¹ Mentioned by Yorke, V. W., Geographical Journal, 8 (1896), 460 f.Google Scholar; with Hogarth, he spent two days at Satala. The other useful accounts are of Taylor, J. G., JRGS, 38 (1868), 287 ffGoogle Scholar, (one day); and of Cumont, F., Studia Pontica 2 (1906), 342–51Google Scholar (three days).

² Epist., 99, 3 f. and 102 f., ed. Deferrari; Jones, A. H. M., Cities of the Eastern Roman Provinces, 428, n. 46Google Scholar; Cumont, loc. cit., 345.

³ I am most grateful to Dr. R. A. Higgins for his kind assistance in the preparation of this article. The text of Biliotti's Report, with the Head of Aphrodite, is reproduced with the permission of the Trustees of the British Museum. Five Byzantine inscriptions, short and garbled, are omitted from the foot of Fig. 13.

⁴ The present population is approximately 1,500.

⁵ The surviving fortifications presumably date from Justinian, who soon after 530 built a new outer wall of great height, and surrounded the fortress with a circular προτέιχισμα, Procopius, , Aed. 3, 4, 2–5Google Scholar. Of the latter traces cannot now be identified on the ground, but may appear to the north-east of the main walls of the fortress in Cumont's plan (Fig. 8).

⁶ Near the north wall is a huge block of rubble, vitrified by fire and enclosing an unburnt brick.

⁷ Biliotti's plan (Fig. 7) is altogether preferable to that of Cumont. The only one published to date, the latter gives an uncharacteristically misleading impression of the layout of the fortress, but confirms the placing of Biliotti's towers (Fig. 8). My own plan is based on the wall cores visible in 1964 (Fig. 9). There is now no evidence to support the curious annex shown by Biliotti in the south-west corner. Indeed the remains of the rubble core prove the continuity of the south wall at its western end. The bath house stood outside it. If the annex existed at all, it must have been a later addition to the original rectangular legionary fortress, presumably to take in part of the canabae. In late Roman times, Satala was evidently a town of some size: the distinction between town and fortress, occupied perhaps by a depleted legion, was no doubt blurred.

⁸ That a substantial spring existed among these structures in Roman times is shown by the terracotta pipes, 23 cm. wide by at least 16 cm. deep, and set in mortar, that can be seen leading down the hillside above the fortress (Fig. 10). They drew from the cistern, lined with ashlar blocks and measuring 15 × 11 m. which was discovered “about 60 years ago after a dream” (1964), and provided water for the fortress. The cistern cannot be identified in Biliotti's account, and was evidently revealed after his visit. There is now no trace of the building that he describes: its place appears to have been taken by a large, shallow-dredged pool, dammed on the lower side by a long, roughly constructed wall, in which are a few ashlar blocks (Fig. 11).

⁹ The vaulted rooms are still intact. The larger is used as a byre. The smaller serves as the village bakery (Fig. 12).

¹⁰ Compare the Erzincan mosaic, which perhaps came from Satala. It is now in the Erzurum Museum, Schneider, A. M., Arch. Anz. 59–60 (1944/1945), 80Google Scholar, no. 1, with plate 29, 3; and Taylor, loc. cit., 288; Cumont, loc. cit., 350.

¹¹ For the garrison at Satala, see Mitford, , JRS 64 (1974)Google Scholar.

¹² In Armenia Minor, coins were minted only at Nicopolis. I have seen no Roman coins at Satala, but they are said to be recovered in quantity, and are collected for sale primarily in Istanbul. Several Roman and Byzantine coins were offered to Cumont (p. 351).

¹³ The arches, of which only three now stand, without trace of a curve, unquestionably belonged originally to an aqueduct (Fig. 4); an identification first made by Tournefort, , who saw them from the caravan route in 1701 (Rélation d'un voyage du Levant, Paris 1717, p. 168f.Google Scholar). Each arch has a span of more than five metres. The remains were used as a quarry during the reconstruction of public buildings in Erzincan after the disastrous earthquake of 1784; and in 1866, Taylor saw only “the remains of seven arches forming one side of a semicircular building–probably a bath–with opposite corresponding buttresses, at a distance of 11 paces from the former”, loc. cit., p. 289; cf. Cuinet, V., La Turquie d'Asie, 1 (Paris 1892), p. 212Google Scholar. Yorke, loc. cit., p. 461 noted “the remains of an aqueduct, of which five arches are still standing”. Cumont inferred that the aqueduct made a curve, more pronounced than he could observe in the four arches which remained in 1900, loc. cit., p. 349, and n. 1; see also Fig. 13. The ruins are called kırkgöz harabe, that is, “forty eyes–or arches–ruins”, on the Turkish survey: a name no doubt preserved from long ago. Kırk (forty) is commonly used to describe an indefinitely large number. The original length of the aqueduct may be preserved in a widely expressed local tradition, according to which “milk-pipes of earthenware or porcelain” can be seen at Derekorusu Köy, some eighteen miles south of Satala, near the headwaters of the Sadak Çay. The aqueduct supplied water to a level below the fortress, which drew from the cistern on the hill above, and confirms the existence of a rich and well populated settlement below the walls.

¹⁴ Inscriptions, once found, have a short life at Satala. Biliotti's five Byzantine stones have by now disappeared. Of fifteen inscriptions recorded by Hogarth and Yorke in 1894, six were sighted by Cumont in 1900, and only three survived in 1972.

¹⁵ The bronze head appears on stylistic grounds to be a work of the late Hellenistic or early Roman period. Dr Higgins suggests that it may be a cast from a mould made in c. 150 B.C., whether a Greek or Hellenistic original, or a Roman copy. Civic life in the cities of eastern Pontus reached its highest development in the second and third centuries, and the head is likely to have been imported to Satala at the same period. It is normally assigned to Aphrodite: an attribution to Anaitis is, at Satala, wholly implausible. The head is 38 cm. high: see in particular BM Catalogue of Bronzes (1899), no. 266 (Fig. 3).

¹⁶ For the position of Satala, at the crossing of the caravan route from Persia to the Aegean, and of the frontier road from the Euphrates valley to Trapezus, see Mitford, , JRS 64 (1974)Google Scholar, and Figs. 14 and 15.

¹⁷ 124 miles in Peutinger, Route 98 (ed. Miller, 1916): compare in reverse sequence Ravenna Cosmography 2, 12 (ed. Schnetz, , 1939). 130Google Scholar miles in the Antonine Itinerary, 216, 4 (ed. Cuntz, , 1939Google Scholar). By the modern road, the distance from Satala to Trabzon is some 115 miles, striking confirmation of the accuracy of Biliotti's measurement.

¹⁸ Dio 55, 23, 5; compare Notitia Dignitatum, Oriens, 38, 13, praefectus legionis quintae decimae Apollinaris, Satala; and Antonine Itinerary, 183, 5, Satala, leg. XV Apollinaris; and see Mitford, loc. cit.

¹⁹ The stone used at Satala for inscriptions and building blocks is soft and quickly eroded.

²⁰ See note 5.

²¹ This opinion is not necessarily valid. There is now no knowledge at Satala of the discovery of bronzes. The villagers regard April and May as the excavation season but their object is building material rather than antiquities.