1. For recent studies on Roman cavalry, see: Hyland, A., Equus: the Horse in the Roman World (London, 1990)Google Scholar, part 2, esp. chapters 10, 11, 14; Training the Roman Cavalry (Stroud, 1993)Google Scholar; Connolly, P. and Driel-Murray, C. van, ‘The Roman Cavalry Saddle’, Britannia 22 (1991), 33–50CrossRefGoogle Scholar; Dixon, K. and Southern, P., The Roman Cavalry (London, 1992)Google Scholar. I would like to thank Professor John Wilkes for reading an early draft of this paper and making numerous valuable suggestions.

2. Tacitus, , Agricola 35Google Scholar: ingens victoriae decus citra Romanum sanguinem bellandi.

3. Cheesman, G. L., The Auxilia of the Roman Imperial Army (Oxford, 1914), 104Google Scholar, repeated by H. Furneaux and J. G. C. Anderson, Comelii Taciti de Vita Agricolae (Oxford, 1922), 134.

4. Campbell, B., ‘Teach Yourself how to be a General’, JRS 77 (1987), 13–29Google Scholar. Though by the midfirst century A.D., all auxiliaries are potential citizens themselves.

5. Burn, A. R., Agricola and Roman Britain (London, 1953), 152Google Scholar; Hyland, , op. cit. (1990), 166Google Scholar sees the auxiliaries as simply ‘cannon-fodder’. W. Liebeschuetz has repeated the argument most recently, ‘The End of the Roman Army’, in War and Society in the Roman World,. Rich, J. and Shipley, G. (edd.) (London, 1993), 268Google Scholar. When discussing the use of barbarian troops in the late Roman army, he states that ‘it was, of course, nothing new that the Romans should assign to auxilia of one kind or another the most exposed position in battle, and thus preserve the lives of legionaries’. In a recent newspaper report, the director of excavations at the Roman fort at Baginton, Coventry, blamed ‘ill-trained foreign auxiliary soldiers’ for the construction of an apparently senseless curved wall which had been excavated, The Daily Mail, August 23rd 1995. The assumption is that well-trained and disciplined (Roman citizen) legionaries would have done a proper job.

6. Ogilvie, R. M. and Richmond, I. A. (edd.), Cornelii Taciti De Vita Agricolae (Oxford, 1967), 272Google Scholar. The examples given are Annals 12. 17.2 and Germania 33.1. However, of these two examples, the latter expresses the pleasure of seeing German tribes fighting amongst themselves rather than against the Romans, so Roman soldiers are not involved at all, and in the first example there were no casualties because the enemy surrendered (to a Roman army that seems to have consisted almost entirely of auxiliary units). The scenes from Trajan's Column which appear to show these tactics are 24, 66, and 72 (as in note 42). Earlier commentators on the Agricola do not seem to see anything out of the ordinary about Agricola's dispositions, or the reasons for them.

7. Annals 3. 39 and 14. 23.

8. Argued most recently by Speidel, M. P., ‘The Pay of the Auxilia’, JRS 63 (1973), 141–7Google Scholar and Speidel, M. A., ‘Roman Army Pay Scales’, JRS 82 (1992), 87–106Google Scholar.

9. Ps.-Hyginus, de Munilionibus Castrorum, § 1 and 25. M. Lenoir suggests this was because the auxiliaries did not camp with the same precision as the legionaries, and that legionary centurions were given a greater proportion of their century's allocation of space than the auxiliaries. He does not suggest that auxiliaries received less space because they were of lower grade than the legionaries (Pseudo-Hyginus, , des Fortifications du Camp [Paris, 1979], 43–4)Google Scholar.

10. Alston, R., ‘Roman Military Pay from Caesar to Diocletian’, JRS 84 (1994), 113–23Google Scholar.

11. Richmond, , ‘Gnaeus Iulius Agricola’, JRS 34 (1944), 34–45Google Scholar; Rainbird, J. S., ‘Tactics at Mons Graupius’, CR 19 (1969), 11–12Google Scholar. For Cerialis' battle against Civilis, see Histories 5. 14ff.

12. Annals 12.31.

13. Idistaviso (A. D. 16), Annals 2.16; Lucius Apronius (A.D. 23), Annals 4. 73.

14. This was the usual method of disposition throughout the Republic with the legions holding the centre, and the allies, then cavalry, on the wings. One of the principal responsibilities of the forces on the wings was to prevent and execute flank attacks. The same formation, using cohorts instead of maniples, is recorded from at least the time of Caesar: B.G. 1. 52, against Ariovistus, ; B.C. 3Google Scholar. 88–9 and Appian B.C. 2. 75ff, at Pharsalus, ; B. Hisp. 29–31Google Scholar, at Munda, where both sides deployed in this manner.

15. Annals 2. 52, against Tacfarinas; Annals 13. 38, against Tiridates, ; Annals 14Google Scholar. 34, against Boudicca. In the first two examples, no battle was fought, but this does not affect the argument since the troops were deployed in this formation, ready to engage if necessary.

16. Acies contra Alanos 1–9. Although the auxiliary cavalry would be used to prevent an outflanking manoeuvre and would lead the pursuit of the fleeing enemy, it is clear that Arrian intended his legionaries to bear the brunt of the battle, and therefore the casualties.

17. Onasander 15.

18. Epitoma Rei Militaris 3. 9.

19. Onasander 31. Arrian proposed to engage the Alani under specific topographical circumstances, fighting in a horned formation with both wings held by detachments stationed on rising ground. He does not say what he intended to do if the Alani decided to engage him where there was not a convenient pair of hills that he could anchor his battle line on, but the nature of the terrain is fundamental to the effectiveness of his dispositions. Similarly, one of Caesar's legates, Sabinus, wished to avoid battle with Viridovix, partly because of the size of the enemy army, ‘unless he was on favourable ground, or an advantageous situation arose’, B.G. 3. 17.

20. B. Hisp. 29.

21. Annals 1.63–8.

22. Annals 1. 68: exprobrantes non hic silvas nee paludes, sed aequis locis aequos deos. That is, fair for the Romans, who would of course have a major advantage on level ground because of their heavy infantry.

23. Epitoma Rei Militaris 3. 14.

24. B. Alex. 60.

25. See B.G. 5. 49, B.C. 2. 34, where battles are avoided or delayed because of the unfavourable nature of the terrain. The Roman setback at Gergovia is blamed on the unfavourable ground, B.G. 7. 53, and at Cremona Otho's battle line was broken up by the nature of the ground which was full of pits and trenches, Plutarch, , Otho 12Google Scholar. During the debate on the granting of a triumph, the proconsul Gnaeus Manlius Vulso was criticized for engaging his troops on unfavourable ground against the Celts in Galatia in 188 B.C. despite gaining an overwhelming victory, Livy 38. 46.

26. Arminius and Civilis, for example, had both served as commanders of auxiliary units and would have had personal experience of the strengths and weaknesses of the Roman forces.

27. Furneaux, , The Annals of Tacitus Vol. II (Oxford, 1896), 306Google Scholar.

28. Since the period of deployment from line of march to line of battle was one of the most vulnerable for any army, there was usually a close correlation between the two formations. There seems little doubt that Germanicus intended to engage these auxiliaries first.

29. Apronius must have seen the advantages of using light armed infantry whilst governor of Africa in A.D. 20, when he sent his son with cavalry, auxiliary infantry, and the most mobile legionaries (quis velocissimos legionum) against Tacfarinas and won a victory, possibly after taking the Numidians by surprise {Annals 3. 21).

30. See Hassall, M. W. C., ‘Batavians and the Roman Conquest of Britain’, Britannia 1 (1970), 131–6CrossRefGoogle Scholar.

31. Cheesman, G. L., op. cit., 104Google Scholar; cf. Onasander 18, where he advises the general to send light armed troops (probably slingers and archers) against the enemy, when their heavy infantry have deployed on high ground, because of their agility. Parker, H. M. D., The Roman Legions (Oxford, 1928), 258Google Scholar. Furneaux and Anderson consider Tacitus', explanation to be ‘unworthy’, op. cit., 134Google Scholar.

32. Bell, M. J. V., ‘Tactical Reform in the Roman Republican Army’, Historia 14. 1965, 404–22Google Scholar.

33. Appian, , B.C. 3. 24Google Scholar; Deiotarus', ‘legions’, B. Alex. 38–40Google Scholar. In 25 B.C. when Deiotarus bequeathed his kingdom to Rome, these two units were united to form the regular legion XXII Deiotariana.

34. B.G. 2. 24; his legate Crassus posted his auxiliary infantry in the centre of his line of battle between the legions because of lack of confidence in them, B.G. 3. 24.

35. B.Afr. 19.

36. B.Afr. 59–61; 71.

37. B.Afr. 78; they are usually referred to as expediti.

38. Carter, J., The Civil War, Books I and II(Warminster, 1991), 191Google Scholar.

39. E.g., B.C. 3. 75, where the antesignani are assigned to protect the rear-guard of Caesar's marching column from attack by Pompey's cavalry.

40. Aside from Vegetius, there are no references to antesignani in the literature of the Principate. There is some scrappy epigraphic evidence for legionary troops in Africa called antesignani; Cagnat, R., I'Armée romaine d'Afrique el I'occupation militaire de I'Afrique sous les Empereurs (Paris, 1913), 495Google Scholar; AE 1978, 471. Although Speidel is willing to argue that some ranks of the imperial legion were armed and trained for different tasks, The Framework of an Imperial Legion, the Fifth Annual Caerleon Lecture (1992), 14–21, the evidence is scanty and there is nothing to indicate that this was a continuous, or Empire-wide, situation.

41. In the Late Republic auxiliary units are referred to as expeditae cohortes, leves cohorles, and levis armatura. For the imperial period, Seneca refers to auxilia et armaturae leves, de Vila Beata 8. 2, and when describing Germanicus' forces at Idistaviso, Tacitus mentions various units including auxiliares Galli Germanique… levis armalura… ceieraeque sociorum cohortes, all of which seem to be auxiliary units, Annals 2. 16. Josephus, refers to some auxiliary units as όπλīται (B.J. 2. 263; 3. 110)Google Scholar and others as γυμνήται (B.J. 2. 508) or simply πεζοί (B.J. 3. 110), but provides no further information on the type or weight of their equipment. There is no suggestion in the literature of the Early Principate that auxiliary units were considered inferior militarily.

42. Trajan's Column, scenes 167 and 177 show slingers, though cf. scene 60, of an auxiliary clubman with no signs of body armour at all.

43. Bishop, M. C. and Coulston, J. N., Roman Military Equipment (London, 1993), 208Google Scholar. Both Livy and Plutarch make this contrast: see below 61–2. M. Feugere agrees that legionaries in lorica segmenlata were best suited to fighting in close formation, Les Armes des Romaines (Paris, 1993), 133Google Scholar. Mail armour does not seem to have been cheaper to produce in terms of labour requirements than lorica segmentata, so it is unlikely that auxiliaries were being issued with ‘second-rate‘, cheap protective armour.

44. Again, , expeditae cohortes, Annals 4Google Scholar. 25, though legionaries too could operate successfully in wooded country, according to Germanicus, , Annals 2Google Scholar. 16.

45. Leves cohortes, Annals 3. 39. Note that Corbulo had sent local Iberian soldiers against the Mardi, sending troops who were no doubt familiar with the moutainous terrain and could operate there with ease, as well as having superior knowledge of the fighting techniques of their enemy, Annals 14. 23.

46. Livy, 22. 18; Plutarch, , Sert. 12Google Scholar.

47. E.g., Ostorius Scapula used both legionaries and auxiliaries to attack Caratacus' forces who were holding a hill, Annals 12. 35 (A.D. 51).

48. Quintus Veranius had claimed back in A.D. 58 that with two more years of campaigning he would have conquered the province, Annals 14. 29, so there is no reason why Agricola should not want to boast of having completed the conquest.

49. W. Hanson suggests that Domitian favoured an expansionist policy in Britain because he lacked the military background of his father and brother, ‘Tacitus’ ‘Agricola‘: an Archaeological and Historical Study‘ ANRW 33. 3 (1991), 1741–8Google Scholar.

50. Agricola 25–6: attacks on forts and a night attack on one of Agricola's columns.

51. Burn in Dorey, T. A. (ed.), Tacitus (London, 1969), 56Google Scholar. Onasander recommends this tactic, 6.

52. Livy 38. 49.

53. Agricola 16 on Suetonius' victory, described as uniusproelii fortuna veteripalienliae restituit.

54. Agricola 39: cf. Panegyricus 16 where Pliny contrasts Domitian's military activities unfavourably with another's, praising Trajan for his cerium triumphum across the Danube, compared with the falsae simulacra vicloriae of Domitian.

55. Agricola 41.

56. Annals 14. 37.

57. Polybius 10. 26, records the speech of a Macedonian orator which notes that the first troops in battle to face danger were the light armed who were the most active part of the force, whereas the credit for the result goes to the phalanx and heavy armed. Hans van Wees has recently demonstrated that the light armed troops had a greater and more important role in hoplite warfare than most of the literary sources give them credit for, ‘Politics on the Battlefield‘, in Powell, A. (ed.), The Greek World (London, 1995), 153–78CrossRefGoogle Scholar. Although light armed troops appear with regularity in Roman historians, it is again the heavy armed troops of the legion who frequently get the credit, Mons Graupius being a prime example.

58. Carrié, J. M., ‘The Soldier’, in Giardina, A. (ed.), The Romans (Chicago, 1993), 100–37Google Scholar.