New evidence for church decoration in the early ninth century

This paper examines church decoration during the Transitional period (ca. 650–850), focusing on unpublished elements from the Fatih Camii (Hagios Stephanos?) in Zeytinbağı (Trilye). Located on the south shore of the Sea of Marmara, the church may be securely dated to the early years of the ninth century. Among the variety of decorative details uncovered during an unauthorized restoration in 1997 are more than a dozen fragments of opus sectile. The variety of these panels expands the corpus of geometric patterns known from the period; in addition, opus sectile seems to have been used both on the walls and the floor.


Introduction
The ninth century, according to most scholars, was a period of cultural revival in Byzantium, but it continues to pose difficulties to art historians. In spite of the many references to renewal and reconstruction in ninth-century Constantinople, for example, none of the monuments discussed in the texts survives. 1 Without monuments, of course, it is difficult to write art history. Although few monuments survive from the so-called Transitional period (ca. 650-850), it was indeed a period of transition, both in architecture and in its decoration. In church design, there is a distinct shift from the grand, woodenroofed basilicas of the Early Christian period, to the small, domed, centrally-planned churches of the Middle Byzantine period. 2 In architectural decoration, figured, tesselated floor mosaics seem to have been abandoned in favor of non-figural marble panels and opus sectile, with figural decoration relegated to the walls and vaults. 3 As scholars have realized, Bithynia seems to have held an important role in this period, from which a variety of monuments survived well into the twentieth century. 4 For architectural developments, the Fatih Camii at Trilye (Zeytinbağı) is one of the best preserved examples, although its architectural form has received more attention than its decoration ( fig. 1). The purpose of this short note is to present some unpublished elements of the original architectural decoration of the Fatih Camii, and to offer a few suggestions as to what it might mean for our understanding of church decoration during the Transitional period and into the early Middle Byzantine period.
Picturesquely set at the center of the historic town of Trilye-Zeytinbağı on the south shore of the Sea of Marmara, the Fatih Camii, perhaps originally dedicated to Hagios Stephanos, has long been recognized as a significant early example of the cross-in-square church type. 5 It can now be safely placed in the early ninth century by dendrochronology, with a post quem date of 793 for the wood analyzed from the building. 6 This date accords well with the historical evidence, which suggests that the foundation is associated with St. Stephen the Confessor, who suffered persecution under Leo V. 7 The study by Sacit Pekak has clarified many of the architectural details and provided a good set of architectural drawings -on which my own are based ( fig. 2). 8 The naos is close to square in overall plan, with a dome just under 5 m (ca. 15 Byzantine feet) in diameter, raised on a tall drum above four columns. The crossarms are covered by barrel vaults. The corner compartments are somewhat uneven, isolated by projecting pilasters and covered by ovoid domical vaults. The pastophoria were quite large-the diakonikon is now missing-with their lateral walls projecting beyond the width of the naos. The bema has an extra bay before the apse, which was curved on the interior and polygonal on the exterior, opened by three windows. The pastophoria each included a setback before the apse, which was semicircular on both interior and exterior. To the west is a broad, barrel-vaulted narthex, preceded by a colonnaded portico.

Lateral arcades
Exposed remains of architectural sculpture and additional marbles littering the site suggest that the original building was lavishly outfitted. Much of the sculpture, including the capitals of the naos and closure panels, is reused from the fifth or sixth century, although some, including the capitals of the lateral arcades and some of the cornice patterns, may be ninth century ( fig. 3). The building was subjected to an unauthorized and undocumented restoration during the winter of 1995-1996. Since then, the interior has been repainted, and the exterior masonry dramatically repointed with raised pink mortar, with the bricks garishly painted red, obliterating much of the historical fabric.
The restoration of the mid-90s was also destructive, although it had the good fortune of revealing many original elements of the ninth-century building. For example, the rubble fill was removed from the lateral transept walls, exposing arcades in the north and south. In each, two column shafts with cubic capitals supported stilted arches ( fig. 4). The barrel vaults of the crossarms above, still blocked, were most likely filled by lunette windows. These provide early comparanda for several Constantinopolitan churches of the early Middle Byzantine period, such as the Atik Mustafa Paşa Camii (ninth century?) or the North Church of Constantine Lips (ca. 907). 9 At the Fatih Camii, the central intercolumniation is wider that the openings to either side, measuring 125 cm across on the south façade, compared to 85-87 cm for the lateral spaces.
The lateral spaces were closed by marble closure panels, 1.07 m high, topped by a marble coping and casement windows, all cut with curved surfaces to join the columns. The casements are divided into two lights and topped by a cavetto cornice. Both the closure panels and the capitals of the arcades are decorated on both surfaces, as described below. 10 The central opening on both the north and south facades is now filled with post-Byzantine masonry. Its original form is unclear, although it may have contained a door frame. The western portico once had a door frame in its central intercolumniation, as recorded by Hasluck, and doors set between columns are known from several monuments in Constantinople: St. John Stoudion, the Hagia Sophia Baptistery, and the Cho- 9 Ousterhout, Eastern medieval architecture, figs. 11.12b, 15.2. These are briefly noted by Sacit Pekak, Trilye, 133-135, but they were revealed only after the survey conducted for his 1991 dissertation. 10 V. Sacit Pekak, Trilye, 101-102. Similar forms of closure are likely in several early Middle Byzantine churches, but these do not survive. The decorative elements within each arcade are significant as well. The capitals are large and cubic in form, and they may be partially recarved sixth-century pieces. On the exterior, they are plain with a circular medallion set within a trapezoidal field (see fig. 5). On the interior, however, the capitals are lavishly covered with a shallow relief of intertwined circles containing a radiating leaf pattern (see fig. 4). The capitals have been cut down at the lower surface in order to fit the comparatively slim columns, and this suggests reuse. The patterns conform to those elsewhere in the building, and the recarving is most likely from the ninth-century construction period. 13 13 Sacit Pekak, Trilye, 101-102, identifies these as Middle Byzantine, while M. Dennert, Mittelbyzantinische Kapitelle. Studien zu Typologie und Chronologie, Bonn 1997, no. 170, suggests comparisons of All four of the closure slabs are decorated on both surfaces. Those on the south side have a lavishly molded diamond pattern with small leafy details on both surfaces. Those on the north side are relatively plain on the exterior, with a simple cross mounted on a globe, while the interior features diamond patterns similar to those on the south side ( fig. 6). There are similar panels elsewhere at the site -one loose, another incorporated into the eleventh and twelfth centuries, and on this basis proposes a later date for the church. the fountain west of the church, all spolia from the late fifth or sixth century. 14 The diamond pattern is relatively common on closure panels, and a number of these have been documented elsewhere in Bithynia. 15 The quantity of early spolia at the Fatih Camii led Pekak to the intriguing but unfounded suggestion that the present building represents a reconstruction of a sixth-century 14  church. 16 Above each panel, the window sill is also reused, plain on the exterior but decorated on the interior with a pattern of alternating rings and leaves surrounding a bullnose molding. 17 In sum, the details from the north and south crossarms are significant, as they are preserved in greater de-16 Sacit Pekak, Trilye, 147-164, suggests the lower lateral walls and lower main apse walls predate the ninth century; note in particular his figs. 25-26. 17 Cf. ibid., no. 41b.

Opus sectile
The interior of the Fatih Camii was originally decorated with mosaics, the presence of which was noted during the period of Greek occupation in 1920-1922, when the building was briefly reconverted to a church. Tryph-on Evangelides noted at that time, "after the whitewash had been scraped off the walls, there appeared wonderful mosaics, which I deeply regret I did not photograph for lack of film". 18 Unfortunately he is no more specific about the mosaics, so that we know neither where exactly they were located, nor what they represented, nor what kind of mosaics they were. Were they tessellated or opus sectile?  Today, opus sectile in a simple grid of oversized tesserae of black, white, and, yellow stone is visible in the reveals of the south arcade and east windows (see fig. 4). 19 The crustae are relatively large, measuring between 2.7 and 2.9 cm square. To my knowledge, the appearance of opus sectile within the reveals of arches is exceptionally rare within the context of Byzantine architecture, although something similar is known from Hagios Demetrios in Thessalonike, probably from the seventh century. 20 In addition to the crustae in the window reveals of the Fatih Camii, a small area of floor mosaic survives in the prothesis apse composed of hexagons and triangles of white and black stone ( fig. 7).
During the winter of 1996-1997, the restoration exposed quantities of opus sectile, primarily from the floor. Regrettably these were removed and discarded without documentation. This is all the more unfortunate because of the dearth of securely dated opus sectile from the period of the seventh through tenth centuries, as will be discussed below. A dozen fragments were salvaged from the Fatih Camii, which I was able to study, and during a cursory visit to the site in 1997, I noted several additional fragments.
Among the fragments I studied, none bore evidence for a covering of plaster or paint, suggesting the majority of them came from the floor. Many, however, had a splat- tering of cement and latex paint, apparently from the restoration. Most were set into a thick bedding of pink mortar, sometimes up to 9-12 cm thick. Some preserved sand or small stones or irregular impressions on the reverse side that suggest they came from the floor. One fragment incorporates glass tesserae, which would have been impractical -although not unknown -in a floor covering. In addition, a few fragments bear patterns similar to those in the window reveals, so some of the pieces discussed here may have come from the superstructure.
Most surprising is the variety in the patterns of the opus sectile. Several fragments preserve clear edges, and at least some of these patterns must have formed decorative bands between large panels of marble, as was common in the opus sectile of later centuries. In addition, several hexagonal crustae are large, measuring ca. 10 by 20 cm, and a floor fragment noted at the site indicates they were enveloped by patterning composed of smaller stones.
What follows is a descriptive catalogue, followed by a brief discussion of the significance of this material.
[1] A rectangular fragment, measuring 12 by 8.5 cm, with mortar 3-5 cm thick with exposed aggregate on the lower surface ( fig. 8). The pattern is very fine, consisting of squares of a hard, yellow limestone, measuring 3 cm across, with an inset pattern of rotated white tesserae, now quite corroded. These are surrounded by triangles primarily of green porphyry, but they also include red stone, porphyry, and blue glass. A similar border 1.5 cm wide, of rotated squares and triangles. Considering the delicacy of this fragment, it is difficult to imagine it as part of the floor decoration.
[2] A triangular fragment, measuring 17 by 14 cm, with the mortar 5-6 cm thick, with a flat lower surface  . 9). An original edge is preserved on the long side, with the impressions of two larger panels and the mortar joint between them. The pattern was set diagonally to the larger panels. It consisted of large black squares, measuring 2.5 cm square, framed by white bands, ca. 1 cm thick, with small red squares at the corners. All the stone has flat color; I assume none is marble.
[3] An irregular fragment, measuring 11.5 by 9 cm, with the mortar 5 cm thick, with a flattish lower surface ( fig. 10). The pattern is composed of black hexagons, 3 cm wide, framed by white triangles. These join a black diamond, 5 cm wide, joined by a yellow ochre triangle. All are flat in color, presumably with no marble. Not enough of the pattern is preserved to determine how it was developed over the field -perhaps the joint between two dissimilar patterns.
[4] An irregular fragment, measuring 19.5 by 14 cm, with the mortar 6.5 cm thick, with exposed aggregate on the lower surface ( fig. 11). The pattern consists of radiating bands arranged in a complicated repeat pattern. Every other band alternates green porphyry (or occasionally green Breccia) with red or striated purple marble. The other bands alternate yellow limestone, green porphyry, and a corroded white stone. The two different alternations come together to form crosses of green, with red at the center and alternating white and yellow at the corners. The radiating pattern suggests this might have been the frame of an omphalion, perhaps with a rota at its center, set beneath the dome or within the apse.
[5] Another irregular fragment, measuring 32 by 18.5 cm, with the mortar 5-9 cm thick, with irregular imprints on the lower surface ( fig. 12). This fragment has a notable convex curve, with a recession of ca. 1 cm over its length. Either it came from an irregular area of the floor or perhaps from the inner surface of the apse. The pattern consists of regular rows of circles, 3.4 cm in diameter, divided into halves, with one half black and the other half alternating between white and yellow ochre. Small red squares are set diagonally between the circles.
[6] An irregular squarish fragment measuring 26.5 by 26 cm, with the mortar 9 cm thick, with irregular imprints of brick or stone on the rear surface ( fig. 13). The pattern is a basket-weave composed of rectangular black bands, ca. 4.5 by 1.4 cm each, framing white or ochre squares, measuring 1.5 by 1.5 cm each, organized in groups of four. There is a flat edge to one side. The patter is similar to that appearing in the south window reveals, and thus could have come from one of the windows. The imprints on the rear surface may encourage this identification. [7] The trapezoidal fragment measures 26 by 17 cm, with the mortar 9 cm thick, with irregular imprints of brick or stone on the reverse (fig. 14). The pattern is identical to no. 6, although the two fragments do not join.
[8 & 9] Two fragments that may be joined together; individually they measure 15 by 14.5 cm, and 24.5 by 14.5 cm; together they measure 37 by 14.5 cm ( fig. 15). The mortar is ca. 9 cm thick. The pattern is rather complex, with a grid formed by yellow limestone bands, measuring 4.0-4.2 by 1.0-1.3 cm, with small green porphyry squares at the intersections. Within the grid are two different patterns. The first has a rotated square of green Breccia, ca. 2.8-3.0 cm on each side, the corners filled by triangles of porphyry or red stone framing a central triangle of white limestone. In alternating fields, the rotated square is of porphyry or of red marble, and the surrounding triangles are of green porphyry or Breccia, framing a triangle of white limestone. There is a distinct edge (on the bottom in the image), indicating this was a band framing a large panel of marble.
[10, 11 & 12] Three hexagonal marble crustae, all measuring 10-10.5 cm by 20-20.5 cm, with a covering of white mortar on the rear surface ( fig. 16). One is a white marble with purple veins, ca. 2 cm thick; the second of Proconessian marble, 2.5 cm thick; the third of a crystalline white marble, broken into two pieces, 3 cm thick.
In addition to these fragments, I also noted several loose pieces littered around the site.
[13] A roundel of grey-white marble, ca. 20 cm in diameter and ca. 10 cm thick, with an irregular lower surface ( fig. 17). Could this have been the center to an omphalion (see no. 4)?
[14] A fragment of a checkerboard pattern, with a slight suggestion of radiating pattern, similar to no. 4, measuring 13 by 9.5 cm ( fig. 17, lower left).
[15] A fragment measuring 27 by 18 cm (fig. 17, lower right). It includes the settings for two hexagonal crustae, like nos. 10 to 12 above. Between these are patterns of black and white triangles.
[16] A fragment measuring 16 by 25 cm, with traces of a circular pattern matching that of no. 5 above, but with almost all the stone lost ( fig. 17, upper right). Traces of the setting bed are preserved.

Commentary
The unique nature of the Fatih Camii opus sectile fragments requires some additional comments. Opus sectile was one of the most lavish treatments used in ancient and Byzantine interior decoration. 21 The continuation of the ancient tradition of opus sectile has been the subject of important articles by Guiglia Guidobaldi,Peschlow,and Pedone. 22   ing decorative patterns or figures depended on the availability of marble and other luxury stones, most of which were not quarried after the seventh century. In the surviving Middle Byzantine examples, one normally finds large panels framed by decorative bands, all composed of reused materials. The large panels could be rectangular or circular, or in more complex arrangements. 23 The pieces from the Fatih Camii would seemed to have conformed to this general pattern. Although the large panels have disappeared, the hard edges to several of the fragments suggest they framed panels; moreover, the fragments of radiating patterns and the single surviving disk suggest that omphaloi were also included in the floor decoration. The substitutions and mixtures of different types of stone, however, and the occasional lack of precision distinguished the examples from Trilye. These are the product of a workshop with limited resources, perhaps imitating the style of the capital.
As with so many other arts, the period of Iconoclasm seems to have instigated a transition in the development of Byzantine floor mosaics. Figural tessellated floors are rare after this period, and opus sectile increases in popularity -a medium which could communicate meaning not through images or narratives but through the luxury of its materials. As Henry Maguire has em- phasized, after Iconoclasm, figural pavements would have competed for prominence with the wall and vault decorations in a small building. 24 The fact that opus sectile could be easily created from spolia may be significant in a period when colored stones were no longer readily available -not to mention a period in which figural imagery was banned from churches. But, again, as Maguire emphasizes, the shift from tesselated floors to opus sectile must have been result of a conscious, aesthetic choice, rather than an ideological one, or a practical one -that is, it did not come about because of a lack of skilled artisans or materials. 25 In his study of the development of opus sectile, Peschlow emphasizes the importance of Bithynia, and more generally, the hinterland of the capital. Indeed, the closest comparisons for the Fatih Camii are close at hand, for example at Imralı Adası, perhaps from the 780s; at Değirmenaltı by Tuzla, and at Yakacık on the Gulf of Izmit, both generally dated to the ninth or tenth century. 26 The monastery at Kurtköy near Pendik belongs to this group but is not securely dated. 27 Other examples have been noted along the south shore of the Sea of Marmara, most of which no longer survive, as at the Pantobasilissa at Trilye, the Medikion monastery, St. 24 Ibid., Ibid. 26 31 The omphalos on the floor of Hagia Sophia has also been assigned to the restoration of Basil I. 32 The lovely fragment from a chapel at Hagia Euphemia may be assigned a post quem date of seventh-century. 33 Alas, all of the Constantinopolitan floor lack secure dating.
Although the overall organization of the interior decoration at Trilye is unknown, several of the patterns and systems of organizations are unique -or at least unusual. For example, the use of large hexagonal crustae is fairly common both earlier and later, but normally they are used to form a rectilinear grid. At the Fatih Camii, the pattern is more complex, apparently with some of the hexagonal crustae set on the diagonal. In addition, designs common in later centuries, such as a matrix of interlocking circles and squares, are absent. Similar patterns appear at Nicaea, but many resist comparison. 34 For example, I have found nothing in opus sectile to compare to 30 S. Y. Ötüken, 1993Yılı Demre, Aziz Nikolaos Kilisesi Kazısı, in: XV. Kazı Sonuçları Toplantısı II, ed. İ. Eroğlu, F. Bayram et al., Ankara 1995O. Feld, Die Innenausstattung der Nikolaoskirche in Myra, in: Myra. Eine lykische Metropole in antiker und byzantinischer Zeit, ed. J. Borchhardt, Berlin 1975, 394-397. 31  the halved circles, the basket-weave pattern, or the filled squares.
In sum, the new information from the Fatih Camii opus sectile decoration may help us to form a link, now however incomplete, between the Late Antique tradition and its Middle Byzantine successors. Perhaps most importantly, the Fatih Camii begins to provide us with a concrete example of the richness of architectural decoration achieved in the early ninth century.

Architectural sculpture
For the sake of completeness, I include examples of the disjecta membra that were salvaged following the restoration. Many correspond to elements still in situ at the Fatih Camii.
[1] A cornice fragment of crystalline white marble, 21 cm long by 10.5 cm deep by 7 cm high ( fig. 18). The exposed surface has a 2-cm fillet and a 7-cm chamfered decorated surface. The low relief pattern alternates an arrow and a double wicket. This corresponds to the upper naos cornice in the Fatih Camii and must have come from there. 35 [2] Fragment of a white marble cornice broken from a corner, 14.5 by 16.4 by 11 cm tall ( fig. 19). This has a 2.7cm fillet at the top, with an 8-cm decorated chamfered field, then a 2.4-cm setback and a 1.5-cm lower fillet. The pattern has five-lobed leaves, joined by a stem, with alternating leaves inverted. This matches the lower naos cornice. 36 [3] Fragment of an attenuated marble capital from a pilaster or mullion, 24 cm high, with maximum widths 35 Cf. Sacit Pekak,Trilye,104. 36 Oddly omitted from ibid., but visible above the capitals of the lateral arcades, 101.  (fig. 20). The decoration extends over two surfaces of the piece, with bulbous necking rings at the based a lily-like motif spreading above it. Nothing similar is found in or around the building. It may have come from the templon.
[4] Fragment of a marble panel, 21 by 17 by 6.2 m thick, with a 2.5-cm band at the bottom ( fig. 21). The pattern is composed of a three-branched-leaf set within a stem that loops around it. Each measures 4.5 cm across and is cut to a depth of ca. 1 cm. These are organized into rows with the pattern inverted in alternating rows. There is no corresponding motif in situ, however the pattern fits well with the decorative repertory of the building. The rear surface was left rough. An identical fragment was built into the wall of the Greek School at Trilye.

Commentary
Most of these fragments should date from the time of the construction of the church -that is, early ninth century. The crispness and exactitude of the lower cornice corresponds with that observed at the North Church of Constantine Lips in Constantinople (ca. 907). 37 The only exception may be the upper cornice, for which Pekak tentatively proposes an early Byzantine date. 38 The suggestion that these pieces were spoliated is encouraged by the 37 C. Mango awkward cuttings of the pattern at the corners of those pieces in situ.

Conclusions
With its date now secure and with new evidence of its interior decoration, the Fatih Camii offers an important example of the transition from Late Antiquity to the Middle Byzantine period. With the exception of Hagia Sophia, its casement windows may be unique in their survival, and they offer a model of closure for the less well-preserved churches of Constantinople. The architectural sculpture -a mix of reused pieces and newly carved ones -accords with what we find in the capital as well. Finally, the fragments of opus sectile offer new and necessary information for the transformation of church interior decoration between the sixth and the tenth centuries.