^{page 12 note 1} I wish to thank Madame Rivoira and the Clarendon Press for lending blocks from Rivoira, Roman Architecture, here Figs. 2, 3 and 4; the British and American Archaeological Society of Rome for allowing me to copy Parker's valuable photographs, here Figs. 1, 6, 7 and 8; Generale Di Giorgio for granting permission to study remains inside the fortress, and Tenente Bitocco for much kindness. Mr. M. A. Sisson's help has been invaluable and untiring. Dr. Ashby also provided me with much information about literary and other sources, for which I am deeply grateful.

^{page 12 note 2} Cass. Dio, lvii. 19: cf. Suet. Tib. 37.

^{page 12 note 3} This is pointed out by Jordan, , Roman Topography, I.³ p. 386Google Scholar, n. 32.

^{page 12 note 4} Bonn, : Klein, , Korr. blatt des W.D. Zeitschrift, viii. 88Google Scholar. Novaesium: Bonner Jahrbücher, 1904: Koepp, Römer in Deutschland, Pl. X.

^{page 12 note 5} Bull. Com. IV. (1876), p. 178Google Scholar: cf. F.U.R. Sheet Xl. The upper storey may be either for arms or men. The question is complicated by the uncertainty of the strength of the garrison. No trace now remains of the painted wall-plaster and star spangled ceiling which Parker saw (cf. Fig. 1).

^{page 13 note 1} Cf. Appian, , Bell. Civ. v. 3Google Scholar, and Tac. Hist. ii. 93.

^{page 13 note 2} Zos. ii. 17.

^{page 13 note 3} There were four issues of this type, all from the Rome mint—the first in A, the rest in A and Æ; see Mattingly, Coins of the R.E. in the B.M., I. Pl. 36, Nos. 4, 12, 13, 15, 16, 26.

^{page 13 note 4} Vespignani, , Bull. Com. I. (1872), pp. 103–110Google Scholar, Pl. XL.: F.U.R. Sheet XI. is not to be trusted.

^{page 13 note 5} This contradicts Hyginus, c. 56: it has caused much confusion; e.g. Cagnat, s.v. praetoriae cohh., in Daremberg-Saglio, and Hülsen-Jordan get matters wrong. The via principalis must, as always, run straight through. Then we get a fortress with Porta Decumana out of centre, as often, e.g. Haltern, Oberaden.

^{page 13 note 6} Tac., Hist. iii. 84Google Scholar.

^{page 13 note 7} Tac., Ann. xii. 36Google Scholar. Here doubtless was the tribunal shown on Nero's adlocutio coins.

^{page 14 note 1} Rivoira notes these as early tile relieving-arches: their real function was to hold the concrete firm as it set. Architettura R. p. 53, Fig. 48.

^{page 14 note 2} A schematic view of the N. gate appears in Daremberg Saglio, s.v. Praetoriae cohh.

^{page 14 note 3} The plan given is traditional: the earlier versions differ; cf. Canina and Ciconetti-Parker: everything, in fact, is masked by late-Roman brickwork.

^{page 14 note 4} Vitruvius, de Arch. i. 5, 4Google Scholar.

^{page 15 note 1} Architettura R. p. 53; Lanciani, , F.U.R. Sheet XIGoogle Scholar.

^{page 15 note 2} Tac., Hist. iii. 84Google Scholar.

^{page 15 note 3} Dio lxxii. 4, 6.

^{page 15 note 4} Herodian, ii. 6–12 = Vit. Maxim. & Balbin. 10.

^{page 15 note 5} C.I.L. vi. 30876: Antoninianae is erased.

^{page 15 note 6} See n. 2, p. 16.

^{page 16 note 1} E.g. Canina, copied by Durm, , Baukunst, iiGoogle Scholar.

^{page 16 note 2} Presumably this wall was reached by stairs parallel to the wall, and set at intervals on top of the contubernia.

^{page 17 note 1} This is the facing in the lowest cross-hatched division on the left-hand tower of Pl. VII., c: cf. also Pl. VI.

^{page 17 note 2} Rossini, Le porte antiche e moderne del recinto di Roma, Pl XXXIV.: when this is compared with Canina's production, no praise is too high for it.

^{page 17 note 3} Reber, Ruinen Roms, p. 521.

^{page 17 note 4} Rivoira, Arch. R. Fig. 279. Gauckler, , Install. Hydraul. R. en Tunisie, Vol. II. Part I. pp. 67–68Google Scholar, Fig. 20, figures beaks that are unquestionable buttresses, as here. Ivanoff, , Bull, de la Societé Archéologique Bulgare, VII. (1919–1920), p. 90Google Scholar, gives two triangular towers from Kustendil of this period.

^{page 19 note 1} Mau, , Katalog, i. 279Google Scholar. The artist is unknown.

^{page 19 note 2} See above, n. 2, p. 17.

^{page 19 note 3} Zos. ii. 17.

^{page 19 note 4} Lanciani's estimate of 3.50 m. (N.d.S. 1888, p. 734) is too large.

^{page 19 note 5} Lanciani, ibid.: also C.I.L. xv. I, 1578aGoogle Scholar, 6, of(ficina) s(ummae or -ummarum), of(ficina) Dom(itiana) Decemb(ris). See also facts about these tiles noted below, p. 21.

^{page 20 note 1} See also Discovery, vi., No. 68, August 1925.

^{page 20 note 2} Hist. Augg., Aurelian. 21.

^{page 20 note 3} The type is given by the Porta Nomentana; see Discovery, August 1925.

^{page 20 note 4} Deman, E. van, A.J.A. XVI. (1912), p. 424 ff.Google Scholar, Figs. 8, 9 and 10. Aurelian's work, however, goes with 8 rather than with 9 or 10. To avoid a misjudgment, 9 must be visualised with the missing mortar.

^{page 21 note 1} Lanciani, , Bull. Com., XX. (1892), p. 92Google Scholar: C.I.L. xv. 1, 502, 1315c: also Mon. Ant. i. p. 513, where Lanciani says more: in connection with the dating of posterns it is noteworthy that this one had travertine imposts.

^{page 21 note 2} As e.g. between Porta Asinaria and Anfiteatro Castrense; see Suppl. Pap. Amer. School in Rome, i. 1905Google Scholar.

^{page 21 note 3} That was Aurelian's way of building; see Lanciani, Bull. Com., 1892, p. 93.

^{page 21 note 4} This fact lias been obscured by the presupposition that no work took place on the wall between Aurelian and Honorius.

^{page 21 note 5} Aurel. Victor, xl. 26.

^{page 21 note 6} Mommsen, , Chronograph. Abhandl. d. K. Sächs. Ges. d. Wissensch., ii. p. 648Google Scholar.

^{page 21 note 7} This would be the practical sequence.

^{page 21 note 8} Cohen, , Méd. Imp. vii.Google Scholar, Maxence, 20–40, etc.: cf. Lactant. de mort. persecut. 26.

^{page 22 note 1} Tac., Hist. iii. 84Google Scholar.

^{page 22 note 2} C.I.L. xv. 2. 7240, A.D. 175.

^{page 22 note 3} Contributed by the Aqua Marcia to the Marcia-Iulia Tepula castellum near by.

^{page 22 note 4} For this repair must have been made the tiles stamped CASTRISPRAETORIA/G (C.I.L. xv. I, 3 = a bad figure in Daremberg-Saglio, s.v. Praetoriae cohh.). The type is not unique, for the bearded head, described in C.I.L. xv. I. 381 as of Minerva or Roma (sic), occurs on tiles from the Figlinae Oceanae Minores under Caracalla. The period of the type is A.D. 200–222, vide loc. cit. Dressel's dating, in C.I.L. xv. 1.3, is unwarranted.

^{page 22 note 5} C.I.L. xv. 2. 7237.

^{page 22 note 6} Ibid. 7238.

^{page 22 note 7} C.I.L. vi. 30876 = Terme 185(321), Cat. 1922, p. 119: both accounts omit to note two trial S's, carved on the right hand side of the stone.

^{page 22 note 8} Roscher, Lexik. d. G. & R. Mytholog. s.v. restitutrix.

^{page 22 note 9} Herodian, vii. 11–12.

^{page 22 note 10} Zos. ii. 17.

^{page 22 note 11} Hist. Augg. Did. Iulian. 2.