PLATE LIII.
Myrelaion. The South Side.

Myrelaion.
The South Side.

Myrelaion. The Narthex, looking north.

Myrelaion.
The Narthex, looking north

To face page 196.

Architectural Features

The building is on the 'four column' plan. The dome, placed on a circular drum, is supported on four piers, and divided into eight concave compartments, with windows in the alternate compartments. The arms of the cross, the chambers at the angles, and the bema are all covered with cross-groined vaults that spring, like those in the chapel of the Pammakaristos (p. 151), from the vaulting level. The apsidal chambers have dome vaults, a niche on the east recessed in an arch to form the apse, and a niche both on the north and the south rising above the vaulting string-course. In the lowest division of the south wall stood originally a triple arcade with a door between the columns. The arcade has been built up, but the moulded jambs and cornices of the door, and the arch above it, now contracted into a window, still show on the exterior, while the columns appear within the church. Above the column string-course is a range of three windows, the central window being larger than its companions; higher up in the gable is a single light. The interior of the church has been much pulled about and cut away. The narthex is in three bays, separated by strong transverse arches, and terminates at either end in a high concave niche that shows on the outside. The central bay has a dome vault; the other bays have cross-groined vaults. The church had no gynecaeum, although Pulgher indicates one in his plan. A striking feature of the exterior are the large semicircular buttresses that show beyond the walls of the church—six on the south side, one on either side of the entrance on the west, and two on the east, supporting the apsidal chambers. In the last case, however, where entire buttresses would have been at once too large and too close together, the buttresses are only half semicircles. The apses project with three sides. The northern side of the church and the roof are modern, for the building suffered severely in 1784 from fire. 320 The church stands on a platform, built over a small cistern, the roof of which is supported by four columns crowned by beautiful capitals. Hence the Turkish name of the mosque, Bodroum, signifying a subterranean hollow. Gyllius 321 is mistaken in associating this church with the large underground cistern situated lower down the slope of the hill close to the bath Kyzlar Aghassi Hamam.

PLATE LIV.
Myrelaion. The Interior, looking east.

Myrelaion.
The Interior, looking east.

Myrelaion. The south-west cross Angle.

Myrelaion.
The south-west cross Angle.

To face page 198.

Since the above was in print, the church has, unfortunately, been burnt in the great fire which destroyed a large part of Stamboul on the 23rd July 1912 (see Plates II., III.).

Note

Gyllius (De top. C.P. iii. c. 8) places the Horreum, the statue of Maimas, the house of Craterus, the Modius, and the arch bearing the two bronze hands, after passing which a criminal on the way to punishment lost all hope of reprieve, near this church; basing that opinion on the statement of Suidas that these buildings stood near the Myrelaion. But there was a Myrelaion also (Codinus, De aed. p. 108) in the district in which the Shahzadé mosque is situated. The buildings above mentioned were near this second Myrelaion. On the other hand, the Chrysocamaron near the Myrelaion mentioned by Codinus (De signis, pp. 65-66) stood near the church under our consideration, for it was close to the church of S. Acacius in the Heptascalon. So also, doubtless, did the xenodocheion Myrelaion (Du Cange, iv. p. 160), possibly one of the many philanthropic institutions supported by Helena (Theoph. Cont. p. 458), the daughter of Romanus Lecapenus and wife of Constantine VII. Porphyrogenitus.


Plan of the Church and Longitudinal Section.

Figs. 66 and 67

308 De top. C.P. iii. c. 8.

309 Banduri, iii. p. 48.

310 Ibid. ut supra.

311 Theoph. Cont. p. 402.

312 Scylitzes, in Cedrenus, ii. p. 649.

313 Theoph. Cont. p. 404.

314 Ibid. pp. 461, 757.

315 Scylitzes, ut supra, pp. 648-49.

316 Theoph. Cont. p. 402.

317 Ibid. p. 420.

318 Ibid. p. 473.

319 Ibid. pp. 403-4.

320 Chevalier, Voyage de la Propontide et du Pont Euxin, vol. i. p. 108.

321 De top. C.P. iii. c. 8, 'habens inter se cisternam, cujus camera lateritia sustinetur columnis marmoreis circiter sexaginta'; cf. Die byzant. Wasserbehälter, pp. 59, 222-23. The bath of Kyzlar Aghassi Hamam may represent the bath built by the eunuch Nicetas, in the reign of Theophilus, and was probably supplied with water from the cistern beside it (Banduri, vi. p. 133).


CHAPTER XII

THE CHURCH OF S. JOHN THE BAPTIST IN TRULLO, ACHMED PASHA MESJEDI

The identification of the church of S. John the Baptist in Trullo (Μονὴ τοῦ ἁγίου προφήτου προδρόμου Ἰωάννου τοῦ ἐν τῷ Τρούλλῳ) with the mosque of Achmed Pasha Mesjedi is based on two reasons: first, because of their common proximity to the church of the Pammakaristos, 322 now Fetiyeh Jamissi; secondly, on the ground of the tradition current in the Greek community on that point. The latter reason is in this case particularly strong, seeing the church of the Pammakaristos was the patriarchal cathedral almost immediately after the Turkish conquest, and retained that honour until 1591. 323 The highest Greek ecclesiastical authorities were therefore in a position to be thoroughly acquainted with the dedication of a church in their close vicinity. In 1578 the protonotarius of the patriarch showed Gerlach the site of the Trullus close to Achmed Pasha Mesjedi. 324

The church is mentioned in history only by Phrantzes, 325 who informs us that when the Patriarch Gennadius transferred the patriarchal seat to the monastery and church of the Pammakaristos, certain nuns previously accommodated in that House were removed to the neighbouring monastery of S. John Baptist in Trullo. Phrantzes explains the designation of the church, 'in Trullo,' as derived from a palace named Trullus which once stood in the vicinity to the north of the Pammakaristos. It was the palace, adds the historian, 326 in which the Council of Constantinople, known as the Concilium Quinisextum (Πενθέκτη), or the second Concilium Trullanum, assembled in 692, in the reign of Justinian II. But the palace Trullus, in which the first Concilium Trullanum met in 680, was one of the group of buildings forming the Great Palace 327 beside the Hippodrome, and there the second Concilium Trullanum also held its meetings. 328 Phrantzes is therefore mistaken in associating the Council of 692 with a palace in the vicinity of the Pammakaristos and Achmed Pasha Mesjedi. But his mistake on that particular point does not preclude the existence of a palace named Trullus in the neighbourhood of the Pammakaristos. In fact, the existence of such a palace in that district is the only possible explanation of the attachment of the style 'in Trullo' to a church on the site of Achmed Pasha Mesjedi. Nor is it strange to find a name pertaining primarily to a building in the Great Palace transferred to a similar building situated elsewhere. The imperial residence at the Hebdomon, for example, was named Magnaura after one of the halls in the Great Palace. 329 There was an Oaton or Trullus in the palace of Blachernae, 330 and in the palace at Nicaea. 331 Consequently, a palace known as the Oaton or the Trullus might also be situated near the Pammakaristos, to command the fine view from that point of the city. Mordtmann, 332 indeed, maintains that the building to which Phrantzes refers was the palace at Bogdan Serai, the subsequent residence of the Moldavian hospodar in Turkish days, and that the church of S. John in Trullo was not Achmed Pasha Mesjedi, but the church of S. John in Petra (Kesmé Kaya) beside that palace. This opinion, however, is at variance with the statements of Phrantzes and Gerlach. Furthermore, the designation 'in Petra' was so distinctive a mark of the church of S. John near Kesmé Kaya, that the church could scarcely have been recognised under another style.

PLATE LV.
S. John in Trullo, from the south-west.

S. John in Trullo, from the south-west

Balaban Mesjedi. Interior view.

Balaban Mesjedi (page 265).
Interior View.

To face page 202.

Details from the Church—Details from the Pammakaristos—Details from the Pantepoptes.

Fig. 68.

Architectural Features

S. John in Trullo belongs to the ordinary 'four column' type of church building, and has a narthex. Its three apses are semicircular both within and without, presenting the only instance in Constantinople of apses semicircular on the exterior. The central apse projects m. 3 beyond the body of the building, and was lighted by a large but low window, divided into three lights by two pilasters crowned with carved capitals (for details see Fig. 68); the diaconicon has been built up to form the mihrab of the mosque; the prothesis, to the north, has a barrel vault.

Details from S. Andrew in Krisei—Details from the Chora.

Fig. 69.

The drum dome is octagonal, with eight ribs and as many windows. It seems large for the size of the church, and is lower than usual inside. The windows do not cut into the exterior cornice of the dome. Originally the dome arches rested on four piers or columns, but these have been removed in the course of Turkish repairs, and the dome arches are now supported by beams running across the church, under the impost of the arches.

The arms of the cross to the north and south have barrel vaults, and the walls are pierced by triple windows. Two capitals built into the exterior face of the northern wall, and marked with a cross, were doubtless the capitals of the shafts which divided the northern window into three lights. The western arm of the cross is covered by the roof of the narthex, and lighted by a small round-headed window above it. The small narthex is in three bays, covered with cross-groined vaults.

It is not probable that the church was converted into a mosque before 1591, when the patriarchal seat was removed from the Pammakaristos to S. Demetrius beside the Xyloporta. Nor could the conversion have been later than 1598, the year in which Achmed Pasha—who converted the building into a mosque—died. 333

Plans and sections of of S. John in Trullo and S. Thekla.

Figs. 70 and 71.

322 Phrantzes, p. 307.

323 Patr. Constantius, p. 80.

324 Tagebuch, p. 456. On the way eastwards from the residence of the Moldavian agent (Bogdan Serai), says Gerlach, 'Auf diesem Spazier-weg hat mir Theodosius auch den Trullum auf der Seiten des Patriarchats gegen dem Sultan Selim gewiesen. Welches vor diesen ein sehr weiter Platz gewesen, nun aber nichts mehr da als ein rundes getäffeltes Haus, wie ein kleines Kirchlein ist.' Cf. his statement reported by Crusius in Turcograecia, p. 189: 'Patriarchatui contiguum est monasteriolum Joannis Baptistae a Graecis sanctimonalibus inhabitatum.'

325 Phrantzes, p. 307; cf. Turcograecia, p. 189.

326 It was also styled Ὠάτον, 'the Oval,' after the form of its roof or of the body of the building itself (Synax., Sept. 14). Vita Stephani. For the Ὠάτον, see Labarte, Le Palais impérial de Cons'ple, pp. 62, 121, 122, 186.

327 Vita Stephani Junioris, Migne, P.G. tom. 100, col. 1144 ἐν τῷ ἱερῷ παλατίῳ, ἔνθα ἐπιλέγεται ὁ Τροῦλλος ὅπερ ἡμεῖς Ὠάτον καλοῦμεν.

328 Balsamon, vol. i. col. 501 ἐν τῷ Τρούλλῳ τοῦ βασιλικοῦ παλατίου.

329 Theoph. p. 541.

330 Pachym. i. p. 405.

331 Acta et diplomata Graeca, iii. p. 65; cf. Paspates, Great Palace, p. 248, Metcalfe's translation.

332 Proceedings of Greek Syllogos of C.P., Archaeological Supplement to vol. xvii. p. 8. His principal reason seems to be the fact that a company of nuns occupied some of the cells in the old monastery of S. John in Petra when Gerlach visited the city. But, according to Gerlach, another sisterhood was at the same time accommodated in the small convent of S. John the Baptist near the patriarchate.—Turcograecia, p. 189.

333 Cf. Paspates, p. 304.


CHAPTER XIII

THE CHURCH OF S. THEKLA, TOKLOU IBRAHIM DEDÉ MESJEDI

In the quarter of Aivan Serai, a few paces to the rear of the Heraclian Wall, stands a small mosque known as Toklou Ibrahim Dedé Mesjedi, the architectural features of which proclaim it at once to be an old Byzantine chapel. There is no decisive tradition in regard to the identity of the building. The Patriarch Constantius is uncertain whether it should be recognised as the church of S. Nicholas or as the church of S. Thekla, two sanctuaries situated in the quarter of Blachernae. It cannot have been the former, inasmuch as the site of that church was near the Holy Well, still venerated by Christians and Moslems, 334 in the area enclosed between the Wall of Heraclius and the Wall of Leo the Armenian, now a picturesque Turkish cemetery. One argument for regarding the building as the church of S. Thekla, in this part of the city, is the striking similarity of its Turkish name Toklou to the Greek name Thekla, rendering it exceedingly probable that the former is a corruption of the latter, and a reminiscence of the original designation of the edifice. 335 Turkish authorities, however, have their own explanation of the name Toklou. In the Historical and Geographical Dictionary of Achmed Rifaat Effendi, we are told that a certain Toklou Dedé was the guardian of the tombs of the companions of Khaled, who took part in the first siege of Constantinople (673) by the Saracens. 'His real name was Ghazi Ismail; Dogulu was his nickname. Now Dogh is the Persian for a drink named Airan (a mixture of curds and water), and he was called Dogulu Dedé because during the siege his business was to distribute that drink to the troops. At his request a Christian church near Aivan Serai was converted into a mosque. The church was formerly named after its founder, Isakias.' 336 Another Turkish explanation of Toklou derives the epithet from the rare Turkish term for a yearling lamb, and accounts for its bestowal upon Ibrahim Dedé as a pet name given in gratitude for his services to the thirsty soldiers engaged in the siege of the city. 337 In keeping with these stories is the tradition that the cemetery in the area between the Walls of Heraclius and Leo V. the Armenian, is the resting-place of Saracen warriors who fell in the siege of 673. But have we not here the fancy-bred tales which Oriental imagination weaves to veil its ignorance of real facts? When etymology or history fails, romance is substituted. We may as well believe the tradition that the body of Eyoub, the standard-bearer of Mahomet, lies buried at the head of the Golden Horn, in the mosque of Eyoub, where the Sultan girds the sword on his accession to the throne. No Moslem graves could have been tolerated between the lines of the city's fortification in Byzantine days. The cemetery between the old walls near Toklou Ibrahim Dedé Mesjedi must therefore be later than the Turkish conquest. And as soon as Moslems were laid there, it was almost inevitable that a church in the immediate neighbourhood should either be destroyed or converted into a mosque. By what name that mosque would thenceforth become known was, of course, an open question. The new name might be purely Turkish. But when it sounds like the echo of a name which we know belonged to a Byzantine building in this quarter of the city before Turkish times, it is more reasonable to regard the new name as a transformation of the earlier Greek term, than to derive it from fine-spun etymological fancies and historical blunders. The identification, therefore, of Toklou Ibrahim Dedé Mesjedi with the church of S. Thekla, on the ground of the similarity of the two names, has a strong presumption in its favour.

PLATE LVI.
S. Thekla. North side, from the north-west.

S. Thekla.
North Side, from the north-west.

S. Thekla. East end.

S. Thekla.
East End.

To face page 208.

NOTE ON THE CHURCH OF S. THEKLA
(Chapter XIII.)

On page 209, note 3, I have said that if the mosque Aivas Effendi (more correctly Ivaz Effendi), which is situated behind the Tower of Isaac Angelus within the old area of the palace of Blachernae, could be proved to stand on the site of a church, the argument in favour of the identification of the Church of S. Thekla with Toklou Dedé Mesjedi would be weakened. Since this book went to the press, my learned friend Mr. X. A. Siderides has shown me a passage in the historical work of Mustapha Effendi of Salonica, published in 1865, where the mosque of Ivaz Effendi is described as a church converted into a mosque by a certain Ivaz Effendi who died in 1586, at the age of ninety. In that case we should have a Christian sanctuary whose position corresponded strictly with the position occupied by the Church of S. Thekla "in the palace of Blachernae," an indication not exactly accurate in regard to Toklou Dedé Mesjedi. In view of the late date of Mustapha Effendi's work, and the absence, so far as I can judge, of Byzantine features in the structure of the mosque, it is difficult to decide if the arguments in favour of the identification of the Church of S. Thekla with Toklou Dedé Mesjedi are entirely overthrown by the statement of Mustapha Effendi.

A second consideration in support of this identification is the statement made by Achmed Rifaat Effendi, that before the church became a mosque it was known by the name of its founder, 'Isakias.' For it is a matter of history that the church of S. Thekla was restored by the Emperor Isaac Comnenus 338 in the eleventh century. The association of his name with the building was therefore perfectly natural, if the building is indeed the old church of S. Thekla, otherwise it is difficult to account for that association.

There is, however, one objection to this identification that must not be overlooked. According to Byzantine authorities, the church of S. Thekla stood in the palace of Blachernae (ἐντὸς τῶν βασιλείων; ἐν τῷ παλατίῳ τῶν Βλαχέρνων 339). That palace occupied the heights above Aivan Serai, on which the quarter of Egri Kapou and the mosque of Aivas Effendi now stand, within the walls that enclose the western spur of the Sixth Hill. Toklou Ibrahim Dedé Mesjedi, however, does not stand within that enclosure, but immediately to the north of it, on the level tract that stretches from the foot of the Sixth Hill to the Golden Horn. If the reasons in favour of regarding the mosque as S. Thekla were less strong, this objection would, perhaps, be fatal. But the strip of land between the northern wall of the palace enclosure and the sea is so narrow, and was so closely connected with the life of the imperial residence, that a building on that tract might with pardonable inaccuracy be described, as 'in the palace.' 340

The church is mentioned for the first time in the earlier half of the eighth century as a chapel (εὐκτήριον) which Thekla, the eldest daughter of the Emperor Theophilus, restored and attached to her residence at Blachernae. 341 The princess was an invalid, and doubtless retired to this part of the city for the sake of its mild climate. To dedicate the chapel to her patron saint was only natural. As already intimated, the church was rebuilt from the foundations, in the eleventh century, by Isaac Comnenus, in devout gratitude for his escape from imminent death 342 in the course of his campaign against the barbarous tribes beside the Danube, when he was overtaken at the foot of the Lovitz mountain by a furious tempest of rain and snow. The plain on which his army was encamped soon became a sheet of water, and many of his men and animals were drowned or frozen to death. Thunder, lightning, and hurricane combined to produce an awful scene, and there were moments when the whole world seemed on fire. The emperor took shelter under a large oak, but, fearing the tree might be thrown down by the furious wind, he soon made for open ground. Scarcely had he done so when the oak was torn up by the roots and hurled to the earth. A few moments later the emperor would have been killed. This narrow escape occurred on the 24th September, the festival day of S. Thekla, and, therefore, attributing his deliverance to her intervention, Isaac rebuilt and greatly beautified the old sanctuary dedicated to her in Blachernae, and frequently attended services there in her honour. Anna Comnena 343 speaks of the restored church in the highest terms. According to her it was built at great cost, displayed rare art, and was in every way worthy of the occasion which led to its erection. Zonaras 344 is not so complimentary. He describes the church as a monument of the niggardliness of Isaac Comnenus. In any case, it was pulled down and rebuilt in the following century by the Emperor John Comnenus in splendid style, and dedicated to the Saviour. 345 As the beauty and wealth of a Byzantine sanctuary were exhibited in the lavish adornment of the interior, it is possible that the church of S. Thekla, though small and outwardly plain, may have been a beautiful and rich building in its latest Christian character. It had then the honour of seeing among the worshippers before its altar Anna Dalassena, the mother of the Comneni. For, when charged with the government of the Empire during the absence of Alexius Comnenus from the capital, that able woman came often to pray in this church, 'lest she should be immersed in merely secular affairs.' 346

Architectural Features
(For Plan see p. 206)

The building is an oblong hall, m. 13.55 by m. 5.4, divided into three compartments. It is now covered with a wooden roof, but the arrangements of the breaks or pilasters on the walls indicate that it had originally a dome. At the east end is a single apse, the usual side-apses being represented by two niches. The western compartment served as a narthex. During the repairs of the mosque in 1890, frescoes of the eikons which once decorated the walls were brought to view. On the exterior the apse shows three sides, crowned with a corbelled cornice. The central side is pierced by a window of good workmanship, divided by a shaft into two lights, and above the window are two short blind concave niches. High blind concave niches indent the other sides of the apse. In the northern wall are the remains of a triple window, divided by shafts built in courses. Above this is a row of three small windows.

334 Ancient and Modern C.P. p. 46.

335 Paspates, p. 359.

336 For this information I am indebted to Rev. H. O. Dwight, LL.D., late of the American Board of Missions in Constantinople.

337 Paspates, p. 357, note.

338 Anna Comnena, vol. i. p. 168.

339 Scylitzes, p. 647 (Cedrenus, vol. ii.); Zonaras, iii. p. 672.

340 If the mosque Aivas Effendi could be proved to stand on the site of a church, the argument against the identification of Toklou Dedé Mesjedi with the church of S. Thekla would be stronger.

341 Theoph. Cont. p. 147.

342 Anna Comnena, vol. i. p. 168.

343 Ibid. vol. i. p. 168.

344 Zonaras, iii. p. 672.

345 Ibid. ut supra.

346 Anna Comnena, vol. i. p. 169.


CHAPTER XIV

THE CHURCH OF S. SAVIOUR PANTEPOPTES, ESKI IMARET MESJEDI

The reasons which favour the identification of the mosque Eski Imaret Mesjedi, which is situated on the heights above Aya Kapou (Gate of S. Theodosia), with the church of S. Saviour Pantepoptes, the All-Seeing (παντεπόπτης), are the following: first, the tradition to that effect, 347 which in the case of a building so conspicuous can scarcely be mistaken; secondly, the correspondence of its position to that of the Pantepoptes, on a hill commanding an extensive view of the Golden Horn; 348 and finally, the architectural features which mark it to be what the church of the Pantepoptes was, a building of the Comnenian period. The church of the Pantepoptes was founded or restored by Anna Dalassena, 349 the mother of Alexius I. Comnenus (1081-1118), one of the most remarkable women in Byzantine history, combining to a rare degree domestic virtues with great political ambition and administrative ability. For twenty years she was associated with her son in the government of the Empire, and was the power behind the throne which he owed largely to her energy and devotion. About the year 1100 she laid aside the cares of state, and without renouncing altogether her royal style retired to rest in the monastery she had built, until her death, five years later, at an advanced age. 350 There is nothing of special importance to record in the annals of the House. Its inmates were occasionally disturbed by the confinement among them of some dignitary who had offended the Government, or by the theological disputes that agitated the ecclesiastical circles of the capital. 351 But for the most part life at Pantepoptes was quiet and peaceful. Only once does the monastery stand out conspicuous before the eyes of the world. When the Venetian ships under Henrico Dandolo, with the army of the Fourth Crusade on board, lined the shore of the Golden Horn from Ispigas and the church of S. Saviour the Benefactor to Blachernae (i.e. from Jubali Kapoussi to Aivan Serai) on Easter Monday, 12th April 1204, the Emperor Alexius Murtzuphlus established his headquarters beside the Pantepoptes. There he pitched his vermilion tent, marshalled his best troops, and watched the operations of the enemy. And thence he fled when he saw the walls on the shore below him carried by storm, and Flemish knights mounted on horses, which had been landed from the hostile fleet, advancing to assault his position. So hurried was his flight that he left his tent standing, and under its shelter Count Baldwin of Flanders and Hainault slept away the fatigue of that day's victory. 352 During the Latin occupation the church passed into the hands of the Venetians, and was robbed of many of its relics for the benefit of churches in the West. 353 Upon the Turkish conquest it served for some time as an imaret or refectory for the students and teachers of the medressé, 354 then in course of construction beside the great mosque of Sultan Mehemed. Hence its Turkish name. After serving that purpose it was converted into a mosque later in the reign of the conqueror.

PLATE LVII.
S. Saviour Pantepoptes. The door leading from the outer to the inner Narthex.

S. Saviour Pantepoptes.
The Door Leading from the Outer to the Inner Narthex.
View looking north.

S. Saviour Pantepoptes. The Dome, looking west.

S. Saviour Pantepoptes.
The Dome, looking west.

To face page 212.

Sketches from the Church.

Fig. 72.

Architectural Features

In plan the church belongs to the 'four column' type, and has two narthexes. The dome, placed on a drum, circular within and twelve-sided without, is carried on four piers which the Turks have reduced to an irregular octagonal form. It is divided into twelve bays by square ribs, and is lighted by twelve semicircular-headed windows. The cornice-string is adorned with a running leaf spray of a pleasing and uncommon design. The arms of the cross have barrel vaults, while the chambers at its angles are covered with cross-groined vaults. The apsidal chambers are small, with shallow niches on the north, south, and west, and a somewhat deeper niche on the east where the apse stands. These niches are carried up through a vaulting string-course, carved with a repeating leaf ornament, and combine with the groined vault above them to produce a charming canopy. The southern transept gable, though much built up, still displays the design which occurs so frequently in Byzantine churches, namely, three windows in the lunette of the arch (the central light rising higher than the sidelights), and three stilted arches below the vaulting string-course, resting on two columns and containing three windows which are carried down to a breastwork of carved marble slabs between the columns. The floor of the church is paved with square red bricks, except in the apses, where marble is employed. The gynecaeum, above the inner narthex, is divided into three bays separated by broad transverse arches. The central bay, which is larger than its companions, is covered with a dome vault, and looks into the body of the church through a fine triple arcade in the lunette of the western arm of the cross. The smaller bays are covered with cross-groined vaults. As elsewhere, the vaulting-string in the gynecaeum is decorated with carved work. The inner narthex, like the gynecaeum above it, is divided into three bays covered with cross-groined vaults, and communicates with the church, as usual, by three doors. Its walls seem to have been formerly revetted with marble. In the northern wall is a door, now closed, which gave access to a building beyond that side of the church. The exonarthex is also divided in three bays, separated by transverse arches, and communicates with the inner narthex by three doors and with the outer world by a single door situated in the central bay. That bay has a low dome without windows, while the lateral bays have groined vaults. Turkish repairs show in the pilasters and the pointed arches which support the original transverse arches. The doors throughout the building are framed in marble jambs and lintels, adorned in most cases with a running ornament and crosses. In the case of the doors of the exonarthex a red marble, brèche rouge, is employed, as in the exonarthex of the Pantokrator, another erection of the Comnenian period. On the exterior the building is much damaged, but nevertheless preserves traces of considerable elaboration. The walls are of brick, intermixed with courses of stone, and on the three sides of the central apse there are remains of patterned brickwork. On the buttresses to the southern wall are roundels with radiating voussoirs in stone and brick, and if one may judge from the fact that the string-course does not fit the face of the wall, parts of the exterior of the church were incrusted with marble. The round-headed windows of the dome cut into its cornice. Under the church is a cistern 355 which Bondelmontius deemed worthy of mention. 356 Until some twenty years ago extensive substructures were visible on the north-east of the church, affording homes for poor Greek families. 357 They were probably the foundations of the lofty monastery buildings whose windows commanded the magnificent view of the Golden Horn that doubtless suggested the epithet Pantepoptes, under which the Saviour was worshipped in this sanctuary.