¹ E.g. by Belon and Delia Valle. For a full bibliography and discussion of the accounts of travellers in the Troad, to which this paper is indebted, see Cook, J. M., The Troad, an archaeological and topographical study, Oxford, 1973Google Scholar, Ch.2 passim. Since such travellers' accounts will form the basis of much of this discussion, a detailed bibliography of relevant descriptions and illustrations has been set out in an appendix, and references in the text by name only are to the (usually quite brief) passages cited there.

² Op. cit. (see appendix), p.44 f.

³ Ephesus: Maccanico, Rosanna, “Ginnasi Romani ad Efeso”, Arch. Class. XV, 1963, p.32 ffGoogle Scholar, with references. Sardis: Hanfmann et al., reports in BASOR every year from no.154, April 1959 onwards; for a recent plan see no.211, October 1973, p.16, Fig.l (= Fig.10 here). Hanfmann, G.M.A. and Ramage, N.H. have since published a fuller plan (Sculpture from Sardis, Cambridge, Mass., 1978, Fig.4)Google Scholar which contains nothing unexpected. For the name “gymnasium” see Keil, , JOAI XXV, 1929, B 25–7Google Scholar, and Hanfmann, , BASOR 154, April 1959, p.16Google Scholar (but contrast Johnson, , BASOR 158, April 1960, p.10Google Scholar).

⁴ Rooms or areas of the gymnasium at Troas will be referred to throughout by their designations on Koldewey's plan, Pl. II, p.36 (on which Fig.1 here is based).

⁵ Page 44 f.

⁶ Designations of rooms in the East Gymnasium are those of Miltner's plan published in JOAI XXVIII, 1933, B 7–8Google Scholar, Fig.2 (here Fig.5). It is strange that Koldewey trusted the second volume of Antiquities of Ionia (London, 1797)Google Scholar for Ephesus but chose to discount its evidence for Troas; in spite of the fact that Pars's plate XXXIX is a generally accurate view of the remains still visible at Ephesus, Koldewey condemned his views of the gymnasium at Troas (plates LII, LIII, here Pls.IVd and VIa) as “malerisch” and “unzuverlässig” (p.36, n.1). Krencker more consistently mistrusts AI for the Ephesian building as well, since he obviously thinks it depicts the Vedius Gymnasium (Krencker, D., Krüger, E., Lehmann, H. and Wachtler, H., Die Trierer Kaiserthermen, Augsburg, 1929, p.287Google Scholar), but Koldewey is making AI the basis of his discussion of the similarities between the two buildings, and yet denying its validity as a source for one of them. The reliability of Pars's drawings of Troas is discussed below.

⁷ According to Wood there were half-columns 8 ft 6 ins (2·6 m) in diameter; according to Clarke they were columns 8 ft (c. 2·45 m) in diameter. Koldewey restores his “Röhrenpfeiler” (p.42 f., Pl. III, 3 and 4) as piers with two semicolumnar faces containing two vertical pipes, 2·8 m wide and 3·2 m across, but it is equally possible to restore the fragment he illustrates as an engaged half-column containing one pipe and with a diameter of 2·8 m. He sets up his piers on the foundations at a, but since the calidarium wall stood here this cannot have been the case – he offers no evidence which might compel one to give the possibility more serious consideration. These columns are too massive to have formed part of an engaged interior order like that of the “frigidaria” of the imperial thermae at Rome, and none of the travellers mentions or depicts any sign of such columns having been attached to the calidarium wall; no-one mentions their material, but it seems unlikely to have been decorative marble – although the use of stucco cannot of course be discounted. Their diameter is almost as great as the width of the central pair of piers of the east wall of the calidarium, and it is possible that they were attached to or part of the corresponding piers to the west – in which case if the columns had normal proportions they must have been on the exterior; this is where their association with Clarke's stair would tend to put them in any case (for the terrace of the Theatre Gymnasium see F. Miltner, Ephesos, Stadt der Artemis und des Johannes, Vienna, 1958, p.73 and Fig.62). There is however no parallel for a monumental architectural treatment to the exterior of a calidarium either in Asia Minor or elsewhere (Krencker is probably right in assigning the Wiltheim drawing of the “Barbarathermen” in Trier to the exterior of the frigidarium rather than the calidarium, op. cit. p.245 f., Fig.361b; for the contrary view see Wightman, E. M., Roman Trier and the Treveri, London, 1970, p.84Google Scholar). Where a large order occurs on the exterior of the “frigidarium” of baths of the imperial type, as in those of Caracalla (Krencker, op. cit., Fig.400), it is smaller than the interior order of the “frigidarium” and in any case acts as a backdrop to the natatio: only in the early Baths of Capito at Miletus does a natatio occur in the palaestra of a thermal establishment in Asia Minor (von Gerkan, A., Krischen, F., Milet I, 9, Berlin, 1928, p.25Google Scholar, Fig.27, Pl. I), and there, as at Rome, it is at the other side of the building from the calidarium. In the Sardis and Vedius gymnasia (above, note 3), below, note 29) the “Kaisersaal” faces onto the palaestra, but the effect is muted by the colonnade in front; better parallels for a monumental façade can be seen in the baths of the Maeander area, at Aphrodisias (Collignon, M., CRAI 1904, p.703 ff.Google Scholar; Mendel, G., CRAI 1906, p.158 ff.Google Scholar, Erim, K. T., TAD XIX – 1, 1970, p.75Google Scholar, Fig.32), Hierapolis (Krencker, op. cit., Figs. 428, 429, 431, 434a, 434c), Nysa (Diest, W. v., Nysa ad Maeandrum, Berlin, 1913, Pl.VIIGoogle Scholar) and apparently also at Laodicea ad Lycum, but this is a wholly distinct group from that under discussion here. There are therefore no close parallels which can lead to a better understanding of the position and function of these “Röhrenpfeiler”, and they are omitted from the reconstruction attempted here. It may be hoped that some evidence that will eventually clarify this issue survives on the site.

⁸ The surviving block (Pl.Vd) is at least of the same general type as that appearing in Koldewey's Pl. III, 7.

⁹ Hobhouse, describing his visit of 1810, mentions “three lofty portals … fronting the west”; this is unquestionably a description of the calidarium wall. He thus implies that it was still standing, but immediately says “in this quarter the earthquake had been most destructive; … I am at a loss for several portions of the stately ruins which have been mentioned by those who preceded us”. Prokesch-Osten and Texier unquestionably mention only one arch to the west, probably the same as Napier's “archway”, thus between 1801 (Clarke) and 1826 most of this wall collapsed, and the 1809 earthquake is the obvious cause. Hobhouse's account mentions great destruction but otherwise describes the building virtually as it is in Clarke's version; in view of the many similarities between these two accounts it would greatly simplify matters to assume that Hobhouse based his version largely on Clarke's – he certainly knew the literature on the site.

¹⁰ If Pars (Pl.VIa), Borra (Pl.VIb) and Le Bruyn (Pl. IVb) got the relative heights of the arches in P even remotely right it is clear that the arch must have been demolished and its jambs obscured by the fallen rubble, rather than buried.

¹¹ Both Pars's and Borra's drawings (Pl.VIa and b) show these arches blocked, quite clearly so in Borra's case; yet both Pococke and Correr describe them as open, and Pococke on his plan (Fig.3) clearly distinguishes blocked and unblocked arches. If there ever was a blocking wall it vanished early; on Pars and Borra see below.

¹² See note 6 above.

¹³ Koldewey is concerned to establish a parallel between this building and the lowest floor of the south stoa in the agora at Assos; hence the entrance must be on the opposite side of the building to the “bath cubicles” in room W (p.45 f.). Krencker tried to make Koldewey's plan approximate to the type of the imperial thermae (op. cit., p.285 f., Fig.425); this is a distinct improvement on Koldewey's restoration – room D for instance is rightly identified as a calidarium – but suffers from over-reliance on examples foreign to Asia Minor.

¹⁴ This plan is obviously the result of an incomplete set of measurements; when drawn to scale (Fig.2) it shows the calidarium wall and room P at the correct size, as well as the three main eastern niches. XVW and U are contracted together and given the dimensions of U alone, while the ranges of rooms XSG and RNF are outlined full-size outside the schematic representation of the northern arcade. The overall dimensions correspond closely to those of the actual building (the north wall must of course be 249 feet long and not 224, a figure derived from incomplete addition); in spite of its superficial appearance to the contrary the plan therefore shows nothing outside the building as planned by Koldewey.

¹⁵ See AI I, London, 1821, p.iiGoogle Scholar.

¹⁶ To take an extreme example, his wholly imaginary plan of Pessinus (I, pl.62) has caused much confusion, in spite of Perrot's detailed critique of it (Perrot, G. and Guillaume, E., Exploration archéologique de la Galatie et de la Bithynie, Paris, 1872, I, p.211 ffGoogle Scholar. Cf. also Le Bas's comments on Texier's work, published in Rev. Arch., III ^e S. no. XXXI, 1897, p.386Google Scholar.

¹⁷ Op. cit., pp.19 and 200 note 4.

¹⁸ Hutton, C. A., “The travels of ‘Palmyra’ Wood in 1750–51”, in JHS XLVII, 1927, p.102 ffCrossRefGoogle Scholar.

¹⁹ Wilton, Andrew, “An Appreciation of William Pars” in Chandler, Richard, Travels in Asia Minor, ed. Clay, Edith, London, 1971, p.xxiiiGoogle Scholar; the drawings for AI II plates LII and LIII are catalogued on p.xxxvii and illustrated in plates I and II.

²⁰ See note 6.

²¹ On Pars and Borra see Wilton, op. cit., p.xvii f.

²² See note 13.

²³ See JOAI XXVII, 1932, B 19–20Google Scholar, Fig.9.

²⁴ Koldewey gives the heights of the spring of the arches in W and S as the same in his table on p.44, nos. 9 and 11.

²⁵ Page 41.

²⁶ The fact that the arches in W and S spring at the same height (note 24) should in any case indicate that their spans were about the same; since Koldewey himself gives a difference of only 0·2 metres in the radius of the arches in W and S (being the height of the spring subtracted from the height of the top) it follows that the difference in span can only have been 0·4 metres – the figures derived thus from his table are 5·60 metres (W) and 6·0 metres (S). On his plan, however, he gives the span of the arches in W as 4·60 metres, which is the more accurate figure; this internal inconsistency serves to cast further doubt on k as the only unfixed point amid this confusion.

²⁷ Nos. 12 and 13 on Koldewey's table, p.44.

²⁸ The position of these doors on the Dilettanti plan (Fig.4) would allow them to fit between the piers of a blind arcade, but on this occasion Koldewey's plan is more accurate. There is a parallel for an arcade on only one side of a gallery in the “Mosaiksaal” of the “Varius Bath” at Ephesus; Vetters, Hermann, JOAI L, 1972–1975Google Scholar, Grabungen 1971/72, p.35, Figs. 24–5.

²⁹ A plan of the Vedius Gymnasium is given in JOAI XXVI, 1930, B 19–20Google Scholar, Fig. 6 (= Fig.9 here).

³⁰ Designations as in Figs.9 and 10; for Sardis see note 3.

³¹ See Koldewey, Pl. III, 2.

³² It is not impossible that the arches in room P were originally walled up, their real or imagined function being to take the load of the vaults to the piers.

³³ See von Blankenhagen, P. H., “The Imperial Fora”, in J. Soc. Archit. Historians XIII, 4, Dec. 1954, p.21 ff.CrossRefGoogle Scholar; p.24.

³⁴ This was at any rate Koldewey's interpretation of these pipes, more or less (p.46); Ward-Perkins would also seem to favour such an interpretation of wall-pipes (JRS XLIII, 1953, p.212Google Scholar). It is possible however that they simply carried rainwater down from the roof, which above Q and O seems to have been lower than over P and U, at least (cf. Brödner, E., Untersuchungen an den Caracallathermen, Berlin, 1951, p.32 fGoogle Scholar. and Fig.10); in which case there would be no direct evidence for basins in Q and O.

³⁵ Basins could have been accommodated in the “tepidarium “ here (as well as in the other buildings of this type) on the model of those inserted in the Hadrianic baths at Lepcis Magna; see Bartoccini, R., Le Terme di Lepcis, Bergamo, 1929Google Scholar, Fig.65 and Pl. IX–X, areas X' and X (see Fig.11 here).

³⁶ All these buildings can be classified as belonging to Krencker's “kleine Kaisertyp” (op. cit. p.180), although they form a clear sub-type of their own. The “Kaisertyp” is an elaboration of the “Ringtyp” (op. cit. p.178), and every such building has access from the frigidarium to the central tepidarium; Krencker duly provided such access in his own restoration of the gymnasium at Troas (Fig.425, interpreting U as the frigidarium proper, and P as an exedra projecting from it). Connections between O and T/I, and I and B, are also necessitated by this classification.

³⁷ See Koldewey, p.47 and Pl. III.

³⁸ Or rather, its facing voussoirs are visible. This arch is not shown on Fig.4.

³⁹ Apart from the similarity of plan, the similarity of construction in the calidarium walls has also been mentioned above. One might also point to such shared features as the string course of bricks running tangentially to the arches in the rear gallery of each building (cf. Koldewey Pl. III 1 and Pl. IIb here for Troas); an obvious difference is that more brick is used at Ephesus than at Troas, but this is probably due to variations in its availability.

⁴⁰ Cook twice refers to this building as “the Baths of Herodes” (op. cit., pp.200 note 4 and 201), perhaps after Leaf, , Strabo on the Troad, Cambridge, 1923, p.238 fGoogle Scholar. For Herodes Atticus and Troas see Graindor, P., Un milliardaire antique, Hérode Atticus et sa famille, Cairo, 1930, pp.32 f., 57 ffGoogle Scholar.

⁴¹ Vit. Soph. II, 1, 3Google Scholar.

⁴² Page 46.

⁴³ Koldewey rules out all question of contemporaneity between the East and Troas gymnasia, p.45; but for the discrepancy in date between the East Gymnasium and its palaestra see Keil, in JOAI XXVII, 1931–1932, B 28 and 31 ffGoogle Scholar, where he also formulates the hypothesis that the layout of the galleries can be used to date these buildings.