Some Observations on the Praetorian Guard

SOME OBSERVATIONS ON THE PRAETORIAN GUARD This paper is concerned principally with two questions: first, the numerical strength of the praetorian cohorts in the second and third centuries and, second, the length of· service for guardsmen in the third century. The first question was discussed in 1938 by M. Durry whose conclusions have been widely, but not universally, accepted as the standard guide. The aim of Part I of this paper is to challenge his enumeration and to suggest an alternative. Part II requires no comment at this juncture. Part I. Two literary references give us evidence for the strength of the early Guard. Both refer to the situation in the first century although written much later. From Tacitus we learn that Augustus kept 9 cohorts of praetorians and, later, that Vitellius formed a new Guard of 16 cohorts, each 1000 strong 1 . Corroboration of this figure is apparently provided by Cassius Dio who informs us that Augustus had l 0,000 praetorians in 10 divisions 2 . Epigraphic evidence confirms that the Guard was raised from 9 to 12 cohorts by Gaius or Claudius 3 • Both this and the subsequent Vitellian increase were abandoned by Vespasian who reverted to 9 cohorts 4 • Domitian subsequently added one 5 , setting 10 as the figure for the remainder of the Guard's history 6 . We know, therefore, that Dio was mistaken regarding the number of cohorts in the Augustan period; was he also wrong regarding their effective? Durry believed so, claiming that Dio had attributed to an earlier period the situation before his own eyes in the early third century A.D. 7 . This is a scarcely credible accident. The change in effective which Durry adduces had only occurred in Tac., Ann. IV 5 and Hist. II 93. Dio LVI 24.6. ' CIL VI 2762-5; V 2513; III 4838 (coh. Xl); V 7003; 7162; 7258; VI 2766-8 (coh. XII). Coh. X inscriptions are confused with those of the later cohors X. 4 CIL XVI 21 (A.D. 76): in cohortibus novem praetoriis. 5 M. DURRY, Les cohortes pri!toriennes, Paris 1938, p. 81. 6 CIL XVI 81 (A.D. 89/134), 98 (150), 135 (208), 155 (254) and 156 (298): in cohortibus decem. 7 Op. cit., p. 86. 1 2 276 D.L. KENNEDY PRAETORIAN GUARD the later years of the second century A.D. 8 when Dio was not only at Rome, but a senator, and well-placed to know the facts 9 . It is more likely that Dio, knowing at first hand of the Severan change, attributed the Antonine situation, well-known to him, to the Augustan period in origin. Durry believed Dio's effective wrong, but had to re-interpret Tacitus, too, in order to support his beliefs 10 . The statement of Tacitus is re-emphasized to read as «non seulement on enrolait seize cohortes pretoriennes et quatre urbaines, mais encore chacune devait avoir 1000 hommes » 11 . He therefore concludes cohortes quingenariae in the Julio-Claudian period as the alternative to the new milliary cohorts of Vitellius. I will return to this point below. The important point to note now is that by far the likeliest explanation of Dio is that he was reporting the Guard as it was during his own early life under the Antonine emperors: ten cohorts, each of milliary strength. Durry's interpretation of the literary evidence substantiated his conclusion about the effective based on the laterculi praetorianorum. From these he concluded that praetorian cohorts were quingenaria until 193 A.D., at which time they were raised to milliary strength 12 . Passerini was unconvinced by Durry's thesis, but his own arguments have generally been found unconvincing 13 • The most recent statement on the effective is that by Neumann, who notes the conclusions of both Durry and Domaszewski 14 in favour of 1000 until Severns and 1500 thereafter 15 . In order to disentangle the conflicting opinions and the general uncertainty on the matter, it is necessary to examine the evidence Below, App. I for possible new dating of the Guard increase. Dio LXXV 1.4 and 2.4-5. 10 Op. cit., p. 82. 11 Tac., Hist. II 93. 12 Op. cit., p. 84-9. Durry continued to back this conclusion in RE XXII 2 (1954), col. 1607ff. 13 A. PASSERINI, Le coortie pretorie, Roma 1939, p. 58-67. R.E. SMITH, The Army Reforms of Septimius Severus, Historia 21 (1972), p. 488 and note, accepts it while E. BIRLEY, Septimius Severus and the Roman Army, Epigraphische Studien 8 (1969), p. 64, follows more hesitantly. On the other hand, G. WATSON, The Roman Soldier, Bristol 1969, p. 16ff., believes in quingenary cohorts, but says that, after Domitian, «there was no significant change in establishment until ... A.D. 312». 14 Die Rangordnung des romischen Heeres, 2nd edition by B. DossoN, Koln-Graz 1967, p. ix and 20. 15 Der Kleine Pauly 4 (1972), col. l l 16f. 277 adduced by Durry for his conclusions. From the laterculi praetorianorum, Durry calculated that an average of 5 men per century were discharged each year in the Antonine period and 11 in the Severan period. By a simple calculation the cohorts were, therefore, quingenary (5 men x 6 centuries x 16 years service fin an ideal situation l/16th of the Guard would retire annually)) = 480) and milliary (11 x 6 x 16 = 1056), 'fluctuations' accounting for the imprecision. Before looking at the individual figures and the premises on which the calculations are based, it is worth examining the nature of the 'fluctuations'. It is clear that if ea. 30 men are discharged from each cohort annually, at least that many, on average, must be enlisted annually to keep the effective static. The problem is, how many recruits were required in order for ea. 30 to survive their service for discharge? The question of mortality rates is a difficult one. We do have some evidence for calculating mortality rates in the legions and among civilian populations, which may be used as a guide. Seven dedications to the emperor by the newly discharged soldiers of individual legions are extant. As far as can be deduced from these rather less than half the recruits to the legions survived to ｾ＠ discharged 25/6 years later 16 . How can this be applied to the Guard? First, it should be borne in mind that by virtue of their situation on the frontiers, the legions would be in action more often and far sooner than the Guard, which usually only accompanied the emperor or a member of the imperial family to war 17 . Against this, however, we may set the clear evidence that the auxilia bore the brunt of minor campaigns and, perhaps, even of wars 18 . In addition, the Guard was stationed in Rome which, by all accounts, was an unhealthy city 19 . 8 9 16 J.F. GILLIAM, The Veterans and Praefectus Castrorum of the II Traiana in 157, AJPh 77 (1956), p. 359 n. 3 for a summary of the conclusions. 18 E.g. Tac.,.Ann. I 24; D10 LXXIX 23; 37.4; Herodian VII 11.2 and VIII 5.9. Tac., Agne. 35 tells us that the auxilia were placed in the front line with the legions in reserve, « ... victory would be vastly more glorious if it cost no Roman blood». In the event the battle was won without the legions being engaged. The auxzlw were more easily recruited and less costly than the citizen legions whose fortresses were often far behind the limes, where no river formed the frontier. 19 Juvenal, Sat. III for fires, 'muggings', collapsing buildings, poor quality accomodat10n and tension among the inhabitants. Tac., Ann. XV 38-44 for the Neronian fire. Livy XXXV 9.2-3, XL 28.4 for Tiber flooding. Tac., Hist. II 93 for unhealthiness. The flooding of the river itself and the narrow winding streets of much of Rome were sources of disease; plague was by no means uncommon (J.F. GILLIAM, The Plague under Marcus Aurelius, AJPh 82, 1961, p. 225-251). ａＮｾ［＠ 278 D.L. KENNEDY PRAETORIAN GUARD Despite the many problems in method and interpretation, the rate of mortality among civilian populations, discussed below, seems clearly, as one would have expected, to have been rather higher at Rome than anywhere else investigated. It should also be borne in mind that some emperors of the second century, like their third-century successors, were often absent from Rome on campaign or touring; elements, at least, of the Guard would accompany them 20 • It would not be unreasonable, therefore, to assume that life-expectancy in the Guard was little superior to that in the legions. The difference in length of service would be, to some extent, negated by earlier recruitment for the legions (and hence a shorter life-expectancy) and by .their more active life 21 . A number of investigations have been made over the last fifty years into the questions of life-expectancy and mortality in the ancient world. The most popular method has been via the ages at death given on epitaphs from various parts of the Empire. Various other approaches have been attempted: through census data from Egypt and through physical anthropological studies - determination of the age at death of the skeletons in burial grounds 2 2 . There are a great many problems associated with each of these methods which appear to be insuperable - infant mortality, mortality among the aged, and the class and status of those commemorated 23 . The caveats raised by Hopkins and Kajanto 24 do not apply with the same force to the figures for the younger adult male population. Since we are dealing here with soldiers who die in service, we may assume that, given the availability of the corpse, each man was likely to receive an epitaph from friends/relatives or comanipuli regardless of wealth or status 2 5 . We can therefore examine the tables prepared by demographers for information on life-expectancy within the age brackets which concern us. Both Hopkins and Kajanto have criticised the tables prepared by Burns, and both have serious reservations about the value of epitaphic evidence as a basis for calculating mortality rates. For my purposes here, they are quoted rather to show trends and for the purposes of comparison than as reliable absolute statistics. It can be readily seen from the tables that, first, life-expectancy in Rome was rather less than in the provinces and approaches closest to that for the military on the Danube. Second, in the approximate age brackets within which most praetorians served their contract, over 50 % of 20 E.g. Dio LXXVI 6.6 for the Guard at the battle of Lugdunum. Herodian VII 11.2, VIII 5.9, for the bulk of the praetorians with Maximinus. Dio LXXIX 23.2 for praetorians with Julia Domna at Antioch; LXXIX 37.4 with Macrinus in Syria. 21 As the imperial guard and an elite force, recruits to it would be men of good physique and health and thus, almost certainly, young adults. Durry believed that recruitment to the Guard was from men who had served their tiro stage elsewhere, «C'est it dater de Trajan que le passage des cohortes urbaines dans les pretoriennes. sans doute apres trois ans de service, devient la regle ... » (op. cit., p. 245). What appears to be the principal evidence for this belief is given later in a quotation from Dositheus: Petente quodam ut militaret, Hadrianus dixit: 'Ubi vis militare?' Illo dicente, 'in praetorio', Hadrianus interrogavit: 'Interim in urbana militia, et si bonus miles Jueris, tertio stipendio poteris in praetorium transire'. (Sent. Hadr. 2). I have been able to find little to support the belief in such service being common, never mind the rule. Durry notes (op. cit., p. 245 n. 3) C!L X 3733; IX 5839 (and 5840) and 5650 as examples. I have been unable to find any others. On balance, it seems most likely that, while such transfers were possible, they were rewards and that the point to be gained from Dositheus was that young, physically immature men were not accepted for the Guard. An examination of the 172 praetorians who died in service and whose epitaphs record their length of service and age at death, allowed the approximate age of recruitment to be calculated. This can be tabulated as follows. Number of recruits. Deduced age of Average length of enlistment. service. 20 18 7.6 121 18-22 8.6 22 23-27 9.7 11 28-32 9.9 Even allowing for the smaller number of examples in the final age group, it can be seen that, up to a certain point, the older a recruit was at enlistment, the more chance their was of him surviving to complete his service. In other words, younger men were less likely to complete their service than those enlisting at a greater age. 279 22 For a concise and very useful recent summary of all these methods, see A.E. SAMUEL et al., Death and Taxes: Ostraka in the Royal Ontario Museum (American Studies in Papyrology, 10), Toronto 1971, p. 5-17. 23 Ibid., p. 5-17. I. KAJANTO, On the Average Duration of Life in the Roman Empire, Annales Academiae Scientiarum Fennicae, Series B, 153.2 (1968), p. 1-30. Kajanto concludes that epitaphs do not allow us to calculate life, expectancy since most reflect the lower middle classes and exclude the masses too poor to afford an epitaph, and that life-expectancy was much the same throughout the Empire. Cf. HOPKINS, On the Probable Age Structure of the Roman Population, Population Studies 20 (19661967), p. 263-264, who concludes that life-expectancy at birth in the Empire must have been at least 20 (to avoid difficulties in population replacement in a static population) and probably under 30 (to conform to standards in other pre-industrial societies). 24 HOPKINS, op. cit., p. 263f. See previous note. 25 The reference in P. Gen. Lat. I to a pay stoppage as 'ad signa' has been taken to refer to a payment to a burial club. Although Mommsen disputed this, there can be no doubt about the use of the saccus undecimus as a fund for burial (Vegetius II 20). 281 D.L. KENNEDY PRAETORIAN GUARD one group of civilians in Rome, who had reached the age of 20, died before the age of 35; while over 1/3 of European provincials, reaching the age of 17, died before the age of 37. Despite the caveats raised regarding the figures, the conclusions may be regarded as approximately reflecting the pattern of mortality. rates among the younger adult males in the lower middle classes in Roman society. Mortality rates among the masses of poorer people was certainly higher still. The most recent, and by far the most interesting study has again been for Egypt. In this case, the conclusions were based on tax receipts recorded on ostraka 29 . The study involves the use of statistical methodology and formulae and is therefore difficult for the non-expert to check 30 . The method seems to me reliable, however, and avoids the problems involved in epitaphic studies. What these results show can be summarized in the following figures. 280 Fig. 1 (after Kajanto - Age 15 20 25 30 35 Males 297 288 264 236 203 Fig. 2 (after Burn - Europe (Central)) Age 17 22 27 32 37 Males 426 357 283 213 171 Fig. 3 (after Burn - Age 17 22 27 32 37 Rome) Lambaesis; military) Males 90.9· 84.9 74.8 61.8 53.4 Fig. 4 (after Burn - Age 17 22 27 32 37 Danube; military) Males 284 268 233 187 158 The study based on Egyptian census returns shows that of 148 males surviving to the age of 20, 76 were dead by the age of 40 (over 50 %) 2 6 . The difficulties involving acceptance of these data were recognised by the authors, but they agreed that corrections to it were most likely to increase mortality rates in the early groups 2 7 . Physical anthropological studies are, as yet, largely untried, and the results are unreliable if only because the sample is so small. It is interesting to note, however, that the single study published gives a male life-expectancy of 42. l years; agreeing broadly with calculations based on epitaphic evidence 2 8 • 26 M. HoMBERT - CL PREAUX, Recherches sur le recensement dans l'Egypte romaine (Papyrologica Lugd. Bat., V) Leiden 1952, p. 156-160. 27 A.E. SAMUEL, op. cit., p. 14-17 and M. HOMBERT-CI. PREAUX, op. cit., p. 159-160. 28 J.L. ANGEL, 'The Length of Life in Ancient Greece', Journal of Gerontology 2 (1947), p. 18-24. Fig. 5 (after Samuel) Number of persons surviving from pseudo-age 18 Pseudo-age Money Tax All Taxes 18 20 25 30 34 35 101 86 56 39 27 26 131 114 78 57 41 40 The 'pseudo-age' at death is the age at which an individual last appears on an ostrakon, i.e. equal to or greater than his age at death. The authors state as their conclusion that, « ... beginning from any particular age, on the average, each member of the population will live for 14.4 years; expressed differently, it means that a man of 15 can expect to reach 29 .4 years of age . . . The population is halved roughly every l 0 years ... » 31 . Fig. 5, above, shows that, beginning at the age on which most praetorians were inducted, between 2/3 and 3/4 would have died before reaching the age at which those same soldiers would have been discharged. The approach has the advantage of eliminating the incalculable deaths in infancy and childhood by beginning only with those who have survived to reach the age of 15. The disadvantage from my view-point is that the sample concerns a rural Egyptian community whose standard and way of life was far different from that of praetorian guardsmen. *** 29 3 A.E. ° Cf. SAMUEL, op. cit., p. 19ff. JEA 59 (1973), p. 273f. for a review of this work. The reviewer does not question the figures, but admits that he is in the hands of the statisticians. 31 Op. cit., p. 25ff. 282 D.L. KENNEDY Several conclusions emerge from all these studies. First, life was short in the ancient world. The most conservative estimate allows the tentative conclusion that over one-third of all lower middle class males, who reached the age of 18, were dead by the age of 35. Mortality rates in Rome for the same class over the same age range were more likely to have been considerably higher. Life-expectancy among the urban and rural masses throughout the Empire will certainly have been even higher (following the evidence for preindustrial societies) and shortest of all among the mass of poor in the capital. A number of factors must be borne in mind when attempting to apply these data to the Guard. The men chosen for service will, logically, have been of apparent good health and physique 32 . This standard will have been at least maintained by the availability of medical facilities in the army and by the removal of the problem of seeking food, clothing and shelter; physical fitness will, theoretically, have been kept high by the rigours of military discipline 33 and exercise. In these respects, the life-expectancy of praetoriani will have been increased vis-a-vis the respective classes from which the recruits originated. On the debit side, the very rigours of discipline and service, involving sickness and injury, will, to a not inconsiderable extent, have affected the health of many, and shortened their lifeexpectancy. It should be clear from the foregoing that any precision on the question of mortality in the Guard, through demographic studies, is impossible. The conclusions, however unreliable, can, nevertheless, be used to provide a guideline. If the conclusions of Samuel were applicable to the Guard at Rome - namely, that at any given age, each male could expect a further 14.4. years of life - then well in excess of I /2 the Guard would have died in service. These studies are all, however, concerned with death by natural causes, i.e. not in warfare. The Guard was, of course, an operational military unit and as such did participate in warfare although normally only at the side of the emperor or a member of the imperial family. Consequently, losses in action will have been far lower than among frontier units 34 . However, the very process of accompanying the 32 33 34 Herodian IV 7.3. Cf. Tac., Ann. I 17 for legionaries' views of life in the Guard. Dio LXXVI 6.6 for the Guard in battle. PRAETORIAN GUARD 283 emperor on journeys would increase the mortality rate because of the unusual hardships, climate and fatigue - the same factors which claimed more lives in armies than enemy action up until quite recent times 35 . A further factor must be taken into account. In order to arrive at some conclusions about the men being discharged in the laterculi, it is essential to appreciate that these men represent those who enlisted in a particular pair of years, less al/ losses to the effective. Such losses were not merely by death - natural, or in war - but also from injury or on the grounds of ill health. The soldier who suffered any serious physical injury or disability was just as much a loss to the army as a dead soldier. The proportion of dead to wounded in ancient warfare is liable to have been less than in modern times when soldiers invalided out of the forces far exceed in number those killed outright on active service. Since this was because more wounded would die in ancient times, the effect is the same where as here, we are dealing with losses to the strength. The actual ｉｯｳ･ｾ＠ are of course incalculable but, because of their existence, we should best think of service-expectancy rather than life-expectancy, of loss-rate rather than mortality-rate, when assessing the proportion represented by the laterculi. To return to Durry's calculations. His loss to the effective strength from the 500 men supposedly recruited, down to the 480 he calculates an impossibly low figure. He thus takes as discharged, is 4 % ｾ＠ virtually no account of losses over the 16/17 years of service. It is clear, therefore, that his conclusion that each Antonine cohort was quingenary, must be rejected. I have examined the laterculi available and counted the discharges per century. The laterculi list by century, and each is divided between the two groups who enlisted 25 and 26 years earlier, now being discharged. All the laterculi are badly mutilated, and often the centuries we do have are incomplete. I have compiled my figures on the basis of (a) complete centuries and (b) incomplete centuries where a minimum figure can be reached. These counts can be summarized as follows: 35 For praetorians buried in the provinces, CIL III 5043 5073 5105 5596 6046 ' ' ' ' ' 6085; II 2!02. 284 D.L. KENNEDY PRAETORIAN GUARD Fig. 6 Laterculi showing those men recruited before AD. 193. Number of centuries extant shown in brackets. Plus ( +) figures refer to evocat1. therefore, are more likely to be reliable. The pattern for the two sets may well suggest that if we had a sample with a greater number of complete centuries, then the figures for the enlistment years would tend to approximate more closely 37 . If we apply Durry's calculations for the effective strength to these, then we arrive at (taking 17 years as the real length of service in the Antonine Guard) 38 : 6.2 x 17 x 6= 632; 4.54 x 17 x 6= 463; Mean: 5.295 x 17 x 6 = 540 The same revision is required for the third-century guard. (a) complete centuries 1 2 CIL, VI, 32515 32516 32518 32519 32520 Total: (13) (1) ( 1) (4) (24) (43) Mean per century: (181 : 43) 4.209 Both columns: Mean: (b) incomplete centuries 46 5 3 18+2 109+8 181+ 10 (10) (2) (1) (3) (20) (36) 67 11 5 13+ I 124+6 220+ 7 (220 : 36) 6.11 (79) 401 (401 : 79) 5.076 3 4 (3) = 27 ( 1) = 3 (2) = 6 (I)= 2 (2)=4 (9)=42 (5) = 38 (42: 9) = 4.66 (3) = 14 (2) = 11 (6) = 25+ 1 (16) = 88+ 1 (88: 16) = 5.5 (25) = 130 (130 : 25) 5.2 A close examination of the actual numbers discharged from individual centuries shows that the figure fluctuates between 0 and 13; only in 29 of the 79 complete centuries (columns 1 and 2) does the number fall between 4 and 6 inclusive, i.e. less than 40%. It is obvious from the above that the mean discharge is highly erratic: even if we take the mean for the columns in pairs (1&2 and 3&4) the incomplete centuries still give a higher mean than the complete. The main problem is that the second to fifth laterculi have far too few centuries complete to provide a reliable average. These three, therefore, together with the evidence in columns 3 and 4, must all be abandoned as too small or incomplete. If we recalculate for 32515 and 32520 only, we get (average only): 1 32515 32520 (13) (24) 3.538 4.54 2 (10) (20) 6.7 6.2 Mean= 4.9 Mean= 5.295 Although the two sets of figures here reflect a similar pattern, they are not close enough to allow us, legitimately, to fix a mean for the sets. Since 32515 reflects the situation for a smaller sample of centuries and the means for the two enlistment years are much further apart, it is justifiable to regard them as less reliable. 32520, on the other hand, has many more complete centuries, and the means, 36 I have followed Durry in excluding those laterculi referring to periods involving abnormal warfare. These laterculi are in any case fragmentary and have too few centuries for accurate calculations. 285 Fig. 7 Laterculi recording men recruited after AD. 193 Annual discharge. (a) complete centuries C/L, VI, 32533 32536 32624 32625 32627 32628 32639 32640 (b) incomplete centuries (4) 40 + 1 (7) 75 + 4 (4) 61 (3) 7+ 1* (2) 22 (2) = 22 (2) = 27 (5) = 51 (2) = 11 + 2 * (2) = 8 + 1 * (I)= 16 (4) = 54 (3) 40 + 1 (3) = 35 (* = Too small a sample to be regarded as normal: not included in the mean). Total: (20)247 + 7 (17) = 205 + 3 Mean per century: 12.35 12.06 The samples, though again small, are consistent. Precision on a mean is impossible, but it seems clear that 12.35 should be regarded as a minimum. On this basis the Guard would have an effective of at least 889 (12.35 x 12 x 6) 39 . We have seen how the Guard in the pre-Severan period was formed in cohorts of 1000 men 40 . The revised calculations for discharges show that 31.77 men on average would survive in the Guard to be discharged after 17 years service. The rate of loss over 17 years would, therefore, be ea. 46 %. In view of the shorter period of service in the Severan Guard, we would expect a smaller loss rate: 32.4 % would be 12/17 ths of 46 %, given a constant death rate. From this, it is calculable that a Severan cohort 37 The discharges do not suggest that, as with cohors XX Palmyrenorum, recruitment was in every alternate year (Dura Final Report, V. !, p. 36). 38 The practise of discharge in alternate years makes the mean length of service ea. l7 years rather than 16 in the period before A.D. 193. Cf. Table 2 for 17.9 years although this includes rounding up. 39 See below, Part II, for this figure of 12 years' service in the Severan Guard. 40 Above p. 2. 286 D.L. KENNEDY would have a strength of ea. 1315 men (assuming the figure of 889 to represent a loss of 32.4 %). What can we make of units with an effective of 1315 men? Since all the figures here are approximate and, in some cases, minimum, we may well have the 1500 strong cohorts proposed by Domaszewski: the same size, in fact, as the cohortes urbanae. The gap between 1315 and 1500 is bridgable if we remember that the figure of 12.35 discharges per century is a minimum and was probably rather higher. Second, the intensive and extensive warfare of the third century, often under the emperor's personal command, will have led to far more casualties among the Guard than it would have suffered in the Antonine period 41 . It seems obvious, therefore, that the size of the Guard in the Severan period was well in excess of 1315 men per cohort. A straight increase to 1500, to bring it into line with the cohortes urbanae, is most attractive, giving a body of 15,000 in 10 cohorts 42 . The most obvious objection to this conclusion is that we have no evidence to suggest that the castra praetoria was either enlarged or provided with an annex at any time. On the contrary, it is clear that the Tiberian barracks provided the ground plan for all susequent remodellings. This casern was 440 x 380m. (41.2 acres= 16.67 hectares), sufficient for not much more than 6000 men 43 . At the time of its construction, the Guard had not yet been increased by the additions of Gaius/Claudius. It still stood at the figure established by Augustus of 9 praetorian and 3 urban cohorts 44 ; all 12 were stationed in the new barracks of Tiberius 45 . Because of the area, it would be reasonable to assume that we were dealing with units of quingenary size at this stage (12 x 500 = 6000 (+ evocati?]). The creation of three new praetorian cohorts need not have led to overcrowding since the urban cohorts could have moved out; all three urban and three of the praetorian cohorts had formerly been stationed in billets in Rome sine castris 46 • 41 D.J. BREEZE, The Career Structure below the Centurionate during the Principate, in Aufstieg und Niedergang der romischen Welt II I, Berlin-New York, 1974, p. 439, claims that increased promotion opportunities will have increased the flow of praetorians to posts beyond the Guard. 42 Dio LV 24.6 «six thousand in number organised in four divisions ... ». Cf. H. FREIS, Die Cohortes urbanae, Epigraphische Studien 2 (1967), p. 38-42. 43 M. DURRY, op. cit., p. 47. 44 Tac., Ann. IV 5: ... tres urbanae, novem praetoriae cohortes .... 45 Dio L VII 19. Concentrated by Sejanus to promote discipline. 46 Suet., Aug. 49. PRAETORIAN GUARD 287 As we have seen above, the literary evidence does not allow us to reach any effective conclusion on the strength of the first-century Guard. It may have been milliary from the first 47 or quingenary raised to milliary by Gaius/Claudius or Vitellius. The sole firm point, as I have argued, is that the cohorts were certainly milliary under the Antonine emperors 48 . Then, at least, we should have expected an increase in the area of the barracks. The explanation for its non-appearence when the Guard was 1OOO strong comes in a report of 1876 49 noted by Richmond 50 : the barracks were of more than one storey. Building upward probably reflects the desire to minimize the scale of military installations in the Capital. Not only did the barracks go upwards, but barrack rooms were built within the thickness of the walls of the casern itself 51 . Richmond also noted the heightening of the outer enclosure wall at various times, one of which he dates to, possibly, Caracalla; perhaps to overtop the highest internal building 52 • It should be clear that if the castra praetoria of only one storey was adequate for ea. 6000 men, then two storeys were certainly adequate for ten milliary cohorts. It is debatable whether a two storey barracks, as well as the chambers in the walls would be sufficient for 15,000 men in the third century. I ｢･ｬｩｶｾ＠ not, but can see no reason why there should not have been a third storey too, if only in part. The barracks would thus, in many ways, have resembled the tenements of Rome and Ostia 53 . Again, pressure on accomodation may have been relieved with the right of conubium granted by Septimius to serving soldiers and the growth of canabae; married soldiers may simply have lived outside the barracks 54. 47 I.A. RICHMOND, PBSR 10 (1927), p. 12. Above p. 2. 49 Bull. Com. 4 (1876), p. 178. 50 I.A. RICHMOND, op. cit., p. 12. 51 Ibid., p. 13 plus fig. l showing a contubernium in the wall found in the J9th century 52 lb "d ' . 1 ., P: 14. RIC I, pl. V, no. 88 shows a tall building ('temple') inside the castrapraetona. Cf M. DURRY, op. cit., p. 50. 53 . J.E. PACKER, The lnsulae of Imperial Ostia (Memoirs of the American Academy m Rome: XXXI), 1971, suggests that the barracks of the vigiles at Ostia may have had a third storey. 54 Th . . e intensive. warfare of the third century, involving the emperors themselves ｾｵｳｴ＠ have greatly increased the casualty rate in the Guard. The extent of this is mcalculable, but it should be noted that the 1500-strong cohorts of the third century are by no means as secure as the Antonine milliary cohorts. A possible alternative of cohorts 2000 strong is difficult to accept. Alternatively, we may consider the 48 288 D.L. KENNEDY PRAETORIAN GUARD Conclusions. Part II. •1 .. , . . . . :Manlius Q. f. Vtl ! ｴｲｾｬ｡ｵｳ＠Ｎ (sic) ＱｾｴＮ＠ atti;i. :KV pient.1SSuno Ｇｉｾ＠ Cam. I Severns I Alba Pompeia veter. I v. a. XLII m. II ex ｾ･ｧＮ＠ XXII I Primig. in praet. chor. (sic) VII / in I 11ussus honesta missione J Q. Mattlius Epaphroditus I I bene merente fecit et sibi I et suis paterisq. eorum s1. ｭＮＢｾﾷ＠ origin suggests an early date for his service. The legion Ol\e of those which supported Vitellius and this man must, ＴＱｴ･ｾｯｲＬ＠ be a transferee of A.D. 69 58 . He was clearly retained by Vespaaian. The most interesting point to arise from this inscription it ｴｨ｡ｾＮ＠ ｡ｾ＠ the time of his discharge, whatever had been the length ｾＮＬＯｬ･Ｘｩｯｮ｡ｲｹ＠ service, he had served only 15 years as a praetorian. ｾﾷ［＠ 2:X ILS 2036. 1 ｾｬｩｦｳ＠ I C. f. Rom. I Crescens I militavit leg. IV I annis VIIII I traiectus mi praetorium I cob. V pr. I militavit attn. III. Tack of any imperial gentilicia, The legion (IV Macedonica?) Vitellius on the Rhine, making .. 9 • He had served 9 years in the ffi ·the Guard. Ａｾﾷｉｌｂ＠ the inscription is likely to was one of those which 69 the most likely date of legion but died after only 2034 add. C. ＺﾷＱＧＭＮｴ･ｮｾ＠ ｾＮ＠ In A.D. 69, Vitellius rewarded his army by opening the ranks of his new Guard to them all, legionaries and auxilia alike. The practice was abandoned by the Flavians but reverted to by Septimius Severus following: a milliary cohort was probably ea. 960 strong and a 'doubled' cohort was actually in the ratio of 6 to 10 (R.W. DAVIES, A Note on a Recently Discovered Inscription from Carrawburgh, Epigraphische Studien 4 (1967), p. 112-113) giving a unit of 1600 men. The unit would be unparalleled, but it is worth noting that Caraca!la formed a Macedonian phalanx of 16,000 men (Dio LXXVII 7 and LXXVIII 18); a Roman unit of this size is unattested and it is possible that the new body in which he took a great personal interest, and which accompanied him on campaign, was modelled in size on the existing Guard. 55 Dio LV 24.8. -f established recruitment from the legions as the regular method to the. Guard. I have found 23 inscriptions referring to . . . . • tlm Guard from the legions. Seven of these are incomplete information or are mutilated and therefore of no value for .,...... d y 56 . ｾＺｮｴｳｵ＠ ｾｴ｟Ｌ＠ From an examination of available demographic studies, I have argued that the loss rate would have been sufficiently high for the effective strength of the Guard to have been well in excess of the quingenary cohorts argued by Durry. No impediment prevented acceptance of the milliary cohorts which are strongly arguable from the literary evidence. The Guard of the third century, as the laterculi reveal, had been augmented and, from calculations based on the increased discharges and an examination of the likely loss rate, it seems probable that the new cohorts were of 1500 men each. If my deductions are correct, then we may see a loss rate over the period of service of ea. 46 % in the Antonine Guard and ea. 41 % in the Severan Guard (over a shorter period of service); figures which are compatible with the substantially higher mortality rates calculated for ancient societies, as well as with those we would expect in preindustrial societies and with the more prolonged warfare of the third century. (I have excluded the evocati from the count in the above calculations: they appear on the laterculi, but in view of Dio's testimony that they constituted a special corps, it is unlikely that they counted towards the effective of the centuries ss. They cannot, in any case, have been particularly numerous and their inclusion, if required, would increase the figure for discharge means.) *** ｾ＠ 289 C. f. I Qui. Moderatus Antio I milit. in leg. XVI Gal. a. X I (su:) m. coh. IX pr. I in qua milit. ann. VIII I missus honesta missione I 1 ab ｾｊｩ＾ﾷ＠ fact. evoc. Aug. I arcitect. (sic) armament. imp. I evoc. ﾷｾｬｴ｡ｯ＠ Aug. Germ. I[ ....... . tn Gallica was at Mainz, where it surrendered to Civilis espasian cashiered it and formed a new one, XVI Flavia. ｴｾ＠ is certainly a transfer from the Vitellian vexillations at ［ｷｾ＠ f!!l .• . > ·. transferred rather late in life and presumably to tap 566, 2758, 2803; VIII 9391; IX 1609; X 532 and AE 1940, 117. ·. SS, op. cit.,. has ｴ･ｾｰ､＠｡ to date praetorian epitaphs on the basis ｾ＠ Ｎｾ｡･＠Ｌ used. I will note m. Clauss, A.D. 100.180. .!M. Clauss, c. A.D. 50. his dates only where they disagree with my own. 290 D.L. KENNEDY PRAETORIAN GUARD d' f ages then he will his talents. If we allow for the froun . ｭｧ｡ｾｨ＠ time ｾｦ＠ his honesta be · his l 6- l 8th year o service . m of the practice of discharge in alternate years). His mrssw (ben ecause subsequent service was as an evocatus. (probably) just prior to the battle of Issus when it apparently changed sides and was rewarded with the titles felix constans, given here 61 . The most likely occasion for his transfer would be after Issus when the legion was honoured and replacements for the Guard would have been required. Thus, at most, Valerianus had completed 5 years in the legion followed by I 3 in the Guard before being discharged. The former figure is likely to be too high, however, as a result of numbers being rounded up. His enlistment would, in this case, have been some time after 9 January, 190, so that on the discharge date of 9 January, 206 he had not completed 16 years service. 7. !LS 2037. ｨｾｶ･Ｎ＠ , 7 Marfci natio 4. CIL VI 2697. Aur. Eliaseir. coh. I cammpid?ct?r coh_. I VIIII praet. et in praet. I I . I v1x. . an. LX (e)t miliftav1t m leg1one I ann. X D. Pan.m. qm · l'k I onian origin make it most I e y The ｩｭｾｲ｡ｬ＠ gentilicium an ｯｴｾ･＠ third century. His extremely that this. man was a ｴｲｾｳｨＺｶ･＠ been because of his specialist talents "d t Much of his service long service of 35 years w1 (cf. No. 3, above) as the cohort camp1 oc or. will have been as an evocatus. ann. XXV. fit/ d h p 5 /LS 2044. . .· ( ·)XIII nat I ·1 h r (sic) I VI pret. (sic) 7 BhclSI stup. sic . . D. m. s. Aur. ｖｾｲｯ＠ m1 . c o . a rtio Vico Budalia qui (vi)xit an. I XL m. Pannon. pede S1rmese ｰ｡ｧｾ＠ [ ... ] ｾ＠ I . . t (sic) III (or IIII) Aur. III d. XV mil(i)t. in l(eg10ne) ｾｴｵｦｮ｣･＠ (sic) s ｵｾｯｭ｡ｮｩｰｬ＠ sui I re ipsius Iustinus Aur. Florinus Val. I Av1tianus et omnes . b. m. f. *milibus. f . third century date is suggested by the incidence o Once .agam, . a . . o note that after three years in a legion, Aurehus. It is mterestmg t I . "ble that his service was h d 14 years in the Guard. t is poss1 e serve . d d his burial by comanipuli suggests. enrolled comparatively late in inn e l:gions, he certainly old enough to have completed is term of service. ｾｲｭ｡ｴ･ｨ｜Ｚ［ｮｳＢ､＠ ｴｾ･＠ ｾ｡ｳ＠ ｾｦ･＠ . · t 6 C/L VI 210. . !em defensorem I gemo centunae ex vo o Pro salute dd. nn. Augg. _I ｈ･ｲｾ＠ K itoliade stip. XVIII I mi!. coh. X posuit I L. Domitius Valenanus I_ omo ｾｰ＠ ex le VI Fer. f. c. I . p· 7 FI Caralitani lectus m praetono dd. nn. I . g. pr. p. v.honesta. m1ss10 . . ne I VII Idus Ianuar. dd. nn. I imp. Antomno 10 missus Aug. III et I Geta nobilissimo Caes. II cos. . . d 9 J 208 Valerianus had enlisted m legw Date anuary, . Dur assumes that he not earlier than 9 January, 190: . ry h d R e6o the Guard in 193 when Septlmms reac e om .. however, was m . the East and adhered to Pescenmus 60 M. DuRRY, op. cit., 247. VI Ferr_ata passed mto His legion, N' r until tge 291 D. m. I C. Maccenio Vibio mi!. coh. I pr. 7 I Primitivi vix. ann. L V mi!. in leg. X i Gem. ann. VIIII ind. tra(ns)l. in pr. an. I XIIH Ulpia Valentina c:u{n)iugi / karissimo b. m. f. cura agente I demandatum fratri ex tesltamento ｍｾ｣･ｮｩｯ＠ Crispino I evok. Aug. n. Vibius' legion was stationed in Pannonia superior in the second century, and his wife's name, 'Ulpia' suggests an original grant of citizenship by Trajan. The most likely date for a Pannonian to be transferred is after 193 when recruitment, from the Danubian legions in particular, became the rule. The inscription does not make it dear whether he died in service or not, or even if he was an evocatus like his brother. His service of up to 23 years, however, would have made him ea. 32 years old on enlistment in the legion if' he died in service. Even allowing for rounding up of years or for guessing at his final age, it is most likely that he had long since ｾＨｉｲ･､＠ by the time of his death. If so, mention of his honesta missio I?J also omitted and we may legitimately presume that his service as an evocatus may have been included in the 14 years in the Guard, without being mentioned. Since he died at a comparatively ripe age and. was survived by his brother, it is possible that the brother was Y<i:qnge,r, l!dditional reason for believing that Vibius too was an ･ｾＬＮｶｯ｣ｧｴｵｳＺ＠ His service may be reconstructed as follows: 9 years · kg. X Gemina: miles praet. : 8/9 years honesta missio evocatus: 4/5 years .; konesta missio. 61 RE XII 2 (1925), col. 1593. 292 D.L. KENNEDY 8. CIL VI 2785. D. m. I Val. Sarmatio civis Filipopulletanus militabit (sic) in legilone prima Italica annis I 1111 in pretoria (sic) annis I duo vixit annis XXIIII Sudilcentius frater Genitus fecit I mimoriam (sic). PRAETORIAN GUARD 293 II'"•x'C:t.L VI 260 l. F.,Jtito eq. cor. (sic) VI pr. I natione Trax cives (sic) Filopolpulitanus • • '1 .XXXV qui I mil. an. XVII sic: in legione I Itialica an. II in cor. •,(sit) I munifex an. XIIII factus I eq. mil. menses n. X fratri I Val. Aulusa/nus pret. (sic) incomparabili. The legion I Jtalica was raised in 67, supported Vitellius, and was subsequently sent to Novae in Moesia. His Thracian nationality rules him out as one of the recruits to the newly raised legion, but fits well with a post-193 transfer. He died soon after. 'nly third century. Bito is one of the many Thracians lis in the third-century Guard 64. The inscription does to determine whether he died in service or not. The 9. CIL VI 2673. [Aur. T]ertio b. f. praef. coh VIII pr. El[qui] vix. ann. XLV dieb. VII mil. in leg. I [Cl]aud. ann. V inde tra(ns)lat. in pr. I [mi!.] ann. XIIII natione Pannl[on.?] Aur. Quintus frater et Aur. I { ... Ja soror hered. et Iul. matr. I [ ... ]onia co(n)iugi (sic) carissimo b. m. I [fece]runt. Neither of the legiones Claudiae were associated with ViteHius in 69; both were on the Danube in the second century. His nomen and his Pannonian origin suggest a transfer in the third century. As can be demonstrated, most recruits to the Guard were in the bracket 18-22 (possibly even earlier in the legions). If Tertius died in service, his age at enlistment in the legion must have been at least 26. It is more likely that he had been discharged 62 - hence, too, his burial by his wife and not by comanipuli - and that the 5 and 14 should be regarded as 'in his 5th' and 'in his l4th year'. His total service may thus have been as little as 17 years 63 • praet. ·· 16 years 10 months. Aur. Mestro I ex evok. qui mil. in l(egione) I an. VI in cast. Jl1:lt. vico Bitalcost. I reg. Pautaliense [ ...... ] J q. v. ann. XXXVI Vitupaus vet. fratri J b. m. fecit. lllllret•ll'Y· One of the numbers is clearly corrupt. As an ex-evocatus low. However, 21 years in the Guard is very long. 18181,mbhlble that the service should he emended to Xl. In this •l*'l!Bli'i; Service would be: 10. !LS 2038. D. m. s. I Aurelius Vincentius I miles cohrtis (sic) I tertiae praetoriae I centuria Maximini I militavit in legione I undecima Claudia I annis V I in praetoria annis I XI vixit XL I civis Trax I memoria fecerunt I cives de rebus 11 ipsius bene I merenti. The combination of the imperial gentilicium, the base of legio Xl in Moesia and the man's Thracian origin, suggests a third-century date. His age and the place of his burial (Mauretania Caesariensis) may indicate that he had been discharged. His military service will, therefore, have been 5 years in the legion and 11 years in the Guard. Burial by a legal wife implies a discharged soldier or/and a pbst-Severan d<lte. The M. Aurelius M. f Tertiu[s] of the laterculus CIL VI 32523 a 24, was discharg too early (9 January, 204) to enable this man to be identified with him. 62 63 ｬｾｴ｡＠ Ursianus cives Aquileiesis probitus ann. XVIII I in legione X if. an. V in pretoria (sic) an. Ill! I decissit (sic) an. plus minus coniux bene (me)renti I Verginio suo (f)ecit Justa I mi!. tt. (sic) ailiy ' be a transfer of 193 since he belonged to one of •.legions in Pannonia superior. The nomen and the Christian of plus minus suggest a third-century date 65 . I 2799 listing 19 cives Philippopolitanorum in 227. M. Clauss, ea. ,..,:..:, n plus minus is most common on later Christian epitaphs, although - ...we. 294 PRAETORIAN GUARD D.L. KENNEDY 14. CJL VI 37224. D. m. s. I Val. Patprnianus (sic) millix (sic) coh. X prae(t). 7 Boni qui I vixit annis XL m. HI I dies (sic) X militavit in le\gione annis XI in praeltoram (sic) annu (sic) et m. V I natione Pannonica (sic) I pag. Traiani Val. Velrecundus et Iul. Prislcianus (h)eredes huius I b. m. f. The inscription is clearly post-Trajanic (line 9) and, since Paternianus(?) is a Pannonian, a post-193 date is most likely. Because of his age at death and his burial by his heirs rather than comanipuli, it is probable that he had already been discharged. His total service was, therefore, 16 years 5 months 66 . 15. !LS 2045. Val. Tertius militi (sic) I cortie (sic) X pretorie (sic) qui I vixit annis XXXVI mes. III I dies XV militabit (sic) legilone Mesiaca ann(i)s V inlter lanciarios annis XI I in pretoria (sic) ann[is .... ] I 7 Salbi[ ......... I ...... . Since he died at the age of 36 and had already served 16 years before his promotion to the Guard, it is possible that he died in service; mutilation at the end of the text prevents us following his praetorian service. His earlier career, allowing for rounding up of numbers and for exaggeration, was less than 16 years. He is, however, a special case as a lanciarius promoted to the Guard. No Moesian legion supported Vitellius and a cohors X was not formed again until the late first century. It seems probable that the inscription is thirdcentury67. 16. CJL VI 2672. D. m. I Aur. Saturnina eq. coh. VIII pr. 7 I Verissimi qui militavit in leg. selcunda Italica tesserarius an. VI I qui vixit an. XVIU Aur. Optatianus I eq. fratri benemerenti posuit. His cognomen and his service in legio JI Jtalica (formed ea. 165) make his transfer certainly third-century. He only lived to age 28 and patently died in service. 66 By 221 at the latest, discharge was annual rather than biennial allowing soon after the completion of the contract. C/L XVI 139, 140 (for 221, 2), 143 (224, 5, 6), 147-150 (243, 4, 5, 6), all for January 9, make it clear that in the third century (No. 6 above attests a discharge on this date in annual rather than biennial as in the second century. 67 M. Clauss, A.D. 100-180. discharge 189, 142, discharge 208) was 295 Conclusions. · 'fbe most striking conclusion to be drawn from !sqnat in both 69 and after 193, recruits to the these inscriptions Guard from the ｬｾｩｲｳﾷｷ･＠ not expected to serve, as praetorians, for the balance of ihef 25 years for which they had originally enlisted (20 years prior to ｖｴｾｰ｡ｳｩｮＩＮ＠ Nor did they have to serve in the Guard for the 16 year coitfract required of direct recruits between Vespasian and Septimius. ﾷＧｦｾｩｳ＠ fact is clear in No. 3 (and probably No. l) for a ViteBian ttan:sfer, :and in Nos. 6 and 7 (and almost certainly No. 9), probably No's:;'. 5, 10 .and 14 and possibly Nos. 11 and 15. The total amount ｾ･｜ＧＮｩｮ｣＠ is not great, but it is signifcant that not a single inscription tfijkeS: a discordant note. At the worst, the text on a few stones does ll,!)t:pennitofcertainty. It is impossible to determine whether Septimius ｓ･ｾ＠ . knew of the practice adopted by Vespasian towards transferred ｾｬＩｴ｡ｩ･ｳ＠ or if he was applying a rule in force for transfers from ｾｨ･ｲＮ｢ｯ､ｩｳＬ＠ e.g. the urban cohorts. The fact remains that Septimius ｾﾷ｜Ｘｲ｡ｮｴｩｧ＠ .an important concession to many of his legionary ｾｩ､･ｴＵＧ＠ l!enceforward, transferees from the legions enjoyed not only tl're.hi!J!er pay and prestige of the Guard, but were required to serve Ｎ［ＨＩｦｽＺｬｾｴ＠ balance of 16 years over their service in the legions. They ｾ･ｬｩｧ｢Ｎ＠ for discharge on 9 January following their completion ｯｴ｡ｊＹｾｬ＠ of 16 years as a soldier 68 . Ｑ ﾷｾﾮｮｴｬｵｳｩｯ＠ is of direct relevence to Durry's theory of a doubling of:rM .Guard. Passerini objected on the grounds that the 11 discharges Ｎｾｲｯ［｡ｮＱＡｬｨ＠ were not twice the Antonine figure of 5. It is now obvious period of service is reduced in the third century, then mir: •.ｬＩＺｲ＼ｩｮ｣ｾｴｯ＠ of the total body retiring annually will increase : and reduce the contract and the discharge rate will earlier figure. However, as we have seen, the Guard Ｎｗｾｬｊｏｴ＾Ｑ＼［ＺｩＧ･､＠ but increased by, probably, 50%. ｊｾ｣Ｚｬ･ｪｶｵＱ＠ is too slight for any firm conclusions about the mean service for transferees to the Guard 69 . Logic the men chosen for transfer will have completed at least ﾷｬ＿ｾ｜｣［ＺｴｩｭｊＮ＠ The transfer.s of 69 must be ignored since they reflect situation. Those of post-193 are inconsistent: in all 13 cases ·..ﾷＬ｡ｬｾＮｉｩＺｴｲ＠ laterculi omit the year of entry into the Guard indicating a simple matter but varied with the individual. 296 D.L. KENNEDY we are told or can deduce the period of legionary service. No. 4 is a special case and must be excluded, as should No. 7 which is probably corrupt. Among the remainder, No. 14 is so much greater that it too probably reflects an irregular situation. The mean legionary service of the remaining ten is 4.4 years, giving an average of 12 years praetorian service before discharge. Appendix. Since M. Durry demonstrated an increase in the strength of the Guard in the late second century, it has been common, for those who accepted his conclusions, to follow him in attributing the measure to Septimius Severns 70 . There is abundant evidence for the dismissal of the old Guard, by Septimius, in 193 and his replacement of it by a new corps then, and subsequently, drawn from the legions 71 . It was an easy and perhaps logical step to attribute the increase to Septimius again in 193. There is, however, no evidence to support this conclusion. Dio is, of course, in epitome, and we are not entitled to argue from his silence, or from that imposed upon him by the epitomator, that the event did not occur in 193. Similarily, we may draw no conclusions from the silence of the Historia Augusta, Aurelius Victor and Herodian in their coverage of the events of 193, even when the account is as full as that in Herodian. There is, however, an alternative date which is both arguable from the literature and an attractive occasion. The only clear evidence for the increase occurring in the second century is, we have seen, derived from the laterculi praetorianorum. It is from these too, that we derive the termini post et ante quern. The latest second century laterculus (CIL VI 32520 listing men discharged in A.D. 160) demonstrates that at the time of recruitment in A.D. 143-4, the cohorts were still of the smaller size, i.e. milliary. The earliest laterculus to show an increase (CIL VI 32533) for A.D. 209 implies a recruitment date to the enlarged cohorts some 70 E. BIRLEY, Septimius Severus and the Roman Army, Epigraphische Studien 8 (1969), p. 64f.; C.R. WHITTAKER in the notes to the Loeb edition of Herodian, Harvard 1969, p. 352 (but cf. p. 246 n. l); A.R. BIRLEY, Septimius Severus, London 1971, p. 165 (following E. Birley, above); R.E. SMITH, The Army Reforms of Septimius Severus, Historia 21 (1972), p. 487f. 71 Dio (Xiph.) LXXV 1.1-2 and 2-4; Herodian II l3 and 14.5; Aur. Viet., Lib. de Caes. 20.l. 72 Section II above. PRAETORIAN GUARD 297 ＱｾﾷＺＷ＠ years before 72 . The increase, therefore, can be placed between 4':p. 144 and the end of the second century (ea. 193-8). Clearly, ＮｽＬ［｛ｾ＠ ·iS possible and has much to be said for it. In such a broad span O.ftjme, however, we must at least look for and consider alternatives. ()t}.-e ｾｵ｣ｨ＠ ·alternative is available from a passing literary reference. ＼ｅｊｾ｢ｬｩ｡ｮＬ＠ recounting the various plots against Commodus, deals ｱｾ＠ fully with the activities of one Maternus, an ex-soldier, who ｰｴｩｬＺ｜ｾ＠ a large band of fellow deserters and criminals 73 . The extent ｯｴＧｨｾ｣ｮｳ･ｱｵ＠ helium desertorum 74 may have embraced the whole dfl'J3ti;rI and Spain as Herodian claims; it was certainly wide enough ttl•lui.ve involved the military intervention of Pescennius Niger, the ｧｾｹ･Ｎｴｭｲｯｦ＠ Aquitania 75 and, possibly, of Clodius Albinus and Septirn1U;s' Severns, governors of Belgica and Lugdunensis respectively 76. ｓｦｴＭｬＮｾＧｩｳ＠ war is certainly historical, albeit obscure 77 , there is a good ｾｦ＼ｬｲ＠ believing that Herodian, in claiming that Maternus entered ltaly<'wit:h his followers seeking to assassinate Commodus and seize ｂｑｾ＿ｳ［ｲｮ｡ｹ＠ be preserving just such an event. The coinage of the ｴｾ＠ ｾｮＭｬｐｨｴＺｗｩｳ･＠ salus 79 , an inscription testifies to special precautions ＲＱＬｾＢＭ confirmation surely of an emergency involving security W'.hUe ｾｯｴ｢･ｲ＠ refers to an unassignable «victory» so. account is factual, then his concluding remark on assumes a greater significance in relation to the augmentation ｾﾷｩｦＡｏｮ｡ｴｬ＠ He tells us that ｎｾｧＮ＠ 3.4: «For he was sent to apprehend a body of deserters who numbers» (D. MAGIE, Loeb edition). .. , , .. .,.., ..,nn m Loeb edition of Herodian, p. 63 n. 2. U1 ｧｾ･｡ｴ＠ . an inquest in A.D. 186 (AE 1959, 141 = AE 1956, 90) of trouble m the unit - possibly confirmed by the title it soon ans Commoda. rs much less fantastic in the light of the later, better known Italy (note 83, below), the Bacaudae in Gaul (Aur Viet ｾｵｴｲｯｰＮ＠ IX 20.3; Oros. VII 25.2; Zon. XII 31. Their. leade; and _appears on coins as IMP.C.C.AMANDUS P.F.AUG. ｾｑＯｩ＠ m Egypt (Dio LXXII 4. l-2). .nQS. 169 & 178. \i!Y."lii.,a11<:IL,Jides.oＧｾｐＺﾷ･Ｌ＠ ＮｾＬＱｴｩ＾ｨ＠ cit., p. 66f. RIC III, p. 420, nos. 465, 468•-d - con. on coins usually implies that the reverse is true .l1rm1svenLous trouble in the army (ef. Dio [Xiph.] LXXIII 10.1 for 298 D.L. KENNEDY «After his escape from Maternus' plot, Commodus surrounded himself with a stronger guard Ｈ＼ｰｯｵｾＩ＠ and rarely appeared in public, spending most of his time avoiding legal and imperial business away in the suburban districts or on his imperial estates far away from Rome» 81 . In the context of the latter part of the statement one might be tempted to believe that nothing more is implied than an increase in the number of personal bodyguards on duty at doors, in corridors etc., to watch against the individual assassin. This seems improbable for two reasons. First, the plot of Maternus involved the infiltration of himself and his men into Rome disguised as praetorians. Elsewhere, we learn that he had a sufficiently large band to have attacked cities and for them to have been officially recognized 'as enemies and no longer simply as robbers' 82 . Indeed, Maternus even weighed the chances of success in a formal confrontation with the garrison at Commodus' disposal. Presumably, he felt capable of holding Rome by force after killing Commodus. In the circumstances, Commodus required, not a few more men in his presence, but a larger army at his personal disposal in the vicinity of Rome to prevent any repetition of Maternus' action 83 . Second, Herodian regularly renders the term praetorianus by oopucpopos 84 , with the Greek ｣ｲｯﾵｭ＼ｰｵａＮｾ＠ used to translate the concept of a personal bodyguard 8 5 • In the instance quoted above, however, neither of these are used - the term being cppoupa. In itself, the change in terminology probably indicates that the increase was neither in the praetorians alone nor the bodyguards about his person, but rather the guard, in the sense of the army which he disposed of, at Rome. In fact, the term cppoupa is regularly used to render praesidium 86 . We should perhaps regard Herodian I 11.5. Ibid. I I I. I. 83 That he would have been justified was demonstrated by the career of Bulla in Italy itself during the reign of Septimius Severns (Dio [Xiph.J LXXVII 10), cf. Herodian III 13.4 regarding the increased garrison at Rome «which left no external force competent to be a rival» - surely referring to threats from the provinces. 84 E.g. I 10.6; II I.I; 4.4; 5.1 and 14.5; VIII 7.7 but especially V 4.8 oi cr@µaw<puA.aKi;i; Kai liopuqi6p0l, oui; 81) 1tpatt@p1avoui;. Cf. Dio LY 10.10 !lmipx01 'tfilV oopuqi6p@v. It is the regular rendering of praetorian elsewhere too, see H.J. MASON, Greek Terms for Roman Institutions, Toronto 1974, s.v. oopuqi6pe<;. 85 E.g. IV 13.1 and esp. V 4.8 contrasting cr@µaw<puA.aKei; Kai oopu<p6p01. The term probably refers in most cases to the equites singulares. 86 H.J. MASON, op. cit., p. 98, s.v. <ppoup<i citing Appian, Hann. 33 (137), 6 tfi<; !lv Min:atµovticp <ppoupai; Tjyeyrov and Diod. Sic. XXII 1.3. Cf. Dio ｾｉ＠ 24. I tro; o' innt@v 060 wui; upicrwui; tfii; nspi cre <ppoupdi; apxs1v. In Herodtan, <ppoupa 81 82 PRAETORIAN GUARD 299 tile increase, therefore, as involving the entire military body available imperial security: praetorians, urbanicii, equites singulares and yigfkfs$ 7 • Michael Speidel has shown that the equites singulares retnail,led at the unchanged figure of I ,OOO during the second and third ;oont\iries8 "', but individual studies have revealed that increases in ｾｬｦ･ｩｶ＠ in two of the other three corps did occur in the late ｾｮ､＠ century 89 • : iA·later passage in the history (III 13.4), however, presents a strong ｾ｢ｪ･ｯｮＮ＠ Herodian, in the context of Septimius' final years, claims .. that' the troops in Rome had been quadrupled. This is usually to mean that Septimius had been the agent of the increase 90 . . ｾｮｴ･ｲＮｰ､＠ ＮＧｦｨｾｴ＠ may indeed be translated as «had just been quadrupled» 91 . . ＾ｻｦｾＧＮｗ･ｴ＠ true, then the Guard increase is best attributed to Septimius. Ｇｊ［ｨ･ｲｾ＠ is, buwever, a difficulty in this translation. No-one has yet ﾷｾ＠ a;ble to enumerate the troops needed to give this four-fold ＮＺｩｴｾＭ･＠ under Septimius. Durry (op. cit., p. 88), lists 11,500 before .alid'.24,000 after, exclusive of II Parthica. He is followed by Smith ｾｧＺ､｣＠ aideyn, Allowing for my re-.cakulated size of the Guard, the figures are 16,500 and 29,000 respectively. We know that there were Moorish troops in Rome 93 , and there may have been others, :f(fl'., ＨｐｬＺ｜ＱｾＰｴＦＬ＠ ｾＢＧ＠ frequently (II 5.3; III 11.6; IV 4.2). Where the word guard is required in sense, the usual rendering is ＼ｰｵａＮ｡ｾ＠ (III 12.l and 8 - «guards of the and guards of the emperor's quarters). In IV 1.5, <ppoupti is used in guard corps - «Both (i.e. Caracalla and Geta) established their own d (<ppoupa) ... ». Cf. III 5. 7 « Albinus had increased the size of his guard .». H.l MASON, op. cit., p. 98, s.v. <ppoupapxoi; = praefectus praesidi Sttat. S.21) or praefectus castrorum (Dio LV 33.2). 0 veralt ·strength of the garrison could have been increased at this time br more components unchanged. "'"'"J.n:,i.., Die equites singulares Augusti, Bonn 1965, p. 10-15. cohortes urbanae, Epigraphische Studien 2 (1967), p. 38-42. M. DURRY, praetorians, but see now above, Part I. The vigiles are believed, with the other units in Rome, to have been doubled from 3500 the same period. It is impossible to determine which changes may ow rather than under Septimius - if indeed any occurred then. .tbat the praetorian cohorts would have had priority in any list of ened. ﾷＮｾﾻｕｴｮｯｩＡｵｳ＠ op. cit., p. 385; R.E. SMITH, op. cit., p. 487f.; C.R. WHITTAKER, Herodian, note to III 13.4. op. Cft., p. 487ff.; E. BIRLE,Y, op. cit., p. 64f. p. 65 notes equites itemque pedites iuniores Mauri (!LS 1356), dated to pre-Alexander Severns (Les carrieres, p. 808f.). Cf. Herodian receiving javelin instruction from Moorish troops at Rome. 300 D.L. KENNEDY but their numbers will have been far from sufficient to supply the deficit. Two explanations are possible. First, the statement may be regarded as rhetorical exaggeration and we need regard it as indicating no more than the large increase of which we know. Second, Herodian may be referring not to the reign of Septimius, but, generally, to the period covered by his History (ea. A.D. 180-238). In this eventuality, quadrupling is possible. By the time at which he was writing, the emperor, when he was at Rome, brought units of the embryonic field army with him 94• Even if the large force of Osrh0eni had been dispersed after their abortive revolt against Maximinus 9 5 , other troops took their place 96 • In addition, Pupienus kept large numbers of auxiliaries from the German army at Rome as a counterwight to the restless praetorians 97 • These cannot have been much inferior in numbers to their rivals. In the circumstances the troops at Rome may well have been four times more numerous than at the death of Marcus 98 • The passage of Herodian need not, therefore, be regarded as a serious check to my suggestion since it can be plausibly explained otherwise. The evidence available is insufficient for certainty. It does, Herodian VII 8.9; VIII 7.7-8; 8.2. Herodian VII l.9-11 and note 2. 96 Herodian VIII 6.6; 7.8; 8.2; 8.5 and 8.7. 97 Regiments of catafracts and numbers of Parthian mercenaries and deserters entered Italy in Maximinus' field army along with Orsh6enian archers (Herodian VIII 1.3). 98 Marcus Pupienus Praetorians 10,000 15,000 Urban Cohorts 6,000 2,000 vigiles 7,OOO 3,500 eq. singulares 1,000 1,000 German auxilia ea. 15,000 alae of catafracts 1,500.(say, 3 alae) Moors say, 2,000 say, 1,000 Osrh6eni Parthians say, 1,000 16,5000 49,500 There were almost certainly other units attached to the new field army and the notional figures for the given additions may well be higher. (Cf. SHA, Car. 6.7, for equites extraordinarii under Caracalla, and Dio LXXIX 5.5 for Scytbians with Caracalla in Syria.) Much helpful criticism and advice has been given by Dr. A.K. Bowman, Dr. A.J. Graham, Professor S.S. Frere and Dr. 0. Murray. l am most grateful to them all, although the responsibility for the final text is mine. 94 9' PRAETORIAN GUARD 301 , meet the requirements of the situation. Should we no,t , consider the possibility that the increase in the Guard under Commodus (ea. A.D. 187/8)? D.L. KENNEDY