¹ cf. Davies, R. W. in Epigraphische Studien iv, 110 f.Google Scholar

² Hyginus, de munitione castrorum, 21. This writer is sometimes referred to as Pseudo-Hyginus since this tract has been wrongly assigned to the known writer Hyginus the gromaticus. Lenoir's Budé edition of PseudoHygin (Paris, 1979)Google Scholar, which discusses some of the problems raised in this article, came to my hands too late for use.

³ Vegetius, epitoma rei militaris ii, 6 and 8. Vegetius has a first cohort of double strength, but his detailed breakdown is peculiar to himself and in places demonstrably wrong. It seems clear that, as elsewhere in his work, Vegetius was supplementing scanty information found in his sources with calculations of his own, designed to fill in the details as best he could. No reliance should be placed on them.

⁴ H. von Petrikovits, Die Innenbauten römischer Legionslager wahrend der Phnzipatszeit (1975), 38–43, 119 ff. (cited below as Innenbauten).

⁵ Writing of events in Africa in A.D. 20 Tacitus (Annales iii, 21) records a victory won by a vexillum veteranorum ‘not more than 500 in number’. In A.D. 15 after various abortive reforms Tiberius had fixed legionary service at 20 years to be followed by a further 5 sub vexillo, with duties limited to combat only. A. von Domaszewski (Rangordnung 79 f.) identified these as missicii, quoting CIL iii 2037 (cf. 8579): Quartus Iuventius missicius leg(ionis) xi anniprum) xxxv h.s.e. Dying at 35, Quartus could not have reached the age of final discharge. Keppie, however (‘Vexillum Veteranorum’, PBSR xli (1973), 8–17, esp. p. 10)Google Scholar, shows that other missicii died at very advanced ages, and he quotes evidence that the term missicius most probably indicated merely a soldier in retirement. Thus this term is irrelevant to the problem.’ The word missici on the Adamklissi funerary altar (CIL iii p. 2316⁵⁰, cf. Dacia N.S. v (1961), 358, fig. 3B), for instance, though at first sight hopeful, seems to appear among the auxiliaries and is best taken, as Keppie suggests, as an abbreviated nomen. It certainly does not form a sub-heading.

⁶ CIL iii 6178 (v Macedonica); AE 1955 No. 238 (ii Traiana); CIL iii 14507 (vii Claudia). Cf. Breeze, D., JRS lix (1969), 50 ff.Google Scholar The evidence is circumstantial. The inscriptions list soldiers discharged on single occasions. In the first (v Macedonica) at least 40 names appear under cohort i, 17 under cohort ii, at least 14 under cohort iii, at least 10 under cohort iv and at least 12 under cohort ix. In the third inscription (vii Claudia) 47 men leave cohort i, 22 leave cohort ii, and 18 leave cohort iii. The implication is clear that the first cohort in each of these legions was approximately twice as strong as the rest. Of course it is dangerous to rely on only two inscriptions; but it would be odd if the only surviving evidence all turned out to relate to exceptional conditions. AE 1955, No. 238 is less conclusive as regards numbers but still shows more men leaving cohort i than any other. This inscription may, and CIL viii 18072 (A.D. 253) does, show that the first cohort had only five centurions; and so as late as the mid third century it had an organization different from the others.

⁷ His views were first briefly published in Actes du deuxieme congres international d'épigraphie grecque et latine Paris 1952 (Paris, 1953), 234Google Scholar, and later were set out in slightly greater detail in Corolla memoriae E. Swoboda dedicata (Römische Forschungen in Niederösterreich, Bd v, Graz-Köln, 1966), 57.

⁸ The MS corruptly reads getati. This was emended to Getae by Schelius in 1660 and Getae is accepted in Gemoll's Teubner edition of 1879; von Domaszewski's edition of 1887 reads Gaetuli, quoting Mommsen (Hermes xv (1884), 223)Google Scholar who describes the reading Getae as ridiculous on two grounds: (a) that the Getae do not appear in technical Latin and (b) that they were identical with the Daci. The most recent edition of the text, by Grillone (Teubner, 1977) reads gaesati; but see note 20 below.

⁹ Pflaum, , Libyca iii (1955), 135 ff.Google Scholar; AE 1956, No. 124. The text will also be found in Pflaum, , Carrieres procuratoriennes equestres (Paris, 1960), No. 181Google Scholarbis and in Saxer, , Epigraphische Studien i (1967), No. 68.Google Scholar

¹⁰ e.g. Alfoldy, G., Situla 14/15 (1974), 204Google Scholar; Dobson, B. and Breeze, D. J., Epigraphische Studien viii (1969), 123.Google Scholar

¹¹ Papiri della Societa Italiana ix (1929), No. 1026Google Scholar; CIL xvi, Appendix, p. 146, No. 13 (see note 58).

¹² Starr, Chester G. Jr., The Roman Imperial Navy (New York, 1941), 188Google Scholar; G. Alfoldy, Die Hilfstruppen der römischen Provinz Germania Inferior, Epigraphische Studien vi (1968), 55 f.; Hind, J. G. F., Britannia v (1974), 287, n. 14.Google Scholar

¹³ Road-building may have been undertaken by classiarii in Trajan's first Dacian War if the men thus occupied in Cichorius, Taf. 67 have been correctly identified (Petersen, , Dakische Kriege ii, 46)Google Scholar; but this is not certain (Starr, op. cit. (note 12), 159, n. 31).

¹⁴ CIL xvi, No. 108.

¹⁵ BRGK 45 (1964), 161–2. The same opinion is expressed in PIR², L 439.Google Scholar

¹⁶ ILS 9200. For commentary on this inscription see Dobson, B., Die Primipilares (Koln, 1978), 216 f.Google Scholar For a different view see Visy, Z., Ada Arch. Hung, xxx (1978), 49 ff.Google Scholar

¹⁷ CIL viii, 1026 (=ILS 2127).

¹⁸ Tacitus, , Historiae ii, 58–9.Google Scholar

¹⁹ CIL iii, 13795 ( = ILS 8909).

²⁰ Gaesati is in fact read by Grillone in his 1977 Teubner text; but it appears improbable that it can be correct, since Gaesati were not a Natio but troops armed in a certain way (i.e. ‘javelin-men’).

²¹ CIL xiii, 1041 (=ILS 2531).

²² CIL xvi, No. 68; Margaret Roxan, Roman Military Diplomas 1954–77, No. 17.

²³ AE 1933, No. 205.

²⁴ Midrash Genesis lvi, II, quoted by Callies, , BRGK 45 (1964), 165.Google Scholar See Midrash Rabbah (English translation by H. Friedman and Maurice Simon, London, 1939) i, 502.

²⁵ The Getae, pace Mommsen, are mentioned officially in connection with the campaign of M. Crassus (PIR² L 186) of 28 B.C. in the Ada Triumphorum for 27 (CIL I², p. 50) and occur in Pliny NH iv 40, although later (80) he identifies them with the Dacians.

²⁶ Dio lxvii, 6; cf. Mócsy, , Historia xv (1966), 511 ff.Google Scholar

²⁷ Strabo vii, 3, 10; cf. Pliny, NH, iv 41.Google Scholar

²⁸ Caesar, , Be Hum Africum xxxii.Google Scholar

²⁹ CIL v, 5267 ( = ILS 2721).

³⁰ Dio xxii, 6–7; Strabo vii, 3, 13; cf. Mocsy, , Historia xv (1966), 511 ff.Google Scholar

³¹ Mocsy, , Pannonia and Upper Moesia (London, 1974), 66Google Scholar, citing a diploma of 71 issued to a man named Dacus (CIL xvi, No. 13), who served in the fleet of Misenum.

³² Baatz, Kastell Hesselbach (1973), 78.

³³ Dessau, , Klio xx (1926), 227.Google Scholar

³⁴ Tettius Iulianus was legate of Upper Moesia (Syme, , Danubian Papers (Bucharest, 1971), 213).Google Scholar

³⁵ Cichy, B., Das römische Heidenheim (Heidenheim, 1971), 77.Google Scholar

³⁶ CIL xiv 2287 ( = ILS 2211).

³⁷ Kennedy, , Britannia viii (1977), 249 f.CrossRefGoogle Scholar; Sasel, , Arheoloski Vestnik xxviii (1977), 334–8.Google Scholar

³⁸ CIL xvi, No. 56.

³⁹ Birley, Corolla, 55; CIL vi 3720 (see Ephemeris Epig. iv 759 ( = ILS 1418)) with Pliny, Ep. vii, 31. P. Holder, BAR Int. Ser., 70, 14–15, has suggested that in some cases the title Augusta was substituted for Domitiana after 96 by units which had been honoured by Domitian.

⁴⁰ CIL xvi, No. 108.

⁴¹ ‘Hilfstruppen…’, Epigraphische Studien vi (1968), 14.Google Scholar

⁴² cf. Bogaers, in Archeologie en Historie (Brunsting Festschrift) (Bussum, 1973), 251–62.Google Scholar

⁴³ CIL iii, 4378.

⁴⁴ For air-photograph and plan see Yadin, Yigael, Bar-Kokhba (London, 1971), 46–9.Google Scholar

⁴⁵ It is worth recalling the inscription of C. Sulpicius Ursulus (AE 1935, No. 12 = 1926, No. 88) who was praefectus Symmachiariorum Asturum belli Dacici, presumably under Trajan. Dessau (Klio xx (1926), 228)Google Scholar considered that this supported von Domaszewski's early second-century date for Hyginus, whose text (as presciently emended by Mommsen (Hermes xix (1884), 223 ff.Google Scholar)) also mentions symmacharii, a type of ally otherwise totally unattested save (under Hadrian) in Arrian (Ektaxis 7, 14) who mentions symmachoi. Richmond, (PBSR xiii (1939), 18 and n. 1)Google Scholar was inclined to connect symmacharii with the scenes on Trajan's Column depicting men irregularly armed yet fighting on the Roman side. Clearly such levies existed under Trajan and Hadrian, and thus probably earlier also under Domitian; but there is no sign of them later on. Dobson, indeed (Epigraphische Studien viii (1969), 122 ff.Google Scholar, cf. idem, Die Primipilares (Koln, 1978), 271 ff.), has suggested a Commodandate for Ursulus; but his argument partly depends upon Birley's dating of Hyginus. The only point of substance in it is the absence in the inscription of mention of Ursulus's filiation and tribe which ‘for a man who had risen to be praefectus legionis would be noteworthy at the beginning of the second century’. These are real difficulties, but they must be balanced against the unlikelihood of Asturians still serving in a tribal levy in the late second century when regular auxiliary alae and cohortes had been raised among them from at least as early as the reign of Nero (and in one case certainly earlier).

⁴⁶ This measurement excludes the three further buildings which lie between these barracks and the principia.

⁴⁷ For Gloucester, see Antiq. Journ. liv (1974), 19Google Scholar, and for Kingsholm, ibid, lv (1975), 266–76. The Gloucester plan, as restored, shows accommodation for six centurions in this cohort, but no parts of the relevant officers' blocks were recovered. It is possible that the excessive length of these barracks (100 m against 72) is to be accounted for by the presence of larger officers' houses differently arranged.

⁴⁸ Bogaers, J. E. and Haalebos, J. K. in Studien zuden Militargrenzen Roms (Köln, 1977), 105 f.Google Scholar; Oudheidkundige Mededelingen uit het Rijksmuseum van Oudheden te Leiden lviii (1977), 149.Google Scholar

⁴⁹ Britannia v (1974), 34 f.Google Scholar

⁵⁰ There is a tombstone of an eques of the Ala Afrorum datable stylistically to c. 110 and burials within the area of the former fortress which are datable at latest to c. 150. I owe this information to Dr C. B. Riiger.

⁵¹ Innenbauten 44, Bild 4, 7.

⁵² It is possible, alternatively, that in the penultimate period some of the barracks of the first cohort lay on the left side of the principia (Baatz, , Germania 55 (1977), 268)Google Scholar, below the later bath-building.

⁵³ See note 6.

⁵⁴ JRS lix (1969), 53.Google Scholar

⁵⁵ Keppie (loc. cit. in note 5), 13–15.

⁵⁶ Keppie, ibid., 15–16.

⁵⁷ The recall of veterans to the standards was one of the measures taken by Vespasian in 69 at Berytus (Tacitus, Historiae ii, 82)Google Scholar; but the context shows that it refers to soldiers already in retirement.

⁵⁸ Parker, , Roman Legions (Cambridge, 1958), 214.Google Scholar For a hint that the years of service beyond 20 were still (under Hadrian) considered to be somehow different, see the papyrus cited in note II… cum militaverimus, domine, in classe praetoria Misenensis (sic) et ex indulgentia divi Hadriani in leg(ionem) Fr(etensem) translatis (annos) super xx.… Codex Just, xii 35, 2 shows the same under Caracalla: idem A(ug.) mililibus cohortisprimae: Viginti stipendia si implestis, sordida munera militiae vobis non indicentur. See also Digest xxvii 1, 8, 2 and other references cited by Keppie (loc. cit. in note 5), 16, n. 56.