fMg 
 
DUKE 
 
 UNIVERSITY 
 
 LIBRARY 
 
Inthrraitjj ^tidies 
 
 Publtalfeb by % Hnthpraity of (Eutrimtatt 
 
 Series II. MARCH-APRIL, 1907. Vol. Ill, No. 2 
 
 THE ROMAN COHORT 
 CASTELLA 
 
 GEORGE H. ALLEN, Ph. D., 
 
 Assistant Professor of Latin, University of Cincinnati. 
 
 Issued Bi-Monthly. 
 
 University of Cincinnati Press, Cincinnati, O. 
 
 Entered February 24, 1906 at Cincinnati, Ohio, as second class matter, under 
 Act of Congress of July 16, 1894. 
 
CONTENTS 
 
 SIM 'll 
 C. 514S 
 S^t, X. 
 
 THE ROMAN COHORT CASTEELA 
 
 1. Choice of Location 9 
 
 2. Form and Area, considered in Connection with 
 
 the Size and Character of the Garrisons 13 
 
 3. Internal Arrangements 19 
 
 4. Fortifications 35 
 
 5. Commanders 
 
 a. Command of the “Cohort Castella” 39 
 
 b. Command of the “Numerus Castella” 42 
 
 PLANS 
 
 1. General Scheme of the Relationship of the Base 
 
 Lines and Central Building of the Castella to 
 their Front 22 
 
 2. Plan of the Central Building at Wiesbaden 30 
 
 3. Plan of the Central Building at Theilenhofen ... 31 
 
 4. Plan of the Central Building at Arnsburg 32 
 
 5. Cross Sections of the Ramparts at Urspring, rep- 
 
 resenting them as before and after the Erec- 
 tion of the Stone Wall 36 
 
 6. Sectional Representations of the Walls at Worth 
 
 as they were found lying in 1882 and 1887 and 
 as they must have stood originally 38 
 
 7. Plan of the Roman Caslellum at Chesters (Cilur- 
 
 num) 44 
 
 8. Map of the Limes in Germany and Rhaetia 45 
 
 9. Plan of the Roman Castellmn “Saalburg” 46 
 
 10. Southwestern Corner of the Fortifications of the 
 
 Saalburg (restored) 47 
 
 11. South Front of the Saalburg with Porta Prae- 
 
 toria (restored) 47 
 
 12. Porta Praetor ia of the Saalburg (restored) 48 
 
 13. Sacellum or Shrine at the Saalburg (restored)... 48 
 
Digitized by the Internet Archive 
 in 2016 with funding from 
 Duke University Libraries 
 
 https://archive.org/details/romancohortcaste01alle 
 
THE ROMAN "COHORT CASTELLA”. 
 
 The Romans were accustomed from the earliest times to fortify 
 their camps with ramparts or palisades whenever an army made a 
 halt, in time of war, if even for a single night. The arrangement 
 of these temporary camps, large enough to accommodate entire 
 consular armies, is described for the republican period by Poly- 
 bius (VI, 27-32), for the imperial epoch by Hyginus in a special 
 treatise ( Liber de Munitionibus Castroruni ) . 
 
 When Rorrrtan authority was extended over conquered lands 
 outside of Italy, the necessity arose for distributing the forces of 
 the standing army in accordance with the requirements of defense 
 and of maintaining military frontiers with garrisons stationed at 
 convenient intervals. This led to the erection of fortified camps 
 for permanent occupation. These are to be distinguished from 
 the temporary camps mentioned above. 
 
 The permanent camps varied in size according to the numerical 
 strength of the bodies of troops which they were intended to 
 shelter. The combination of large camps for the legions with a 
 series of smaller forts for the alae and cohorts is the arrangement 
 which was characteristic of the distribution of the forces on all 
 the boundaries under the empire. The immediate protection of 
 the frontier was regularly entrusted to the auxiliary troops, the 
 cohorts and alae, which were levied and recruited among the sub- 
 ject communities of the empire. The legionary camps, on land 
 frontiers at least, were always situated some distance to the rear 
 of the actual boundary. 
 
 The word ccistra was applied to the larger camps, both those 
 intended for armies on the march, as described by Polybius and 
 Hyginus, and these designed for permanent occupation by the 
 legions. The latter were distinguished as hiberna or castra 
 stativa. The smaller forts for the auxiliary cohorts and alae were 
 called castella. This term may also include still smaller fortresses 
 erected for the numeri 1 or for detachments (vex illationes) . It is 
 convenient to designate as “cohort castella,” or cohort forts, these 
 
 iWith regard to the composition and organization of the “numeri” 
 
 see note 3 , page 5 . 
 
 3 
 
fortified camps which were garrisoned by single bodies of the 
 auxiliary troops. It is the purpose of the present paper to describe 
 the. characteristic features of the “cohort castella.” But these 
 fortresses were erected, as we have observed, on the borders of 
 the empire, and formed part of a comprehensive system of defen- 
 sive works. It will be of advantage, therefore, by way of intro- 
 duction, to consider briefly the development of the fortified 
 boundary as a whole with special reference to the chronology of its 
 different portions. Our attention must be confined, for the most 
 part, to Britain and Germany, which contain nearly all the extant 
 remains of the “cohort castella.” 2 
 
 The development of a fortified land boundary in Germany 
 commenced with the permanent extension of Roman dominion 
 east of the Rhine. As early as the reign of Vespasian the Romans 
 had carried forward their outposts from Strassburg (Argentora- 
 tum) and Windisch (Vindonissa) to the valley of the Neckar. 
 Castella at Waldmossingen and Sulz date from this period. At 
 the same time an advance was made in Rhaetia to the north of 
 the Danube, where the castellum at Heidenheim goes back to the 
 first century. 
 
 But the decisive step leading to the establishment of a perma- 
 nent land boundary east of the Rhine was taken by Domitian as a 
 result of the war against the Chatti in 83 A. D. At that time the 
 valley of the Main was occupied by the Romans as far as Hanau 
 and the barbarians were driven from their fastnesses in Mt. 
 Taunus. The limes, or boundary, traced at that time, extended 
 from the Rhine near Neuwied to the Main at Hanau, following, 
 as far as possible, the summit of the Taunus range. It was repre- 
 sented, at first, merely by a road or cutting through the forest. 
 The cohorts and alae which were appointed for garrison duty in 
 the new territory were stationed in castella situated, not on the 
 actual boundary line, but some distance to the south of it. Wies- 
 baden, Hofheim, Heddernheim, Okarben and Friedberg are the 
 sites of the “cohort castella” of this period. On the limes itself, 
 at convenient intervals, smaller forts consisting of earthworks 
 
 2In the following outline I present mainly the conclusions of Krueger 
 ( Die Limesanlagen in nordliehem England, Jahrhuecher des Vereins von 
 Altertumsfreunden im Rheinlande, Heft 110, 1905) on the chronology of 
 the works in Britain, and of Febricius {Die Entstehung der roemischen 
 Limesanlagen in Deutschland, Trier, 1905) for those in Germany. 
 
 4 
 
were erected for the detachments which were detailed from the 
 castella for sentinel duty on the actual boundary. 
 
 Early in the reign of Trajan the boundary was extended along 
 the course of the Main from the termination of Domitian’s line as 
 far as Worth. From there the limes was carried across the Oden- 
 wald to a point on the Neckar near the mouth of the Jagst. The 
 river Neckar formed the continuation of the new line to a point 
 near Cannstadt, where the land boundary commenced again, and 
 was extended across the country as far as the northwest corner 
 of the province of Rhaetia. On this new line the “cohort cas- 
 tella” were erected close to the boundary and not in the interior. 
 Through the Odenwald there were “cohort castella,” such as 
 Oberscheidental and Neckarburken, and smaller forts of earth- 
 works corresponding to those of Domitian’s line. Seckmauern 
 is the only one of these earlier forts of earthworks which was not 
 re-erected in stone at a later period. 
 
 With the extension of this new line the legionary camp at 
 Windisch (Vindonissa) and the castella at Sulz and Waldmos- 
 singen were abandoned. These sites were henceforth in the 
 interior of the province, and would have had no strategic signifi- 
 cance. At this time, probably, the limes was continued along 
 the northern frontier of Rhaetia as far as the Danube at Eining. 
 Weissenburg and Pfiinz were erected as early as this, though 
 perhaps as earthworks. 
 
 Under Hadrian the cohorts and alae, which had been stationed 
 in the older forts between the Taunus range and the Main, were 
 transferred to the boundary and provided with new castella. 
 Most of these newer fortresses had defenses in stone. Hadrian 
 fortified the whole line of the limes from the Rhine to the Danube. 
 The continuous barrier which he erected consisted of a palisade 
 of stakes sunk into the ground to the depth of 1.4 meters and at 
 least 2 meters high above the earth. The upright stakes were 
 secured by stones placed against them in the ground as wedges 
 and tied together by posts fastened to them in a horizontal posi- 
 tion. 
 
 During the reign of Antoninus Pius, shortly after 142 A. D., 
 bodies of Brittones, organized as numeric made their appearance 
 
 3 In the wider sense of the word, “numerus” denoted any regular 
 body of soldiers under the command of a single officer. But, during 
 
 5 
 
along the limes in the Odenwald. These were probably Britons 
 who had been conquered in the campaign of Antoninus Pius when 
 the boundary in Britain was carried from the Tyne-Solway line 
 to the Forth and Clyde. The inscriptions in the Odenwald, be- 
 tween the Main and Neckar, offer evidence for a considerable 
 building activity at the time of the appearance of these new lines, 
 especially in 145-146 A. D. At this time the watch towers and 
 most of the castella of the smaller class on the line between the 
 rivers mentioned were re-erected in stone. 
 
 One of the “numerus castella,” if we may coin such a term to 
 denote the forts of this smaller class, was erected with defenses 
 in stone 200 meters east of the “cohort castellum” at Neckar- 
 burken. 
 
 About 155 A. D. a new boundary was laid off to the east of 
 the Odenwald-Neckar line. It left the Main at Miltenberg and 
 extended in almost a straight line to Welzheim near the northwest 
 corner of Rhaetia. At this time the cohorts on the inner line 
 were transferred to corresponding castella on the outer line, leav- 
 ing only the numeri in the former positions. The fortresses on 
 the new Miltenberg- Welzheim line were erected from the first in 
 stone. In fact, during all of this period, stone walls were grad- 
 ually taking the place of earthworks along the entire line of the 
 limes. 
 
 Under Commodus the Odenwald-Neckar line was given up 
 and the Brittones transferred to the corresponding positions on 
 the outer line. Additions to the “cohort castella,” or else smaller 
 forts near the larger ones, were erected to shelter the Brittones. 
 As a result we find fortified annexes, as at Osterburken, and 
 double “castella,” as at Welzheim and Oehringen. 
 
 The frontier fortifications in Germany reached the final stage 
 in their development with the erection of a stone wall, the so- 
 called “Teufelsmauer,” along the boundary in Rhaetia, and the 
 construction of the earth rampart with accompanying ditch, the 
 “Pfahlgraben,” on the German line. 
 
 the first three centuries of the empire, the term was rarely employed 
 in this general sense. A narrower meaning developed, denoting bodies 
 of troops which, being neither alae nor cohorts, had no special desig- 
 nation. In practice the numeri were distinguished from the alae and 
 cohorts by the fact that they were armed and organized in accordance 
 with the manner of warfare of particular nations embraced in the em- 
 pire. They were recruited in the localities where the influence of 
 Roman civilization had been least felt. 
 
 6 
 
The above outline furnishes the following- general data for 
 determining the chronology of the castella in Germany and 
 Rhaetia : 
 
 1. The castella situated at some distance behind the boundary 
 were erected in the latter part of the first century. Those of this 
 class which were situated north of the Main were abandoned 
 under Hadrian. 
 
 2. The “cohort castella” on the Odenwald-Neckar line were 
 erected during the early years of the reign of Trajan and con- 
 tinued in use until about 155 A. D. 
 
 3. The smaller earth forts through the Odenwald were prob- 
 ably erected during the early part of the reign of Trajan. They 
 were replaced, for the most part, about 145-146 A. D., by stone 
 castella which were occupied by numeri of Britons. The 
 “numerus castella” on the Odenwald-Neckar line probably con- 
 tinued in use until the reign of Commodus. 
 
 4. The castella along the Rhaetian frontier are not earlier 
 than Trajan. Most of the extant remains are still more recent, 
 the castella having been rebuilt from time to time. 
 
 5. The castella on the frontier between the Rhine and Main 
 are not earlier than Hadrian. The extant remains do not usually 
 go back to Hadrian because most of the castella, like those in the 
 preceding group, were rebuilt at different times. 
 
 6. The castella on the Miltenberg-Welzheim section of the 
 limes are not earlier than 155 A. D. 
 
 In Britain the earliest system of defenses corresponding with 
 those just described, which endured for any time, was the so- 
 called “Vallum of Hadrian,” which was carried across the island 
 from the Solway Firth, near Bowness, to the estuary of the Tyne. 
 In reality it consists of two parallel mounds or earth ramparts 
 with a fosse or ditch running between them. The remains of 
 about a dozen castella have been discovered to the south of the 
 “vallum.” Two of these are in stone; the others in earthworks. 
 
 These forts must have been the stations for the cohorts and 
 alae which were to defend the frontier. 
 
 Antoninus Pius erected a new barrier in Britain about 143 
 A. D. It extended from the Clyde to the Forth and consisted of 
 a rampart constructed of turf and a fosse. This new line, usu- 
 
 7 
 
ally called the “Antonine Vallum,” was provided with castella at 
 convenient intervals. But these forts appear to have been aban- 
 doned after about forty years. 
 
 Septimius Severus either gave up himself the more northern 
 boundary of Antoninus Pius or at least proceeded on the assump- 
 tion that its earlier abandonment was to be permanent. For he 
 re-enforced or replaced the existing works on Hadrian’s earlier 
 line by the construction of a new turf rampart and a ditch. The 
 course of this new vallum lay to the north of Hadrian’s barrier. 
 Although the interval between the two lines is usually only 25 to 
 50 meters and never more than 800 meters, new castella were 
 erected for the later line. While the earlier forts were separated 
 from Hadrian’s vallum by an interval, at times as great as one- 
 half or three-quarters of a mile, the later castella of the time of 
 Septimius Severus were so situated that the new barrier fell into 
 line with one of their sides. 
 
 At a later period the turf rampart was replaced by a stone 
 wall. The stone wall occupies, for the most part, the place of 
 the earlier turf rampart. The same castella served for the vallum 
 of Septimius Severus and the stone wall, but at the time of the 
 erection of the latter some of the forts seem to have been ex- 
 tended towards the north so. as to project for about a third of 
 their length beyond this later fortified line. 
 
 In view of the comparatively frequent changes involving the 
 whole system of defenses in Britain we may assume, perhaps, that 
 the remains of each fort go back to the time of erection of the 
 series of fortifications of which they form a part. For the neces- 
 sity of extensive reconstruction would scarcely have arisen during 
 the periods which intervened between the erection of the several 
 fortified barriers. In view of this supposition the British cas- 
 tella , in so far as they form a part of the frontier works, may be 
 divided, chronologically, into the following classes: 
 
 1. The forts connected with Hadrian’s vallum. We can 
 not determine whether these continued to be occupied after the 
 “Antonine Vallum” was laid out or not. 
 
 2. The forts on the “Antonine Vallum.” 
 
 3. The castella in line with the vallum of Septimius Severus. 
 
 4. The extensions of some of the preceding to the north of 
 the line of the stone wall. 
 
 8 
 
The castella in Germany and Britain may be conveniently dis- 
 cussed under the following headings: I. Choice of location; 
 
 2. Form and area, considered in connection with the size and 
 character of the garrisons ; 3. Internal arrangements ; 4. Forti- 
 fications ; 5. Commanders. 
 
 1. CHOICE OF LOCATION. 
 
 Both Hyginus (op. cit. 56) and Vegetius (III, 8) advise as 
 location for a camp a site sloping gently in the direction of the 
 front, which the former writer defines in this passage as the side 
 nearest the enemy. This rule is followed in most of the castella, 
 but there are at least six exceptions in which the surface of the 
 area enclosed by the forts descends in a direction opposite to that 
 of the enemy. 4 
 
 Vegetius advises also the selection of sites not commanded by 
 nearby elevations. But this principle was not strictly adhered to, 
 as is shown by the notable instance of the Saalburg, which is com- 
 manded on the southeast by the Gickelsberg. 5 
 
 The general policy seems to have been to choose positions en- 
 joying a good prospect without being inaccessible. The castella 
 lying near the boundary were usually placed where a section of 
 the limes fell within the field of vision. 
 
 The Romans almost invariably selected points where a copious 
 supply of good drinking water could be obtained from springs or 
 wells, either in the immediate neighborhood or within the walls 
 of the camps themselves. 
 
 An important factor in determining the location of the castella 
 which lay near the boundary was the position of the roads which 
 crossed the limes. For one of the most important purposes of 
 the barrier was to control communication to and fro. The bar- 
 barians were permitted to cross l-he limes only when disarmed and 
 under surveillance and upon payment of custom dues. The ex- 
 change of wares was limited to a few fixed points along the line. 6 
 
 •^Kernel, Hesselbach, Oberscheidental, Seckmauern, and the western 
 fort at Oehringen, and Niederberg. In general, sites seem to have 
 been selected which were as nearly level as possible. 
 
 SHolzhausen, Feldberg, Arzbach, Urspring, Heftrich, Lorch, Hunzel, 
 Pfuenz and Seckmauern might be mentioned in the same category. 
 
 OTac. Hist. IV, 64, Germ. 41. 
 
 9 
 
It was expedient, therefore, to erect the forts at or near the 
 points where roads crossed the limes and offered access to the 
 territory of the empire. On the barriers in Britain subsequent to 
 that of Hadrian the only openings giving an opportunity for com- 
 munication across the boundary line seem to have been the gates 
 of the forts, while in Germany the roads passed at the side of the 
 castella. 
 
 The intervals at which the castella were erected along the 
 boundary varied on different parts of the limes. The distance 
 from the Rhine to Welzheim on the limes in Upper Germany is 
 375 kilometers, including the section where the Main forms the 
 boundary. There were probably 36 castella connected with this 
 line. The average distance between them was, therefore, approx- 
 imately 10 kilometers. The Notitia Dignitatum , Occidentis, 
 chap. 38, ed. Booking, page 144, gives 23 forts on the later boun- 
 dary in Britain which was 117 kilometers in length. The average 
 interval would, therefore, be 5 kilometers. 7 
 
 On the line in Caledonia we find 19 castella for a distance of 
 59 kilometers, at an average interval of about 3 kilometers. 
 
 In considering the conditions which determined the selection 
 of sites for the castella it is of particular importance to notice their 
 location with reference to the limitcs 'or boundaries. For the 
 latter were all eventually fortified with continuous ramparts or 
 barriers of some sort with which most of the castella stood in 
 such intimate relationship as to form with them a common system 
 of defensive works. 
 
 The forts erected in connection with the vallum of Hadrian in 
 Britain were situated at some distance to the rear of its course, in 
 some instances as far as one-half to three-quarters of a mile. 
 
 In selecting the points where the castella on the “Antonine 
 Vallum” were to be erected a different practice was adopted. 
 They were placed in juxtaposition to the line of the new barrier, 
 in such a manner that the northern wall of each fort fell into line 
 with the course of the vallum. 
 
 7 It is possible that the average distance was really greater, for there 
 is some reason to believe that only eighteen of the twenty-three sta- 
 tions mentioned in the Notitia Dignitatum, were immediately connected 
 with the wall, the others being supporting forts to the north and south 
 of it; cf. Bruce, The Hand-hook to the Roman Wall, London, 1895, 
 page 27. 
 
 10 
 
The same method is exhibited in the works on the Solway- 
 Tyne line which we have ascribed to Septimius Severus. For the 
 northern walls of the stone forts built on the line of his turf 
 vallum were in alignment with the course of this frontier barrier. 
 
 This practice, which was followed in Britain after Hadrian, 
 was not observed in Germany, where Gross-Krotzenburg is the 
 only fort which touches the boundary rampart. While there is 
 great variety with regard to the distance of the other castella from 
 the limes , some degree of uniformity may be observed in the rela- 
 tive proximity or remoteness of the forts on particular sections of 
 the boundary. 
 
 We have already observed (page 4 ) that the “cohort castella” 
 north of the Main which were erected during the early period of 
 occupation were not located with special reference to the line of 
 the limes. 
 
 The distance from the limes of forts like Wiesbaden, Hof- 
 heim, Heddernheim and Okarben would be determined, therefore, 
 largely by fortuitous or purely local circumstances and would have 
 no significance for the present discussion. Moreover, some of 
 the castella south of the Main, as Heidenheim, Urspring, Wald- 
 mossingen and Sulz, clearly fall into the same class. 
 
 The circumstances connected with the abandonment of Ma- 
 rienfels seem to indicate, approximately, for the region north 
 of the Main, the distance from the limes beyond which castella 
 were no longer considered as having an organic connection with 
 the boundary system. For at the time of Hadrian’s reorgan- 
 ization of the frontier arrangements Marienfels, at a distance of 2 
 kilometers from the boundary line, was given up for Hunzel, 
 which was only 200 meters from the same line. It would 
 seem, therefore, that forts which were separated from the limes 
 by an interval of more than 2 kilometers, if not abandoned at this 
 time, must have owed their subsequent occupation to the strategic 
 significance of their own particular positions, independently of 
 the course of the limes. For this reason Niederberg (7 kilo- 
 meters from the limes), Heddesdorf and Ober-Florstadt (2500 
 meters) will not be taken into consideration in the present dis- 
 cussion. Besides, these three forts date from the early part of the 
 reign of Trajan when the arrangement of Domitian for the dis- 
 position of the castella was still in force, and before the limes was 
 provided with a continuous barrier. 
 
 11 
 
With these exceptions the castella on the line between the 
 Rhine and Main, so far as their distances can be ascertained, are 
 separated from the limes by intervals varying from 42 (Arzbach) 
 to 1200 meters (Arnsburg and Echzell). 8 
 
 On the water boundaries formed by the Main and Neckar the 
 castella are, for the most part, as near the rivers as the character 
 of the banks would permit. 
 
 On the Odenwald line Oberscheidental is 24.5 meters from the 
 limes. The “cohort castellum” at Neckarburken stands at a dis- 
 tance of about 260 meters from the line. The eastern castellum 
 at the same point, which was built for a numerus, is only about 
 25 meters from the boundary and the other “numerus castella,” 
 Schlossau, Hesselbach, Wiirzberg, Eulbach, Vielbrunn and 
 Lutzelbach, on the same line, are equally near it. Seckmauern, 
 which dates from the earliest period of Roman occupation in this 
 region, is traversed bv the limes. 
 
 On the outer line, which extends from near Miltenberg to 
 Welzheim, the castella stand at distances varying from 232 to 
 1500 meters. 9 
 
 The castella in the neighborhood of the Rhaetian boundary 
 fall into two classes with regard to their distance from the limes. 
 The first contains forts which are distant from 40 meters (Hal- 
 heim) to 2200 meters (Theilenhofen), the second a series of forts 
 at distances of from 6 to 14 kilometers. 10 
 
 In Britain there exist the remains of many detached forts be- 
 tween the vallum of Hadrian and that of Antoninus Pius and even 
 north of the latter, but in Germany only one instance occurs of a 
 fort erected beyond the frontier, the eastern castellum at Welz- 
 heim. 
 
 8 The distances, reckoned in meters, are as follows: Arzbach 42, 
 
 Hunzel 200, Holzhausen 70, Kernel 200, Zugmantel 396, Heftrich 50, 
 Feldberg 120, Saalburg 220, Capersburg 61, Langenhain 300, Arnsburg 
 1,200, Echzell 1,200, Altenstadt 1,000, Markoebel 200 and Rueckingen 200. 
 
 9Miltenberg lies 326 meters from the Main, which forms the bound- 
 ary at the point nearest its location. The distances of the other forts 
 are as follows: Wallduern 330, Osterburken 500, Jagsthausen 1,250, the 
 
 eastern fort at Oehringen 232, the western fort at the same place 1,300, 
 Mainhardt 626, Murrhardt 1,500, and the western fort at Welzheim 355. 
 
 iOThe nearer forts are: Lorch 1,200 (meters), Schierenhof 1,400, 
 
 Unterboebingen 1,000, Buch 1,200, Halheim 40, Ruffenhofen 2,100, Dam- 
 bach 100, and Theilenhofen 2,200. 
 
 The more distant castella are: Aalen 10 (kilometers), Weissenburg 
 
 6, Pfuenz 9, Pfoering 9, and Koesching 13. 
 
 12 
 
2. FORM AND AREA, CONSIDERED IN CONNECTION 
 WITH THE SIZE AND CHARACTER OF THE 
 GARRISONS. 
 
 The castella, as has already been mentioned, were garrisoned 
 by single alae, cohorts or numeri. No instance is known where 
 two alae, two cohorts, two numeri or a garrison consisting of the 
 combination of an ala and a cohort occupied a single castellnm. 
 But sometimes a numerns was stationed in company with a cohort 
 or ala. 11 
 
 The alae, or squadrons of horse, usually consisted of 500 men 
 ( alae quingenariae) . A few contained 1000 men ( alae mil- 
 iariae ). The cohorts, or divisions of foot, received a similar ap- 
 portionment of strength. About one quarter of the known cohorts 
 are cohortes miliariae. About the same proportion of all the 
 cohorts, miliariae and quingenariae together, are known to be 
 cohortes equitatae, mixed divisions of horse and foot. But the 
 real proportion was probably much greater. The cohortes mil- 
 iariae equitatae are said to have contained 760 infantry and 270 
 cavalry, the corresponding numbers for the cohortes quingenariae 
 being 380 and 120. 12 
 
 The numeri are said to have varied in numerical strength from 
 300 to goo. 
 
 Thus there was great diversity in the strength of the bodies 
 of troops assigned as garrisons to the different castella. We 
 should naturally expect the size of the garrisons to be the chief 
 factor in determining the area of the forts. But it was not suf- 
 ficient to establish a single prescribed area for bodies of 1000 men 
 and another for those of 500. It is obvious that cavalry must 
 have required more space than infantry, so that not only would 
 alae require larger camps than cohorts of the same numerical 
 strength, but cohortes equitatae would demand a greater space 
 than the cohorts which were made up entirely of foot soldiers. 
 When we consider further the numeri of all sizes and the fact 
 
 nin these instances, as was mentioned on page 6, accommodation 
 was usually provided for the numerns in smaller subsidiary forts, or in 
 fortified annexes attached to castella of the larger class. We find the 
 combination of a large and small fort at Neckarburken, Oehringen and 
 Welzheim, annexes at Camelon, Castlecary, Osterburken and other 
 points. 
 
 i2Cf. Hyginus, op. cit. 26 and 27. 
 
 13 
 
that the character of the sites chosen must have exercised some 
 influence in determining- the extent of the space which it was ex- 
 pedient to fortify, we shall not be surprised to find the greatest 
 variety in the area of the castella. 
 
 It will be instructive to compare the sizes of the castella with 
 the character of the bodies of troops which formed their garrison, 
 so far as these are known , 13 with a view to ascertaining the rela- 
 tion existing between the area of the camps and the numerical 
 strength of the bodies of troops for which they were intended. 
 
 In the following table I have arranged the castella in the order 
 of their size on a descending scale. It seemed expedient to sep- 
 arate the British castella from those in Rhaetia and Germany, for, 
 as will appear, the relation between area and numerical strength 
 in Britain was quite different from that which prevailed in the 
 other two provinces. 
 
 CASTELLA AND AREA GARRISONS, 
 
 reckoned in square meters. 
 
 1. 
 
 Aalen 
 
 60,740 
 
 
 2. 
 
 Okarben 
 
 57,966 
 
 
 3. 
 
 Echzell 
 
 54,891 
 
 
 4. 
 
 Heidenheim 
 
 52,845 
 
 
 5. 
 
 Niederbieber 
 
 50,905 
 
 
 6. 
 
 Welzheim (west) 
 
 42,748 
 
 
 7. 
 
 Pforing 
 
 38,774 
 
 Romanorum. 
 
 8. 
 
 Ruffenhofen 
 
 37,527 
 
 
 9. 
 
 Butzbach 
 
 33,345' 1 
 
 
 10. 
 
 Saalburg 
 
 32,593 
 
 
 civium Romanorum. 
 
 13 The statistics and other data concerning the German castella, both 
 in this present connection and in general throughout the paper, are 
 taken from the official reports of the excavation carried on under the 
 oversight of jthe Imperial Limes-Commission, as published under the 
 title, Der Obergermanisch-Raetische Limes des Roemerreiches im AuR 
 trage der Reichs-Limes-Eommission, Heidelberg, 1894—. The references 
 will be made to the different numbers, or Lieferungen, in which the 
 publication is appearing. For information regarding castella which are 
 not yet included in this work, I have had recourse to Von Cohausen, 
 Der Roemische Orenzwall in Deutschland. In the following table my 
 authority for the apportionment of the garrisons is Cichorius in his 
 articles on Ala and Cohors in Pauly-Wissowa, Real-encyclopaedie, Vol. I, 
 columns 1224-1270, and Vol. IV. columns 231-356. In a few instances 
 his views have been revised in the light of the more recent discoveries 
 as published in the reports of the Limes-Commission. 
 
 14 At an earlier period the fort contained only 27,262 square meters. 
 
 14 
 
GARRISONS. 
 
 CASTEEEA AND AREA 
 reckoned in square meters. 
 
 11. 
 
 Markobel 
 
 32,538 
 
 
 12. 
 
 Eangenhain 
 
 31,903... 
 
 igum. 
 
 ( cohors IX Batavorum equitata 
 
 13. 
 
 Weissenburg 
 
 30,945... 
 
 ( ala I Hispanorum Auriana. 15 
 
 14. 
 
 Obernburg 
 
 29,800... 
 
 tata civium Romanorum. 
 
 15. 
 
 Arnsburg 
 
 29,538... 
 
 
 16. 
 
 Marienfels 
 
 28,500... 
 
 
 17. 
 
 Ober-Florstadt 
 
 28,365... 
 
 
 18. 
 
 Heddesdorf 
 
 28,364... 
 
 
 19. 
 
 Niederberg 
 
 27,993... 
 
 
 20. 
 
 Theilenhofen 
 
 27,440... 
 
 
 21. 
 
 Pfiinz 
 
 27,390... 
 
 
 22. 
 
 Mainhardt 
 
 27,217... 
 
 
 23. 
 
 Oberscheidental 
 
 20,694 
 
 24. 
 
 Miltenberg 
 27,200' 0 
 
 25. 
 
 Riickingen 
 
 26. 
 
 Gorch 
 
 
 (28. 
 
 27. 
 
 Bockingen I 
 19, 876. | 29. 
 
 
 1 
 
 30. 
 
 Wiesbaden 
 
 31. 
 
 Benningen 
 
 21,856 
 
 cohors I Sequanorum et Rau- 
 
 racorum equitata. 
 
 25,200 cohors III Delmatarum quin- 
 
 24 684 genaria [as shown by the 
 
 mention of a prefect in an 
 inscription, cf. page 39.] 
 
 Oehringen (west) 'J 
 
 OehringVnW) [ Cohors 1 Helvetiorum " 
 22,289 J 
 
 22,687 
 
 32. Murrhardt 
 
 21,905 ...cohors XXIV voluntariorum 
 civium Romanorum. 
 
 33. 
 
 Hofheim 
 
 21,873 18 
 
 
 34. 
 
 Niedernberg, 
 
 21,766 
 
 
 35. 
 
 Wahlheim 
 
 21,666 
 
 .cohors I Asturum (3). 
 
 36. 
 
 Gross-Krotzenburg 21,525 
 
 .cohors IV Vindelicorum. 
 
 37. 
 
 Dambach 
 
 21,505 
 
 
 38. 
 
 Neckarburken 
 
 39. Osterburken • 
 
 
 (west) 20,777 
 
 21,423 .... 
 
 .cohors III Aquitanorum civi- 
 um Romanorum. 
 
 40. 
 
 Buch 
 
 20,950 
 
 
 41. 
 
 Waldmossingen 
 
 20,507 
 
 
 15. These two bodies occupied the fort successively (cf. Lieferung XXVI, page 25). 
 
 16. When the outer line from Miltenberg to Welzheim was traced under Antoninus 
 Pius, the cohort which had been stationed at Oberscheidental was transferred to Milten- 
 berg. In general, the garrisons of the inner line (excepting the numeri ) were moved to 
 the nearest stations on the outer line. I have placed the corresponding castclla together, 
 for the sake of comparison, whenever their garrisons are known to have been the same. 
 
 17. There exists evidence pointing to an earlier occupation of Bockingen by cohors 
 V Delmatorum. 
 
 18. The later stone castellum. 
 
 IS 
 
GARRISONS. 
 
 CASTEEEA AND AREA 
 reckoned in square meters. 
 
 42. 
 
 Schierenhaf 
 
 20,504 
 
 
 43. 
 
 Unterbobingen 
 
 20,042 
 
 * 
 
 44. 
 
 Zugmantel 
 
 18,096 
 
 
 45. 
 
 Urspring 
 
 17,874 
 
 
 46. 
 
 Sulz 
 
 17,500 
 
 
 47. 
 
 Capersburg 
 
 16,401 
 
 
 48. 
 
 Welzheim (east) 16,360 
 
 
 49. 
 
 Holzhausen 
 
 14,300 
 
 orum. 
 
 50. 
 
 Neckarburken (east) 9,775 
 
 sium. 
 
 51. 
 
 Walldiirn 
 
 8,134 
 
 
 52. 
 
 Worth 
 
 7,682 
 
 
 53. 
 
 Hunzel 
 
 7,484 
 
 
 54. 
 
 Arzbach 
 
 7,389 
 
 
 55. 
 
 Heftrich 
 
 7,240 
 
 
 56. 
 
 Feldberg 
 
 7,240 
 
 
 57. 
 
 Kernel 
 
 7,159 
 
 
 58. 
 
 Heldenbergen 
 
 7,050 
 
 
 59. 
 
 Halheim 
 
 6,694 
 
 
 60. 
 
 Vielbrunn 
 
 6,214 
 
 
 61. 
 
 Seckmauern 
 
 6,094 
 
 
 62. 
 
 Wiirzberg 
 
 5,905 
 
 
 63. 
 
 Hesselbach 
 
 5,905 
 
 
 64. 
 
 Schlossau 
 
 5,789 
 
 tiensium. 
 
 65. 
 
 Eulbach 
 
 5,730 
 
 
 66. 
 
 Trennfurth 
 
 5,574 
 
 
 67. 
 
 Eiitzelbach 
 
 5,189 
 
 
 BRITISH CASTEELA AND GARRISONS. 
 
 AREA IN SQUARE YARDS. 
 
 1. Camelon 28,800 
 
 2. Chesters (Cilurnum) 27,400 ala II Asturum. 
 
 3. Housesteads (Borcovicium) 24,000 cohors I Tungrorum miliaria. 
 
 4. Castlecary 17,690 
 
 5. Greatchesters (Aesica) 16,600 cohors I Asturum. 
 
 The area of the following' British forts on the line of the 
 Roman wall can only be given approximately in acres (on 
 the authority of Bruce, op. cit.). 
 
 Wallsend (Segedunum) 3)4 cohors IV Lingonum equitata. 
 
 Rutchester (Vindobala) 3)4 cohors I Frisiavonum. 
 
 Halton Chesters (Hunnum) 4 )4 ala Sabiniana. 
 
 Carrawburgh (Procolitia) 3)4 cohors I Batavorum. 
 
 Chesterholm (Vindolana) 3% cohors IV Gallorum equitata. 
 
 Caervoran (Magna) 3)4 cohors II Dalmatarum. 
 
 Birdoswald (Amboglanna) 5)4 cohors I Aelia Dacorum. 
 
 16 
 
[The names of the castella in Britain along the boundary on 
 the Solway-Tyne line, per linearn vaUi, are given in the Notitia 
 Dignitatum. Occidents, chap. 38, page 114. The application of 
 the names to the forts as far as Birdoswald (Amboglanna) is 
 established on the basis of epigraphic evidence. West of Birdos- 
 wald (Amboglanna) the correlation between the ancient names 
 and the sites of the camps, which are known, cannot be determined 
 on account of the lack of inscriptions.] 
 
 Considering first the castella on the limes in Germany and 
 Rhaetia, we notice that they fall into five general classes with 
 respect to their extent. The first of these divisions contains five 
 castella , varying in area from Aalen with 60,740 square meters to 
 Niederbieber with 50,905. It is evident from the garrison of 
 Heidenheim that these forts were intended for alae miliariae. The 
 next class, consisting of three forts, was doubtless the class of 
 castella for the alae quingenariae. These varied in size from 
 42,748 down to 37,527 square meters. In considering the camps 
 of the cohorts it is evident from a comparison between Ober- 
 scheidental and Miltenberg, which were occupied in succession by 
 a cohors equitata. that we must make a distinction between the 
 practice of the earlier and later period. For the period of the 
 occupation of the outer, Miltenberg-Welzheim, line, the difference 
 in size between Miltenberg and Riickingen may be taken, per- 
 haps, as representing the point of division between the class of 
 castella intended in general for cohortes equitatae and those 
 erected for the cohorts which contained no cavalry, although the 
 first of these classes contains Weissenburg, which, with respect 
 to the relation of its area with the size and character of its garri- 
 sons, is quite exceptional. For it was occupied successively by a 
 cohors equitata miliaria and an ala. If this division be warranted 
 on the basis of the evidence at hand, the larger of these classes 
 contains fifteen castella, with areas which vary from 33,345 down 
 to 27,200 square meters. The next class, that of the ordinary 
 cohorts, would then extend from 25,200 to I4,300. 19 
 
 iSThe only inscriptional evidence which we have pointing to the 
 presence of soldiers at the eastern fort at Welzheim mentions the 
 Brittones Exploratores, a numerus, although the fort seems too large to 
 belong to the numerus class. It has been suggested (Lieferung XXII, 
 page 28) that the Cohors Antoniniana was divided, one part being sta- 
 tioned at Holzhausen, the other at the neighboring castellum Zugmantel. 
 
 17 
 
In Britain the camps intended for the alae seem to have had 
 an area of from about 4)4 acres (23,490 square yards), as at 
 Halton Chesters (Hunnum), to about 5)4 acres (27,400 square 
 yards), as at Chesters (Cilurnum). They were evidently much 
 smaller than the forts of the corresponding- class in Germany. 
 We find in Britain, moreover, evidence for the size of the castella 
 which were garrisoned by cohortes miliariae. Housesteads (Bor- 
 covicium), with an area of approximately 24,000 square meters, 
 was the camp of the cohors I Tungrorum, which was known to 
 have been a division of 1,000 men. The cohors I Aelia Dacor- 
 um , stationed at Housesteads (Amboglanna) [area 5)4 acres, or 
 28,710 square yards] was almost certainly a cohors miliaria, since 
 its commanders were tribunes. 20 
 
 The evidence at hand shows no distinction in size between the 
 forts of the cohortes quingenariae equitatae and those of the same 
 numerical strength which contained only infantry. They seem to 
 have had, generally, an area of a little over 3 acres. 
 
 The usual form of the castella is that of a rectangle, more or 
 less elongated and with rounded corners. 21 
 
 The castella , of which the dimensions are known, exhibit a 
 considerable range of variation in the extent of their elongation, 
 or, in other words, in the ratio of the width to the length. At 
 Housesteads (Borcovicium), where the oblong form is most pro- 
 nounced, the length of the shorter axis is 60 per cent, of that of 
 the longer one. At Osterburken the width is 61 per cent, of the 
 length. These are the extreme examples. Hyginus (op. cit. 21) 
 says that the camp should have a width equal to two-thirds of 
 its length. An investigation, however, into the relation of the 
 width to the length of the castella will show that the practice was 
 not uniform. Omitting Halheim, which is square, and Waldmos- 
 singen, which is not rectangular, we may examine the castella 
 mentioned in the list above (pages 14-16) with regard to the rela- 
 
 20See page 39. 
 
 2iThe earlier earth forts at Hofheim and Waldmoessingen had the 
 form of irregular polygons. The later fort at Waldmoessingen, although 
 a quadrilateral, was far from being a parallelogram. I have over- 
 looked slight irregularities in the form of some of the other castella. 
 Halheim and Kesselstadt were square in plan, but Kesselstadt, owing 
 to its excessive area (140,625 square meters), can not be included in 
 the list of “cohort castella.” It must have been a legionary camp 
 (cf. Lieferung X, Kesselstadt, page 6). 
 
 18 
 
tion between the length of their axes. The breadth of 5 of the 
 65 castella to be considered is from 60 per cent, to 70 per cent, 
 of their length. In 17 others it lies between 70 per cent, and 80 
 per cent. In 23 it is between 80 per cent, and go per cent., and in 
 21 between go per cent, and 100 per cent. 
 
 The castella were usually provided with four gates, one open- 
 ing in each side. But some of the smaller “numerus castella” had 
 only three gates, namely. Liitzelbach, Vielbrunn, Eulbach, Wiirz- 
 berg, Hesselbach and the eastern fort at Neckarburken, while one, 
 Halheim, has but two. Mention has already been made of the 
 extension of some of the fortresses in Britain, making them pro- 
 ject beyond the line of the mural barrier which crossed the island. 
 These larger castella have six gates, two of them being on each of 
 the sides which would be cut by the stone wall if it were extended. 
 Chesters (Cilurnum) is a good example of the forts of this class. 22 
 
 The probable names of the gates may be more conveniently 
 discussed in connection with the next topic. For, as will appear, 
 the identity of the several gates can be fixed only by determining 
 first the names of the main roads of the camp which connect them. 
 But the roads cannot be treated apart from the consideration of 
 • the whole interior arrangement with which their position and 
 direction is intimately connected. 
 
 3. INTERNAL ARRANGEMENTS. 
 
 Before attempting to distinguish the various parts of the cas- 
 tella it will be convenient to discover a standard or criterion for 
 determining with certainty the direction in which each fort is to 
 be considered as facing. For the identity of the front not only 
 gives us a key to the general interpretation of the interior arrange- 
 ment but determines the assignment of the names to the different 
 gates. 
 
 It is usually supposed that the front of the castella was the 
 face exposed to the enemy, and observation has shown that the 
 side opposed to the enemy is usually one of the short sides. But 
 this opinion is based upon statements of our ancient authorities 
 which have reference to the larger camps, the castra; yet it is of 
 
 22See the plan of this castellum on page 44. It will be observed that 
 the stone wall, or continuous barrier, reaches the ramparts of the fort 
 just south of the more northern gateways on the long sides. 
 
 19 
 
fundamental importance, throughout the present discussion, to 
 distinguish clearly between the large, temporary camps ( castra ) 
 and the smaller, permanent castella. 
 
 Since we have no direct literary or inscriptional evidence to 
 guide us in our enquiry concerning the front of the castella we 
 naturally have recourse to the well-known descriptions of the 
 castra with a view to establishing the front of the smaller camps 
 on the basis of the apparent analogy of their arrangements with 
 those of the larger fortresses. But the fact is generally ignored 
 that the testimony of the ancient authors is not in agreement with 
 regard to the front of the castra. The divergence of their ac- 
 counts might suggest that the practice followed in the orientation 
 of the castra was not uniform, at least as regards their depend- 
 ence upon external circumstances. 
 
 The camp is variously represented as facing the east (Hygi- 
 nus, de Limitibus Constituendis , page 69; Nissen, das Templum, 
 page 11 ; Vegetius I, 23), the enemy (Vegetius I, 23 ; Hyginus, de 
 Munitionibus Castrorum , 56) and the direction from which water 
 and provisions could be most conveniently brought (Polybius VI, 
 27, 3). 23 
 
 But fortunately the accounts given by the ancient authorities 
 themselves concerning the disposition of the area enclosed within 
 the ramparts of the castra offer sufficient data for determining 
 the front independently of any conditions external to the fort- 
 resses themselves by establishing a principle of uniformity in its 
 location based upon a fixed relationship which it bears to some of 
 the interior parts. The castra like the castella were rectangular 
 in form. Two lines, called the cardo maximus and decumanus 
 maximus , which crossed at right angles, served as base lines for 
 the orientation of the camp as well as for the plotting of the space 
 within it (Polybius VI, 28). 
 
 The course of the cardo was represented by a principal road 
 which crossed the camp in a direction parallel to the front (Poly- 
 bius VI, 27, 7 to 30, 3; cf. Hyginus, Liber de Munitionibus Cas- 
 trorum , 14). This road was called the via principalis (Livy, X, 
 33, 1). The presence of a main road traversing the area of the 
 
 23Yet in his account of the siege of Carthagena, Polybius (X, 9, 10) 
 represents Scipio as placing his camp to the north of the town, with its 
 front toward the south and the defenses of the enemy. 
 
 20 
 
camp, therefore, reduces the problem of determining the front to 
 the selection of one of the two sides with which this road is par- 
 allel. The other main road, the via praetoria, which corre- 
 sponded with a section of the course of the decumanus, did not 
 traverse the whole camp, but what is especially significant about 
 its position, it extended from the via principalis to the front line 
 of the castra (Polybius VI, 28). Its presence, therefore, iden- 
 tifies as the front of the camp the side with which it is in contact. 
 But to fix with greater precision the relative position of the front 
 we may consider, briefly, the location of the praetorium or head- 
 quarters. The orientation of this space, which was rectangular 
 in plan, was so adjusted with reference to the car do and decu- 
 manus (or they to it) that its front fell into line with the margin 
 of the via principalis or else extended in a direction parallel to it 
 at a distance of 50 feet (Polybius VI, 27, 1-7). The praetorium 
 faced the line of the via praetoria and the front of the camp (Hy- 
 ginus, de Munitionibus Castrorum, 12; Polybius VI, 28). Thus 
 the front of the camp can be determined by the orientation of the 
 praetorium as well as by the position of the via praetoria, while 
 the course of the via principalis is of importance in establishing 
 these two determining factors. 
 
 The castra had four gates, one at each of the extremities of 
 the two base lines (Hyginus. de Limitibus Constituendis, page 
 180; Josephus III, 5, 2). The gates at the extremities of the via 
 principalis were called the portae principals and distinguished as 
 dextra or sinistra according to their position on the right or left 
 side of the camp (Hyginus, de Munitionibus Castrorum, 14; Livy 
 IV, 19, 8, XXXIV, 46, 9 and XL, 27, 4.) The gate which 
 opened in the rear of the camp was called the porta decumana 
 (Livy, X, 32, 9). The remaining gate, the porta praetoria, and 
 the porta decumana are often mentioned as opening, the former 
 in the side exposed to the enemy (cf. Hyginus, de Munitionibus 
 Castrorum, 56; Festus, Ep., page 233; Tac. Hist. IV, 30), the 
 latter in the side furthest removed from the enemy (cf. Caesar, 
 Bellum Gallicum, III, 25, 2; Tacitus Ann. I, 66). These state- 
 ments, in view of the fact that the porta praetoria is evidently the 
 front gate, being opposite the porta decumana, of which the posi- 
 tion at the rear is known, are proof that the castra usually faced 
 the enemy. Yet, as has been shown, we may reasonably assume 
 
 21 
 
Front. 
 
 CO 
 
 P 
 
 P 
 
 C 
 
 § 
 
 P 
 
 o 
 
 p 
 
 Q 
 
 CARDO 
 
 CARDO 
 
 bo 
 
 a 
 
 rt 
 
 a 
 
 co 
 
 CO 
 
 p 
 
 p 
 
 S 
 
 P 
 
 a 
 
 p 
 
 P 
 
 General Scheme of the Relationship of the Base Lines and Central 
 Building of the Castella to their Front. 
 
 22 
 
that this practice was not uniform in all cases, but that exceptions 
 were made as expediency demanded. 
 
 We may now turn our attention again to the castella. These 
 follow, in their general arrangements, the plan of the larger 
 castra , and, in particular, invariably display in their interior the 
 same essential features, two main roads and the rectangular space 
 corresponding to the praetorium in the same relative positions 
 which we have observed in the castra. It seems unquestionable, 
 therefore, that these parts bear the same relation to the front of 
 the castella that the corresponding parts of the larger camps have 
 been shown to bear to their front. No argument can be brought 
 against this assumption except that in some instances its applica- 
 tion would clash with the popular notion that the orientation of 
 the castella was determined by the position of the enemy. But 
 the supposition that the castella always faced the enemy is based, 
 as we have seen, upon a statement (cf. Vegetius, I, 23; Hyginus, 
 de Munitionibus Castrorum , 56), which applies primarily to the 
 castra. I have shown, moreover, that this assertion cannot be ac- 
 cepted as an infallible rule, even when applied to the castra. 
 Therefore, its assumption as the standard for the universal prac- 
 tice in the castella is not justified on the ground of analogy. I 
 have shown, on the other hand, that the front of the castra can 
 always be determined on the basis of its relationship to the via 
 praetoria and praetorium, and we are justified in assuming that the 
 same relationship holds in connection with the corresponding 
 parts of the castella. 
 
 It is doubtful to what extent we are warranted in applying to 
 the castella the names of the corresponding parts of the castra. 
 But in view of the fact that the roads and gates of the smaller 
 camps agree so closely, in their general position, with those of the 
 larger ones, it is expedient to follow the usual custom and apply 
 to the former the authorized designations of the latter. 24 
 
 In connection with the more practical problem of determining 
 the front of the castella of which remains are extant, it is often 
 necessary to ascertain the position of the via principalis as pre- 
 
 24An inscription has been found mentioning the erection of the 
 porta praetoria at Kutlovica in a castellum (C. I. L. Ill, 7,450). It is 
 reasonable to suppose, therefore, that the names of the three remain- 
 ing gates in the castella were the same as those of the corresponding 
 gates in the castra. 
 
 23 
 
requisite to establishing the identity of the via praetoria. For a 
 third road, probably called the via decumana, leaves the rear line 
 of the castella at right angles and, like the via praetoria, extends 
 in the direction of the center of the camp. It is obvious, there- 
 fore, that in castella in which the remains of the central building 
 are not sufficient to indicate the direction in which it faces and 
 where only vestiges of a short section of the via praetoria are ex- 
 tant, and these lying near the porta praetoria, a confusion might 
 arise regarding the identity of the via praetoria and via decumana 
 which would render it impossible to distinguish between the front 
 and rear of the castellum. But this confusion would be readily 
 dispelled by ascertaining the location of the via principalis. For, 
 unlike the via praetoria, the via decumana does not reach the via 
 principalis but terminates at the rear line of the space reserved for 
 the central building. 25 
 
 A discovery made at Hofheim may be cited as evidence to 
 show that the main roads in the castella, as in the larger camps, 
 represented the courses of the lines upon which the orientation 
 and interior division of the enclosures was based, and, conse- 
 quently, that their direction was of equal importance in determin- 
 ing the position of the front. In the stone fort at Hofheim (cf. 
 Lieferung VII, Hofheim, page 4) two clearly defined furrows were 
 found, crossing at right angles in front of the central building 
 and extending under the main roads. These must have repre- 
 sented the cardo and dccumanus. The same phenomenon was 
 noticed at Heddernheim, Okarben and Zugmantel. 
 
 Proceeding to apply the method developed above for identify- 
 ing the front of the castella we may consider, a little more in de- 
 tail, the relative location of their parts. But first it will be of 
 interest to determine how nearly the supposed rule that the cas- 
 tclla face the enemy approximates to the general statement of 
 fact. Thirty-eight of these forts may be enumerated on or near 
 the limes in Britain and Germany having sufficient remains of the 
 central buildings, which correspond to the praetoria in the castra, 
 and of the main roads, or at least the gates by which the position 
 
 2oln the castra, as restored from the accounts of Polybius and 
 Hyginus, the via decumana either does not exist, or appears as a name- 
 less road of insignificant length. 
 
 24 
 
of the roads may be determined, to establish, in the case of each 
 fortress, the identity of the front . 26 
 
 Twenty-six of these face in the general direction of the enemy, 
 although their fronts are not always parallel with the course of 
 the limes . £7 
 
 One fort, Walldiirn, presents an angle to the limes, the direc- 
 tion of which may be taken, in general, as representing the direc- 
 tion of the enemy. The other eleven present either a side or their 
 back to the enemy . 28 
 
 The statement is usually made that the front of a fortified 
 camp is one of its short sides. This is substantially correct, as 
 applied to the castella, for the fort at Castlecary is the only ex- 
 ample of a castellum of which the breadth exceeds the length. 
 
 In the castra described by Polybius and Hyginus the via decu- 
 niana is laid out in such a position in relation to the front and rear 
 as to divide the space enclosed by the ramparts of the camp into 
 parts containing, respectively, about one-third and two-thirds of 
 the entire area. But in the camp of the earlier period the via 
 principalis was nearer the porta decnmana than the porta prae- 
 ioria. In the camp of the imperial period we find this relation 
 reversed. It will be of interest to examine the castella with a 
 view to ascertaining what practice was more often followed with 
 respect to this feature of their interior arrangement. 
 
 Limiting our investigation to forty-four forts , 29 in which the 
 
 26Heidenheim, Urspring, Wiesbaden, Hofheim and Okarben are 
 omitted, although the direction of the front is apparent in each. Their 
 location, at a considerable distance from the limes, makes doubtful the 
 particular direction which we should consider as that of the enemy at 
 these localities at the time of the erection of the forts. Camelon is 
 also excluded from the present discussion. It lies outside the line of 
 the limes in Caledonia. 
 
 27They are: Castlecary, Chesters (Cilurnum), Arzbach, Hunzel, 
 Feldberg, Arnsburg, Ober-Florstadt, Markoebel, Rueckingen, Gross- 
 Krotzenburg, Niedernberg, Obernburg, Woerth, the western fort at 
 Neckarburken, Boeckingen, Benningen, Miltenberg, Osterburken, Main- 
 hardt, Murrhardt, Welzheim, Schierenhof, Unterboebingen, Theilen- 
 hofen, Pfuenz and Pfoering. 
 
 28The Saalburg, Weissenburg and Niederbieber face in a direction 
 opposite to that of the enemy. Housesteads (Borcovicium) , Butzbach, 
 Niederberg, Holzhausen, Buch, Heltrich and Aalen offer their left side 
 to the enemy, Langenhain its right side. 
 
 29The thirty-eight considered in the discussion above (pages 24-25), 
 with Heidenheim, Urspring, Wiesbaden, Hofheim, Okarben and 
 Camelon, which were purposely omitted (cf. Note 26). 
 
 25 
 
position of the central buildings, main roads and front is known, 
 we find that in thirty-five the via principalis is nearer the porta 
 praetoria , 30 in four it is nearer the porta decumana 31 and in five it 
 is central or practically central. 32 
 
 Mommsen, in an article on the practorium ( Hermes , Yol. 
 XXXV [1900], p. 442), to which we shall have occasion to refer 
 later, points out what he considers a ready means for identifying 
 the direction in which the castclla face. He observes that these 
 forts are usually oblong in shape, the sides being longer than the 
 ends, and that the gates in the side walls are placed at a point 
 twice as far from the back as from the front. If this were true 
 in all cases, the front could be identified readily enough, of course, 
 as the end to which the portae principales are nearest. But we 
 have already seen that the relative position of the side gates and 
 via principalis is not uniform. The application of Mommsen’s 
 rule for determining the front would result in confusion in the 
 five forts where the position of the via principalis is central 33 and 
 lead to an erroneous conclusion in the four examples where the 
 via principalis is nearer the porta decumana. 
 
 Although the via principalis is more often nearer the porta 
 praetoria than the porta decumana , such is not the case with the 
 central buildings. Examining the same castella from which the 
 data were drawn for the previous enquiry with regard to the posi- 
 tion of the via principalis , 3i we find only five instances in which 
 the central building is nearer the porta praetoria , 35 
 
 In sixteen castella the central building is central or very nearly 
 
 30Camelon, Castlecary, Housesteads (Borcovicium), Hofheim, Okar- 
 ben, Niederbieber, Butzbach, Niederberg, Arzbach, Hunzel, Holzbausen, 
 Feldberg, Saalburg, Langenhain, Arnsburg, Markoebel, Gross-Krotzen- 
 burg, Niedernberg, Woerth, the western fort at Neckarburken, Boeckin- 
 gen, Benningen, Miltenberg, Wallduern, Osterburken, Mainhardt, Murr- 
 hardt, the western fort at Welzheim, Urspring, Schierenhof, Unterboeb- 
 ingen, Weissenburg, Theilenhofen, Pfuenz and Pfoering. 
 
 siChesters (Cilurnum), Heftrich, Rueckingen and Aalen. 
 
 32Wiesbaden, Ober-Florstadt, Obernburg, Heidenheim and Buch. 
 
 33Cf. note 32. 
 
 34Cf. note 29. 
 
 35Housesteads (Borcovicium), Niederbieber, Saalburg, Butzbach and 
 Aalen. 
 
 36Castlecary, Hofheim, Okarben, Hunzel, Langenhain, Arnsburg, 
 Markoebel, Gross Krotzenburg, Niedernberg, Woerth, Boeckingen, Oster- 
 burken, Murrhardt, Theilenhofen, Pfuenz and Pfoering. 
 
 26 
 
In twenty-three it is nearer the porta decumana than the porta 
 praetoria . 37 
 
 We would naturally suppose that the porta praetoria was the 
 most important of all the gates of the castella. For it offered a 
 means of exit either in the direction of the enemy or the source of 
 supplies. Moreover, to those entering the castellum it gave ac- 
 cess to the road which led directly to the front of the central 
 building. The construction of the gates of many of the camps 
 testifies to this importance of the front gate, as will appear from 
 the following consideration. 
 
 The gates were either single or double. In eight of the cas- 
 tella of which the identity of the front is certain 38 all the gates are 
 double . 39 In seventeen examples all the gates have single por- 
 tals . 40 But in thirteen of these fortresses the gates are not uni- 
 form in this respect . 41 In these thirteen castella the distinction 
 between double and single portals doubtless points to a corre- 
 sponding difference in the importance of the gates. Three of 
 them have double portae praetoriae with the other gates single . 42 
 In six we find all the gates double except the porta decumana . 43 
 At Theilenhofen the portae principals are double, the other gates 
 single. At Niederberg the porta principalis de.vtra is double, the 
 others single. Langenhain has a double porta principalis sinistra 
 with the others single, while at Arnsburg both the porta praetoria 
 and the porta principalis dextra are double and the other gates 
 single. 
 
 37Camelon, Chesters (Cilurnum), Wiesbaden, Arzbach, Niederberg, 
 Holzbausen, Heftrich, Feldberg, Ober-Florstadt, Rueckingen, Obernburg, 
 Miltenberg-, the -western castellum at Neckarburken, Benningen, Wall- 
 duern, Mainhardt, the western fort at Welzbeim, Heidenbeim, Urspring, 
 Schierenbof, Unterboebingen, Buch and Weissenburg. 
 
 38The forty-four forts mentioned in connection with the discussion 
 on page 25. 
 
 39Housesteads (Borcovicium), Chesters (Cilurnum), Obernburg, the 
 western fort at Welzheim, Aalen, Buch, Pfuenz and Pfoering. 
 
 40Camelon, Castlecary, Niederbieber, Wiesbaden, Arzbach, Hunzel, 
 Heftrich, Ferdberg, Ober-Florstadt, Woerth, Rueckingen, the western 
 castellum at Neckarburken, Benningen, Miltenberg, Wallduern, Oster- 
 burken, Mainhardt. 
 
 4iThe character of the gates in the remaining six of the forts, 
 where the front is known, cannot be determined. These are Heiden- 
 heim, Markoebel, Gross-Krotzenburg, Niedernberg, Schierenhof and Un- 
 terboebingen. 
 
 42The Saalburg. Holzhausen and Murrhardt. 
 
 43Hofheim, Okarben, Urspring, Butzbach, Boeckingen and Weissen- 
 burg. 
 
 27 
 
Thus in the thirteen castella, where the gates are not uniform, 
 there are ten double portae praetoriae, nine double portae prin- 
 cipales dextrae, eight double portae principales sinistrae, but not 
 a single example of a double porta decumana. Therefore, while 
 the porta praetoria seems to be the most important gate, it is still 
 more obvious that the porta decumana is the least important. 
 
 The application to the central building of the “cohort castella” 
 of the term praetorium, which denotes the space allotted to the 
 tent of the commander-in-chief in the larger temporary camps, is 
 probably inaccurate. Von Domaszewski 44 and Mommsen 45 both 
 called attention to this and their views have been generally ac- 
 cepted; yet it is difficult to avoid the use of the term praetorium, 
 since it has come to be commonly used in this connection. 
 
 The term praetorium from its original significance as the com- 
 mander’s tent came to have the more general meaning of the 
 commander’s presence as in the expression in praetorio militate. 
 It came also to be applied to residences and villas intended for the 
 use of the emperor or provincial governors. In inscriptions the 
 reference is usually to the villas of the latter. 
 
 But the central buildings of the castella are not residences, nor, 
 if they were, could the term praetorium be appropriately used for 
 edifices intended for the use of subordinate officers such as the 
 commanders of the auxiliary cohorts and alae. According to von 
 Domaszewski the central building was called the principia. An 
 inscription found in Scotland in 1903, subsequently to voft Dom- 
 aszewski’s article, may be cited as adding to his evidence to 
 establish the correctness of the term principia in this connection. 
 This inscription, published in the Westdeutsche Zeitschrift 
 (Korrespondenzblatt, 1903, page 202), mentions the dedication 
 of the principia by cohors VI N erviorum. 
 
 Forty-four forts have been enumerated in which the remains 
 are sufficient to determine the location of the central building. 46 
 But we may omit from the following discussion nine of these 
 castella where the vestiges of the central buildings are too slight 
 to offer any serviceable data for ascertaining their plan and in- 
 
 uNeue HeideTberger Jahrbueclier, XI (1SS9), page 142 ff. 
 
 45 Hermes , XXXV (1900), pages 437-442. 
 
 46Mentioned in connection with the discussion on page 25. 
 
 28 
 
terior arrangement. 47 However, we should include Waldmos- 
 singen, although it was excluded from the previous investigation 
 of the relative positions of the central building and main roads on 
 account of the irregularity of its general plan. Our examination, 
 therefore, into the form and structure of the central buildings 
 will extend to thirty-six examples. 
 
 In twenty-four instances it has been possible to discover the 
 dimensions of these buildings, and their areas are found to vary 
 between 400 and 3,951 square meters. 48 
 
 Some of the earlier central buildings seem to have been erected 
 in wood. 49 We also find examples of a combination of wood and 
 masonry in different parts of the same building even in the later 
 period. 50 At Urspring the remains point to a method of construc- 
 tion resembling a technique which is very much in vogue in Ger- 
 many to-day, the central building being what is known as a 
 Fcichwerkbau; cf. Lieferung XXIV, Urspring, page 11. At in- 
 tervals of a meter perpendicular wooden supports were set up in 
 the ground. These were afterwards imbedded in a wall of rough 
 stones and mortar so that their outer faces alone appeared as a 
 framework dividing the masonry into regular sections or panels. 
 The impression of the posts is still visible in the remains of the 
 wall, although the wood has disappeared. These vertical sup- 
 ports were probably secured in position by horizontal tie-pieces 
 at suitable intervals. 
 
 Most of the central buildings known to us were built with 
 walls of solid masonry. 
 
 47The castella to be omitted are Camelon, Castlecary, Housesteads 
 (Borcovicium), Hofheim, Okarben, Gross-Krotzenburg, Benningen, Mil- 
 tenberg and Wallduern. 
 
 48The areas, reckoned in square meters, in the order of their ex- 
 tent, are as follows: Heidenheim 3951, Aalen 3708, western fort at 
 
 Welzheim 2758, Butzbach 2172, Obernburg 2060, Pfoering 2052, Ober- 
 Florstadt 1892, Saalburg 1880, Niederberg 1654, Pfuenz 1618, western 
 castellum at Neckarburken 1554, Theilenhofen 1539, Weissenburg 1360, 
 Langenhain 1350, Arnsburg 1302, Murrhardt 1283, Buch, 1088, Cilur- 
 num 972, Waldmoessingen 907, Wiesbaden 807, Urspring 806, Feldberg 
 672, Hunzel 541, Heftrich 400. 
 
 49Remains of the wooden central building at Seckmauren have been 
 found (cf. Lieferung XIX, Seckmauern, page 3.) 
 
 soThe greater part of the central building at Feldberg was in wood, 
 only a small portion in stone. Upright posts a foot square were sunk 
 into the ground to the depth of one meter at intervals of 2.50 to 3 
 meters. A siding of boards was probably nailed to these supports (cf. 
 Lieferung XXV, Feldberg, page 8.) 
 
 29 
 
The central buildings contained either one or two open, rec- 
 tangular courts. 
 
 In camps where two courts are found their relative position 
 correspond with that of the atrium and peristyle in a Roman 
 house. The first court is often called the atrium in the modern 
 literature dealing with the subject. In general the earlier forts 
 had central buildings with only one court , 51 and the development 
 of these buildings is from a simple to a more complicated plan. 
 
 Plan of the Central Building at Wiesbaden. 
 
 The central building at Wiesbaden may be taken as a fair 
 example of the earliest type of these buildings, the one in the 
 Saalburg as representing their characteristic form in the latest 
 period . 52 
 
 siMurrhardt is the only known example of a castellum subsequent 
 to the reign of Trajan with a single court. Aalen and Pfoering, in 
 Rhaetia, have double courts, although they go back to the period of 
 Trajan. But in both of them the second court may be a later addition. 
 
 52Compare the plans of the central building at Wiesbaden with that 
 of the Saalburg indicated on the map of the fort on page 46. 
 
 30 
 
The central building at Wiesbaden, which is nearly square in 
 plan (26.27x30.75 meters), consists of a rectangular court with 
 a series of five small rooms opening upon it at the rear. This 
 court was once surrounded by a portico. The bases of the twenty- 
 four columns which supported the roof have been discovered. 
 Substantially the same plan is found in the castella at Urspring 
 and Waldmossingen and probably at Pfiinz. Besides, remains at 
 Weissenburg point to an earlier arrangement of the central build- 
 
 Plan of the Central Building' at Theilenhofen. 
 
 ing which would conform to this general type. At a later period 
 small rooms or long halls, serving, perhaps, as armories, were 
 laid out along the sides of these single courts. Theilenhofen is a 
 good example of this later development of the building with single 
 court where series of rooms have been added. 
 
 A comparison of the central building at Theilenhofen with the 
 one at Arnsburg will point to what was perhaps the line of devel- 
 opment which led from the simpler to the more complicated form 
 
 31 
 
of these buildings. At Theilenhofen the two passages which open 
 into the rear part of the court suggest the alae of a Roman house 
 in connection with the atrium. The hall or open space which 
 crosses the central building at Arnsburg represents, perhaps, the 
 enlargement of these two passages in combination with the rear 
 part of the earlier single court. In any case, the transverse hall or 
 space, as at Arnsburg, was the prototype of the rear court in the 
 buildings with two courts. For all the inner courts are oblong 
 in plan, with their longer sides extending in a direction parallel 
 
 
 
 1 
 
 1. 
 
 R 
 
 I 
 
 i n 
 
 I u 
 
 s 
 
 0 
 
 Cl 
 
 \J 
 
 1 
 
 LJ 
 
 n" 
 
 Si 
 
 sj 
 
 
 Plan of the Central Building at Arnsburg. 
 
 with the front of the central buildings. Chesters (Cilurnum) 
 shows the simplest form of central building in which there are no 
 rooms at the sides of the courts. (See the map of this castellum 
 on page 44). In most instances the narrow spaces at the sides of 
 the outer court are occupied by rooms or long halls. 
 
 The buildings thus far mentioned have been without heating 
 apparatus, apses or corresponding rectangular projections and 
 cellars. 
 
 In the central buildings of the later period one or more rooms 
 were usually provided with hypocausta, that is, floors which were 
 
 32 
 
elevated sufficiently to permit of the passage of hot air beneath 
 them. These hollow spaces were connected with furnaces. The 
 rooms which were warmed in this manner were usually the guard 
 rooms and offices. 
 
 A chamber at the rear of the central building was regularly 
 employed as a sacellum or shrine. In the earlier castella, those 
 in which the central buildings resembled the one at Wiesbaden, 
 the shrine was undoubtedly the central chamber in the series of 
 five rooms which opened on the court from the rear. It occupied 
 a corresponding position at the rear of all subsequent castella. 
 But in the earlier period the sacella were not distinguished archi- 
 tecturally from the other rooms in the series. Later they were 
 often constructed with rounding or rectangular projections reach- 
 ing beyond the rear wall of the central building. The apse is the 
 commoner form of this projection . 53 In two castella (Buch and 
 Saalburg ) 54 the shrines, although rectangular in plan, extend for 
 a part of their length beyond the rear wall of the central building. 
 In one instance (Heftrich) the whole shrine forms an extension 
 to the central building, while in three (Pforing, the western cas- 
 tellum at Welzheim and Osterburken) the walls of the shrine are 
 erected in the form of an apse within but with no outer projection. 
 
 The military standards were placed in the sacellum together 
 with the statues of the military deities. 
 
 It was customary for the soldiers to deposit their savings with 
 the signiferi , probably because the latter had charge of the signa 
 to which divine honors were paid. The expression employed in 
 this connection, deponere ad signa , is significant. For the money 
 was actually placed in the presence of the symbols of the military 
 deities. In the later period cellars u r ere provided for the security 
 of the savings together, probably, with the military chest. These 
 cellars were usually under the shrine. They have been found in 
 at least eight castella. 
 
 In the temporary camps a space in front of the praetorium, at 
 the junction of the via praetoria and via principalis, was set aside 
 
 53lt occurs in fifteen examples among the thirty-six castella under 
 consideration in connection with the central buildings. Murrhardt is 
 the only one of these fifteen central buildings which is known to have 
 only one court. But Murrhardt is an exception, chronologically, in 
 having a single court, since it was not erected before 155 A. D. 
 
 54The former has only one court in its central building. 
 
 33 
 
for assemblies of the soldiers, the exercise of military jurispru- 
 dence and the execution of penalties. The excavations at 
 Urspring, one of the oldest forts in Rhaetia, have revealed the 
 arrangement of the corresponding space in the castella of the 
 earliest period of the permanent defensive works on the northern 
 frontier. In the fortress mentioned a space of 520 square meters 
 (40 x 13 meters) in front of the central building was enclosed by 
 a wooden barrier or fence, 55 so as to form an open court. This 
 enclosure could be entered from the front by means of a large 
 gateway 56 which opened onto the via praetoria. The barrier 
 enclosed the central section of the via principalis, but smaller gates 
 opened at each end of it, giving access to the portions of this road 
 which were not included within the enclosing fence. 
 
 At a later period we find the corresponding spaces in front of 
 the central buildings occupied by covered halls. These are prob- 
 ably the basilicae mentioned by Vegetius as intended for military 
 exercises in wet weather. They must, like the earlier uncovered 
 spaces, have served also as the places of assembly for the soldiers. 
 
 In several of the castella there may have been assembly halls 
 erected in perishable material which have completely disappeared. 
 But the wooden basilica at the Feldberg is the only one of which 
 the outline can still be determined from the remains. Its area 
 was 20.30 x 8.85 meters and its walls were constructed with up- 
 right posts sunk into the ground to which a siding of boards was 
 probably nailed (cf. Lieferung XXV, p. 8). 
 
 The remains of ten basilicae are known to us which were 
 erected in stone. The area enclosed by their walls varies from 
 478 to 1225.8 square meters. 57 The basilicae really form part of 
 the central buildings having common walls on the side where they 
 adjoin them. They are always oblong in plan, their longer sides 
 being parallel with the front of the central buildings and the direc- 
 tion of the carao. 
 
 The permanent stone basilicae are generally of the later period, 
 although four ( Weissenburg, Murrhardt, Buch and Theilen- 
 
 55Cf. Lieferung XXXIV, Urspring. 
 
 56Massive stones to support the gate posts have been found. 
 
 57The areas of the basilicae in these nine camps, reckoned in square 
 meters, are as follows: Aalen 1225.8, Theilenhofen 1140, Butzbach 
 
 1007.5, Welzheim 978.2, Buch 699.75, Unterboebingen 624.58, Pfoering 
 569, Weissenburg 553.5, Saalburg 520, Murrhardt 478. 
 
 34 
 
hofen) out of the ten, which are known to us, are found in con- 
 nection with central buildings which have only single courts. But 
 in the case of one of these four, Weissenburg, the basilica is 
 known to be a later addition 58 and the same may be true of 
 others . 59 
 
 Six of the basilicae mentioned are longer than the front of 
 the central buildings to which they are connected and project 
 beyond them . 60 
 
 At the Saalburg five doors open from the basilica (area 40 X13 
 meters) into the covered portico surrounding the atrium or fore- 
 court. On the right hand side of this court, as one enters it from 
 the basilica, there is a long, narrow hall, supposed to have served 
 as the armamentarium or armory. The corresponding space on 
 the west side of the court is occupied by four small rooms, the 
 purpose of which is not known. 
 
 The spaces on both sides of the shrine at the rear of the inner 
 court are occupied by porticoes where statues and altars of the 
 deified emperors were erected. On the east and west sides of the 
 inner court were small chambers, heated by hypocausts. One of 
 these was probably intended as an excubitorium, or room for the 
 soldiers on guard at the shrine. Others may have been occupied 
 by the tabularium or account and record office of the camp. 
 
 The central building was the most essential and most char- 
 acteristic structure in the castcllum. In several camps the re- 
 mains of the horrea or storehouses have been found. In general 
 no remains have been found which could be identified as buildings 
 intended as lodgings for the soldiers, and it is probable that in 
 most camps the troops were sheltered in tents or wooden struc- 
 tures. But stone barracks have been found at Housesteads (Am- 
 boglanna) and the Chesters (Cilurnum) in Britain. 
 
 4. FORTIFICATIONS. 
 
 It is usually believed that the earlier castella were fortified by 
 means of earth ramparts and that a general feature of their later 
 
 58Cf. Lieferung, XXVI, page 14. 
 
 59But, as has been pointed out already (in the note on page 30), 
 Murrhardt is an exception to the general practice that the buildings 
 of the later period had two courts. 
 
 60Theilenhofen, Buch, Weissenburg, Pfoering, Butzbach and Welz- 
 heim. 
 
 35 
 
development was the gradual substitution of stone walls for the 
 defenses in the more primitive material (cf. page 6). 
 
 At Hofheim and Waldmossingen stone forts replaced earlier 
 ones in earthworks, as has been shown by the excavations. 
 Marienfels, an earth fort, was abandoned at the time of Hadrian 
 (Lieferung XX, Marienfels, page 7). Seckmauern (Lieferung 
 
 Cross Sections of the Ramparts at Urspringr, representing them as 
 before and after the Erection of the Stone Wall. 
 
 XIX, Seckmauern, page 4) , an earth fort in the Odenwald, must 
 have been abandoned at an early period, since the limes (as repre- 
 sented by Hadrian’s palisade) crosses it. But it probably repre- 
 sents the type of earth fort which preceded the stone “numerus 
 castella” at Liitzelbach, Vielbrunn, Eulbach, Wurzberg, Hessel- 
 bach, Schlossau and Neckarburken, the other “numerus castella” 
 
 36 
 
on the Odenwald line. 61 In Britain stone castella were probably 
 not erected earlier than the reign of Antoninus Pius, 62 and most of 
 those on the line of the vallum in Caledonia were probably of 
 earth. 63 
 
 The remains at Urspring are a striking illustration of the 
 transition from earth to stone. It appears from the excavations 
 that the castellum was originally defended by a massive earth 
 mound or rampart, the top of which, having a width of 4.20 
 meters, served as a platform for the defenders. On the inner side 
 the mound sloped from the elevation of the platform down to that 
 of the interior of the fort, but on the outside it presented a per- 
 pendicular elevation, being faced with boards which were sup- 
 ported and held in place by upright posts set in the ground at 
 intervals of 1.50 meters. These vertical posts were anchored by 
 being fastened to slanting beams and placed within the mass of 
 earth. Later a stone wall, about two meters in thickness, was 
 added on the outside as a facing or revetment to this structure of 
 earth and wood. 
 
 We should not suppose, however, that stone castella were 
 never erected in the earlier period. For the stone forts at Hof- 
 heim, Sulz, Heidenheim and Okarben must have been constructed 
 before the time of Hadrian, if, as is commonly believed, they were 
 abandoned under his reign. 
 
 The stone walls either stood alone or were backed by earth 
 embankments, as at Urspring in the later period. The latter prac- 
 tice was probably more common, although in several instances it 
 can be shown that no earth mound could have existed. Thus at 
 Okarben (Lieferung XVI, Okarben, page 4) ashes, broken pot- 
 tery and other refuse is found quite near the inner side of the 
 walls at the level of their base. 
 
 6iSee Lieferung XIX, Seckmauern, page 2, and Lieferung XXIII, 
 Luetzelbach, page 7. 
 
 62Two of the forts lying to the rear of the vallum of Hadrian (cf. 
 page 14) are fortified in stone, Aesica and Magnae, but their fortifica- 
 tions may have been restored in the better material at a later time. 
 For these two castella appear to have been occupied as late as the time 
 of the Notitia Dignitatum (cf. Occidentis, chap. 38, ed. Boecking, page 
 114). 
 
 63Castlecary on the Antonine vallum with stone walls is probably 
 exceptional. Camelon, 1100 yards north of the vallum, had earth ram- 
 parts 40 feet thick at the base. 
 
 37 
 
The walls vary in thickness, ranging from about one to about 
 two meters. For a long time their height was in doubt. But a 
 fortunate discovery made in connection with the examination of 
 the site of the castellum at Worth has established, approximately, 
 the height of the fortifications at that point. In 1882 a section 
 of the wall was uncovered lying on the inner slope of the ditch 
 where it had fallen. All except the upper part lay, as it had for- 
 merly stood, as a single, cohesive mass of concrete with stone 
 
 Sectional Representations of the Walls at Worth as they were found 
 lying- in 1882 and 1887 and as the 3 T must have stood originally. 
 
 facing. After a foundation layer of 70 centimeters followed 32 
 courses of squared blocks of stone with a length, altogether, of 
 4.15 meters. Beyond this the stones forming the concrete had 
 lost their adhesion and stretched in confusion for a distance of 85 
 centimeters. The appearance of the structure as a whole was 
 such as to suggest the conjecture that the loose stones at the end 
 represented the breastworks and that the real height of the wall 
 with accompanying embankment of earth was 4.15 meters. 
 
 A similar investigation was made at Worth in 1887. At the 
 point where the cutting was made the wall had lost its facing of 
 stones for a distance of 2.30 meters. Beyond this point the facing 
 stones were intact on the outer side (lying underneath as it fell) 
 for a distance of 3.20 meters. At that point a slab of stone was 
 
 38 
 
found forming- part of a cornice which must have marked the 
 summit of the wall and the foot of the breastworks. These 
 dimensions, if we allow 70 centimeters for the foundation, point 
 to an elevation of 4.80 meters for the platform at the top of the 
 wall, apparently disagreeing with the results of the earlier exca- 
 vation. But the previous results are more reliable, for the part 
 of the wall investigated in 1887 seems to have broken in falling, 
 so that the upper, more intact portion may have slipped further 
 down the side of the ditch than would otherwise have been the 
 case. 
 
 The walls of the Saalburg (as restored) have a height of 4.80 
 meters, including the breastworks and battlements. But the plat- 
 form for the defenders on the earth mound has an elevation of 
 only 2.20 meters. 
 
 The fortifications of the castella were re-enforced on the out- 
 side by one or more ditches. In Scotland forts of this class have 
 been excavated with as many as five and six ditches. In England 
 and in Germany, with the exception of Wiesbaden, no castella are 
 known which had more than two ditches. But Weisbaden had 
 three ditches on three sides, although the fourth side seems to 
 have had no ditch at all (cf. von Cohausen, op. cit. page 170). 
 The ditches were rarely more than 2 meters in depth. 
 
 5. COMMANDERS. 
 
 A. COMMAND OF THE “COHORT CASTELLA.” 
 
 Since each fort was garrisoned by a single ala or cohort, it 
 follows that the praefects and tribunes, who commanded these 
 bodies of auxiliary troops, were by the nature of their position, 
 commanders of the “cohort castella.” The officers in command 
 of the alae bore the title of praefect ( praefectus alae) , those in 
 charge of the cohorts of 1,000 men were usually tribunes 
 ( tribunus militant cohortis), of the others praefects ( praefectus 
 cohortis) . 64 
 
 A phenomenon which appears in many of the castella along 
 the boundaries of the empire should be taken under consideration 
 as throwing further light upon the administration of these fort- 
 resses. In many of the castella we find inscriptions indicating 
 
 64Cf. the articles of Cichorius as already cited in note 13. 
 
 39 
 
the presence of legionary centurions. The fact is striking, for 
 the majority of all inscriptions, which point to the presence of 
 legionary centurions in the provinces outside the camps of their 
 legions during their period of service, are found in castella gar- 
 risoned by the cohorts and alae. It is reasonable to suppose that 
 a legionary centurion who had been detailed from his legion and 
 stationed for some time in the camp of a division of the auxiliary 
 forces would be second in command of the castellum, taking rank 
 in this respect after the tribune or praefect of the cohort or ala. 
 I have attempted to exclude from this discussion all inscriptions 
 which are probably contemporaneous with the date of construc- 
 tion of the castella or mural barriers, such as the stone wall in 
 Britain. For the presence of legionary centurions at the time of 
 the construction of such works may be explained by the fact that 
 these officers were frequently summoned from a distance to super- 
 intend building operations. Their presence, therefore, at such a 
 time, would not imply a share in the administration of the 
 fortress. 65 After making this omission the following inscriptions 
 may be cited to prove that it was a common practice to station 
 legionary centurions in the camps of the auxiliary troops even 
 under normal circumstances. I give in each instance the site and, 
 so far as known, the name of the ala or cohort forming the garri- 
 son of the castellum. 
 
 I. BRITAIN. 
 
 1. Benwell (Condercum). Ala I Hispanorum. C. I. L. 
 VII, 503. A votive offering to the dens Antenociticus by a cen- 
 turion of legio XX Valeria Victrix. C. I. L. VII, 506. The 
 rendering of a vow to Dolichenus in behalf of Antoninus Pius by 
 a centurion of legio II Augusta. The fragment C. I. L. VII, 
 514, is similar. 
 
 2. Littlechesters (Vindolana). Cohors IIII Gallorum. C. 
 I. L. VII, 702. Dedication by a centurion of legio VI Victrix. 
 This inscription cannot be dated, but the dedication to Fortuna 
 makes it probable that the centurion was regularly stationed at 
 this point (cf. Jahrbb. d. Ver. v. Altertumsfr. im Rheinlande, 
 vol. 60, page 52). 
 
 65See G. H. Allen, Centurions as Substitute Commanders of Auxili- 
 ary Corps, University of Michigan Studies, Humanistic Series I, pages 
 354 - 356 . 
 
 40 
 
3. Greatchesters (Aesica). Cohors II Asturum. A cen- 
 turion of legio VI erects a sepulchral stone. 
 
 4. Corvoran (Magnae). Cohors I Hamiorum. A centurion 
 of legio II Augusta dedicates to Fortuna. 
 
 5. Manchester, probably the site of the camp of Cohors I 
 Frisiavonum. C. I. L. VII, 21 1. Dedication to Fortuna by a 
 centurion of legio VI Victrix. 
 
 6. Bowes. Cohors I Thracum. Restoration of a building 
 under the direction of a centurion of legio VI Victrix. 
 
 7. Whitley Castle. Cohors III Nerviorum civiurn Roman- 
 orum. Dedication to Hercules by a centurion of legio VI Vic- 
 trix. 
 
 II. GERMANY. 
 
 1. Ems, known to be the site of a castellum, although the 
 body of troops forming the garrison has not been identified. The 
 rendering of a vow to Fortuna by a centurion of legio VIII 
 Augusta. 
 
 2. Wiesbaden. Cohors II Raetorum. Brambach, C. I. Rh., 
 1529. A legionary centurion, c(enturio) leg(ionis) VII [1], 
 dedicates to Apollo. 
 
 3. Seligenstadt. Cohors I civiurn Romanorum. C. I. Rh,. 
 1406. A centurion of legio XXII Primigenia dedicates an altar, 
 204 A. D. 
 
 4. Miltenberg. Cohors I Sequanorum et Rauracorum. 
 Jahrbb. d. Ver. v. Altertumsfr. im Rhcinlande, vol. 60, page 52. 
 
 6. Osterburken. Cohors III Aquitanorum. Lieferung II, 
 page 38. Dedication to Minerva by T. Atlonius Forlio . . . 
 
 c(enturio) ex comic (ulario et Placidius Placidinus c(enturio) . 
 The inscription does not state, it is true, that these men were 
 legionary centurions. But a comparison of this inscription with 
 the one cited above at Miltenberg and another found at Oehr- 
 ingen (Lieferung V., Oehringen, page 27) will show that the 
 first of the two individuals was probably a legionary centurion, 
 also the further consideration that carnicularii were regularly ad- 
 vanced to the position of legionary centurions. 60 Further ex- 
 
 66This is shown by the following inscriptions: C. I. L. Ill, 3846; 
 
 VI, 414, 1645; VIII, 702; IX, 5338; X, 1763; XI, 3108, 5693; XIII, 1832; 
 C. I. Rh. 1304, 1559. 
 
 41 
 
amples of the same nature appear in Rhaetia ; C. I. L. Ill, 5876, 
 '5937; in Dacia; C. I. L. Ill, 1334, 7858; in Cappadocia; C. I. L. 
 Ill, 242; and in Africa ; C. I. L. VIII, I759I. 07 
 
 Most of the inscriptions mentioned refer to thank offerings such 
 as officers might be expected to make at the successful termination 
 of a definite portion of their service. Five are dedications to For- 
 tuna. It is known that the tribunes and prefects of equestrian 
 rank, who were commanders of the bodies of auxiliary troops, 
 commonly made such offerings, especially to Fortuna, at the com- 
 pletion of their year of command. We naturally infer that the 
 centurions also, after the termination of the period of appointed 
 service in the same localities, paid the vows which they had made 
 upon their arrival. It seems, therefore, to have been the policy 
 under the empire to station in many of the castella legionary cen- 
 turions as well as officers of equestrian rank. 08 
 
 These centurions, then, were second in command of the cas- 
 tella where they were stationed, acting, perhaps, as chiefs-of-staff, 
 to the prefects and tribunes. Their position might be compared 
 to that of the praefecti castrorum in the legionary camps. 
 
 B. COMMAND OF THE "NUMERUS CASTELLA/'’ 
 
 As has already been observed (page 40), the smaller castella, 
 those which contained an area of less than 9,775 square meters, 
 were assigned to the numeri. The area of these castella varied 
 greatly, and according to Mommsen, Hermes XIX (1884), page 
 288, there was a corresponding diversity in the size of the bodies 
 of troops which were assigned to them as garrisons, the numeri 
 ranging in strength from 300 to 900. 
 
 Both centurions with the title praepositus (C. I. L. VIII, 
 2494, 9745, 18007, 18008, Brambach, C. I. Rh., 1739; Cagnat, 
 
 670ne might cite here the inscriptions C. I. Rh. 1559; C. I. L. Ill, 
 8484; III, 141472, 141474; Ephem. Epig. VII (1892), No. 1071; West- 
 deutsche Zeitschrift, XI (1892), Korrespondenzblatt SI, all of which 
 mention dedications made by auxiliary corps at their headquarters 
 under the supervision of legionary centurions. But there is some 
 reason to suppose that these centurions were in actual command of the 
 bodies of troops mentioned. 
 
 6SA striking example is found in Tacitus, Annals XII, 45-46; partic- 
 ularly significant are the words — Castellum Gorneas, tutum loco 
 acpraesidio militum, quis Caelius Pollio praefectus, centurio Casperius 
 praeerat. 
 
 42 
 
D Annie Eftigraphique, Revue Archiologique , Vol. XXVI 
 (1895), page 275, No. 20, Vol. XXX (1897), No. 118, Vol. 
 XXXVII (1900), page 510, No. 197 ( and officers of equestrian 
 rank with the titles jqraefectus (C. I. L. VIII, 9906, 9907; C. I. 
 Rh. 991; C. I. G. 67713), praefositus (C. I. E. II, 1180; VIII, 
 9047, 9358, 9962; XI, 3104; Cagnat, LA Annie Ivfiigrafthique, 
 Revue Archiologique , XIV (1889), page 443, No. 187) and 
 tribunus (C. I. E. VIII, 9381, 11,343; Ephem. Epig. VII 
 (1892), No. 1092) are found in command of the numeri. 
 
 No chronological distinction between the command of numeri 
 by centurions and by officers of equestrian rank can be made on 
 the basis of the evidence at hand. But in view of the variety in 
 rank and dignity of the commanders, which seems to point to a 
 wide range of diversity in the importance of the commands, and 
 of the great difference in size of the “numerus castella,” it seems 
 probable that the essential factor in determining the choice and 
 title of the commanders was the numerical strength of the numeri 
 themselves. 
 
 43 
 
CILURNUM 
 
 Plan of the Roman Castellum at Chesters (Cilurnum), from 
 Bruce, Handbook to the Roman Wall. 
 
 44 
 
Nr. 5 KasteU Hunzel liegt sudlioh von Punkt 5. Nr. 61 » Kastell Sulz sudlich von Rottonburg, Nr. 61 i Kastell Waldmossingen sQdwestlich von Rottenburg. 
 
 obergermanlschen und raetlschen Limes. 
 
 • feil^oslellleo, Ovormuthete# r6ml«chM Kastell. 
 
Plan of the Roman Castellum “Saalburg''’. 
 
 46 
 
By permission of "Records of the Past”. Washington, D. C. 
 Southwestern Corner of the Fortifications of the Saalburg (restored). 
 
 By permission of "Records of the Past”. Washington. D. C. 
 South Front of the Saalburg' with Porta Praeturia (restored.) 
 
 47 
 
By permission of “Records of the Past' 1 . Washington. D. C. 
 
 Fort a Praeioria of the Saalburg- (restored). 
 
 By permission of “Records of the Past'', Washington. D. C. 
 
 Sacellum or Shrine at the Saalburgf (restored). 
 

 
 
 / 
 
L