¹ We are grateful to Mr. Hall for his invitation to join the survey team, to the Oxford University Craven and Meyerstein Funds and to Merton College, Oxford, for financial assistance, and to successive Turkish Government representatives, especially Bay Edip Özgür (see below p. 22) for inestimable assistance. We should also like to thank Mr. Ian Gardiner of Hydraulics Research Ltd., Wallingford, Oxon., for taking so much trouble in answering a number of complex technical questions, often with inadequate evidence to work on. No firm line separates the contributions of the two authors, but the descriptions as a whole, and the discussion of the inverted siphon system and its built aqueduct, are primarily due to J. J. C., while the discussion of the other elements is mainly the work of E.C.S.

² Inscriptions published by Hall, A. S., AS 27 (1977) 193–7Google Scholar (cf. SEG 27 (1977) 930–2Google Scholar; J., and Robert, L., REG 91 (1978) 475–6, no. 462Google Scholar.

³ Inscriptions published by Heberdey, R., Kalinka, E., Bericht über zwei Reisen in südhwestlichen Kleinasien (Denkschr. Akad. Wien, phil.-hist. Kl., 45, 1896) nos. 77–8Google Scholar; the relief above no. 77 shows a spring gushing into an urn.

⁴ An inscription discovered close to the sanctuary in 1983 contained a dedication to Apollo, so invalidating an earlier suggestion (Coulton, J. J., PCPS NS 29 (1983) 2Google Scholar), that the sanctuary might belong to Asklepios.

⁵ The building on the southwest hill, with small rooms along its north and east sides (Coulton, J. J., PCPS NS 29 (1983) 17, fig. 7Google Scholar), is paralleled by other late forts (Harrison, R. M., in Actes du colloque sur la Lvcie antique (Bibl. Inst. Fr. d'Et. Anat. d'Ist. 27, 1980) 112, fig. 2Google Scholar).

⁶ For Corinthian tiles on Early Christian basilicas see Orlandos, A., Xylostegos Paleochristianiki Basiliki 2 (1954) 395–6Google Scholar.

⁷ On the terminology see Hodge, 1983, 174–80.

⁸ Dropping spring levels: Burger, A., Dubertret, J. (ed.), Hydrogeology of Karstic Terrains (1975) 9, 11Google Scholar. Weber, 1899, 22, 1904, 95, refers to springs which have sunk at Smyrna and Hierapolis and attributes the lower level of the former to earthquakes, another possible factor at Oinoanda. Sens: Germain de Montauzan, 143–5, Grenier, , Manuel 4, 174–5Google Scholar.

⁹ On dams and river intakes see Fahlbusch, H., VAGRW, 31–33Google Scholar, and the bibliography of Hodge, A. T., Classical Views 27.3, NS 2.3 (1983) 323 fGoogle Scholar.

¹⁰ Thompson, H. A., Hesperia, 27 (1958) 147CrossRefGoogle Scholar. For other examples see Grenier, , Manuel 4, 89Google Scholar (Pont du Gard system, Nîmes); Wodociagi Rzymskie, fig. 121 (Sofia).

¹¹ Wodociagi Rzymskie, 275.

¹² Fabricius, E., AthMitt 9 (1884) 171–3Google Scholar. For other Greek spring boxes see Südhoff, K., Kos und Knidos (1927) 240–5Google Scholar (Burina, Kos); H. A. Thompson, loc. cit. (n. 10), (Areopagus, Athens); also Kienast, H., WA Hellas, 50 (general)Google Scholar.

¹³ H. Kienast, loc. cit. (n. 12), followed by Fahlbusch, H., VAGRW, 28–9Google Scholar, stresses the importance of concealment; Merckel, C., Ingenieurtechnik, 518Google Scholar, refers to protection against infiltration of water from the surrounding soil, though in a Roman context. Modern spring boxes are sealed against pollution and locked against vandals, while storage tanks are often underground to increase solidity and keep the water cool. Cairncross, S., Feachem, R. G. A., Small Water Supplies, Ross Institute Bulletin 10 (London School of Hygiene and Tropical Medicine, 1978) 8–10, 50Google Scholar.

¹⁴ Wasser. nach Köln, 56–63. For other Roman spring boxes (many ill-preserved and some of doubtful interpretation) see: Wasser. nach Köln, 146 (Miesenheim); Engelhardt, R., Bingen am Rhein: die römische Wasserleitung in Bingen (1978) 12 (Bingen)Google Scholar; Weber, 1899, 22 (Smyrna; spring chamber); Gräber, , AvPerg, 406 (Kaikos line, Pergamon)Google Scholar; Forchheimer, P., FiE 3 (1923) 229Google Scholar (Ephesos; very fragmentary), 238–9 (Ephesos, Artemision); Ingenieurtechnik, 518 (Tusculum)Google Scholar; Van Deman, , Building, 72 (Aqua Marcia)Google Scholar; Grenier, , Manuel 4, 179–80Google Scholar (Sens), 214 (Bavai); Germain de Montauzan, 143–5 (Sens), 146–9 (Zaghouan, Carthage; but this also uses infiltration galleries); Rakob, F., Aqueducs romains, 316–17 (the same)Google Scholar; H. Vertet, ibid., 357–8 (Rusicade, Algeria); J. Lassus, ibid., 211 (Antioch; but this seems to be a clearing basin), Wodociagi Rzymskie, figs. 78, 114 (Arčar), figs. 88, 141 (Nikopolis), fig. 89 (Dolno Botevo); and in general Fahlbusch, , Hist. Wasser, 2Google Scholar; id., VAGRW, 30.

¹⁵ Using a main extraction tunnel with branch adits. The first option seems to occur at Alabanda (Özis, Ü. et al. , Hist. Wasser, 2Google Scholar); the second in the Mont d'Or system, Lyon (Germain de Montauzan, 151–5). Many infiltration galleries of all dates have branches, e.g. Peirene, Corinth (Hill, B. H., Corinth 1.6 (1964) 54–63CrossRefGoogle Scholar), near the church of Aghios Georghios, Pergamon (Gräber, , AvPerg, 410–12Google Scholar); Benden, Hausener (Wasser. nach Köln, 52–5)Google Scholar. For other Roman infiltration galleries see: Forchheimer, P., FiE 3 (1923) 228–9Google Scholar (if this is what he means by Hohlgang); Fahlbusch, H., Hist. Wasser, 30 (Aksu line, Pergamon, Antoninus Pius line, Athens)Google Scholar; Wodociagi Rzymskie, 275, fig. 117 (Obzor), fig. 118 (Starozagorskite Mineral Baths), fig. 119 (Sviščov), fig. 120 (Gigen); Wasser. nach Köln, 17–18 (Gleuel), 64–7 (Grüner Putz); Grenier, , Manuel 4, 56 (Cimiez), 144Google Scholar (apparently; Saintes); Peletier, A., Aqueducs romains, 294–6 (Vienne)Google Scholar; Germain de Montauzan, 140–2 (Rome), 146–9 (Zaghouan), 149–51 (La Brevenne, Lyon), 154 n. 1 (Africa); Rakob, F., Aqueducs romains, 316–17Google Scholar, id., RömMitt 81 (1974) 41–88 (Zaghouan); Crouch, D. P., Studia Palmyrenskie 6–7 (1975), 151–8, 160–70 (Palmyra)Google Scholar. Greek infiltration galleries are discussed by Stenton, E. C., Greek Fountains: design and function (Oxford D.Phil, thesis, 1984) 32–42Google Scholar.

¹⁶ As apparently at Alabanda and the Grüner Putz (see n. 15 above).

¹⁷ Loc. cit. (n. 15). Feed lines to the aqueduct to Arles sometimes unite at a basin, but sometimes discharge directly into it. However, this is not a piped system.

¹⁸ Greek examples: Roux, G., L'Architecture de l'Argolide aux IVe et IIIe siècles av.J.-C. (1961) 285–9Google Scholar; Pace, B., Arte e Civilta della Sicilia Antica 2 (1938) 430–1 (pot)Google Scholar; Isler, H., Samos 4 (1978) 52 (pot)Google Scholar; Wiseman, J., Hesperia 38 (1969) 74CrossRefGoogle Scholar (depression); Kienast, H., WA Hellas, 54Google Scholar (general); detailed discussion in E. C. Stenton, op. cit. (n. 15) 114–17, 122–7. Roman examples: Karo, G., AA 1936, 175 (clay bowls)Google Scholar; Garbrecht, G., Holtorff, G., WAAP: die Madradağ-Leitung (LIW Mitt 37, 1973) 68–72 (small tank)Google Scholar; Wasser. nach Köln, 62 (small tank); Wodociagi Rzymskie, fig. 47 (stone basin), fig. 71 (pot); Grenier, , Manuel 4, 153Google Scholar (depressions under manholes); Rakob, F., Aqueducs romains, 315 (depressions)Google Scholar; L. Jeancolas, ibid., 191 (depression); Fahlbusch, H., Hist. Wasser, 10 (general)Google Scholar.

¹⁹ Clearing basin is here used generically, of all containers intended to remove any kind of sediment from the water; silt trap is used for clearing basins serving only to remove coarse sediment, while sedimentation tank is used for clearing basins serving to purify and clarify the water by removing the finest sediment.

²⁰ Ingersoll, A. C., McKee, J. E., Brooks, N. H., Proc. American Soc. of Civil Engineers 81Google Scholar, sep. no. 590 (Jan. 1955) 9, 21, 27; Clements, M., Proc. Institute of Civil Engineers 34 (June 1966) 171, 177, 196Google Scholar. E. C. S. is preparing a short study of ancient clearing basins.

²¹ Clearing basins do often double as manifolds (E. C. Stenton, op. cit. (n. 15) 131–3, 137–8), but at Oinoanda there are no side openings. The pipe joints of the Hellenistic Madra Dağ gravity line were extremely water tight (Garbrecht, G., Holtorff, G., WAAP: die Madradağ-Leitung (LIW Mitt 37, 1973) 86–94Google Scholar).

²² This is the minimum recommended height for the outlets of spring boxes in the Developing World (S. Cairncross, R. G. A. Feachem, op. cit. (n. 13) 8. However, the Samian outlet seems rather too high, since it is above the eye of the spring, and so might allow it to be blocked by silt (ibid).

²³ F. Rakob, loc. cit. (n. 18), seems, wrongly, to regard many small basins as equivalent to one large one.

²⁴ Not adequately published. Fullest accounts: Martin, R., Metzger, H., BCH 73 (1949) 317–23, 349–50CrossRefGoogle Scholar; Judeich, W., Topographie von Athen (1931) 190–1Google Scholar.

²⁵ Compare the rock cuttings for the Marnas supply line to Ephesos (Forchheimer, P., FiE 3 (1923) 226–7Google Scholar).

²⁶ Conze, A., Schuchhardt, C., AthMitt 24 (1899) 105–7Google Scholar; Gräber, , AvPerg, 381–3Google Scholar; Garbrecht, G., Aqueducs romains, 152–3Google Scholar; Kienast, H., WA Hellas, 147Google Scholar; Fahlbusch, H., VAGRW, 109Google Scholar.

²⁷ On the problem of entrained air, see Hodge, 1983, 197–9. He suggests (205) that air release valves may have been used on these rises, following Gräber, , AvPerg, 378Google Scholar.

²⁸ We are grateful to Mr. I. Gardiner for this information.

²⁹ As noted by H. Fahlbusch, loc. cit. (n. 26).

³⁰ As in a modern serpentine sedimentation tank (see S. Cairncross, R. G. A. Feachem, op. cit. (n. 13) 37, figs. 25, 26).

³¹ Gräber, , AvPerg, 393–4Google Scholar.

³² Germain de Montauzan, 72–6, 104–5, 118–22; Wasser. nach Köln, 153; A. T. Hodge, 1983, 180–1. For other examples of header and receiving tanks see Weber, 1898, 3, 57 (Laodikeia); Gräber, , AvPerg, 393–4, 396–7 (Pergamon)Google Scholar; Fahlbusch, H., WA Hellas, 144–5, 147 (Pergamon)Google Scholar. Also, Ingenieurtechnik, 580 (basins at ends of aqueduct, Metz).

³³ See above. Other Asiatic examples: Weber, 1898, 3–4. (city tank, Laodikeia; several chambers), 57 (header, Laodikeia; 2 ch.); id., 1899, 12 (water tower, Smyrna; several ch.); Gräber, , AvPerg, 396Google Scholar, Fahlbusch, H., WA Hellas, 145Google Scholar (clearing basin/header, Pergamon; 4ch.); Hülsen, J., Milet 1.5 (1919) 4, 8–9Google Scholar (clearing basin?, nymphaion, Miletus; 2 ch.). Examples elsewhere, several with two storeys, are similar: Ingenieurtechnik, 540–2; Germain de Montauzan, 306–8; Ashby, , Aqueducts, 84, 133, 174, 277, 281, 226Google Scholar; Grenier, , Manuel 4, 104, 137Google Scholar; Audin, A., Aqueducs romains, 13–18Google Scholar; Wodociagi Rzymskie, 280.

³⁴ As in the Hellenistic header tank at Pergamon (see above p. 23). For other examples of screens in various contexts see: Lang, M., Waterworks in the Athenian Agora (Picture Bk. 11, 1968) fig. 22Google Scholar; Eschebach, H., Aqueducs romains, 88–9 (?)Google Scholar; Ingenieurtechnik, fig. 242; Wodociagi Rzymskie, 280.

³⁵ Weber, 1898, 7, attributes a two-socketed elbow block to the junction between the header and stone mains at Laodikeia; see n. 36, below. Gräber, , AvPerg, 393Google Scholar, refers to stone outlet “pipes” at the header tank of the Great Aqueduct at Pergamon, but this is in the context of a clay line, and he gives no details.

³⁶ Certain examples: Gräber, , AvPerg, 396Google Scholar, Fahlbusch, H., WA Hellas, 145 (clearing basin/header, Pergamon)Google Scholar; Weber, 1904, 94 (tank above city, Hierapolis). Probable example: Weber, 1898, 7 (header tank, Laodikeia). Possible examples: Weber, 1904, 91 (fountain?, Aphrodisias); Forchheimer, P., FiE 3 (1923) 250Google Scholar (no evidence of related structure cited). See also n. 73, below.

³⁷ See n. 36 above.

³⁸ 1898, 7. See also, Weber, 1899, 12–13, where a similar hole in a standard stone pressure pipe block near the water tower at Smyrna is interpreted in the same way.

³⁹ Cylinder valves: Squassi, F., L'Arte Idro-Sanitaria (1954) 78–80Google Scholar; Daremberg, C., Saglio, E., Dictionnaire des antiquités grecques et romaines, 2.1 (1892)Google Scholar s.v. epistomium. More detailed studies: Kretzschmer, F., Revue Archéologique de l'Est et du Centre Est 11 (1960) 89–113Google Scholar; Hodge, 1983, 202 n. 76, 203–4.

⁴⁰ Stone bungs: Gräber, , AvPerg, 394 (header), 399 (fountain/manifold, Pergamon)Google Scholar. Wooden bung: Grenier, , Manuel 4, 166Google Scholar (clearing “tower”, nr. Poitiers). Cement bung: Wasser. nach Köln, 154 (header, Lyon, but perhaps permanent). Greek sluices (e.g.): Gruben, G., AD 19A (1964) 68Google Scholar, Amandry, P., BCH Suppl. 4 (1977) 210–11Google Scholar (outlets, reservoirs and basins), 192–3 (diversion in supply channel, date uncertain); Broneer, O., Isthmia 2 (1973) 27Google Scholar (outlet, reservoir; probable); Tomlinson, R. A., BSA 64 (1969) 162Google Scholar (diversion in supply channel; probable). Roman sluices (e.g.): Garbrecht, G., Holtorff, G., LIW Mitt 37 (1973) 60Google Scholar (outlet, clearing basin); Scranton, R. L. et al. , Kenchreai 1 (1978) 25–35Google Scholar (connecting channels, fish ponds); Germain de Montauzan, 153 (outlet, clearing basin), 317 (outlet, manifold), 323 (inlet and outlet, reservoir); Grenier, , Manuel 4, 80Google Scholar (channel, mill basin), 97 (outlet, castellum); Fahlbusch, H., VAGRW, 111Google Scholar (drain, clearing basin); Wasser. nach Köln, 24 (outlets, clearing basin); Lassus, J., Aqueducs romains, 212Google Scholar (diversion channels), etc. See also Hodge, 1983, 203–4 for the idea that inverted siphons were shut down at the header tank.

⁴¹ Weber, 1898, 6–7, suggests that the curved course of the pipeline at Laodikeia was intentional, being more resistant to deformation than a straight one. This explanation works only if dislocating forces come from a single, predictable direction, which in this case is not true.

⁴² A single block of drafted polygonal masonry was noted in 1983 at the Kemerarasi site below Oinoanda. This might have come from the aqueduct, but the problem of thoroughgoing robbing restricted to its northern end remains unresolved.

⁴³ Lanckoronski, C. et al. , Städte Pamphyliens und Pisidiens 1 (1890) 120–4Google Scholar, 179 no. 64h; Ward-Perkins, J. B., BSR 23 (1955) 115–23Google Scholar; Bean, G. E., Turkey's Southern Shore² (1979) 53 (date)Google Scholar; Hodge, 1983, 187–9.

⁴⁴ The account offered here, based on the indications of landform on the sketch map in Lanckoronski, op. cit. (n. 43), facing p. 85, may explain why the aqueduct did not follow a line straight to the city, a question which troubled J. B. Ward-Perkins (op. cit. (n. 43) 118) and Hodge, 1983, 188–9.

⁴⁵ Frova, A., Annuario 23/4 (1961–1962) 650–2Google Scholar; see also Negev, A., IEJ 14 (1964) 237–49Google Scholar, Levine, L. I., Roman Caesarea (Qedem 2, 1975) 30–5Google Scholar, Olami, Y., Peleg, Y., IEJ 27 (1977) 127–37)Google Scholar.

⁴⁶ The economic advantage of arched construction is noted by Fahlbusch, , V AGRW, 58–9Google Scholar.

⁴⁷ Weber, 1905, 205–6; id., 1898, 8. We assume about one metre between the soffit of the arch and the top of the aqueduct.

⁴⁸ For the long support walls, Garbrecht, G., Fahlbusch, H., W AA Perg: die Kaikos-Leitung (LIW Mitt 44, 1975) 27, 46Google Scholar; Hecht, K., Hist. Wasser, 4–5Google Scholar.

⁴⁹ A rubble-and-mortar aqueduct of uncertain date near Smyrna has arches spanning 3·70 m. and piers apparently c. 5 m. wide (Weber, 1899, 9). However, the parallel is not a good one, for this aqueduct is also unusual in having arches which spring from ground level and an abnormal thickness of masonry between the soffit of the arch and the top of the structure. A section of aqueduct at Fréjus with arches springing directly from a high continuous structure (Germain de Montauzan, 225, fig. 104 = Grenier, , Manuel 4, 46, fig. 13Google Scholar) probably shows the city wall heightened to carry the conduit; cf. Grenier, , Manuel 4, 45, fig. 12Google Scholar.

⁵⁰ Weber, 1898, 3.

⁵¹ Lyon: Audin, A., Aqueducs romains, 13–18Google Scholar; Grenier, , Manuel 4, 137. Rome: AshbyGoogle Scholar, Aqueducts, 133; Van Deman, , Building, 109Google Scholar; Ingenieurtechnik, 542. The device has been advocated in modern times: Skeat, W. A. (ed.), Manual of British Water Engineering Practice 3 (4th ed., 1969) 212–13Google Scholar.

⁵² e.g. theatres: Ephesos (fountain), Wilberg, W., FiE 3 (1923) 266–73Google Scholar; Sikyon (fountain), McMurtry, W. J., AJA 5 (1889) 279–80Google Scholar, Fiechter, E., Das Theater in Sikyon (1931) 22, 31Google Scholar; Delphi (water basin), Pouilloux, J., Roux, G., Énigmes á Delphes (1963) 81Google Scholar; Side (channels and basins), Mansel, A. M., Die Ruinen von Side (1963) 128CrossRefGoogle Scholar. Stadia and racetracks: Isthmia (fountain, reservoir and basins), Broneer, O., Isthmia 2 (1973) 24–7, 61–3Google Scholar; Corinth (fountain and basins), Williams, C. K., Hesperia 39 (1970) 91–2Google Scholar; Delphi (fountain), Aupert, P., Fouilles de Delphes 2, Le Stade (1979) 81–2Google Scholar.

⁵³ Lyon: Hodge, 1983, 184, 186 n. 38; Wasser. nach Köln, 153; Laodikeia: Weber, 1898, 2, figs. 1, 3, 7; Pergamon: Garbrecht, G., Aqueducs romains, 153Google Scholar (but note that here he uses a height for the acropolis about 15 m. greater than that shown on the recent map of Pergamon accompanying Nohlen, K., Radt, W., Altertümer von Pergamon 12, Kapıkaya (1978))Google Scholar.

⁵⁴ On these see Fahlbusch, H., Hist. Wasser, 10Google Scholar; id., VA GRW, 112–14, though his contention that they are an indispensable element in any Roman supply system must be questioned and some of his examples were evidently intended primarily as clearing basins.

⁵⁵ As does that at Nîmes, on which, see Germain de Montauzan, 316; Grenier, , Manuel 4, 97–101Google Scholar; Fahlbusch, H., VAGRW, 118–19Google Scholar; also Hodge, 1983, 176 n. 6. On other castella, mainly of indeterminate type, see Weber, 1898, 3–4; Ashby, , Aqueducts, 82Google Scholar; Ingenieurtechnik, 566; Grenier, , Manuel 4, 48, 84, 104, 184Google Scholar; Wodociagi Rzymskie, figs. 48, 147, 152. More general discussion of distribution, including the Vitruvian scheme: Fahlbusch, H., Hist. Wasser, 11Google Scholar; id., VAGRW, 116–19; Wasser. nach Köln, 113–16; Germain de Montauzan, 309–11; Ashby, , Aqueducts, 45Google Scholar.

⁵⁶ Eschebach, H., Aqueducs romains, 87–90.Google Scholar

⁵⁷ See e.g., Fahlbusch, H., VAGRW, 118–19Google Scholar; Hainzmann, M., Untersuchungen zur Geschichte und Verwaltungen der stadtrömischen Wasserleitungen (1975) 21, n. 1Google Scholar; Wasser. nach Köln, 113.

⁵⁸ H. Fahlbusch, loc. cit. (n. 57), seems to dispute this because of the shallowness of the basin, but what matters is the relative amounts of inflow and outflow.

⁵⁹ Germain de Montauzan, 316–18.

⁶⁰ Its size also suits this interpretation. There is no evidence for Germain de Montauzan's assertion (Germain de Montauzan, 318) that the small scale of manifolds results from their use in combination with large reservoirs; compare n. 54 above.

⁶¹ Hülsen, J., Milet 1.5 (1919)Google Scholar. For other possible examples of combined fountains and manifolds, see Fossel, E., Langmann, G., JÖAI 50 (1972–1975) 301–10Google Scholar; Ingenieurtechnik 545; Germain de Montauzan, 318; Grenier, , Manuel 4, 56, 66Google Scholar.

⁶² Gräber, , AvPerg, 399–401.Google Scholar

⁶³ Gräber, , AvPerg, 401Google Scholar.

⁶⁴ Priene, 71–2.

⁶⁵ These are poorly preserved, but essentially the same as the header tanks; see Hodge, 1983, 184 and n. 32 above (on headers).

⁶⁶ Bammer, A., JÖAI 50 (1972–5)Google Scholar Beibl., 381–2 fig. 1. In the Madra Dağ pressure pipeline at Pergamon (Ingenieurtechnik, 507–8; Gräber, , AvPerg, 377–81Google Scholar; Garbrecht, G., Aqueducs romains, 153–5)Google Scholar there are no sockets for pipe ends, and Garbrecht, ibid., fig. 9, shows a suggested anchoring scheme with pipe ends outside the blocks.

⁶⁷ Fahlbusch, H., WA Hellas, 144Google Scholar (Attalos line, Pergamon); id., VAGRW, 80–1 (general); Gräber, , AvPerg, 386Google Scholar (line to Greek Gymnasium, Pergamon; he calls this a stone main, but the blocks have sockets at both ends), 388–91 (Great Aqueduct pipeline, Pergamon); Ingenieurtechnik, 563 (the same); Weber, 1899, 9, 11, 18 (various blocks at Smyrna with sockets, but no spigots, which thus presumably do not belong to the stone pipeline); id., 1904, 91 (two-socket elbow, Aphrodisias); id., 1905, 202–3 (? from terracotta pressure pipeline, Magnesia ad Sipylum); Wodociagi Rzymskie, figs. 44, 45, 137 (Plovdiv and Madara).

⁶⁸ On the effects of pressure and momentum on bends in pipelines, see Hodge, 1983, 200–4, or, for a more technical treatment, Cairney, , Hydraulics, 76–82Google Scholar (basic equations and use of concrete anchor blocks at bends). Biernacka-Lubańska, M., Wodociagi Rzymskie, 281Google Scholar, refers to stone elbows as “pressure intensifies”, which seems singularly inapposite. Other examples of such blocks: Stillwell, R., Antioch on the Orontes 2 (1938) 18Google Scholar, fig. 16, 19, fig. 17, 122, fig. 42; Lassus, J., Aqueducs romains, 219 (Antioch)Google Scholar; Crouch, D. P., Studia Palmyrenskie, 6–7 (1975), 173–6, esp. figs. 14, 15 (Palmyra)Google Scholar; Wodociagi Rzymskie, figs. 46, 148 (?) (Čatalka); Weber, 1898, 5 (?) (Laodikeia); id., 1899, 20 (Smyrna).

⁶⁹ The elbow is formed by the stone block, as at Oinoanda, not by a clay elbow encased in stone, as suggested by Landels, J. G., Engineering in the Ancient World (1978) 45Google Scholar; cf. Hodge, A. T., Scientific American 252.6 (June, 1985) 117CrossRefGoogle Scholar.

⁷⁰ R. Stillwell, op. cit. (n. 68) 122. (The other advantages which he sees in the use of stone distribution blocks are dubious or unclear). Compare the use of stone basins and clay pots as small manifolds: Wodociagi Rzymskie, 281, E. C. Stenton, op. cit. (n. 15) 131–3, 137.

⁷¹ Double sockets: Weber, 1899, 11, fig. 5 (Smyrna, pressure pipe block, bevelled steps), fig. 6 (Smyrna, two-socketed block), fig. 7 (Smyrna, two-socketed block with one single and one double socket); id., 1905, 202 (Magnesia ad Sipylum). Funnel sockets: Gräber, , AvPerg, Beil. 97.2, 3, 5, 7Google Scholar (socket and funnel combined), 97.6 (with pipe still inserted), 97.13 (Great Aqueduct, Pergamon); Wodociagi Rzymskie, fig. 42 (Sviščov).

⁷² On the blocks at Antioch, see: R. Stillwell, op. cit. (n. 68) 122. This interpretation clearly cannot apply to the vertical holes in stone pipe blocks discussed below (p. 50), but there is no reason why the two sorts of hole must have had the same function.

⁷³ Fountains at Priene, (Priene, 76–80Google Scholar) and near the Tholos, Athens (Thompson, H. A., Hesperia (Suppl. 4, 1940) 96–8Google Scholar; Glaser, F., Antike Brunnenbauten in Griechenland (1983) 91–2Google Scholar and esp. fig. 172) incorporated vertical elbows in their bases; a separate stone elbow block with two branches at right angles to each other lies near the Roman fountain on the Sacred Way at the Asklepieion, Pergamon (personal observation). A cylindrical tube is bored in the base of the Statue Fountain in the Asklepieion, Epidaurus (Kavvadias, P., Praktika tis en Athinais Archaiologikis Etaireias (1922–1923) 24–5Google Scholar), and in each pillar of the street fountains of Pompeii (Eschebach, H., Pompeji (1978) 41Google Scholar). Semicircular furrows for separate pipes run up the parapets of basins in the Round Building of the Asklepieion, Pergamon (personal observation). Three small vertical offtakes in a stone reinforcement block from a double clay line at Palmyra (Crouch, D. P., Studia Palmyrenskie, 6–7 (1975) fig. 15Google Scholar) are arranged in a triangle and may have fed three spouts at the apex of a triangular fountain basin similar to that in her fig. 14, though its inlets are not visible. Finally, Weber, 1898, 5, attributes an elbow block with two outlets in the same face to a fountain because of dowel holes in another face. However, these may result from re-use, and anyway would imply a horizontal channel. Also in the horizontal plane are the stone elbows near the two fountains at Pella (Makaronas, C., AD 16 (1960) 81Google Scholar, 17B (1961–2) 212), though they must have connected with vertical channels.

⁷⁴ e.g. Southeast and Southwest Fountain Houses, and Roman Nymphaeum, Athens, The Athenian Agora (Guide) (1975) 151–2, 154–6, 169–71Google Scholar; fountain at Magnesia on the Maeander, Humann, C., Kohte, J., Watzinger, C., Magnesia am Maeander (1904) 135–7Google Scholar; and, indeed, the nymphaeum, Miletus, Hülsen, J., Milet 1.5 (1919)Google Scholar.

⁷⁵ As at Pompeii, Eschebach, H., Aqueducs romains, 90–1Google Scholar, or Priene, , Priene, 76–80Google Scholar.

⁷⁶ Eschebach, H., Aqueducs romains, 83 (re-used raised cisterns), 91–5Google Scholar (towers and arches); Hodge, 1983, 175–7; Fahlbusch, H., VAGRW, 90, 119Google Scholar.

⁷⁷ For contradictory interpretations of their function, see Hodge, 1983, 175, n. 5. Biernacka-Lubańska, M., Wodociagi Rzymskie, 281Google Scholar, figs. 46, 148, refers to blocks with vertical holes as “pressure regulators” or “reducers", but the Polish text makes it difficult to understand how she envisages them working. There are no known water towers in the region (ibid.), and the holes cannot simply have been left open.

⁷⁸ See Eschebach, H., Aqueducs romains, 91–4Google Scholar, Hodge, 1983, 175 n. 5.

⁷⁹ At Aspendos, and Les Tourillons de Craponne, Lyon; see Ward-Perkins, J. B., BSR 23 (1955) 118–19Google Scholar; Wasser. nach Köln, 154–7; Fahlbusch, H., Hist. Wasser, 8Google Scholar; id., VAGRW, 86–92; Hodge, 1983, 185–9, 199, n. 68 (he is sceptical of this interpretation of Les' Tourillons). Other towers of doubtful identity and function: Weber, 1898, 3; id., 1899, 12–13; id., 1904, 99; Forchheimer, P., FiE 3 (1923) 244–5Google Scholar; Özis, Ü. et al. , Hist. Wasser, 3Google Scholar; Ingenieurtechnik, 525, 530; Van Deman, , Building, 140Google Scholar; Grenier, , Manuel 4, 164Google Scholar.

⁸⁰ Ingenieurtechnik, 579. (Fahlbusch, H., VAGRW, 103Google Scholar, says that this is a gravity line, from which air cannot be removed without a pump, but then suggests, rather unflatteringly, that the Romans may not have realized this.)

⁸¹ We are grateful to Mr. Ian Gardiner for confirming this. On pressure surges and surge tanks, see Hodge, 1983, 201–4 (though he does countenance the use of a simple pipe), and for a more technical treatment, Cairney, , Hydraulics, 128–33, 139–41Google Scholar (note especially the size of the surge tank in fig. 9.6b).

⁸² Cairney, , Hydraulics, 141Google Scholar.

⁸³ Weber, 1898, 5; this too has no socket. Compare also two unsocketed blocks with large, tapering holes, at Smyrna, Weber, 1899, 13.

⁸⁴ See Cairney, , Hydraulics, 86–7Google Scholar.

⁸⁵ Weber, 1899, 20–1, fig. 30 (a U-shaped channel with side outlet and what seem to be four smaller holes for an attachment). On modern desilting valves, see Cairney, T., Hydraulics, 107Google Scholar.

⁸⁶ Weber, 1899, 18, fig. 24 (clearly a drain, since it still contains a section of pipe), 20 fig. 28, 20–1, fig. 30 (as in n. 85, above).

⁸⁷ McNicoll, A. W., Hellenistic Fortifications from the Aegean to the Euphrates (D.Phil. thesis, Oxford, 1971)Google Scholar; Coulton, J. J., AS 32 (1982) 121Google Scholar.

⁸⁸ Appian, , Bell. Civ. 4.79Google Scholar.

⁸⁹ Wilberg, W., FiE 3 (1923) 156–65Google Scholar; Börker, C., Merkelbach, R., Die Inschriften von Ephesos 2 (IK 12, 1979) no. 402Google Scholar; Meriç, R., Merkelbach, R. et al. , Die Inschriften von Ephesos 7.1 (IK 17.1, 1981) no. 3092Google Scholar.

⁹⁰ Lawrence, A. W., Greek Aims in Fortification (1979) 235–6Google Scholar.

⁹¹ Adam, J.-P., L'Architecture militaire grecque (1982) 116–65Google Scholar.

⁹² The Balboura aqueduct was seen by Falkener in 1844, but remains unpublished; its inscription is IGR 3, 466Google Scholar; cf. also Naour, C., Ancient Society 9 (1978) 166–70Google Scholar.

⁹³ Texier, C., Description de l'Asie Mineure 3 (1849) 204, 233Google Scholar, pl. 207–8, where the site is wrongly identified as Aperlai. For the correction see R. Heberdey, E. Kalinka, op. cit. (n. 3) 17, no. 56. The inscription is IGR 3, 690Google Scholar.

⁹⁴ A. Farrington, personal communication.

⁹⁵ The baths Ml 1 at Oinoanda, together with other Lycian bath buildings, will be discussed in an Oxford doctoral thesis by Mr. A. Farrington of the University of Western Australia.

⁹⁶ Patara, : IGR 3, 659Google Scholar = TAM 2.2, 396Google Scholar (inscription, with a minute plan ibid., p. 142 D). Kadyanda, : IGR 3, 507Google Scholar = TAM 2.2, 651Google Scholar. The date of the Patara inscription is discussed by Balland, A., Xanthos 7 (1981) 3–4Google Scholar.

⁹⁷ Ling, R. J., AS 31 (1981) 31–46Google Scholar.

⁹⁸ Hecht, K., WAAP: Nochmals zwei Aquädukte der Kaikos-Leitung (LIW Mitt 61, 1978) 23–5Google Scholar; id., Hist. Wasser, 19; Weber, 1904, 89–90.

⁹⁹ See provisionally Coulton, J. J., AS 32 (1982) 121–2Google Scholar.

¹⁰⁰ Distribution pipes at ground level, which would generally operate under some pressure, were in contrast very widespread, as Germain de Montauzan, 178, notes.

¹⁰¹ Germain de Montauzan, 207–8; Smith, N. A. F., Man and Water, a History of Hydro-Technology (1976) 89–93Google Scholar; Hodge, 1983, 193–4. Hodge and Smith suggest that inverted siphons were used only where the drop in ground level was too great to be made up by a built aqueduct, i.e. greater than c. 50 m. This may be true in Gaul, but it is not true of the stone inverted siphons discussed here.

¹⁰² Germain de Montauzan, 59–63, 79, 90–3, 101–5, 118–35, 176–220.

¹⁰³ Spain: Hodge, 1983, 190–1. Clay pipes in concrete jacket: Olami, Y., Peleg, Y., IEJ 27 (1977) 132–3Google Scholar; Thompson, F. H., Arch. Journal 111 (1955) 106–28Google Scholar. Clay pipes with stone sleeves: Fahlbusch, , VAGRW, 65, 80–2 and above n. 67Google Scholar.

¹⁰⁴ The following stone pipe blocks are known to us (excluding stone junction blocks):

Western Europe

Arezzo (Poti): Pasqui, A., Notizie degli Scavi 1878, 332, n. 2Google Scholar.

Avenches (Switzerland): Dunant, E., Guide illustré du Musée d'Avenches (1900) 3Google Scholar; Grenier, , Manuel 4, 110Google Scholar.

Lyon: Ingenieurtechnik, 564Google Scholar; but the inverted siphon is denied by Germain de Montauzan, 88–9.

Padua: Garbrecht, G. in WARom, 181Google Scholar.

Rome: Van Deman, , Building, 141Google Scholar.

North Africa

Ruzasus (Azeffoun): Gsell, S., Monuments antiques de l'Algérie 1 (1901) 257Google Scholar.

Thammugadi (Timgad): Lohmann, H. in Wohnungsbau im Altertum (DAI Disk, zur Arch. Bauforschungen 3 (n.d.) 179Google Scholar, figs. 9–10.

Thugga (Belad Zehma): Carton, D., Revue Tunisienne 3 (1896) 544–5Google Scholar.

Balkans

Eretria: in Museum garden.

Istria: Canarache, V., Studii şi Cercetaři de Istoria Veche 2.2 (1951) 61Google Scholar, fig. 7, 16; Condurachi, E. et al. , Histria: Monografie Arheologică 1 (1954) 359–60Google Scholar.

Tomis: Wodociagi Rzymskie, fig. 140.

Asia Minor and Islands

Akmonia: Weber, 1905, 207.

Alabanda/Gerga: Özis, Ü. et al. , Hist. Wasser, 3Google Scholar.

Ankyra: Aqueducts, 4Google Scholar.

Antioch by Pisidia: Weber, 1904, 96–101.

Apameia (Phrygia): Weber, 1904, 89–90.

Aphrodisias: by Odeion.

Aspendos: Lanckoronski, K. et al. , Städte Pamphyliens und Pisidiens 1 (1890) 120–4, 179Google Scholar no. 64h; see also n. 43.

Ephesos: Forchheimer, P., FiE 3 (1923) 250–1Google Scholar.

Kibyra: Aqueducts, 3Google Scholar, citing Falkener.

Laodikeia on the Lykos: Ramsay, W. M., The Cities and Bishoprics of Phrygia 1 (1895) 48–9Google Scholar; Weber, 1898, 1–13.

Methymna: Koldewey, R., Die antike Baureste der Insel Lesbos (1890) 17–18Google Scholar; Malinowski, R., Fahlbusch, H., WAHellas, 209–10Google Scholar.

Oinoanda: above.

Patara: below p. 00, and n. 149.

Prymnessos: Weber, 1905, 208.

Samos: Aqueducts, 4Google Scholar, citing Pococke.

Smyrna: Weber, 1899, 4–25.

Tralleis: Weber, 1904, 89–90.

Trapezopolis: Weber, 1904, 92–3.

Levant

Apameia (Syria): Garbrecht, G. in WARom, 181Google Scholar.

Beth-Yerah (Philoteria ?): Mazar, A., LIW Mitt 82 (1984) 13Google Scholar.

Hippos (Susita): Mazar, A., LIW Mitt 82 (1984) 13Google Scholar.

Jerusalem: Mazar, A., Qadmoniot 5 (1972) 120–5Google Scholar (Hebrew, with detailed contour map); id., LIW Mitt 82 (1984) 12–17 (I am most grateful to Dr. Mazar for an offprint of the latter).

Palmyra: Michaelowski, K., Palmyra (1970), pl. 65Google Scholar; Crouch, D. P., Studia Palmyrenskie 6–7 (1975) 176–80Google Scholar.

¹⁰⁵ The importance of the saddle was identified by G. Weber, 1905, 209.

¹⁰⁶ See n. 102 above. Note also the two stone pressure pipelines used at Laodikeia on the Lykos. Although the stone pressure pipe serving Jerusalem has about one and a half times the cross-sectional area of the individual pipes at Lyon, this is still much less than could be given to a built channel.

¹⁰⁷ Healy, J. F., Mining and Metallurgy in the Greek and Roman World (1978) 61–2Google Scholar.

¹⁰⁸ de Jesus, P. S., The Development of Prehistoric Mining and Metallurgy in Anatolia (BAR Int. Ser. 74, 1980) 64–9, maps 12–13Google Scholar.

¹⁰⁹ Germain de Montauzan, 205.

¹¹⁰ The stone for the Oinoanda pressure pipe would have weighed about 280 tons, but was available within a kilometre, and with no uphill haulage needed.

¹¹¹ This was not always an overriding consideration, for decorative stone was often imported from a considerable distance for prestige buildings.

¹¹² The tension in the cross-section of a pipe wall is given by the formula S = (P.d)/2W, where

S = tension (N/cm²)

P = pressure (N/cm²)

d = internal diameter (cm)

W = wall thickness (cm).

¹¹³ For analyses of pipeline sealant see Garbrecht, G., Fahlbusch, H., WAAP: die Madradağ-Leitung (LIW Mitt 37, 1973) 86–94Google Scholar; Fahlbusch, H., WA Hellas, 147Google Scholar; R. Malinowski, H. Fahlbusch, ibid., 212–18; R. Malinowski, Aqueducs romains, 251.

¹¹⁴ C. R. Cockerell (see below n. 149).

¹¹⁵ Whether these holes should be connected with the notorious collivaria of Vitruvius 8.6.6 is very doubtful. The various explanations of this term are discussed by Fahlbusch, H., VAGRW, 86–91Google Scholar (concluding that it refers to pressure-release towers), and Hodge, 1983, 213–17 (deciding rather for drain cocks).

¹¹⁶ Thompson, H. A., Hesperia 25 (1956) 49–50CrossRefGoogle Scholar; Thompson, H. A., Wycherley, R. E., The Athenian Agora 14Google Scholar: the Agora of Athens (1972) 199Google Scholar.

¹¹⁷ Fabricius, E., Ath Mitt 9 (1884) 175–6Google Scholar; Priene, 73; Robinson, D. M., Graham, J. W., Excavations at Olynthus 8 (1938) 310–11Google Scholar; Robinson, D. M., Excavations at Olynthus 12 (1946) 108–10Google Scholar; Martin, R. et al. , Sicilia Antiqua 1.2 (1980) 406Google Scholar; Shear, T. L., Hesperia 53 (1984) 48–9Google Scholar. These and similar instances are discussed by E. Stenton, op. cit. (n. 15) 185–9.

¹¹⁸ Fahlbusch, , VAGRW, 85Google Scholar. Özis, Ü. et al. , Hist. Wasser, 3, suggest more generally “inspection and cleaning”Google Scholar.

¹¹⁹ On sludge removal see Hodge, 1983, 207 and also n. 85 above.

¹²⁰ This explanation was suggested by Weber, 1898, 6; holes at joints: Laodikeia (ibid.) and Aspendos (Fahlbusch, VAGRW, fig. 49).

¹²¹ The figures are read from the contour plan in Mazar, A., Qadmoniot 5 (1972) facing p. 124Google Scholar.

¹²² Weber, 1904, 89–90; id., 1899, 25.

¹²³ Singer, C. et al. , A History of Technology 1 (1958) 500Google Scholar.

¹²⁴ Forchheimer's, tests in AnzAkadWien, phil.-hist.kl. 35 (1898) 38Google Scholar; those of Telford reported by Mackain, D., Institution of Civil Engineers: Minutes of Proceedings 2 (1843) 136Google Scholar.

¹²⁵ Weber, 1905, 205–6.

¹²⁶ Weber, 1904, 89–90.

¹²⁷ Gräber, , AvPerg, 386–98Google Scholar. Since the bottom storey was about 11 m. high and the inverted siphon about 21 m. high, the built structure here too coped with about one third of the total.

¹²⁸ The Lacedaemonian fountain at Delphi (Herodotos 1.51) must have depended on a pressure supply; see also F. Glaser, op. cit. (n. 73) 152–5.

¹²⁹ The Madra Dağ system: Gräber, , AvPerg, 368–83Google Scholar; Garbrecht, G., Holtorff, G., WAAP: die Madradağ-Leitung (LIW Mitt 37, 1973)Google Scholar; the Attalos and Damophon pipelines: Fahlbusch, H. in WAHellas, 143–6Google Scholar.

¹³⁰ Weber, 1905, 209–10, summarizing the conclusions of his three earlier articles.

¹³¹ Gerkan, A. v., Griechische Städteanlagen (1924) 89CrossRefGoogle Scholar; Van Buren, A. W., RE 8A (1955) 473Google Scholar.

¹³² Fahlbusch, H., VAGRW, 63–93Google Scholar; on the date of the stone pipeline at Patara see below pp. 57–8.

¹³³ Mazar, A, LIWMitt 82 (1984) 13, 16Google Scholar.

¹³⁴ Ward-Perkins, J. B., BSR 23 (1955) 115–23Google Scholar; Bean, G. E., Turkey's Southern Shore² (1979) 53Google Scholar.

¹³⁵ E. Condurachi, op. cit. (n. 104) 361.

¹³⁶ For stone pressure pipes on the Ayasoluk hill of Ephesos see n. 104. Foss, C., Ephesos after Antiquity (1979) 92 n. 96, 183–4Google Scholar, argues that there was no piped supply here before Justinian.

¹³⁷ Using the Manning equation: Q = (A/n).R^2/3.i^1/, where

Q = quantity of water (m.³/sec.)

A = area of flow (m.²)

n = coefficient of friction

R = hydraulic radius (m.) ( = A/P where P = wetted perimeter)

i = gradient.

In this case A = π × 0·0875², n = 0·02 ±, P = π × 0·175, i = 0·00607 (or ·01607), so that Q = (0·0241/0·02) × 0·0414^2/3 × 0·00607^1/2 (or × 0·01607^1/2) = 0·00932 (or 0·01517) m.³/sec = 805 (or 1310) m.³/day.

¹³⁸ In this case, using the same equation, A = π × 0·07², n = 0·013 ±, P = π × 0·14, i = 0·01 (or 0·005), so that Q = (0·0154/0·013) × 0·035^2/3 × 0·01^1/2 (or × 0·005^1/2 = 0·0127 (or 0·00896) m.³/sec = 1095 (or 774) m.³/day.

¹³⁹ Using the equation Q = 1·446 × (h + v²/2g)^3/2 × b, where

Q = quantity of water (m.^3/sec)

h = depth of water at lip of fall (m.)

v = velocity of flow (m./sec)

g = gravitational acceleration (m./sec²)

b = breadth of channel (m.).

In this case h = 0·048, v = 1·446 × h^1/2 = 0·3168, g = 9·82, b = 0·085, so that Q = 1·446 × (0·048 + 0·3168²/19·64)^3/2 × 0·085 = 0·001504 m.³sec = 130 m.³/day. Germain de Montauzan, 149, discusses springs whose discharge had been reduced by deforestation or earth movements.

¹⁴⁰ Multiple pipelines are mainly Hellenistic, e.g. the supplies to Pergamon (n. 129 above) and the Marnas supply to Ephesos (n. 25 above). See above p. 00.

¹⁴¹ Neyses, A., Aqueducs romains, 281–2Google Scholar.

¹⁴² The difficulties associated with the Roman method of measuring water flow in quinaria have been discussed recently by Fahlbusch, H. in WARom, 137–44Google Scholar, and Hodge, A. T., AJA 88 (1984) 205–16CrossRefGoogle Scholar.

¹⁴³ Henschel, A. in Wayner, M. J. (ed.), Thirst (1st International Symposium on Thirst in the Regulation of Body Water, 1964) 19–28Google Scholar.

¹⁴⁴ So S. Cairncross, R. G. A. Feachem, op. cit. (n. 13) 1; but note the variation in consumption in different areas of the world recorded by Saunders, R. J., Warford, J. J., Village Water Supplies (1976) 124Google Scholar.

¹⁴⁵ Fahlbusch, loc. cit. (n. 54), emphasizes the substantial storage facilities in some cities, but Hodge, A. T., AJA 85 (1981) 489–91Google Scholar, is probably right to conclude that taps were not normally turned off.

¹⁴⁶ e.g. Pergamon: 160–190 1./pers.day (H. Fahlbusch, VAGRW, 173, id., WAHellas, 152); Rome: 520–900 l./pers/day (H. Fahlbusch VAGRW, 151, id., WARom, 138); Pompeii: 800 l/.pers/day (Eschebach, H., Aqueducs romains, 100–1Google Scholar); Trier: 1000 l./pers/day (Neyses, A., Aqueducs romains, 278Google Scholar).

¹⁴⁷ Suggested in the first instance by the absence of High Imperial houses (Coulton, J. J., PCPS NS 29(1983) 4, n. 13Google Scholar).

¹⁴⁸ On the relation of Roman water systems to status rather than population see Levau, P., Paillet, J.-L., Aqueducs romains, 231–4Google Scholar, Ward-Perkins, B., From Classical Antiquity to the Middle Ages (1984) 119–27Google Scholar.

¹⁴⁹ This account is based on the following sources, supplemented by a brief visit to the site in 1981. The first and fullest treatment is by C. Texier, op. cit. (n. 92) 192–3, 224, pl. 179 (but his illustration grossly compresses the length of the aqueduct, and exaggerates the steepness of the slope at either end); some further information and an even less accurate illustration (derived from C. R. Cockerell) in Institution of Civil Engineers: Minutes of Proceedings 14 (1855) 206–8Google Scholar. Additional comments and photographs: Malinowski, R., Fahlbusch, H., WAHellas, 208–9, 214Google Scholar, figs. 3.7; Fahlbusch, H., VAGRW, 56, 65, 83.6, 93, fig. 50Google Scholar.

¹⁵⁰ Texier, op. cit. (n. 93) 224, calls these openings corbelled, and shows their sides converging sharply; in fact they converge only slightly.

¹⁵¹ R. Malinowski, H. Fahlbusch, op. cit. (n. 149) 214.

¹⁵² Ibid., 209, fig. 7; Fahlbusch, H., VAGRW, 85Google Scholar, fig. 50b. The block in question differs from the possible sludge outlets just mentioned.

¹⁵³ Numerous fragments of clay pipe were noted by Texier, op. cit. (n. 93) 193, with the suggestion that they belonged either to an earlier phase than the aqueduct or were inserted into the stone pipeline. The latter suggestion has been generally discarded.

¹⁵⁴ The inscription was not observed by Texier, who found the foot of the aqueduct much overgrown, but it has been recorded by M. Wörrle, and is referred to by Malinowski, R., Fahlbusch, H., WAHellas, 208Google Scholar.

¹⁵⁵ The only parallel for a Hellenistic structure above ground is a small bridge where the Madradağ pipeline crosses the west branch of Kemerdere on the way to Pergamon (Garbrecht, G., Holtorff, G., WAAP: die Madradağ-Leitung (LIWMitt. 37, 1973) 54–6Google Scholar). This is a minor structure a long way from the city, whereas the openings in the Delik Kemer aqueduct show that it was on a more or less frequented route. However, even this crossing of Kemerdere West may not be an original feature of the Madradağ system, for a pipe running up the west side of the valley (ibid., figs. 15, 39, 40, interpreted by Garbrecht and Holtorff (ibid., 36) as a feeder pipe) may be the remains of a ground level crossing further up the stream, comparable with that in the east branch of the valley. The small bridge would then have been introduced during later repairs to the system, for the pipeline remained in use even after the larger Roman system was constructed (Fahlbusch, H., WAHellas, 150Google Scholar).

¹⁵⁶ Radke, G., RE 18 (1949) 2557Google Scholar.

¹⁵⁷ See above n. 89.

¹⁵⁸ For the cistern see Bean, G. E., JHS 86 (1948) 57–8Google Scholar, who also comments on the poor water supply at present. Evidence of a lower sea level in antiquity at Aperlai (Bean, G. E., Lycian Turkey (1978) 103Google Scholar) and Phaselis (Schäfer, J. et al. , Phaselis: Beiträge zur Topographie … (1st Mitt Beih. 24, 1981) 29, 70–86Google Scholar.

¹⁵⁹ See above n. 96.

¹⁶⁰ Pliny, , Ep. 10.37Google Scholar.