielding very large and efficient armies, and they rarely suffered

significant defeats in battle. Early in this period, the Roman army had become the most

successful exponent of retreat tactics in the Mediterranean. Thus, in the rare case of the defeat

the Romans could potentially recover from tactical reverses when their opponents could not,

reinforcing their already impressive array of advantages in battle.

The advantages of these superior retreat tactics are demonstrated in several incidents depicted

in Josephus’ account of the First Roman–Jewish War (AD 66–73). At the siege of Gamala in

AD 66, for example, Roman forces had initially penetrated deep into the city but were

eventually routed, leaving Vespasian and a body of men isolated (Joseph. BJ 4.31–4). These

Romans formed a testudo and resisted enemy attacks until the pressure slackened, and then

steadily retreated from the city while still facing the enemy (BJ 4.31–4). Also at Gamala, a

Roman officer named Placidius, who commanded six hundred cavalry, was able to entice a

group of Jewish fighters onto flat ground and carry out a successful feigned retreat, turning

and defeating his enemies once they had spread out in pursuit (4.58–60). Later, in AD 70 at

the outset of Titus’ siege of Jerusalem, an incident occurred in which the majority of a legion

was routed by a ferocious Jewish attack, leaving the general himself isolated with his

personal guard (5.76–7). Titus and his comrades beat back the Jews pressing upon them, and

when the enemy attempted to bypass him in pursuit of the rest of the legion, Titus

counterattacked their flank (5.81–2). The fleeing soldiers eventually rallied and

counterattacked, driving the Jews to flatter ground, where they attempted to make their own

fighting retreat, but were quickly overwhelmed as the Romans attacked from higher ground

(5.90–5).


232

In the extracts regarding Gamala and Jerusalem, the resilience of small groups of Roman

legionaries against more lightly–equipped foes is demonstrated, as these groups were able to

force back potential pursuers and retreat safely.819 In the case of Placidius, it can be see that

the Romans continued to use feigned retreat tactics with cavalry, just as they occasionally had

in the Republic.820 In the Jerusalem example, one can see the inability of Jewish infantry to

replicate effective retreat tactics, as after their first flight they attempted to rally, but when

attacked from higher ground they simply fled once more. With their lack of training, less

developed organisation and tactics, and lack of reserves, it would have been almost

impossible for these Jewish infantry to carry out a fighting retreat against an effective Roman

pursuit. It can therefore be seen that even in a war that featured few large field battles, the

Romans’ superior ability to carry out retreats meant that they could preserve their forces in

potentially disastrous situations, while Jewish fighters that suffered defeat in open ground

were not so fortunate.

Imperial forces continued to use the tactics developed in the Republican period. They used

heavy and light infantry cooperatively during battle, as notable for instance in the joint attack

of ferementarii and milites during the Romans’ final battle against the British chieftain

Caratacus, in AD 50 (Tac. Ann. 12.35.10). It is reasonable to assume that these separate arms

supported each other as closely during retreat and withdrawal as they did in attack. However,

just as Late Republican armies sometimes suffered reverses in retreat and withdrawal, such

also occured in the Imperial period. Prominent examples include Cestius Gallus’ disastrous

withdrawal through the pass of Beth–Horon in AD 66 (Joseph. BJ 2.540–50), and Caesennius

Paetus’ disorderly and costly retreat from Armenia following his defeat near Rhandeia in AD

62 (Tac. Ann. 15.16). As explained (Ch. 6, 217–20), difficult terrain and an unfavourable

balance of forces could prevent a successful retreat or withdrawal in any situation. As late as

the eleventh century AD, the Byzantine army at Manzikert demonstrated that the Romans had

not forgotten how to carry out complex and large–scale fighting retreats, utilising close order,

gradual movement and combined arms.821 The Romans were eventually routed at Manzikert,

primarily due to the fact that their reserves did not support the fighting line when it attempted

to counterattack.822 At Manzikert the Romans attempted to maintain ratio, to use reserves and

819 Cf. Sall. Jug. 97.5; App. Pun. 103. 820 See Ch. 1, 58–9. 821 Haldon, 1999, 178–9, 205. 822 Ibid., 205.


233

to make counterattacks in order to manage enemy pursuit. While the reserves did not perform

their role properly in this instance, the same principles of successful retreats observed in the

Middle and Late Republic evidently still applied to Roman retreat tactics, a millennium later.

Following the seeming culmination of the Roman army’s development of retreat tactics in the

first century BC, perhaps the most significant new challenge to the Roman army came from

the cavalry–based armies of Asia. As demonstrated (Ch. 6, 197, 199–200), the offensive use

of retreat tactics by the Parthians was already giving the Romans trouble in the first century

BC.823 The feigned retreat, a manoeuvre particularly effective with cavalry, could be

combined particularly effectively with mounted archery to keep a force safe from enemy

action whilst inflicting heavy casualties. Because feigned retreats were not common in

Roman tactics, they did not field mounted archers. The Romans quickly attempted to counter

such tactics with large numbers of local auxiliary cavalry, perhaps some of whom utilised

similar equipment and techniques.824

However, the Romans continued to encounter problems with the horse peoples of the

Eurasian Steppe, and in the final years of the Western Roman Empire the Huns repeatedly

demonstrated the incredible power of horse archery and feigned retreats in battle.825 Effective

use of such tactics could overwhelm an infantry–based army, even one well–practised in

carrying out fighting retreats or withdrawals. As a result, the Romans developed their own

horse archer and cataphract units with which they could utilise the same tactics, and they

continued to adapt these cavalry forces in response to the successful tactics of the steppe

peoples and Persians.826

823 Cf. Plut. Crass. 24.5–6, Ant. 41.1–4. 824 Plut. Ant. 37.3. Such auxiliaries could even include Parthians (App. B Civ. 4.59). 825 These were the typical tactics of the steppe (May, 2012, 130). As peoples from the steppe, the Huns were experts in mounted archery (Rance, 2007, 355). The Mongols further evolved the already–sophisticated tactics of the steppe (May, 2012, 130–1), and were all–but invincible in battle for an extended period of time. On Mongolian tactics and training under Genghis Khan see May (2006, 620–1, 622–3, 623–4, 632–5). The first significant battle that the Mongols lost (Ain Jalut, AD 1260) was against the Mamluks of Egypt, who themselves employed similar tactics (May, 2006, 625–7, 631). 826 Response to Parthians: Kennedy, 1996, 87–8. Response to Huns: Rance, 2007, 355. During the final Romano–Persian war of AD 602–628, the Emperor Heraclius spent some time training soldiers as horse archers before launching his decisive counterattack campaign (Gregory, 2009, 172).


234

As the steppe peoples continued intermittently to conquer European and Asian states for

centuries after the fall of the Western Empire, cavalry–based retreat tactics continued to play

a decisive role in warfare, but in a very different context to the experiences of the Roman

Republic. The Romans had painstakingly built their empire in the Western Mediterranean,

where battles between infantry–centric forces were the norm. In this context, the retreat

tactics that the Romans developed were probably never improved upon. The resulting ability

of Roman armies to preserve themselves even after serious defeats was surely a factor in the

army’s endurance during the myriad challenges it faced throughout the Later Empire.

Suggestions for Further Study

There are enough examples of retreats noted by ancient sources to show that the writers were

far from ignorant that such manoeuvres could play a significant role in battle. Given the

evidence, there is scope for further research on the subject. This thesis has only touched upon

the role of retreats and retreat tactics in Imperial Roman warfare, and it has said almost

nothing of Byzantine warfare. Although it would be logical to assume that the Romans

utilised their well–developed retreat tactics throughout the Later Empire, I have not tested

this assumption. I have also not examined in any depth the role of such tactics in the military

actions of Greeks, Macedonians and the Successors, the Carthaginians, or other

Mediterranean powers such as the Gauls and Iberians.827 Nor have I considered in detail the

extent to which Rome and these other powers influenced each other in their retreat tactics.

Nothing has been said of the legacy or new development of these tactics in Medieval warfare,

for although this would be outside the scope of ancient history, the technology and tactics of

the periods are comparable. I have only fleetingly touched upon the famous retreat tactics of

the horse peoples. Furthermore, even in the main focal period of this study, the Middle

Republic to Early Empire, there is room for deeper analysis than I have carried out. A closer

examination of links between changes in the structure of the Roman army and their retreat

tactics could be instructive. Finally, detailed studies of the countless battles and wars

throughout this period have usually been constructed without a significant appreciation of the

827 There are many examples of retreats and withdrawals by Greek and Hellenistic armies (eg. Thuc. 1.105–6; Xen. Hell. 5.3.3–7, 6.2.20–3, 6.5.13–4; Polyb. 2.68, 4.69, 8.14, 11.17; Diod. Sic. 19.43.4–5; App. Syr. 35, etc ). There were several recorded examples of retrograde movements by the Carthaginians in the Second Punic War (Polyb. 11.21, 24; Livy 28.13–5, 30.18; perhaps also Polyb. 15.13). There is a even an example of a fighting retreat by the Helvetii at Bibracte, 58 BC (Caes. BGall. 1.25).


235

important general role that retreats played in Roman–era warfare. In the light of my analysis,

it may be possible to review some aspects not only of Rome’s military history, but that of

other ancient states and peoples.

This thesis opened by examining the prevalent assumption that ancient battles usually ended

with the rout of one army. However, it is evident that the ways in which battles and

campaigns could end were more varied than has generally been recognised. Perhaps most

major battles ended with routs, but many others ended by mutual consent, in confused retreat

or with orderly withdrawal. Likewise, a campaign might climax with a major battle, but could

also end with a withdrawal from the theatre of operations, or surrender due to one army

having been manoeuvred into a position of supply shortage.

Roman armies were trained and organised to perform retreats and withdrawals. Officers knew

how to lead and manage such movements, and good generals will have considered retrograde

movements in their tactical and strategic thinking. The ability of Roman armies to retreat and

withdraw efficiently contributed to their military success by allowing them to escape from

difficult situations without suffering heavy casualties.

No army was invincible in battle. After a reverse at Dyrrhachium in 48 BC, Caesar (BCiv.

3.72) criticised premature celebration by the Pompeians thus:

... they did not recollect the common chances of warfare, how often trifling

causes, originating in a false suspicion, a sudden alarm, or a religious scruple,

have entailed great disasters, whensoever a mistake has been made in an army

through the incapacity of a general or the fault of a tribune ...828

What is important to recognise is that while defeats were common, disaster did not inevitably

follow. In the case of the Romans, disasters in the wake of defeat were perhaps the exception

rather than the rule.

828 Trans. A. G. Peskett, 1914.

236

Appendix A: Table of Basic References to Retreats and Withdrawals Following are the main primary source citations used throughout the thesis. This is not an exhaustive list of all references to retreats and withdrawals in Roman warfare. Other references may be found in some sources that I have not relied upon heavily, and some of the citations provided are difficult to interpret and classify, as retreats and the reasons for retreat, flight or disengagement from battle are not always described or stated by the sources. Some of the below examples of retreats are only inferred to have been such. There are also ancient references in the thesis that do not appear below. Nonetheless, these are the sources I have used for my central arguments.

Retreat from Battle Withdrawal

Successful

Polyb. 3.84, 6, 4.12, 11.24, 18.25; Diod. Sic. 19.43.4-5; Sall. Jug. 97-8; Caes. BGall. 1.25, 6.40, BCiv. 1.45-6, 2.25, 3.46. 3.94-5; BAfr. 15-8, BHisp. 31, BAlex.40; Livy 4.39.4-9 , 25.19.1-8, 28.13, 15, 33.15, 18, 42.59, 44.42.2-3; Plut. Aem. 20.1-6, Ant. 39.5-7; App. Pun. 103 Hisp. 88, BCiv. 2.80. Front. Strat. 1.5.9, 15-6.

Polyb. 3.74, 3.84, 5.86; Livy 4.39.4-9, 27.49, 32.12.2-3, 37.38; Plut. Ant. 41.4-5, 42.1-4, 45.2-3, 49.1-2; Tac. Ann. 15.16; App Hisp. 82; Joseph. BJ 2.540-555. Front. Strat. 1.5.1-8, 12-4, 17-8.

Unsuccessful

Polyb. 2.68, 4.69, 5.86, 11.17; Caes. BCiv. 2.41-2; BAlex. 76; Livy 6.24, 24.16.1-5, 27.49, 30.18, 32.12.2-3, 37.38, 38.25.10-13, 40.40; App. Syr. 6.35, BCiv. 1.90, 3.69, 4.128.

Livy 28.13, 15, 33.18; Caes. BGall. 5.26-31, 37, BCiv. 1.79, 3.94-5; Plut. Crass. 27.5-8, 29; Tac. Ann. 15.16; Joseph. BJ 2.540-555; App. B Civ. 2.80; Dio. 56.19-22.1.

237

Appendix B: Table of Retreats and Withdrawals According to Tactical Context Below are the same citations shown in Appendix 1, but organised according to the tactical context in which the retrograde movement was conducted. I have included examples of retreats from battle that were followed by the reformation of a battleline or a counterattack, even where such movements could also be described as routs, because the fleeing army did retain enough order to reorganise itself. Some citations are repeated under different headings as they fit multiple criteria.


Retreats to High Ground

Polyb. 3.40, 3.86, 11.24; Caes. BGall. 1.25, 2.23, BCiv. 1.45; BAlex 40; Livy 27.41-2, 28.15.8–11; Sall. Jug. 51, 98; App. B Civ. 1.90.

Retreats to Camps

Polyb. 11.21, 4; Caes. BGall. 5.37, 6.40, BCiv. 1.45-6, 2.42; BAlex. 76; Livy 6.24, 24.16, 28.13, 15, 33.15, 37.43, 38.25; App. Syr. 35-6, BCiv. 2.80, 3.69, 4.128-9.

Withdrawals from Camps

Caes. BGall. 2.10-1, BCiv. 1.61-4, 2.43; BAlex. 31; Livy 22.16-7, 28.15.12-16.8, 36.46.7.

Pursuits stopped or prevented by cavalry reserves

Caes BCiv. 1.46; Livy 42.59; App. Mith. 49.

Withdrawals under cover of fire/smoke

Sall. Jug. 75-7; BHisp. 27.4; Plut. Luc. 19; Front. Strat. 1.5.1-3, 8, 16.

238


Feigned retreats Caes. BGall. 6.7-8; Sall. Jug. 50; Livy 6.24, 29.34; Onas. Strat. 21.9.

Mutual retreats Polyb. 3.105; Livy, 10.35.1-3, 25.19.4-5.

Armies disengaging at nightfall/ withdrawing by night:

Caes. BGall 1.50; Livy 4.39.4-9, 9.23, 32, 10.12, 27.2, 38.40, 44.4; Sall. Jug. 98.1-5. Front. Strat. 2.1.13.

Caes. BCiv. 1.27-8; Livy 4.39.4-9, 10.12, 27.2; Plut. Crass. 27.3, 28.1; App. Hisp. 82.

Armies disengaging due to poor weather:

Livy 8.1.4, 28.15.11.

Retreats and Routs followed by rally or counterattack

Pol 3.86, 11.15, 18.22; Caes. BGall. 1.25, 2.23, 5.37, 7.80, BCiv. 1.45, 3.95; Sall. Jug. 51; Livy 27.49; App. Hisp. 58, Mith. 42-4, BCiv. 1.50, 1.90.

239

Appendix C: Table of Routs and Flights from Battle Following is a selection of accounts of routs and flights from battle, for the interest of readers who wish to compare accounts of retrograde movements and accounts of more chaotic flights. This list is by no means exhaustive.

Polybius Livy Caesar Sallust Appian

2.34, 68; 3.84; 4.12, 58, 68; 5.73, 69, 85; 10.12, 31; 11.14, 16-7; 11.21, 24; 14.8; 15.13; 18.25-6.

4.33.11-2; 5.38; 6.8.7, 13; 7.23.7-10; 9.39; 10.14.21, 39.7-9; 22.6, 49, 24.16, 39; 25.21, 34, 37, 39; 27.12, 14, 18, 42, 49; 28.2.9-12, 15.7-10; 29.3, 34; 30.18, 34; 31.37, 41; 32.12.2-3; 33.9-10, 15, 18; 33.36.11; 34.15, 28; 35.5; 36.38; 37.38, 42-3 ; 38.23, 27, 41; 39.31; 40.28, 32, 33, 40; 41.3; 42.59; 43.10; 44.12, 42.

BGall. 1.52-3, 2.11, 2.23, 4.12, 15, 6.8, 7.27-8, 51, 80, 8.13, 19; BCiv. 2.25, 34, 3.37, 46, 64, 69-71, 95; B. Alex. 40, 29, 31, 76.

Sall. Cat. 60; Jug. 38, 52, 54, 56, 58, 74, 99, 101.

BCiv. 1.50, 84, 2.62, 2.80, 3.69, 4.55, 4.128, 5.114. Hann. 7; Hisp. 56, 82; Mith. 42-4, 80, 85, 100; Pun. 102, 124-5; Syr. 35.

240

Appendix D: Detailed References to Retreats and Withdrawals by Chronology, Location and Result Find below a chronological list of the main references to retreats and withdrawals used throughout this thesis, with the name of the battle or campaign where known. Not all examples are Roman: I have included a small handful of Hellenistic examples because they are either referenced in the thesis, or because they are otherwise instructive. Again, this list is not exhaustive, but does include the main source material discussed throughout the thesis. c. 423 BC, Battle of Verrugo Reference: Livy 4.39.4-9. Result: An indecisive battle followed by a night retreat. c. 381 BC, Camillus’ campaign against Satricum Reference: Livy 6.24. Result: The Romans pursue when enemy make feigned retreat, become disordered and attempt their own retreat, but broke into flight. 316 BC, Battle of Gabiene Reference: Diod. Sic. 19.43.4-5. Result: The argyraspides fall back successfully in a square formation. 222 BC, Battle of Sellasia Reference: Polyb. 2.68. Result: The Spartan phalanx is pushed over the summit of a slope, then forced to retreat downhill and finally routed. 220 BC, Battle of Caphyae Reference: Polyb. 4.12. Result: An orderly retreat breaks down when anticipated support does not materialise, but the army escapes to safety in nearby towns. 218 BC, Battle of Stymphalus Reference: Polyb. 4.69. Result: The Eleans make an orderly retreat to higher ground while the troops did not despair of safety, but once they realised they were alone, their formation broke and was annihilated in flight. 218 BC, Battle of the Trebia Reference: Polyb. 3.74. Result: The Roman centre breaks that of Hannibal and marches through the gap, 10,000 men escape the battlefield and withdraw to Placentia.

241

217 BC, Battle of Raphia Reference: Polyb. 5.86. Result: The Seleucid phalanx is broken and flees in disorder to Raphia, but Antiochus is able to withdraw his army to friendly territory. 217 BC, Battle of Trasimene Reference: Polyb. 3.84, 6. Result: Like at Trebia, a large force of Roman infantry break the enemy to their front while the rest of the army is defeated. This force gets to nearby high ground and tries to withdraw, but is forced to surrender. A group of cavalry also escapes to a hill and surrenders. 217 BC, Hannibal’s operations near Casilinum Reference: Livy 22.16-7; Front. Strat. 1.5.28. Result: After being outmanoeuvred by Fabius Maximus, Hannibal decided to withdraw, and to avoid pursuit through nearby defiles he had bundles of sticks tied to oxen and set them alight, then used the mass of flaming animals as a distraction with which to withdraw his army. 216 BC, Carthaginian operations in Spain Reference: Livy 23.26.8-10 Result: The Carthaginian general Hasdrubal Gisco, alarmed by an enemy attack, leads his army onto a steep hill with a river to its front, and fortifies the position. 216 BC, Battle of Cannae Reference: Polyb. 3.115; Livy 22.47.4-10. Result: The Carthaginian centre deliberately retreats from a heavy Roman assault, drawing the Romans into a position where their flanks were open to counterattack. 214 BC, Battle of Beneventum Reference: Livy 24.16.1-5. Result: The Carthaginian line is forced back in confusion, eventually breaking into flight. 212 BC, Battle near Capua between Hannibal and the Consuls Reference: Livy 25.19.1-8. Result: Both armies disengage without incident. 212 BC, Roman counterattack after the Battle of the Upper Baetis Reference: Livy 25.37.15. Result: A Carthaginian force is driven back in confusion after a failed attack on the Roman camp, but when they saw there was no pursuit, they regained order and retreated slowly to their own camp. 211 BC, Hannibal attacks Q. Fulvius Flaccus (cos. 211) Reference: App. Hann. 41. Result: Hannibal attempts to trick the Romans into retreating from their camp to a nearby hill, but the Roman general prevents his army from doing so.

242

208 BC, Battle of Baecula References: Polyb. 10.39; Livy 27.18. Result: Scipio attacks Hamilcar Barca’s rearguard as the latter attempts to depart Spain to assist Hannibal in Italy, and despite some casualties the Carthaginians are able to withdraw successfully. 207 BC, Battle of Mantinea Reference: Polyb. 11.17. Result: Machinidas attempts to retreat with a body of men through surrounding enemies, but when their line of escape is blocked the men flee. 207 BC, Battle of Metaurus Reference: Livy 27.49. Result: The Carthaginians are pushed back and routed with heavy casualties, but some of the army withdraws or flees successfully as the Romans are too exhausted to pursue. 206 BC, Battle of Ilipa References: Polyb. 11.24; Livy 28.13, 15. Result: The Carthaginians attempt an orderly retreat but are overwhelmed and flee back to their camp, which they successfully defend. They then attempt to withdraw further, but their disordered column is caught by pursuers and destroyed. 203 BC, Mago Barca defeated in N. Italy Reference: Livy 30.18. Result: The Carthaginian army retreats in an orderly manner at first, but Mago falls wounded and the army flees. Much of the army apparently withdraws successfully to Africa as it reappears at Zama. 198 BC, Battle of the Aous Reference: Livy 32.12.2-3. Result: The Macedonians are beaten and part of the army flees, part is cut off from retreat, but the ground prevented effective pursuit and the Macedonians were able to reorganise many of their forces and withdraw. 197 BC, Battle of Cynoscephalae Reference: Polyb. 18.25 Result: The Roman left wing forced to slowly retreat under pressure of phalanx. 197 BC, Macedonians defeated near Corinth Reference: Livy 33.15.10-3. Result: A Macedonian army falls back with order at first, but under pressure they broke and suffered many casualties in pursuit. 197 BC, Macedonians defeated near Bargyliae Reference: Livy 33.18.17-8. Result: The Macedonian line retreats slowly, but when attempting to reform to face enemies

243

attacking their flank, disorder occurs and the Macedonians break and flee toBargyliae. 190 BC, Battle of Magnesia Reference: App. Syr. 6.35. Result: The Seleucid phalanx retreats with order at first, but is broken by the panic of elephants sheltering inside the phalanx. 190 BC, Seleucid and Roman cavalry skirmish prior to Magnesia Reference: Livy 37.38. Result: Seleucid cavalry attempt to retreat across a river while under pursuit by Roman cavalry. They suffer a number of casualties, but succeed in returning 4 miles to their camp. 189 BC, Battle of Mt. Olympus Reference: Livy 38.25.10-13. Result: The Roman consul and his guards are attacked by Galatians, and at first retire gradually, but eventually scatter in flight and suffer a number of casualties to pursuers. 181 BC, Battle of the Manlian Pass Reference: Livy 40.40. Result: Celtiberians looked for a place to retreat and found nothing, so they fled. Destroyed while attempting to escape through a defile. 171 BC, Battle of Callinicus Reference: Livy 42.59. Result: Cavalry and light infantry skirmish. Successful retreat, primarily supported by Thessalian cavalry. 168 BC, Battle of Pydna References: Livy 44.42.2-3; Plut. Aem. 20.1-6. Result: Successful retreat as the Romans cause the enemy formation to become disrupted, and then counterattack. 149 BC, Battle of Nephoris Reference: App. Pun. 103. Result: A successful retreat, facilitated by Scipio’s use of cavalry counterattacks. 136 BC, Withdrawal from the Siege of Pallantia Reference: App. Hisp. 82. Result: A disorderly but successful withdrawal, perhaps only due to the weakness of pursuing forces. 134 BC, Cavalry skirmish near Pallantia Reference: App. Hisp, 88. Result: Successful retreat by cavalry.

244

106 BC, Battle of the Muluccha Reference: Sall. Jug. 97-8. Result: A complex and successful retreat organised by Gaius Marius, partly done at night using both infantry and cavalry. 82 BC, Sullan ambush near Praeneste Reference: App. B Civ. 1.90. Result: The Marians are beaten in a sharp contest and try to take refuge on a hill, but fail. 58 BC, Battle of Bibracte Reference: Caes. BGall. 1.25. Result: A successful and long-distance retreat as Helvetian phalanx disengages and re-establishes their fighting line on a hill over a kilometre away. 54 BC, Battle of Atuatuca Reference: Caes. BGall. 5.26-31, 37. Result: Sabinus withdraws his army from winter camp and the Romans are destroyed in an ambush in a defile. 53 BC, Battle of Carrhae Reference: Plut. Crass. 27.5-8, 29. Result: Two withdrawals, one semi-successful but costly in men from the battlefield to Carrhae, and a disastrous one towards Armenia that saw the rest of the army destroyed. 53 BC, Q. Tullius Cicero defends his camp in Gaul Reference: Caes. BGall 6.40. Result: Successful retreat through encircling enemies, but heavy casualties suffered. 49 BC, Skirmishing near Utica Reference: Caes. BCiv. 2.25. Result: Curio falls upon a group of Numidian infantry and cavalry, killing over a hundred and driving the rest into their camp. 49 BC, Battle of Bagradas Reference: Caes. BCiv. 2.41-2. Result: Curio’s exhausted infantry attempt to retreat to nearby high ground, but fail and are destroyed. 49 BC, the Ilerdan Campaign Reference: Caes. BCiv. 1.45-6, 1.79. Result: Caesar retreats successfully from a dangerous position on the slope below Ilerda, after five hours of fighting. The Pompeians later make repeated attempts to withdraw to another location for logistical reasons, but are delayed and cut off repeatedly by Caesar and eventually forced to surrender.

245

48 BC, Battle of Dyrrachium Reference: Caes. BCiv. 3.46. Result: Caesar successfully organises the retreat of a legion from a heavily exposed position, in the face of enemy pursuit, thanks to support from archers and a successful counterattack. 48 BC, Battle of Nicopolis Reference: BAlex.40. Result: The Legio XXXVI is surrounded but successfully retreats to nearby high ground, and withdraws entirely, after the rest of the Roman army is destroyed. 48 BC, Battle of Pharsalus References: Caes. BCiv. 3.94-5; App. B Civ. 2.80. Result: The Pompeian army is beaten in battle and suffers many casualties, but apparently escapes to their own camp successfully. The camp is then overwhelmed by Caesar, and the army retreats to nearby hills with the aim of withdrawing to a nearby town, but Caesar prevents this and forces them to surrender. 47 BC, Battle of Zela Reference: BAlex. 76. Result: The Pontic phalanx attempts to retreat downhill after a failed assault, but is driven into flight and their camp is taken. 46 BC, Battle of Ruspina Reference: BAfr. 15-8. Result: Caesar organises a successful retreat against heavy and well-organised pursuers. He utilises an unconventional battleline, with cohorts facing alternately forward and backward, and also orders repeated counterattacks with both cavalry and infantry. 45 BC, Battle of Munda Reference: BHisp. 31. Result: The Pompeians are beaten and retreat successfully into Munda. 43 BC, Battle of Forum Gallorum Reference: App. B Civ. 3.69. Result: An inexperienced Roman force attempts to retreat from Antony’s victorious veterans, but is overwhelmed and routed. 42 BC, Second Battle of Philippi Reference: App. B Civ. 4.128. Result: The triplex acies of Brutus falls back under enemy pressure, and eventually falls into confusion and is routed. 36 BC, Antony’s Parthian Campaign References: Plut. Ant. 39.5-7, 41.4-5, 42.1-4, 45.2-3, 49.1-2. Result: Antony retreats successfully to his camp after a failed pursuit (39.5-7), and over a period of weeks makes a number of successful retreats and withdrawals in the face of constant pursuit

246

by Parthians. AD 9, the Clades Variana Reference: Dio. 56.19-22.1. Result: Varus is betrayed and ambushed, and attempts to withdraw towards the Rhine but is trapped and killed at Kalkriese, and his army destroyed. AD 62, Battle of Rhandeia Reference: Tac. Ann. 15.16. Result: Paetus allows his forces to become disorganised as they withdraw and they are heavily pursued by Parthians, suffering many casualties. The army is later rescued by Corbulo. AD 66, Battle of Beth-Horon Reference: Joseph. BJ 2.540-555. Result: Cestius Gallus successfully withdraws his army from Judea despite very heavy attacks in unfavourable ground, but suffers heavy casualties, loses his baggage, and has to sacrifice his rearguard.

247

Reference List Ancient Sources Appian. The Civil Wars. Translated by Horace White. Loeb Classical Library edition. New

York: Harvard University Press, 1913.

———. The Spanish Wars. Translated by Horace White. Loeb Classical Library edition. New


———. The Punic Wars. Translated by Horace White. Loeb Classical Library edition. New


———. The Hannibalic Wars. Translated by Horace White. Loeb Classical Library edition.

New York: Harvard University Press, 1913.

———. The Syrian Wars. Translated by Horace White. Loeb Classical Library edition. New


———. The Mithridatic Wars. Translated by Horace White. Loeb Classical Library edition.


Arrian. Anabasis. Translated by E. J. Chinnock. London: Hodder & Stroughton, 1884.

Accessed February 10, 2020. http://www.gutenberg.org/files/46976/46976-h/46976-

h.htm.

Asclepiodotus. Tactics. Translated by the Illinois Greek Club. Loeb Classical Library edition.


Bellum Africum. Translated by A.G. Way. Loeb Classical Library edition. New York: Harvard

University Press, 1955.

Bellum Alexandrinum. Translated by A.G. Way. Loeb Classical Library edition. New York:

Harvard University Press, 1955.

Bellum Hispaniensis. Translated by A.G. Way. Loeb Classical Library edition. New York:


Caesar. Bellum Gallicum. Translated by H.J. Edwards. Loeb Classical Library edition. New


———. Bellum Civile. Translated by A.G. Peskett. Loeb Classical Library edition. New


Cicero, Marcus Tullius. “Pro Caelio.” In M. Tulli Ciceronis Orationes, translated by Albert

Curtis Clark. Oxford: Oxford University Press, 1908.

———. “De Senectute.” In Cicero: De Senectute De Amicitia De Divinatione, translated by

William Armistead Falconer. London: Harvard University Press, 1923.

———. “Pro Lege Manilia.” In The Orations of Marcus Tullius Cicero, translated by C. D.

Yonge and B. A. London. London: Henry G. Bohn, 1856.

http://www.gutenberg.org/files/46976/46976-h/46976-h.htm

http://www.gutenberg.org/files/46976/46976-h/46976-h.htm

248

Dio, Lucius Cassius. Historia Romana (9 volumes). Translated by Earnest Cary. Loeb

Classical Library edition. New York: Harvard University Press, 1914–1927.

Dionysius of Halicarnassus, Roman Antiquities (7 volumes). Translated by Earnest Cary. Loeb

Classical Library edition. New York: Harvard University Press, 1937–1950.

Florus. The Epitome of Roman History. Translated by E.S. Forster. Loeb Classical Library

edition. New York: Harvard University Press, 1929.

Frontinus. Strategemata. Translated by Charles R. Bennett. Loeb Classical Library edition.


Josephus. History of the Jewish War. Translated by H. St. J. Thackeray. Loeb Classical

Library edition. New York: Harvard University Press, 1927.

Livy. Ab Urbe Condita Libri (14 volumes). Translated by B.O. Foster et al. Loeb Classical

Library edition. New York: Harvard University Press, 1919–2017.

Marcellinus, Ammianus. The Roman History. Translated by J.C. Rolfe. Loeb Classical

Library edition. New York: Harvard University Press, 1939–1950.

Maurice. Strategikon. Translated by George T. Dennis. Philadelphia: University of

Pennsylvania Press, 1984.

Maximus, Valerius. Facta et Dicta Memorabilia. Leipzig: Teuber, 1888.

Onasander. Strategikos. Translated by the Illinois Greek Club, edited by W.A. Oldfather. New


Orosius, Paulus. Historiarum Adversum Paganos, Edited by. C. Zangemeister. Accessed

October 20, 2019. http://www.attalus.org/latin/orosius.html.

Pliny. Naturalis Historia. Translated by John Bostock and H.T. Riley. London: Taylor &

Francis, 1855.

Plutarch. Aemilius. Translated by Bernadotte Perrin. Loeb Classical Library edition. New


———. Antony. Translated by Bernadotte Perrin. Loeb Classical Library edition.


———. Crassus. Translated by Bernadotte Perrin. Loeb Classical Library edition. New York:


———. Sulla. Translated by Bernadotte Perrin. Loeb Classical Library edition. New York:


Polybius. Historíai (6 volumes). Translated by W.R. Paton. Loeb Classical Library edition.

New York: Harvard University Press, 1922–1927.

Sallust. Bellum Jugurthinum. Translated by John C. Rolfe. Loeb Classical Library edition.


http://www.attalus.org/latin/orosius.html

249

Siculus, Diodorus. Library of History. Translated by C. H. Oldfather. Loeb Classical Library

edition. New York: Harvard University Press, 1933.

Suetonius. Augustus. Translated by Maximilian Ihm. Loeb Classical Library edition. New


———. Tiberius. Translated by Maximilian Ihm. Loeb Classical Library edition. New York:


Tacitus. Agricola. Translated by M. Hutton and W. Peterson. Loeb Classical Library edition.


———. Annales (3 volumes). Translated by J. Jackson. Loeb Classical Library edition. New

York: Harvard University Press, 1931–1937.

Thucydides. History of the Peloponnesian War. Translated by Rex Warner. Penguin Classics

edition. London: Penguin Books Ltd., 1972 (1954).

Velleius Paterculus, Marcus. Roman History. Translated by Frederick W. Shipley. Loeb

Classical Library edition. New York: Harvard University Press, 1924.

Vegetius, Publius Flavius. Epitome of Military Science: Second Revised Edition. Translated by

N.P. Milner. Liverpool: Liverpool University Press, 1997.

Xenophon. “Anabasis.” In Xenophon in Seven Volumes (vol. 1 and vol. 2), translated by

Carleton L. Brownson. Cambridge: Harvard University Press, 1918 (vol. 1), 1922

(vol. 2).

———. “Hellenica.” In Xenophon in Seven Volumes (vol. 3), translated by Carleton L.

Brownson. Cambridge: Harvard University Press, 1922.

———. “Memorabilia.” In Xenophon in Seven Volumes (vol. 4), translated by Carleton L.

Brownson. Cambridge: Harvard University Press, 1923.

Modern Sources

Anders, Adam O. “The ‘Face of Roman Skirmishing’.’’ Historia 64, no. 3 (2015), 263–300.

Anglim, Simon, Phyllis G Jestice, Rob S. Rice, Scott M. Rusch, and John Serrati. Fighting

Techniques of the Ancient World, 3000 BC - AD 500: Equipment, Combat Skills, and

Tactics. London: Amber Books, 2013 (2002).

“Auf den Spuren der Römerschlacht am Harzhorn.” NDR. Posted 7 January 2015.

https://www.ndr.de/geschichte/schauplaetze/Die-Roemerschlacht-am-

Harzhorn,roemerschlacht100.html.

Austin, N.J.E., and N.B. Rankov. Exploratio: Military and Political Intelligence in the Roman

World in the Second Punic War to the Battle of Adrianople. London: Routledge, 1995.

Baltrusch, E.P. “P. Quinctilius Varus und die bella Variana.” In 2000 Jahre Varusschlacht:

https://www.ndr.de/geschichte/schauplaetze/Die-Roemerschlacht-am-Harzhorn,roemerschlacht100.html

https://www.ndr.de/geschichte/schauplaetze/Die-Roemerschlacht-am-Harzhorn,roemerschlacht100.html

250

Geschichte—Archäologie—Legenden, edited by E. Baltrusch, M. Hegewisch, M.

Meyer, U. Puschner, and C. Wendt, 117–131. Berlin: De Gruyter, 2012.

Batstone, William W. Catiline’s Conspiracy, the Jugurthine War, Histories. Translated by

William W. Batstone. New York: Oxford University Press, 2010.

Bar-Kochva, Baz Alel. The Seleucid Army: Organisation and Tactics in the Great Campaigns.

Cambridge: Cambridge University Press, 1979 (1976).

Baynham, Elizabeth. Alexander the Great: The Unique History of Quintus Curtius. Ann

Arbor: University of Michigan Press, 1998.

Bell, M.J.V. “Tactical Reform in the Roman Republican Army”. Historia 14, no. 4 (1965):

404–422.

Bellón, Juan Pedro, Carmen Rueda, Miguel Ángel Lechuga, and María Isabel Moreno. “An

Archaeological analysis of a battlefield of the Second Punic War: the camps of the

battle of Baecula.” Journal of Roman Archaeology 29 (2016): 73–104.

Benario, Herbert W. “Bellum Varianum.” Historia 35, no. 1 (1986): 114–115.

Bosworth, A.B. Conquest and Empire: The Reign of Alexander the Great. Cambridge:

Cambridge University Press, 1993.

———. The Legacy of Alexander: Politics, Warfare, and Propaganda under the Successors.

New York: Oxford University Press, 2002.

Bugh, Glenn R. “Hellenistic Military Developments.” In The Cambridge Companion to the

Hellenistic World, edited by Glenn R. Bugh, 265–294. Cambridge: Cambridge


De Blois, Lukas. “Army and General in the Late Roman Republic.” In A Companion to the

Roman Army, edited by Paul Erdkamp, 164–180. Oxford: Blackwell Publishing Ltd.,

2011 (2007).

Cagniart, Pierre. “The Late Republican Army (146-30 BC).” In A Companion to the Roman

Army, edited by Paul Erdkamp, 80–95. Oxford: Blackwell Publishing Ltd., 2011

(2007).

———. “Strategy and Politics in Caesar’s Spanish Campaign, 49 B.C.: Variation on a Theme

by Clausewitz.” Ancient Warfare 26 (1995): 29–44.

Campbell, Brian. “Teach Yourself How to Be a General.” The Journal of Roman Studies 77

(1987): 13–29.

Carter, J. M. Julius Caesar: The Civil War, Book I & II. Warminster, Wiltshire: Aris & Philips

Ltd., 1991.

———. Julius Caesar: The Civil War, Book III. Warminster, Wiltshire: Aris & Philips Ltd.,

1993.

https://www.researchgate.net/journal/1047-7594_Journal_of_Roman_Archaeology

http://www.oxfordscholarship.com.ezproxy.newcastle.edu.au/view/10.1093/acprof:oso/9780198153061.001.0001/acprof-9780198153061

251

———. Appian: The Civil Wars. London: Penguin Books Ltd., 1996.

Clark, Jessica H. Triumph in Defeat: Military Loss and the Roman Republic. Oxford, New

York: Oxford University Press, 2014.

Clark, Jessica H., and Brian Turner, eds. Brill’s Companion to Military Defeat in Ancient

Mediterranean Society. Boston: Brill, 2017.

Connolly, Joy. “Virtue and violence: The historians on politics.” In The Cambridge

Companion to the Roman Historians, edited by Andrew Feldherr, 181–194. New

York: Cambridge University Press, 2009.

Crossland, D. “Unearthing the mysteries of the 'battle that created Germany.” DW. Posted 25

July 2017. https://www.dw.com/en/unearthing-the-mysteries-of-the-battle-that-

created-germany/a-39817362.

Cunliffe, Barry. The Ancient Celts. London: Penguin Books, 1999.

Daly, Gregory. Cannae: The Experience of Battle. London: Routledge, 2002.

Davidson, James. “Polybius.” In The Cambridge Companion to the Roman Historians, edited

by Andrew Feldherr, 123–136. New York: Cambridge University Press, 2009.

Dawson, Doyne. The Origins of Western Warfare: Militarism and Morality in the Ancient

World. Boulder: Westview Press, 1996.

Delbrück, Hans. History of the Art of War. Lincoln, Nebraska: University of Nebraska Press,

1920.

De Libero, Loretana. “Surrender in Ancient Rome.” In How Fighting Ends: A History of

Surrender, edited by Holger Afflerbach and Hew Strachn, 29–38. New York: Oxford


De Ligt, Luuk. Peasants, Citizens and Soldiers: Studies in the Demographic History of

Roman Italy 225 BC–AD 100. Cambridge: Cambridge University Press, 2012.

Dobson, Mike. The Army of the Roman Republic: The Second Century BC, Polybius, and the

Camps at Numantia, Spain. Oxford: Oxbow Books, 2008.

Eck, Werner. The Age of Augustus, 2nd Edition. Victoria: Blackwell Publishing, 2007 (1998).

Eckstein, Arthur M. “Bellicosity and Anarchy: Soldiers, Warriors and Combat in Antiquity.”

The International History Review 27, no. 3 (2005): 481–497.

———. Mediterranean Anarchy, Interstate War, and the Rise of Rome. Berkeley: University

of California Press, 2009 (2006).

Edwards, H.J. Caesar: The Gallic War. London: Harvard University Press, 1917.

Engels, Donald W. Alexander the Great and the Logistics of the Macedonian Army. Los

Angeles: University of California Press, 1978.

Erskine, Andrew. “How to Rule the World: Polybius Book 6 Reconsidered.” In Polybius and

https://brill.com/view/title/31960


https://www.dw.com/en/unearthing-the-mysteries-of-the-battle-that-created-germany/a-39817362

https://www.dw.com/en/unearthing-the-mysteries-of-the-battle-that-created-germany/a-39817362

252

his World: Essays in Memory of F.W. Walbank, edited by Bruce Gibson and Thomas

Harrison. New York: Oxford University Press, 2013.

Formisano, Marco. “Fragile expertise and the authority of the past: the ‘Roman art of war’.”

In Authority and Expertise in Ancient Scientific Culture, edited by Jason König, 129–

52. Cambridge: Cambridge University Press, 2017.

Friday, Karl F. Samurai, Warfare and the State in Early Medieval Japan. New York:

Routledge, 2004.

Fuller, J.F.C. Julius Caesar: Man, Soldier and Tyrant. New Brunswick: Rutgers University

Press, 1965.

Gaertner, Jan Felix. “The Corpus Caesarianum.” In The Cambridge Companion to the

Writings of Julius Caesar, edited by Luca Grillo and Christopher B. Krebs, 263–276.

Cambridge: Cambridge University Press, 2017.

Geiger, Joseph. “The Project of the Parallel Lives, Plutarch’s Conception of Biography.” In A

Companion to Plutarch, edited by Mark Beck, 292–303. Chichester, West Sussex:

Wiley-Blackwell, 2014.

Gibson, Bruce, and Thomas Harrison, eds. Polybius and his World: Essays in Memory of F.W.

Walbank. New York: Oxford University Press, 2013.

Gichon, Mordechai. “Cestius Gallus’ Campaign in Judea.” Palestine Exploration Quarterly

113 (1981): 39–62.

Gilliver, Catherine, Adrian Goldsworthy, and Michael Whitby. Rome at War: Caesar and his

Legacy. Oxford: Osprey, 2005.

Gilliver, Catherine. “Battle.” In The Cambridge History of Greek and Roman Warfare, Vol. II,

Edited by Philip Sabin, Michael Whitby, and Hans van Wees, 122–141. London:


Gleason, Bruce P. “Cavalry Trumpet and Kettledrum Practice in the Time of the Celts and

Romans to the Renaissance.” The Galpin Society Journal 61 (2008): 231–239.

Goldberg, Charles. “Decimation in the Roman Republic.” The Classical Journal 111, no. 2

(2015): 141–64.

Goldsworthy, Adrian. The Roman Army at War: 100 BC- AD 200. New York: Oxford


———. The Complete Roman Army. Melbourne: Thames & Hudson, (2003) 2011.

———. In the Name of Rome: The Men Who Won the Roman Empire. London: Phoenix,

(2003) 2004.

———. The Fall of Carthage. London: Weidenfeld & Nicolson, 2006.

———. Caesar. London: Weidenfeld & Nicolson, 2006.

253

Goodyear, F.R.D. The Annals of Tacitus, Volume II (Annals 1.55–81 and Annals 2).

Cambridge: Cambridge University Press, 1981.

Gregory, Timothy E. A History of Byzantium, 2nd ed, Blackwell History of the Ancient World.

Oxford: Blackwell Publishing, 2009.

Grethlein, Jonas. Experience and Teleology in Ancient Historiography: ‘Futures Past’ in

Herdotus to Augustine. New York: Cambridge University Press, 2013.

Großkopf, B., A. Rost, and S. Wilbers-Rost. “The Ancient Battlefield at Kalkriese.” In

“Sickness, Hunger, War and Religion: Multidisciplinary Perspectives”. Edited by

Michaela Harbeck, Kristin von Heyking, and Heiner Schwarzberg. 91–111. RCC

Perspectives 2012, no 3. doi.org/10.5282/rcc/5593

Haldon, John. Warfare, State and Society in the Byzantine World, 565–1204. New York:

Routledge, 1999.

———. The Byzantine Wars. Brimscombe: The History Press, 2008.

Hammond, N.G.L. “The Battle of Pydna.” The Journal of Hellenic Studies 104 (1984): 31–47.

Hanson, V.D. The Western Way of War: Infantry Battle in Classical Greece. London: Hodder

and Stoughton, 1989.

Harris, William V. War and Imperialism in Republican Rome, 327–70 BC. Oxford: Clarendon

Press, 1979.

Harrington, D. “Cassius Dio as a Military Historian.” Acta Classica 20 (1977): 159–65.

Heather, Peter. Empires and Barbarians: Migration, Development and the Birth of Europe.

London: Pan Books, 2010 (2009).

Heckler, Waldemar. The Conquests of Alexander the Great. Cambridge: Cambridge


Hoyos, Dexter. Hannibal’s Dynasty: Power and Politic s in the Western Mediterranean, 247–

183 BC. New York: Routledge, 2005 (2003).

———. “The Age of Overseas Expansion.” In A Companion to the Roman Army, edited by

Paul Erdkamp, 63–79. Oxford: Blackwell Publishing Ltd., 2011 (2007).

Hoyos, Dexter. The Carthaginians. New York: Routledge, 2010.

James, Simon. Exploring the World of the Celts. London: Thames & Hudson, 2005.

Junkelmann, Marcus. Die Legionen Des Augustus: Der römische Soldat im archäologischen

Experiment. Mainz am Rhein: Verlag Philipp von Zanbern, 1986.

Kagan, Kimberly. The Eye of Command. Ann Arbor: University of Michigan Press, 2006.

Keegan, John. The Face of Battle. New York: Random House, 1976.

Kennedy, David, ed. The Roman Army in the East: Journal of Roman Archaeology,

Supplementary Series 18. Ann Arbor: Crushing-Malloy Inc., 1996.

https://doi.org/10.5282/rcc/5593

254

Keppie, Lawrence. “The Making of the Roman Army.” In Republic to Empire. Norman:

University of Oklahoma Press, 1998 (1984).

Kestemont, Mike, Justin Stover, Moshe Koppel, Folgert Karsdorp, and Walter Daelemans.

“Authenticating the writings of Julius Caesar.” Expert Systems with Applications 63

(2016): 86–96.

Koon, Sam. Infantry Combat in Livy’s Battle Narratives. Oxford: Archaeopress, 2010.

Lee, A.D. “Morale and Roman Experience of Battle.” In Battle in Antiquity, edited by A.B.

Lloyd, 199–217. Swansea: Classical Press of Wales, 1996.

Lendon, J.E. “The Rhetoric of Combat: Greek Military Theory and Roman Culture in Julius

Caesar’s Battle Descriptions.” Classical Antiquity 18, no. 2 (1999): 273–329.

———. Soldiers and Ghosts: A History of Battle in Classical Antiquity. New Haven: Yale


———. “War and society.” In The Cambridge History of Greek and Roman Warfare, Vol. I:

Greece, the Hellenistic World and the Rise of Rome, edited by Philip Sabin, Michael

Whitby, and Hans van Wees, 498–516. London: Cambridge University Press.

———. “Historians without history: Against Roman historiography.” In The Cambridge

History to the Roman Historians, edited by Andrew Feldherr, 41–61. Cambridge:


———. “Battle Description in the Ancient Historians, Part II: Speeches, Results and Sea

Battles.” Greece & Rome 64, no. 2 (2017): 145–167.

Levene, D.S. Livy on the Hannibalic War. New York: Oxford University Press, 2010.

Levithan, Josh. Roman Siege Warfare. Ann Arbor: University of Michigan Press, 2013.

Lienemann, J. “Der Oberesch am Kalkrieser Berg.” In Kalkriese-Römer im Osnabrücker

Land: Archäologische Forschungen zur Varusschlacht, edited by W. Schlüter, F.

Berger, H. Buck, and U. Dieckmann, 73–79. Bramsche, 1993.

Lieven, Dominic. Russia Against Napoleon: The Battle for Europe, 1807 to 1814. London:

Penguin Books Ltd., 2016.

Lloyd, A.B. “Philip II and Alexander the Great: The Moulding of Macedon’s Army.” In Battle

in Antiquity, edited by A.B. Lloyd, 199–217. Swansea: Classical Press of Wales, 1996.

Manuwald, B. “Politsches Ungeshick oder vorbestimmes Verhängnis? Cassius Dios Bericht

über die Varusschlacht.” In Romische Prasenz und Herrschaft im Germanien der

Auguseischen Zeit, edited by G. Lehmann and R. Wiegels, 431–449. Göttingen:

Vandenhoeck & Ruprecht, 2007.

May, Timothy. “The Training of an Inner Asian Nomad Army in the Pre-Modern Period.” The

Journal of Military History 70, no. 3 (2006): 617–635.

255

———. The Mongol Conquests in World History. London: Reaktion Books, 2012.

McCall, Jeremiah. The Cavalry of the Roman Republic: Cavalry Combat and Elite

Reputations in the Middle and Late Republic. London: Routledge, 2001.

McDonnell, Myles. Roman Manliness: Virtus and the Roman Republic. New York:


Milne, Kathryn H. “The Republican Soldier: Historiographical Representations and Human

Realities.” Publicly Accessible Penn Dissertations. Paper 64. Accessed April 7, 2020.

http://repository.upenn.edu/edissertations/64.

Moosbauer, G., and S. Wilbers-Rost. “Kalkriese—Ort der Varusschlacht?” In Die

Varusschlacht: Wendepunkt der Geschichte?, edited by R. Weigels, A. Becker, J.

Kühlborn, G. Moosbauer, G. Rasbach, and S. Wilbers-Rost, 23–36. Stuttgart: 2007.

Müller-Scheeßel, N. “Die Fundverteilungen von Kalkriese Oberesch im Rahmen einer

Schlachtfeldarchäologie: Neue Aspekte zur Interpretation des Fundplatzes.” In

Ethnographisch-Archäologische Zeitschrift 53, no. 1/2 (2012): 108–121.

Murdoch, Adrian. Rome’s Greatest Defeat: Massacre in the Teutoburg Forest.

Gloucestershire: The History Press, 2008 (2006).

Oakley, Stephen P. A Commentary on Livy, Books VI to X, vols I–IV. New York: Oxford

University Press, 1997–2005.

Ogilvie, Robert Maxwell. A Commentary on Livy: Books 1–5. New York: Oxford University

Press, 1965.

Östenberg, Ida. “Defeated by the Forest, the Pass, the Wind: Nature as an Enemy of Rome.”

In Brill’s Companion to Military Defeat in Ancient Mediterranean Society, edited by

Jessica H. Clark and Brian Turner. Boston: Brill, 2017.

Overtoom, Nikolaus Leo. “The Parthian Rival and Rome’s Failure in the East: Roman

Propaganda and the Stain of Crassus.” Acta Antiqua 57, no 4 (2018): 1–22.

Paul, G.M. A Historical Commentary on Sallust’s Bellum Jugurthinum. Liverpool: Francis

Cairns, 1984.

Pelling, C.B.R., ed. Plutarch, Life of Antony. New York: Cambridge University Press, 1988.

Phang, Sara E. Roman Military Service: Ideologies of Discipline in the Late Republic and

Early Principate. New York: Cambridge University Press, 2008.

The Packhard Humanities Insitute. PHI Latin Database. Accessed 2019-2020.

https://latin.packhum.org/browse.

Potter, David. “Caesar and the Helvetians.” In New Perspectives on Ancient Warfare, History

of Warfare Series, vol. 59. Brill: 2010.

———. “3: The Roman Army and Navy.” In The Roman Republic: Second Edition, edited by

http://repository.upenn.edu/edissertations/64


https://latin.packhum.org/browse

256

Harriet I. Flower, 54–77. New York: Cambridge University Press, 2014.

Quesada-Sanz, F. “Not So Different: Individual Fighting Techniques and Small Unit Tactics

of Roman and Iberian Armies.” In “L’Hellenisation en mediterranee occidentale au

temps des guerres puniqes”, Actes du Colloque International de Toulouse, edited by P.

Francois, P. Moret, and S. Pere-Nogues. Pallas 70 (2006): 1–25.

Rance, Philip. “Narses and the Battle of Taginae (Busta Gallorum 552: Procopius and Sixth-

Century Warfare.” Historia 54, no. 4 (2005): 424–472.

Rankov, Boris. “Chapter 2: Military forces.” In The Cambridge History of Greek and Roman

Warfare, Vol. II: Rome In the Late Republic to the Late Empire, edited by Philip Sabin,

Michael Whitby, and Hans van Wees, 30–75. London: Cambridge University Press,

2007.

Rawlings, Louis. “Army and Battle During the Conquest of Italy.” In A Companion to the

Roman Army, edited by Paul Erdkamp, 45–62. Oxford: Blackwell Publishing Ltd.,

(2011) 2007.

Rauh, Stanly H. “The Tradition of Suicide in Rome’s Foreign Wars”. In Tapa 145, no. 2

(2015): 383–410.

Richardson, J.S. Appian: Wars of the Romans in Iberia. Warminster: Aris & Phillips Ltd.,

2000.

Riggsby, Andrew M. Caesar in Gaul and Rome: War in Words. Austin: University of Texas

Press, 2006.

Roller, Matthew. “The exemplary past in Roman historiography and culture.” In The

Cambridge Companion to the Roman Historians, edited by Andrew Feldherr, 214–

230. New York: Cambridge University Press, 2009.

Rosenstein, Nathan. Imperatores Victi: Military Defeat and Aristocratic Competition in the

Middle and Late Republic. Los Angeles: University of California Press, 1990.

———. “Aristocratic Values.” In A Companion to the Roman Army, edited by Nathan

Rosenstein and Robert Morstein-Marx, 365–382. Oxford: Blackwell Publishing Ltd.,

2011 (2010).

Rost, A. “The site of the Varus battle at Kalkriese. Recent results In Archaeological

Research.” In Limes XX: XX Congreso Internacional, vol. 3, edited by A. Morillo, N.

Hanel, and E. Martin, 1339–1346. Madrid: Consejo Superior de Investigaciones

Científicas, Polifemo, 2009.

———. “Remains of the Roman Baggage Train at the Battlefield of Kalkriese.” In Roman

Frontier Studies 2009: Proceedings of the XXI International Congress of Roman

Frontier Studies, edited by N. Hodgson, P. Bidwell, and J. Schachtmann, 559–564.

257

Oxford: Archaeopress Archaeology, 2016.

Rost, Achim, and Susanne Wilbers-Rost. “Weapons at the Battlefield of Kalkriese.” Gladius

30 (2010): 117–136.

Roth, Jonathon. The Logistics of the Roman Army at War (264 BC - AD 235). Boston: Brill,

1999.

Rush, Robert Sterling. Hell in Hürtgen Forest: The Ordeal and Triumph of an American

Infantry Regiment. Lawrence: University Press of Kansas, 2001.

Sabin, Philip. “The Face of Roman Battle.” The Journal of Roman Studies 90 (2000): 1–17.

———. “Land Battles.” In The Cambridge History of Greek and Roman Warfare, Vol. I:


Whitby, and Hans van Wees, 399–433. London: Cambridge University Press, 2007.

Sage, Michael M. The Republican Roman Army: A Sourcebook. New York: Routledge, 2008.

Salmon, E.T. The Making of Roman Italy. Ithaca: Cornell University Press, 1982.

Sampson, Gareth C. The Defeat of Rome in the East. Philadelphia: Casemate, 2015 (2008).

Schettino, Maria. “The Use of Historical Sources.” In A Companion to Plutarch, edited by

Mark Beck, 417–436. Chichester, West Sussex: Wiley-Blackwell, 2014.

Schlüter, Wolfgang. “Archäologische Zeugnisse zur Varusschlacht? Die Untersuchungen in

der Kalkrieser-Niewedder Senke bei Osnabrück.” Germania 70 (1992): 307–402.

———. “Die archäologischen Untersuchungen in der Kalkrieser-Niewedder Senke.” In

Kalkriese - Römer im Osnabrücker Land: Archäologische Forschungen zur

Varusschlacht, edited by W. Schlüter, F. Berger, H. Buck, and U. Dieckmann, 13–51.

Bramsche: Rasch, 1993.

———. “The Battle of the Teutoburg Forest: Archaeological research at Kalkriese near

Osnabrück.” In Roman Germany: Studies in Cultural Interaction, edited by J.D.

Creighton and R.J.A. Wilson, 125–159. Dexter, Michigan: Thomson-Shore Inc., 1999.

Seager, Robin. “‘Polybius’ Distortions of the Roman ‘Constitution’: A Simpl(istic)

Explanation’.” In Polybius and his World: Essays in Memory of F.W. Walbank, edited

by Bruce Gibson and Thomas Harrison, 247–254. New York: Oxford University Press,

2013.

Sekunda, “Land forces.” In The Cambridge History of Greek and Roman Warfare, Vol. I:


Whitby, and Hans van Wees, 325–356. London: Cambridge University Press.

Serrati, John. “Warfare and the state.” In The Cambridge History of Greek and Roman

Warfare, Vol. I: Greece, the Hellenistic World and the Rise of Rome, edited by Philip

Sabin, Michael Whitby, and Hans van Wees, 461–497. London: Cambridge University

258

Press.

———. “The Hellenistic World at War, Stagnation or Development.” In The Oxford

Handbook of Warfare in the Classical World, 179–198. Oxford: Oxford University,

Press, 2013.

Slavik, Jordan F. “The Roman Infantryman’s Style of Combat in the Middle Republic.” CJ

113, no. 2 (2018): 151–171.

Sheldon, Rose Mary. “Slaughter in the Forest: Roman Intelligence Mistakes in Germany.”

Small Wars & Insurgencies 12, no. 3 (2001): 1–38.

Sidebottom, Harry. Ancient Warfare: A very short introduction. Oxford: Oxford University

Press, 2004.

Southern, Pat. The Roman Army: A Social & Institutional History. New York: Oxford


Sommer, Michael. “Scipio Aemilianus, Polybius, and the Quest for Friendship in Second-

Century Rome.” In Polybius and his World: Essays in Memory of F.W. Walbank, edited

by Bruce Gibson and Thomas Harrison, 308–318. New York: Oxford University Press,

2013.

Spring, Matthew H. With Zeal and Bayonets Only: The British Army on Campaign in North

America, 1775-1783. Norman: University of Oklahoma Press, 2010 (2008).

Stark, F. “Legionäre konnten wohl der Varusschlacht entfliehen.” Welt. Posted 9 June 2016.

https://www.welt.de/geschichte/article156088468/Legionaere-konnten-wohl-der-

Varusschlacht-entfliehen.html. 2016.

Strachn, Hew. “Training, Morale and Modern War.” Journal of Contemporary History 41, no.

2 (2006): 211–227.

Suolahti, Jaakko. The Junior Officers of the Roman Army in the Republican Period: A Study

on Social Structure. Helsinki: Weilin & Göös, 1955.

Swan, Peter Michael. The Augustan Succession; An Historical Commentary on Cassius Dio’s

Roman History Books 55–56 (9 B.C.-A.D. 14). New York: Oxford University Press,

2004.

Syme, Ronald. The Roman Revolution. Oxford: Oxford University Press, 1960, reprinted

2002.

Tarn, William. The Greeks in Bactria and India. Cambridge: Cambridge University Press,

1951.

Thorne, James. “Battle, Tactics and the limites in the West.” In A Companion to the Roman

Army, edited by Paul Erdkamp, 218–234. Oxford: Blackwell Publishing Ltd., 2011

(2007).

https://www.oxfordscholarship.com/view/10.1093/acprof:oso/9780199608409.001.0001/acprof-9780199608409-chapter-17

https://www.oxfordscholarship.com/view/10.1093/acprof:oso/9780199608409.001.0001/acprof-9780199608409-chapter-17

https://www.welt.de/geschichte/article156088468/Legionaere-konnten-wohl-der-Varusschlacht-entfliehen.html.%202016

https://www.welt.de/geschichte/article156088468/Legionaere-konnten-wohl-der-Varusschlacht-entfliehen.html.%202016

https://fi.wikipedia.org/wiki/Weilin%2BG%C3%B6%C3%B6s

259

Thornton, John. “Polybius in Context: The Political Dimension of the Histories.” In Polybius

and his World: Essays in Memory of F.W. Walbank, edited by Bruce Gibson and

Thomas Harrison, 213–229. New York: Oxford University Press, 2013.

Timpe, D. Arminius-Studien. Heidelberg: Winter, 1970.

———. “Geographische Faktoren und politische Entscheidungen in der Geschichte der

Varuszeit.” In Arminius und die Varusschalcht: Geschichte—Mythos—Literatur, edited

by R. Wiegels and W. Woesler, 13–27. Paderborn, 1995.

Todd, Malcolm. The Early Germans, 2nd ed. Oxford: Blackwell Publishing, 2004.

US Army. Field Manual 7-20, The Infantry Battalion, Chapter 5: Retrograde Operations.

Accessed March 5, 2017.

http://www.globalsecurity.org/military/library/policy/army/fm/7-20/Ch5.htm.

Van Wees, Hans. Greek Warfare: Myths and Realities. London: Bloomsbury Academic, 2004.

Von Schnurbein, S. “Augustus in Germania and his new ‘town’ at Waldgirmes east of the

Rhine.” Journal of Roman Archaeology 16, no. 1 (2003): 93–107.

Walbank, F.W. A Historical Commentary on Polybius, vols I & II. Oxford: Clarendon Press,

1957–79.

———. “A Greek Looks at Rome: Polybius VI Revisited.” In Polybius, Rome and the

Hellenistic World: Essays and Reflectios, edited by F. W. Walbank, 277–292. New

York: Oxford University Press, 2009.

Ward, Allen. Marcus Crassus and the Late Roman Republic. London: University of Missouri

Press, 1977.

Ward-Perkins, B. The Fall of Rome: And the End of Civilisation. New York: Oxford


Walsh, P.G. Livy: His Historical Aims and Methods. New York: Cambridge University Press,

1961.

Waterfield, Robin. Taken at the Flood: The Roman Conquest of Greece. Oxford: Oxford


Welch, Kathryn. “Caesar and his officers in the Gallic War commentaries.” In Julius Caesar

as Artful Reporter: The War Commentaries as Political Instruments, edited by

Kathryn Welch and Anton Powell, 85–110. Swansea: The Classical Press of Wales,

1998.

Wheeler, Everett. (2001) “Firepower: Missile Weapons and the ‘Face of Battle’.” In Roman

Military Studies, edited by E. Dabrowa, 169–184. Krakow: Electrum 5, 2001.

———. “Land Battles.” In The Cambridge History of Greek and Roman Warfare, Vol. I:

http://www.globalsecurity.org/military/library/policy/army/fm/7-20/Ch5.htm

260


Whitby, and Hans van Wees, 186–223. London: Cambridge University Press, 2007.

Wilbers-Rost, S. “The Battlefield of Kalkriese: The Rampart at the Site ‘Oberesch’ During

and After the Battle.” In Roman Frontier Studies 2009: Proceedings of the XXI

International Congress of Roman Frontier Studies, edited by N. Hodgson, P. Bidwell,

and J. Schachtmann, 571–576. Oxford: Archaeopress Archaeology, 2016.

Winkler, Martin M. Arminius the Liberator: Myth and Ideology. New York: Oxford


Wolters, R. “Die Shlacht im Teutoburger Wald: Varus, Arminius und das römische

Germanien.” In 2000 Jahre Varusschlacht: Geschichte—Archäologie—Legenden,

edited by E. Baltrusch, M. Hegewisch, M. Meyer, U. Puschner, and C. Wendt, 3–21.

Berlin: De Gruyter, 2012.

Woodman, A.J. Velleius Paterculus: The Tiberian Narrative (2.94-131). Cambridge:


Zhmodikov, Alexander. “Roman Republican Heavy Infantrymen in Battle (IV-II Centuries

B.C.).” Historia 49, no. 1 (2000): 67–78.

Retreats and Withdrawals in Republican Roman Warfare

Documents

Transcript of Retreats and Withdrawals in Republican Roman Warfare