Report 7: The Temple of Artemis at Sardis (2020), Fikret Yegül

Chapter 4. Architectural Analysis, Design, and Comparisons

Introduction

The design of the Temple of Artemis at Sardis is unusual because it represents a mixture of different planning approaches applied over a considerable period of time. In many aspects, it is experimental. Unlike many temples, even those that took a long time to complete or were ultimately left unfinished, the Temple of Artemis does not represent a scheme that was conceived at one instance and realized in orderly phases. One could argue in the broadest, most traditional terms that there was an original Hellenistic phase and a Roman Imperial phase, and probably many smaller stages throughout the building process. The first major phase (here referred to as such because it was the first significant, identifiable stage in the building) was complete, in a sense, since the cella must have been ready and functional by the second half of the third century BC. It is exceptional, however, because the original design was finished only as far as the cella, including the in antis columns of the western front and the eastern back porches; none of the exterior mantle of columns, which gives visual substance and style to a Greek temple, was put in place during the Hellenistic phase. No building could ever be conceived without an end design in mind, and in our case we are confident that it would have been a dipteros, much like the early Classical Temple of Artemis at Ephesus and the nearly contemporary Temple of Apollo at Didyma.

It may be hard to imagine that a building so grandly conceived and beautifully executed was only partially realized. It remained looking unfinished and strange, we think, for some four hundred years before construction was taken up again in earnest, only to look even stranger and more unfinished centuries later when finally, by the middle of the fourth century AD, its use as a pagan temple was gradually terminated, transformed in substance and spirit with the addition of a Christian chapel at its southeast corner. However, this is what the facts on the ground indicate. Yet colossal temples, like colossal medieval cathedrals, were rarely finished, or they at least took a very long time to complete. This seemingly unending process of construction was actually a source of pride for the community, who often could not come up with the funds to finish the work. Still, continuing work for all to see was an affirmation of collective faith and determination. A temple finished honored the past; a temple under construction honored past and future.

When the building of our temple resumed during the High Roman Empire as a grand pseudodipteros, it was practically a new temple for a new world, with new needs and expectations. Yet, the centuries separating the original building from the finally realized monument did not entirely erase the intentions of the original builders. The core design of the Hellenistic monument was what shaped and inspired the Roman one. What was obvious to see—in ornament, detail, and style—was admired and even imitated. The memory and the intentions of the original builders were probably never fully lost, and they played a significant, even restricting role in shaping the final design and its execution. Technically speaking, the resulting scheme was an anomaly among the pseudodipteroi of Asia Minor: a creative synthesis of Eastern and Western influences in temple design.

The following is a description of the Hellenistic and Roman phases of the temple as conceived, planned, and executed, and an analysis of its design in the larger context of temple architecture in Asia Minor and the greater Roman world. Much remains unknown and unknowable about many aspects of this design and the construction “phases” of the temple; therefore, some of these will be presented as possibilities, probabilities, and hypotheses. There are strong alternate views, and as before, these alternate views, interpretations of archaeological data, and varying hypotheses regarding reconstruction and dating will be respectfully represented.

Hellenistic Phase Design: The Shape of the Sacred

The placement of the Hellenistic temple might have been conditioned and restricted by two venerable architectural elements on site that predated it: a tufa altar (LA 1) dating to the Lydian period (Plan 2, Plan 6); and a sandstone foundation (“basis”) roughly 58–60 m east of the altar, ca. 6 × 6 m in size, centrally placed at a higher ground level inside the original cella (Fig. 3.5). Butler recognized this platform as an early feature, and this was confirmed by Cahill on architectural, ceramic, and numismatic grounds (see pp. 163-164).1 The present location of the temple, as well as its size and footprint (as either a dipteros or a pseudodipteros), might have been largely guided by the desire to retain these structures; given that the “basis” was placed in the center of the original cella to support the cult image, the present layout represents the largest (i.e., longest) temple that could be achieved around two existing stationary points. As Cahill explained, “Had they moved the temple farther away from the altar, the basis would be been left awkwardly at the front of the cella.”2 The solution brought a certain degree of awkwardness; once the temple’s planned peristyle design was put in place, there would be no room left between altar LA 1 and the front of the cella to place the stairs, which were necessary to negotiate the considerable level difference between the two structures. The western front row of columns (of which only the foundations of the southwest corner column 64 was put in place) was virtually jammed up against the altar; any potential design solution, Hellenistic or Roman, would need to incorporate the altar (LA 1 or LA 2) into the body of the stairs, as Butler attempted in a somewhat fanciful reconstruction (see Fig. 3.1 and pp. 153, note 8). Although such a tight relation might seem awkward, let us be mindful that it was not an uncommon arrangement between temples and their altars during the Greek and Roman periods.3

With its cella measuring 23 × 67.50 m, the Temple of Artemis at Sardis (Fig. 4.1: Hellenistic phase) belongs to the category of very large Ionic dipteroi of Asia Minor, such as the roughly contemporary second Artemision of Ephesus (ca. 334 BC), the Temple of Apollo at Didyma (ca. 300 BC), and the fifth-century BC (“fourth period”) Temple of Hera at Samos (Fig. 4.2); all except the Sardis temple were dipteroi, and continuations of earlier, Archaic predecessors. The cella of the Sardis temple is distinctive not only because of its exceptional size, but also for its unusually elongated proportions. Its length is roughly three times its width, a proportion strikingly similar to the temple group mentioned above (Sardis 1 : 2.94; Didyma 1 : 2.99; Samos 1 : 3.0; Ephesus 1 : 3.10). Such a pronounced elongation of the cella is generally considered an Archaic characteristic. Since the design and proportions of these famous Ionic dipteroi were closely derived from their equally famous Archaic predecessors, the Sardian Artemision could also be linked to those distant-plan prototypes, even though it lacks an earlier predecessor, unlike its Ionian cohort.

The Sardis cella is comparable to those of Samos, Ephesus, and Didyma not only because of its overall proportions but also in its design. The cellas of these temples all display frontal emphasis in their deep pronaoses filled with columns between antae; at Samos there are ten columns in distyle pairs, at Ephesus eight, at Sardis six, and at Didyma an impressive hypostyle of twelve in three tetrastyle rows! Other significant similarities among columnar dimensions and proportions that link the Sardian Artemision to the Ionian examples emerged when the Sardis temple received its peristyle during the Roman Imperial period. The lower diameters of their average column shafts are the same or very close (Sardis 1.99 m and 1.86 m; Didyma 2.01 m; Samos 1.90 m), as are the lower diameter-to-height proportion of their columns (Sardis 1 : 8.91, 1 : 9.61, and 1 : 9.90; Ephesus 1 : 9.6; Didyma 1 : 9.9). Although the column sizes of these three temples are quite close, it is noteworthy that there is considerable difference in their axial spacing. For such a large temple, the Sardian Artemision displays an uncommonly tight, uniform axial spacing along its long sides (taking the columns of the long sides: Sardis 4.99 m; Ephesus 5.73 m; Didyma 5.30 m).

The cella design of the Hellenistic Artemision of Sardis displays a particular affinity to the late Classical Artemision of Ephesus: it follows the same cella arrangement, with a square pronaos and an opisthodomos half as deep as the pronaos. Considering the demonstrable cultural and religious origins of the cult of Artemis at Sardis with that of Ephesus, a corresponding architectural affinity is logical (see pp. 10-11 and pp. 205-6). Merely a generation or so earlier than the Sardis and Didyma temples, we see one of the earliest uses of the opisthodomos in the Ionic order in Pytheos’s Temple of Athena at Priene, finished by 334 BC (Fig. 4.2). This famous temple also boasts a highly ordered, modular application of the same organizational principle of the cella: a square pronaos and an opisthodomos half as deep as the pronaos.4 The immense interior spans of the Temple of Apollo at Didyma and probably also the Ephesian Artemision precluded a roof over their cellas, which resulted in hypaethral adytons (in the former it also served to house the sacred grove of Apollo). At Sardis, however, the use of the double row of internal columns allowed for a more conventional roofed cella. The large numbers of marble roof tiles discovered by the Butler expedition (and many fragments by us) assures us that there was a complete roof over the cella.

An important aspect of the cella interior at Sardis is the exceptionally wide central span between the columns in the north–south direction (restored at ca. 9.30 m on axis with a ca. 6.70 m clear span, a distance that could barely be covered by stone, but there were probably wooden trusses, or doubled beams (or a combination), while the spans of the columns in the lateral (east–west) direction (restored ca. 5.40 m on axis and ca. 2.80–2.90 m clear) would have been suitable for stone architraves (as shown in Fig. 4.14). The distance between the east–west colonnades and the long side walls of the cella (ca. 3.30–3.35 m clear, north–south), although appropriate and normal for stone architraves to span (but which then would have required a triple-mitered joint with the east–west architraves of the colonnades), might simply have been spanned in wood. The exceptional width of the cella nave, convenient for this cella’s later, Roman use as a cult chamber housing colossal images, must have been necessary to accommodate the existing base (“basis”); otherwise, the more closely spaced pronaos colonnade (restored at ca. 8.40 m center to center, 5.82 m clear) and the cella colonnades might have been aligned. In any case, this arrangement emphasized the spatial impact of the nave in a way that is rare in Greek temple architecture. It recalls innovative exceptions, such as the wide cella of the Parthenon in Athens.

Hypothesis for a Dipteros: Archaic and Classical Models

Since the first phase of the Temple of Artemis did not go beyond the elongated cella and the pronaos porch columns in antis, strictly speaking, we do not know what the final design would have been. However, in light of the massive size, the unusual proportions and fine workmanship of the finished cella, and the close similarities with some of the nearly contemporary Ionic temples of Asia (especially the Second Artemision at Ephesus and the Temple of Apollo at Didyma), it takes little imagination to say that the Sardian Artemision must have been planned as a dipteros, the going mode of that time and place. Both of these great and prestigious temples must have still been under construction when ground was broken for the Sardis temple.5 The significance of the favorable political climate following the establishment of Seleucid rule in western Asia Minor has already been underscored, as kings and queens strove to win the approval of their newly conquered subjects by patronizing local cults and temples. Thus, we believe that the original intention at Sardis was to build a dipteros following the general mode and typology of the giant Ionic dipteroi and, like them, harken back to their venerable Archaic predecessors. Simply put, an Archaic-style dipteros was the natural choice for very large, very ambitious temples. Like many of the great Ionian temples, the Sardis temple also must have been intended to be raised on a platform (or given a platform effect by artificially lowering the ground around it) surrounded by steps.6 Perhaps only the earth embankment, and not the defining outer walls or steps of such a platform, was created. The religious, programmatic, and architectural relationship between the Temple of Artemis at Ephesus and the temple at Sardis promises to reward future study.7

Noting the 15.60 m distance left between the ends of the west antae and the altar, one imagines that the west end of the hypothetical dipteros allowed just enough space for three rows of columns while the back east end probably had two rows, again like the contemporary Didymaion and Ephesian examples (in the latter, this distance is exactly 15.70 m). The nature of the original topography remains uncertain, especially the level of the land east of the temple toward the Acropolis, and the changing ground levels between the altar and the cella are critical. This relationship, more or less preserved between the ground on the west front of the temple and the cella interior floor (roughly 4.80 m difference; the difference between the ground and the west pronaos porch is 3.20 m), is therefore a rare physical reference point for the placement of the temple in its natural topographic context. The land must have been rising in a general way from low on the west to high on the east, from the banks of Pactolus to the foothills of the Acropolis, with a somewhat steeper rise at the east end of the temple. However, the steep hill close to the east end (the scarp rises about 15–16 m at a distance of 50–60 m from temple east colonnade) was cut down by Butler in a series of four terraces in order to unearth the temple (Fig. 4.3).8 The entire low-earth terrace upon which the temple naturally sits appears to have been created by successive landslides from the Acropolis heights over time, starting from the pre-occupation period. Geological studies and soil tests undertaken by successive geographers—starting with W. Warfield, geographer of the Princeton Expedition, then G. W. Olson in 1970, and B. Marsh in 2007—all indicate repetitive flooding and silting in the area, and under the present temple by gravel and silt sliding down from the Acropolis.9 Marsh and Olson concluded that the landscape around the temple was largely built by gravel, silt, and clay deposits that washed down from higher ground over time as seasonal rains and changing streambeds mobilized “debris flows” until the temple was entirely buried (mainly by the ninth century AD).10 As pointed out by Cahill, in all deep sondages at or around the temple, “the earliest known stratum is the massive deposit of natural, water-laid gravel and sand under the temple and altar, apparently washed down from the acropolis… [reaching] bedrock 5–6 m below the level of the temple floor, with no intervening cultural horizons… there is nothing… later than the later sixth or earlier fifth BCE [when the sandstone basis was built on top of this gravel].”11

Although the temple never received its peripteral columns during its original phase, its west-facing cella must have been sufficiently complete by the middle of the third century BC to make it a facade for a functioning architectural shrine. The image base was completed, presumably with the image of the Artemis of Sardis installed on it, with a terminus ante quem of ca. 240–220 BC (see pp. 163-164). Rows of columns inside the cella and more between the front and back anta walls must have carried a combination of stone and wood architraves and the wooden trusses of the roof. The roof must have been in place, as it is difficult to conceive of a functioning temple without one, and this is also attested by quantities of handsome marble roof tiles found in both Butler’s and our excavations (see Fig. 1.14). A recently found inscribed roof tile of the temple (Sardis inv. S10.14 = IN10.4) was read by G. Petzl as “φυ(λης) Και(σαρείου?)” and identified as a dedication from Sardian tribes with “Sullan” and “Caesarean” names. Based on this evidence alone it may be suggested that the “name and tribe were adopted in recognition of Tiberius’s support after the earthquake of 17 CE.”12 This is supporting but not absolute evidence for a complete overhaul of the building’s marble tiles and its whole roof, or an argument for two distinct phases—Hellenistic and Julio-Claudian—for the entire building (see pp. 184-189). Rather, it indicates a major repair and overhaul of the damaged original marble roof—perhaps the most vulnerable structural feature of a temple without peripheral columns (modern buildings have their roof tiles repaired or overhauled every twenty to thirty years even without the catastrophe of an earthquake). Indeed it would be reasonable to expect that the earthquake of AD 17 had caused significant damage to the roof tiles. Consequently, during a major roof repair some tiles were newly supplied, others newly “inscribed” on this occasion.

Dedications and Votive Monuments

Large numbers of marble stelae, inscriptions, dedications, religious offerings, statuary, decrees, and public monuments dating from the Lydian-Persian period to the third century AD have been found inside and around the temple, or in the sanctuary before the establishment of the temple. The majority, however, belong to the period of the temple. Among the roughly one hundred public, honorific, and votive inscriptions found and published by the Butler expedition, some twenty-six come from the temple or are associated with it.13

From an eye-catching monument of the Imperial era incorporating Archaic lions and eagles on a statue base of ca. 550–520 BC dedicated to Artemis in Greek and Lydian by one Nannas14 to long lists of individual families whose names are recorded as trusted keepers or administrators of the temple through the third and second centuries BC;15 from high officials and leading citizens of Sardis, like Menogenes, whose generosity and services to his homeland during the reign of Augustus were publicly named and acknowledged, to one Polybius, an aspiring Greek rhetor of the second century AD, who was moved to honor Cicero, his distant muse in Rome, by erecting a bust of the great Latin orator in the precinct;16 from a royal dedication to Artemis by the Attalid king Eumenes of Pergamon (probably Eumenes II, ca. 190 BC), to another great ruler who was honored some 350 years later, the emperor Antoninus Pius, “on account of his benevolence” to the city;17 from a curious stone ball dedicated ca. 300 BC to the goddess by a very young noble girl who would be queen, to many curious balls with similar dedications to Artemis by generations of girls who were her priestesses through the second and first centuries BC;18 from dedications honoring the priestesses of Artemis on marble blocks or slabs which might have been structural19 to one special block that supported the statue of one priestess who gave voice to her timeless plea: “O, Artemis, ever preserve Sardis in concord by the prayers of Moschine, the daughter of Diphilos!”20—the sanctuary and Temple of Artemis, from the Hellenistic era into the empire, were the focus of Sardians’ devotions and prayers, and the depository of their gifts. Although firmly anchored as the center of the cult of goddess Artemis, the city’s heterogeneous cultural past and its multilayered sense of religious syncretism gave a deeper, more tolerant sense of sanctity to the land subsumed under the goddess’s broad and inclusive spiritual mantle. That this center of devotion and possibly political and economic power extended into the middle of the second century AD and included the worship of Roman emperors and empresses—and that Artemis’s old temple now received their divine images (with enthusiasm or willy-nilly)—is simply the historical confirmation that royal Sardis, city of the Panhellenion and twice neokoros, upheld its proud place among the constellation of Asian cities subsumed by the empire.

Honor and devotion to the goddess continued unabated through the centuries. Still, the physical presence on the ground, the architectural focus of all that devotion, was less than perfect. If we were to encounter the Sanctuary of Artemis at Sardis in the late Hellenistic or early Imperial era, we could be excused for being startled by the bizarre and uncommon appearance of Artemis’s temple: an oddly proportioned, long, tall marble box with shining white walls accented by discrete, exquisite details, covered by a simple roof of marble tiles with gabled ends—rather like a toy house magically transformed into a magnificent marble palace. This white marble box rose on a gentle earth embankment—its hard-edged form in contrast with the natural, lovely roll of the hilly countryside around and the spiky mountains behind—in repose between the banks of the Pactolus and the craggy Acropolis peak that linked the city to the legendary range of the Tmolos Mountains beyond. Beyond the valley of the Pactolus, the land opened toward the north to the wide Hermus plain with its “thousand mounds” of the Lydian kings and nobles, and the famed golden city of Croesus. Around the box of the temple, softening its stark geometry, there must have been trees and bushes and many small votive monuments, inscriptions, stelae, and statues. Sardians must have known and appreciated that their temple was incomplete without its external columns, although, in the words of a reviewer of this manuscript they might have been “less bothered by the absence of outer columns than an Ionian city would have been. The long hall [with nineteen-meter-high walls] would have provided them with an impressive setting for Artemis.”21 To the followers of the goddess this slightly bizarre and unusual shape set among the hills must have become, over time, acceptable and cherished—indeed, the shape of the sacred. In that sense, there was nothing unusual, nothing bizarre.

The West End and the Northwest Stairs

Today it is easy to be baffled by the clutter of structures occupying the area from LA 1/LA 2 to the west pronaos porch, whose physical, functional, and chronological relationships with each other are confusing and even contradictory (Plan 3). The main question is the nature of the physical connection between the temple and the Archaic and Hellenistic altars (LA 1 and LA 2, respectively) during different periods in their histories and how they relate to the northwest stairs that extend between the northwest anta pier and altar LA 2, some 1.20–1.30 m beyond the outer face of the west peristyle. We will summarize what we know of the original topography of the temple’s west end and the primary level changes and other alterations brought about by Butler’s excavations.22 In the following discussion of the west end I mainly follow N. Cahill’s archaeological interpretation and overall reconstruction, with a few alternative ideas.23 I present this discussion through the Hellenistic and Roman periods in order to retain a more unified picture of this area over time.

The temple cella was built over the course of the third century BC atop a relatively low platform of earth that surrounded the cella, so the relationship between the early altar and the temple’s west end must have been a natural one; the nonexistent peristyle of columns was probably evoked by the gentle slope of the earth platform surrounding the cella, made either by building the ground up, digging it down, or both. The drop in the land to the north and south of the temple, which today looks like a trough, was created by Butler’s deep trenching and does not represent the original topography, which we simply do not know. What is certain is that there should have been some form of a natural pteroma (a sort of earthen walkway whose width at the same level of the euthynteria—i.e., at *100.0—we do not know) along the sides, because the deep foundations of these walls had to be buried. Located on the west end was the Archaic altar (LA 1, ca. 550–500 BC), which almost doubled in size with the Hellenistic era altar (LA 2), approached by wide steps on the west side. By the time work on LA 2 began (ca. 280 BC, coeval with the temple or a little earlier; we have no secure date for it), the ground level in front of it its west-facing steps was at ca. *96.80, ca. 1.40–1.50 m lower than its top, preserved at ca.*98.30.

Thus, the original ground level in front of LA 1/LA 2 was some 3.20 m lower than the platform upon which the temple was built, i.e., the pronaos and pteroma set at *100.0. This high level was probably in response to the natural rise of the land toward the east. The level of the cella proper was still another 1.60–1.70 m higher than its pronaos (ca. *101.70), making it nearly 4.80–4.90 m higher than the original ground in front of the altar (Plan 3). This unusually large difference in level between the cella interior and the natural ground in front of the altar may have been partially dictated by the pre-temple “basis” located inside the Hellenistic cella (or subsumed by the cella)—as well as other architectural and topographical factors that are poorly understood today—and a desire to present the temple in a higher position, at a topographical advantage to its immediate surroundings. Although the total leveling of the temple platform or pteroma at *100.0, for a rise of ca. 3.20–3.30 m over a length of ca. 85 m, was mild if not negligible (roughly resulting in a ca. 4 percent slope) as one approached the temple from the west, this level difference had to be negotiated by stairs and/or ramps. Furthermore, this height difference would have to be negotiated within the restricted space between the back of altar LA 2 and the west face of the new temple along its northwest and southwest antae, a distance of ca. 16.60 m.

The obvious solution would have been to build a broad, monumental flight of stairs across the entire west front of the cella porch, rising the full 3.20 m from the base level of the altar at * 96.80 to the west pronaos porch at *100.0 (Fig. 4.4, proposal 1). This would have been a massive construction with about fourteen risers (risers at ca. 23 cm and treads at ca. 36–36.5 cm) and extending out beyond the temple’s west face some 4.70–4.80 m. The north pteroma, however rudimentary and berm-like, was at *100.0 (since the foundations of the wall would have to be properly buried in the ground at the very least), making the connection between the north pteroma and the west end of the temple, with this 3.20 m level difference, difficult and awkward, especially where they meet at the northwest and possibly the southwest corners. A straight north–south wall (essentially a retaining wall) or a steep berm to demarcate the west end of the north pteroma could be thought of as the obvious and cleanest solution. On the face of it, this arrangement presents two problems: first, the sheer monumentality and abruptness of the stairs in such a confined, low space, would be visually jarring; second, at the northwest corner of the anta (where the north and west faces of the anta pier meet) the foundations of the anta and of the north wall would have been exposed, with blocks hanging some 1.60–1.70 m over the level of the earth. This could have been revetted neatly in marble, but that is not a structural solution. More importantly, such an arrangement lacks precedent in Greek and Hellenistic temple architecture. Frontal schemes in Roman podium temples are common enough, but that phase of our temple was not Roman.24 Yet the unorthodox design, elsewhere argued as the reason for other creative aspects of our temple, may well provide the answer for this unusual but impressive arrangement.

An alternative proposal would be to divide the difference in height in half by partially filling and raising the space between the altar and the temple to a height of about 1.50–1.60 m, or roughly to that of the altar top at ca. *98.30; in essence, this would create some sort of a higher open space like a plaza between them. The advantage of this solution would be to reduce the height of the west front stairs and the wall terminating the west end of the north pteroma to a more manageable 1.70 m (leading up from ca. *98.30 to *100.0).25 A hypothetical reconstruction drawing provides a visual aid for such an arrangement but does not confirm it (Fig. 4.4, proposal 2).

The problem with this proposal is the presence of a stele (number 15, assigned by Butler), apparently in situ, at the lower *96.80 level, against the roughly plastered wall of the altar (Figs. 4.5, 4.6).26 Thus, in order to propose an alternative arrangement for a higher ground level here, one has to assume that stele 15 was not in its original position but was moved there by Butler or by Roman builders of the temple’s west front. Or, alternatively, that LA 2 (whose exact date we do not know) was somewhat earlier than the temple, and its west face was plastered and the stele put in front of this wall before the temple was built, then buried; this is possible but not supported by evidence.27

During the Roman phase, the pseudodipteral plan was adopted and the original west wall of the cella was moved forward, toward the west, to create the double cella; work also started on the six-column pronaos porch of the west end, reflected in the design of the better-preserved prostyle porch of the east end (Fig. 4.1, Roman phases). The area between LA 2 and the west porch—which today is deeply cut up by Butler’s trenches—was filled to erect the porch columns (columns 48, 53, 54, 55, and 49) and the extensive mortared-rubble construction that encased them. A large portion of the debris in the west porch area, the rubble retaining walls between the altar and the pronaos porch, and some of the mortared-rubble construction between columns 54 and 55 were all removed by Butler.

It must be emphasized that in our opinion this major Roman phase was a part of an integrated “running design phase” for the entire temple without specific “early” or “late” Roman phases with hard dates. It was a construction process that must have taken a long time and moved from east to west as a succession of stages. It is in this sense that we see the west end—between the altar and the Hellenistic era steps leading up to the pronaos and then to the cella—as an area where Roman construction might have started almost concurrently with the east end. The north and south pteromas and their peristyle colonnades must have just been begun on the east side but were not yet extended to the west. Since the area between the altar and the extended, columnar pronaos porch was too narrow for a stair that could negotiate the full level difference, a new staircase on the north side of the porch was created on new mortared-rubble foundations (probably using the blocks of the Hellenistic era stairs discussed above).28 This is what we call the “northwest stairs” (Figs. 2.129, 2.154, 4.5). The area must have been filled with earth to the level of the porch floor (ca. *100.0) that was defined and partially retained by rubble walls ca. 2 m high and extending north and south of LA 2. These walls were largely removed by Butler, though they can be seen on his “actual state plan.”29 The east end of the northwest stairs was delimited by a “crude wall” of small blocks and mortared-rubble construction extending north, connecting the northwest anta pier to column foundation 45. It supported the fill of the north pteroma, which must have been in the process of being built in stages from east to west (Fig. 4.7). As Cahill notes, the north peristyle foundations eventually continued to the west end of the temple, and “the [northwest] stairs must have been considered a temporary arrangement.”30

The northwest stairs extend for a length of ca. 16.60 m between the “crude wall” on the east (with a 1.10 m space between the crude wall, which makes an almost ninety degree, 1.20 meter–thick dogleg against the north cella wall and the end of the steps) and the mortared-rubble wall on the northeast corner of LA 2 on the west (Fig. 2.129). The stair on its eastern end is preserved in five courses; on the west, only the bottom two courses remain. They can be reconstructed in seven risers that negotiate the total height of ca. 1.70 m, from the bottom at ca. *98.30 to the west pronaos floor at *100.0. The steps are constructed in white marble blocks varying in lengths of 0.45–2.20 m, and these display excellent workmanship with delicate, “raised” vertical joints. Their fine workmanship and small bar clamps (13–16 cm long, some misaligned and none functional) suggest Hellenistic or early Roman work, but it is clear that they were used here as spolia, reset and refitted, possibly from their original use as the Hellenistic era stairs leading up to the west pronaos porch.31 The stairs are supported (or backed) by mortared-rubble construction similar to what surrounds the foundations of the peripheral columns. On the eastern end of the steps, between the northwest anta pier and column 48 before it, the marble foundations are preserved at two courses below the paving level and show extensive pedestrian use; the entire western portion, in front of missing columns 52 and 59, plunges down some 1.60–1.70 m to the lower ground level at ca. *97.0, which extends as a kind of “deep corridor” between the exposed foundations of the columns of the pronaos porch (columns 53–55) and the great altar. The deep corridor as we see it today is simply a trench excavated by Butler; this is misleading in appearance, as is the “plunge” behind the stairs, which would not have existed when the stairs were operational (Figs. 4.5, 4.6).32

The mortared-rubble “crude wall” was excavated in 2010 and found to be bonded into the stairs at its south end; apparently it served as a retaining wall to support the east end of the north pteroma (Figs. 4.7, 4.8, 4.9). Although preserved in rough-rubble construction, it probably had a more presentable stone or marble veneer on its west face.33 The northern end of the rubble wall was cut by the mortared-rubble encasing of the north pteroma column foundations, mainly that of column 45; the west face of its south end is visible ca. 1.80 m from the east end of the stair. As observed by Cahill, “The peristyle must therefore postdate the wall and stairs, and these features were never used together; when stairs were in use, the peristyle foundations had not been laid here; when they were laid, the space between the peristyle and stairs must have been filled, burying the wall and the stairs.”34

The northwest stairs and the platform created between the west pronaos and the altar was the primary means of access to the west front of the temple during its early Roman phase, before the north peristyle columns had been extended this far west. Another means of access to the west pronaos area seems to have been provided from the south by stone stairs at the southwest end of the temple (a parallel arrangement), now preserved poorly in two steps. The northwest stairs could have been installed at almost any time before the construction of the peristyle column foundations of the west end (numbers 45, 47, and 51, the foundations for column 57 never laid; see Fig. 2.129). The construction of the west-end columns of the north peristyle, which buried the stairs, also could have been undertaken at almost any time between the mid-second century AD and the abandonment of the temple in the late fourth century. Yet, since the northwest stairs should logically be associated with the building of the six-column west pronaos porch, we believe that when work on the north pteroma and its columns (or column foundations) had sufficiently advanced by the later second century, the makeshift arrangement of the northwest stairs became unnecessary and unsupportable. Our explanation for the heavily eroded blocks two courses below the surface between the northwest anta and column 48 is less satisfactory (see Fig. 2.107); perhaps the stairs were dug up in late antiquity, when most of the fine marble blocks of the temple, including those of the stairs, had been robbed and this crude, functional passage merely 2.20 m wide served those who had business among the ruins.

  • Fig. Plan 2

    The Sanctuary of Artemis, Sardis. (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. Plan 6

    Temple of Artemis, Sardis, state plan, drawn 1:100 (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. 3.5

    East cella, column foundations 69–71, and late foundation blocks with basis beyond them (looking southwest). (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. 3.1

    Hypothetical restoration by Butler of the west end of the temple, incorporating the “Lydian Building” (LA 1 and LA 2). (Sardis II.1, ill. 97.)

  • Fig. 4.1

    Temple of Artemis, Sardis, plans of temple phases, 1:200. Hellenistic phase, ca. 280–260 BC; Roman phase, as preserved, ca. mid-second century AD; Roman phase, partially restored; Roman phase, fully restored, with hypothetical peristyle. (Fikret Yegül)

  • Fig. 4.2

    Ionic dipteroi of Asia Minor. Temple of Artemis (“second temple”), Ephesus, ca. 334–300 BC; Temple of Apollo, Didyma, ca. 300 BC and later; Temple of Hera, Samos (phase IV/Polycrates); Temple of Athena, Priene, ca. 334 BC. (FKY and DF redrawn after Gruben 1966, fig 260; Coulton,’ Greek Architects’, 27a)

  • Fig. 4.14

    Restored cutaway perspective by the author, Roman period (east porch, looking west into the east cella). See plate VIII for larger version. (Fikret Yegül)

  • Fig. 4.3

    Excavation of the terraces east of the Temple of Artemis, 1914. (Princeton University, Butler Archaeological Archive, research photograph B.519.)

  • Fig. 1.14

    Marble roof tiles found in Butler excavation. (New York, Metropolitan Museum of Art, accession no. 26.50.10.)

  • Fig. Plan 3

    Temple of Artemis, Sardis, western end with Lydian and Hellenistic altars. (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. 4.4

    Author’s hypothetical reconstructions of the west end. Top: Proposal 1 (Hellenistic). Bottom: Proposal 2 (Hellenistic). (Fikret Yegül )

  • Fig. 4.5

    Area between west porch and LA 2, with steles and stones on the northwest steps in the process of being reorganized by Butler, 1913. (Princeton University, Butler Archaeological Archive, research photograph A.54.)

  • Fig. 4.6

    LA 2, northeast corner with stele 15 (looking southwest). (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. 2.129

    Column foundation 45 (right), 47 (middle), 51 (left), and northwest steps. (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. 2.154

    Northwest anta and column 48 plinth (right), connected by block construction with northwest steps (front, right), with pedestal honoring Aphion Secunda, set there by Butler (looking southwest). (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. 4.7

    View of rubble “crude wall” in north pteroma at eastern end of northwest steps (excavated 2010), column 45 at lower left (looking southeast). (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. 4.8

    Mortared-rubble “crude wall” in north pteroma, with eastern end of northwest steps (looking south). (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. 4.9

    Mortared-rubble “crude wall” in north pteroma (the fluted drums in the distance sit on column foundation 45) (looking north). (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. 2.107

    Block foundations connecting northwest anta and column 48 plinth (looking east). (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

Roman Phase Design: The Shape of Power

The Roman construction of the temple commenced sometime in the Hadrianic era following a loosely conceived, pseudodipteral model with eight columns at the ends and twenty along the sides. Since the impetus for the “new” temple appears to have been to incorporate the imperial cult, the manner in which this was done, obviously, was to divide the cella into two nearly equal parts: the east accommodated the cult statues of the imperial family and faced the Acropolis; the west (further extended some 10.15 m westward), as in the original arrangement, belonged to Artemis. It is a strong possibility that there was an imperial cult altar on the east side, balancing the great altar of Artemis on the west side but far smaller in size, set against the towering east front columns, or possibly set inside the spacious pronaos porch in front of the stairs of the east door; since no remains of such a structure was found, one has to assume that it had shallow foundations and was totally destroyed and demolished during the Early Christian era. With a planned mantle of sixty-four columns, the full project was gigantic and ambitious, arguably the largest “pseudodipteral” temple in the world, and it progressed over centuries. Its peripteros of columns was still largely unfinished when the temple lost its pagan meaning and purpose in the fourth century AD.

At first glance, it is difficult to imagine that the great temple of the Sardian Artemis existed and served as a cella only for some four hundred years, all the way into the Roman Imperial era. Perhaps that is one of the reasons that theories for a major Hellenistic era rebuilding (Gruben’s and Hanfmann’s “second phase”), even in scholarly publication, are attractive and will not go away.35 Furthermore, the desire to explain the present “pseudodipteral” plan of the temple as a development directly influenced and inspired by Hermogenes’s exceptionally successful and timely innovation—hence the desire to connect the Sardis temple with Hermogenes—has been too seductive to resist in some of the published histories of the temple. How could one of the most imposing and renowned pseudodipteral temples of all time not have a connection to Hermogenes, whose magnum opus lies some eighty kilometers southwest of Sardis as the crow flies? Or, uncharitably, to credit the Romans as the creators of this remarkable architectural program and beautiful columns and capitals is too unsettling to accept. Chronologies for the temple’s alleged “late Hellenistic phase” differ from 220–200 BC (Hanfmann) to 190–150 BC (Gruben); anything resembling general agreement among scholars on the chronology of the Anatolian architect’s productive years (i.e., 225–175 BC, or possibly even later, ca. 175–125 BC) is a fairly recent phenomenon.36 Yet, as I have tried to show in this study—through technical and construction manner; archaeological evidence; stylistic, architectural, and functional analyses; historical indicators; inscriptions that speak; and reasonable circumstantial deductions based on the meaning and intent—the evidence precludes the interpretation of the pseudodipteral arrangement at Sardis as anything but Roman Imperial, a scheme causally linked to the double-cella plan created for the imperial cult occasioned by the city’s second neokorate. Our best estimate for this event, which signals the “second major phase” of the temple, is Hadrianic (ca. AD 130–60). As mentioned earlier, there is no cogent or compelling evidence for a major intermediary Hellenistic or early Roman phase, unless we conceive of “phases” as a succession of many construction processes in the Roman period and adopt a somewhat minimalist definition of what constitutes a phase in the history of a building. An important event that is tempting to associate with major rebuilding, primarily the design and partial erection of the peripteral columns, was the earthquake of AD 17 which devastated Asia Minor and according to sources, caused the worst damage at Sardis. The disaster was partly but generously alleviated by cash grants and tax remissions under Tiberius (Tacitus, Ann. 2.47). However, we have no idea how much damage this earthquake caused in the temple except for, surely, its marble roof.37 However, if the earthquake had inflicted serious or even minor damage upon the existing temple, the first order of business, in my opinion, would have been to repair this damage, not to start a ruinously expensive new building phase. Periodic and sundry repairs, routine maintenance work, renovations, and even schemes that were started and soon abandoned must have been normal for a monument with a history that spanned many centuries. Such reparation and renovation work often leaves little or no evidence—or the evidence is too elusive and confusing to follow cogently—and may be indicated as hypothetical possibilities rather than substantive phases in the building history of the temple. When the second-century Roman building commenced to surround the Hellenistic cella with a peristyle, it roughly followed the pseudodipteral arrangement we have, creating a gargantuan building out of an already very large one, its footprint some 2.7 times larger than its Hellenistic predecessor (Fig. 4.10).

Perhaps the exceptionally long time span between the completion of the Hellenistic cella and its commencement toward a proper peripteros is not as surprising as it first appears. For its users, a finished cella housing the sacred image was all that would have mattered in defining the sacred bond between the goddess and her worshippers. As underlined by F. Rumscheid, any classically inspired cella with the cult image inside was a “cult building.”38 The peripteral columns adding conventional grandeur could and would come later. Indeed, by delaying their construction—after having spent so much money for so long on the exquisite marble cella—the community could gain some financial respite and divert its funds to other worthy civic projects, such as a theater, a bouleuterion, or a basilica. O. Bingöl extended this logic to Magnesia—also basing his observations on technical and stylistic considerations of peristyle elements of the Hermogenes’s Temple of Artemis there—and made the bold proposition that even this temple might not have advanced beyond the cella stage during the master’s lifetime, but received its external colonnades sometime in the late first century BC or even later—now made more plausible with the discovery of a donor inscription for these columns datable to the early Imperial period. In this scenario, Vitruvius’s elaborate description of this iconic temple’s proportional system must have been based not on built-site observations, but rather on Hermogenes’s descriptive treatise.39

Complex Interaxial Contractions and the Importance of the New East Front

Introduced during the Roman rebuilding phase of the temple is a special convention of Archaic Ionic design known as multiple or complex interaxial contractions, present in the frontal row of columns where the axial spacing between them (or their columnar axes) regularly increases, starting from the ends toward the center. The eastern front of eight columns employs four different spatial variations (subtlety noted and admired by the Neoclassical architect C. R. Cockerell). Each interaxial pair flanking the center decreases in width from 7.06 m (the widest at the center) to 6.64 m, 5.45 m, and 5.31 m at the ends; the interaxial spacing of the long sides keep a uniform, narrow spacing of 4.99 m.40 The locus classicus of this system is at both the Archaic and subsequent Late Classical Artemisia of Ephesus where, according to recent investigations by A. Ohnesorg, the Archaic (“Croesus”) and Late Classical temples shared identical plans and had multiple interaxial contractions with four gradations on their western fronts.41 The evidence for this is not absolute, but it is convincing and suggests that the Temple of Artemis of Ephesus in its distant Archaic version, but more readily in its Late Classical, provided the model for the sophisticated interaxial contractions of the east peristyle columns of the Sardis temple in Roman times. Since only one of the column foundations (number 64) of the west peristyle had been laid, one cannot know if the Roman builders intended to apply multiple interaxial contractions on the west peristyle as well, though it is probably unlikely. At the Heraion in Samos, interaxial spacing is graded down from 8.40 m in the center, to the next pair on either side at 7.04 m, and the outer pair at the ends at 6.55 m (Figs. 4.1, 4.2). The back of this temple (with nine columns) did not employ this system. The Hellenistic Temple of Apollo at Didyma with ten columns in front and back retains uniform spacing all around; it is impossible to know for sure if its Archaic predecessor (completely covered by the later temple) used this system, but neither reconstruction alternative, by Gruben and more recently by B. Fehr, indicates it (Fig. 4.2).42 Thus this particular feature represented in various forms in the Archaic (and possibly Classical) temples at Ephesus and Samos, and so consummately applied at Sardis, represents a deliberate Roman revival of a patently Archaic model. It is also interesting to consider that the Roman temple builders of Sardis ignored the simpler model of equal spacing front and back that had been applied at the Hellenistic Didymaion in favor of more distant historical models.

It is also interesting that this sophisticated, archaizing system was applied only when the east end of the temple (the opisthodomos of the original cella) had become a new front during the Roman rebuilding. We cannot be sure of if such multiple contractions could have been a consideration for the Hellenistic temple; simply, there are no exterior columns and therefore no evidence. However, if columns had been built, it might have been applied only to the west, which was the proper front of the temple. Among the Archaic and (possibly) Classical examples known to us, this system is always used only on the principal facade of a temple. It is surely an indication of the extraordinary importance of the new east front, created solely for its mid-second century association with the imperial cult, that it received an extraordinary design with its full eight columns and carefully applied, historically conscious, catenated interaxial contractions (and perhaps planned, even if not executed, with a magnificent pediment)—while at the west end, the goddess still awaited a facade. Would the system of complex contractions have been applied also on the west facade if it had been built? Another intriguing and important point: Why did the once-principal west facade leading to the goddess’s chamber remain incomplete, or at best, was a truncated-looking tetrastyle porch? Despite arguments for the apparent topographical complications involved in achieving a full west facade, and of course the problems of cost that the political realities of second-century Sardis might have dictated (perhaps after a fight in the city council?), the architectural precedence enjoyed by the imperial gods was at the expense of the old goddess. This is not to say that every significant public dedication in Roman Sardis (such as the Marble Court of the Bath-Gymnasium Complex) did not start evoking the goddess’s name; but then one might say, as epigraphic formulas go, that talk is cheap. It also furnishes us with one of our stronger arguments for a second-century date for the construction of the east end of the temple rather than a Julio-Claudian date; only the Hadrianic (and later Antonine) admission of the imperial cult into the temple could have supplied the rationale for making the east end a new, primary front. If prestige and primacy were not conferred upon the east side by the admission of the imperial cult in the east cella in the second century, the Roman builders would have had no reason to turn this end into the primary facade.

The Divided Cella and the Exterior Columns

We imagine that by the mid-second century AD, the original west-facing cella of the temple was divided by the east crosswall into two chambers of nearly equal length (west cella interior length is 26.74 m, east cella interior is 25.16 m). In order to create sufficient space for the two cellas, the west wall of the original cella was dismantled and rebuilt ca. 10.10–10.13 m toward the west, thus incorporating roughly two-thirds of the original pronaos (Fig. 4.11). The west crosswall (the new western front wall of the temple) extended between the north and south walls of the pronaos porch, not bonding to them, and obliterated porch columns 79 and 80 (Figs. 2.65, 2.67). As discussed above, this operation also involved moving and rebuilding the cella door and stairs in the new location. Since the pronaos floor was ca. 1.60–1.70 m lower than the cella, the floor of the new addition to the western chamber had to be filled to the level of the original cella and a new roof and support structure devised. These operations, especially relating and adjusting the existing gabled roof of the Hellenistic cella to the much wider roof of a temple with peristyle, probably required dismantling and rebuilding the whole roof, even though much of this peristyle had never progressed beyond foundation stage. The new east and west pronaos porches of the two-cella temple were now equal in depth at 6.0 m, i.e., the depth of the original, east-facing opisthodomos. A monumental door with finely ornamented jambs and lintel was cut into the east wall, the original blank back wall of the temple (Fig. 2.181). Approached by a flight of steps, this door gave access to the new east chamber, which we believe was reserved to house the colossal imperial images and honored the imperial cult (Fig. 3.64). A similar door and stairs must have been made for the new west wall (or moved from their position on the original west wall), but nothing of these elements survived.

Following these operations within the cella (or concurrent with them), there must have been a massive attempt to realize the pseudodipteral scheme by laying out some of the foundations of the peristyle columns. Fifteen, but possibly as many as eighteen columns of the east end, including the eight frontal columns, were completed while only the columns of the pronaos porch of the west end were complete. The foundations and the plinths of five of the six columns of the west pronaos porch have been preserved to varying degrees, leaving the northwest corner column, number 52, missing; it was either never started or completely robbed. All others display smooth, finished tops with construction lines and markings that provide evidence for the columns they carried. Particularly, foundation blocks of columns 53 and 54 provide critical information that these columns had pedestal bases comparable to those of columns 11 and 12; the square plinth of column 53, in situ, shows clear markings of a square upper member, not circular as one would normally expect for the spira (round bottom element) of a regular Asiatic-Ionic base (Fig. 2.157). Corroborating evidence is also provided by C. F. Stanfield’s 1830s watercolor showing a few columns standing in the distant background; these could only be the columns of the west porch (Fig. 1.27). Furthermore, no less than fourteen or fifteen fallen fluted column drums inside and around the west porch were found mainly at the west end and the majority grouped together by Butler between the column foundations 53 and 54.

The appearance of the north and south long sides of the temple must have been uneven and visually disconcerting as well. Of the twenty column positions of the north side, only two (numbers 9 and 15) must have had bases and columns. On the south colonnade only column foundation 18 (third from the southeast corner) is fully preserved; the foundations on either side of number 18 (numbers 14 and 20) probably had columns (Fig. 2.199).43 Of the fifty-two peristyle column positions, only ten or eleven (or, eight or ten) had never been started, even in foundations, which is itself a major undertaking. Almost all of the north and south colonnades were in place, to some degree at least, intended to be completed sometime. Still, by the time the pagan temple was abandoned over the course of the fourth century, only a few foundations had actual column bases, and fewer had columns on them.

We do not know if the construction of the bases progressed concurrently or piecemeal—probably the latter, with several at one time. To erect even one of those massive columns was a laborious and costly task that could have relied on the generosity of a patron or a group of patrons. It is interesting, however, that the epigraphic message of column 4 names no particular donor (nor “public funds”), but rather the temple’s own resources as the source of this exceptional column, possibly the quarries owned by the temple. The erection of the temple columns must have progressed slowly and sporadically as funds allowed. Still, it is curious that none of the finished columns of the temple name a donor; after all, each of these columns must have been conceived as a milestone in the completion of the temple in financial, visual, and symbolic terms. Perhaps the very visible and massive act of constructing a row of columns 18 meters tall, a monumental civic achievement, trumped the naming of donors. It is also true that different cities of ancient Anatolia had different traditions or levels of generosity in public giving; while a city like Aphrodisias or Magnesia presents an unusually rich record of public munificence and donors’ names, Sardis has far fewer.44 However, it is more likely that the temple, a venerable civic and economic entity, owned the Mağara Deresi quarries (or portions of them since they are extensive) and paid for its columns over a period of time. Yet, after the politically significant and well planned east-end columns that heralded the imperial cult chamber were in place (and recorded in a laudatory manner by a “talking column”), and the erection of the six-column west pronaos porch, the minimal architectural gesture of monumental entry for Artemis’s own chamber was achieved, and whatever the reason, enthusiasm for further column erections appears to have waned.

A Pre-Modern Ruin?

The appearance of the unfinished temple, despite nearly two centuries of work since the commencement of the Roman phase, must have been intriguing—at least from some angles, visually awkward. How could the mighty Romans rule Sardis during its wealthiest period in Asia Minor, one might ask, yet allow such an important project to remain unfinished? Some colleagues, such as R. R. R. Smith, have asked—ostensibly seriously and certainly creatively—if Sardis left its “new old temple consciously unfinished.” The question posed is more intellectual than shocking. Was this, indeed, an intentional and sophisticated desire to experience the building as an artistic “ruin”? While we may disagree with the concept of an outright intentional ruin (especially in the ancient world), it may be worth going beyond the heterodoxy of this view and rethink the question in a broader context, in which the explanation “they never got around to finish it” is not good enough.45 The lure and pleasure of ruins, a popular theme in the eighteenth- and nineteenth-century Romanticism of Giovanni Battista Piranesi and Hubert Robert, among others, was mockingly revisited in the Postmodernism of 1970s. In a sense, the imagining of a building’s ruins, or imagining a building as a part of a “ruined composition,” is the opposite of its construction/reconstruction cycle, giving license to a greater play of imagination; it is an exercise in restitution. John Pinto quoted Piranesi, a master restitutor of ancient buildings as imaginative ruins, as remarking, “Speaking ruins have filled my spirit with images that accurate drawings would have never succeeded in conveying.” Piranesi might as well have said, with equal cogency, that an incomplete ancient temple could fill one’s spirit with images in a way that a finished one could never do.46

The motivation behind the visually engaging or simply outrageous modern ruin—such as the set of Best Company stores (especially the showroom at Sacramento, 1977, with its iconic “notched” facade, Fig. 4.12)—founded on whimsical aesthetic and social commentary, was at least partially serious. Although our own explanation for our unfinished temple is mainly based on economy and opportunity (rather than artistic delight in sublunary things), and therefore a lot more practical (and dull), we admire the boldness of such an idea in its capacity to challenge the conventional limits of perception, inspire new horizons of thinking, and start new debates on the nature and meaning of classical art and architecture. Some of our proposals about the temple are impossible to verify on hard evidence, so we, too, rely on just that kind of thinking and reasoning. Some of these concerns and questions have been taken up with penetrating wit by R. Harbison, whose chapter on “ruins” in The Built, the Unbuilt, and the Unbuildable starts: “Ruins are ideal: the perceiver’s attitudes count so heavily that one is tempted to say ruins are a way of seeing.” And in reference to the “bizarre design” of Best stores, Harbison observes, “who would have thought that contemporary American shoppers could entertain simultaneously the consumer’s fiction of… shiny function and fantasies of decay?” As delightfully imagined, if the great Artemision of Sardis was designed as or allowed to be an artistic, premodern ruin, what fiction of function or fantasy of decay could the Sardians entertain?47

The Imperial Outlook: An Unfinished Temple

A late second-century AD view of the temple from the east would have been quite impressive and calculatedly deceptive (Figs. 4.10, 4.13). Eight magnificent columns of the east facade, with their tall, unfluted, powerful shafts and possibly a massive pediment, rose against the broad mass of the mountain behind the necropolis hill to the west. An altar dedicated to the imperial cult would be expected here, though none was found. Although no actual remains of the frieze or the cornice have been found, a hypothetical restoration could consist of a cornice decorated with modillions and dentils and a plain frieze like the Pantheon’s—a familiar combination in imperial architecture. The temple must have been raised on a simple embankment or platform like a berm, already proposed for the Hellenistic cella, then regularized and extended out as a proper pteroma, probably not even paved. A crepidoma of six or eight steps built on a mortared-rubble backing, like the ones preserved for the comparable Ionic temples at Ephesus or Didyma—or the early Imperial Wadi B temple at Sardis—must have been intended but was either not constructed or only constructed at a few locations for limited lengths.48 The platform of mortared-rubble construction of the pteroma intermittently and variably projected beyond the line of the peripteral columns. It is on the basis of these possibilities that “idealized,” hypothetical reconstructions of the temple, both inside and outside, have been attempted on paper (Figs. 4.13, 4.14).

In reality, the apparent and quasi-finished grandeur of the east front was an illusion. Along the north and south sides there were only a few isolated columns, the great majority of their places empty with either just a few bases, unfinished bases, only huge foundation blocks, or possibly, in a few locations, nothing at all. The great pediment and the full-width, gabled roof it fronted, if attempted, awkwardly narrowed down and faded out against the narrower cella roof; this might have happened logically along the north–south alignment of columns 21 and 22, with the east wall of the cella in between them. Pteromas must have existed but were not yet roofed, except, as projected, at the east end of the temple. If one viewed the Roman temple from an unfavorable angle, the outlook would have been even more edgy and abstruse than its Hellenistic single-cella predecessor. In contrast to the fully built east front, the west front had a much smaller pediment, or simply an architrave, that was supported by the four front columns of the pronaos porch, resembling a simple prostyle temple writ large—very large. Colleagues have asked, but it is difficult to say how the roof over this mixed shape was configured; the connection between the elongated gabled roof covering just the cella and the wider, full roof of the east end, which partially covered the unfinished pteromas, would have been awkward—although there are a few acceptable ways that this could be achieved with relative visual and structural grace. I present one technical and aesthetic possibility in my reconstruction drawings (Figs. 4.13, 4.14). The most likely hypothesis is a full pediment that covers the east porch at a higher roofline, then continues over the cella with a break (possibly across the line of the cella east wall, as mentioned before) and a separate, simple gabled roof at a slightly lower level.49 The application (really, grafting) of a larger, wider, but slightly lower Roman era pediment upon the narrow facade of the Hellenistic temple of Demeter at Pergamon demonstrates the solution for a similar problem (Fig. 4.15).50 Furthermore, one should remember that any sense of the visual and structural abstruseness attributed to the unfinished temple as a building is, to a certain extent, a reflection of our own cultural and visual perceptions and prejudices; a believer of the ancient world would have seen something deeper in the sacred nature of the shrine and its impressive and moving geography than its apparent static, physical shell—perhaps, something that merges the magnificence of appearance with the ethics of purpose.51

A recent digital reconstruction of the pictorial qualities and visual effects of the Temple of Artemis at Magnesia by L. Haselberger and S. Holzman analyzes the exceptional qualities of pseudodipteral temples with rich contrasts of light and shade, associating this experience to the notion of asperitas (propter asperitatem intercolumniorum), which Vi-truvius uses in his admiring description of the architectural qualities of Hermogenes’s novel, pseudodipteral arrangement (De architectura 3.3.9).52 The Latin word and the notion it expresses are imprecise and flexible, quite different from more familiar literary implications of “asperity.” As Haselberger and Holzman argue, Vitruvius might have alluded to rich, scenographic effects and the “illusion of relief” created by the well-lit row of columns set against the dark depth of the hall-like pteromas that are typical of pseudodipteral temples. Digital restorations of light and shadow simulations of the west facade of Hermogenes’s Temple of Artemis at Magnesia as a pseudodipteros (which it is), along with comparisons to simulations as a dipteros or a peripteros, recreate varieties of spatial depth and quality, or grandeur and dignity (Fig. 4.16). While this notion (as well as the modern-day drafting styles it engenders) lacks precision, we have always emphasized the spatial qualities of the Sardis pseudodipteros in architectonic and scenographic terms, though without recourse to the Vitruvian asperitas.53 Indeed it is the particular strength (and “magnificence”) of our temple, with its immense, hall-like ambulatories where spatial and scenographic effects would have doubled in intensity and certainly in grandeur, as compared to the experience of Hermogenes’s temple in Magnesia (Fig. 4.16).54 Furthermore, at the east and west ends where the ambulatories wrapped around the cella and merged into porches that (probably) soared to light, the contrasting and changing effects of light and dark would have been stunning (Fig. 4.17). From a distance, the sense of architectural “illusion and grandeur” of this largest of all pseudodipteral temples—whether the Vitruvian asperitas or not—must have been enhanced by the natural grandeur of the craggy Acropolis and the rolling Tmolos Mountains.

The interiors of the double cellas must have afforded a more traditional appearance. On the west, preceded by a simple but monumental and possibly unfinished pronaos porch, the old goddess of the Sardians reigned in her new, spacious cella; and judging by the number of votive monuments of Roman date, she continued to enjoy the love and respect she was accustomed to under the new regime. What she might have lacked in terms of a proper temple frontage, she must have made up by the prestige of her venerable altar, which we believe was integrated into the west porch across an elevated “plaza” as in the old days, or the spacious enclosure of the six-column pronaos porch. On the opposite side facing east (Fig. 4.13), the new imperial gods and goddesses enjoyed all the advantages and privileges of politically expedient newcomers inside their traditionally arranged, columnar cella (Figs. 3.63, 3.64). Lords and masters in matters practical, and acknowledged as such in the archaistic and erudite but sometimes bombastic language of Asia’s learned philosophers of the Second Sophistic, their cult and iconic presence were heralded by a suitably impressive temple facade, a shape that served as an effective reminder of the power and circumstance of the new Roman Sardis, even if the facade and what it represented (to some) was an illusion.

If we look at the bigger picture, the apparently untidy impression caused by the centuries-long Roman building and rebuilding of an “always unfinished” edifice, becomes a phenomenon with some positive civic and religious connotations. The roughly textured, unfinished surfaces contrasting against and alternating with the finely finished and polished moldings and joints would have conveyed the image of a powerful, even willful sense of rustication and enhanced the readability of the temple as an integrated whole of variegated marble masses and planes. Furthermore, such a bold and mannered display of architecture in various stages of finish—always “in progress”—could have invoked an underlying sense of appreciation for the skill and effort imbued in the process of construction, hence a sense of inherent monumentality orchestrated to the scale of time. As observed by D. Favro in reference to the frenzied state of construction of Rome under Augustus and other emperors, so, too, the Sardian Artemision—from the slopes of the Acropolis to the valley of the Pactolus—must have looked and sounded like a massive jobsite: networks of scaffolding and cranes; piles of building materials; the constant din of moving, heaving, and lifting; the rhythmic rattle and clatter of carving, shaping, chipping, and scraping—and marble everywhere—all a perennial show of labor and progress to bolster the drama of civic achievement in the service of the city and its gods, celestial and imperial.55 Every step in the process marked the passage of time, measurable in detail but immeasurable and eternal in wholeness. Every impeccably carved detail, every perfectly rounded molding teased out of roughly shaped stone by master masons, reflected the community’s homage to its deities and became the visible, tangible proof of it.

Such a chaotic state of construction was a common scene in Rome and in Roman cities, especially in second-century Asia Minor, and a symbol of messy vitality. The signifier of progress is process, which the leaders of Roman Sardis must have been keen to sustain, although there is little evidence for major construction at the temple beyond the late Antonine era, except perhaps the addition of a foundation or two for the columns of the long peristyle. A finished project was indeed cause for pride and celebration (and repose of arrival); but the perpetual state of construction of an exceptional building, polishing and finishing, must have been cause for perpetual celebration in joining the past to the present.

  • Fig. 4.10

    Massing comparison of Hellenistic- and Roman-phase temple structures. (Fikret Yegül)

  • Fig. 4.1

    Temple of Artemis, Sardis, plans of temple phases, 1:200. Hellenistic phase, ca. 280–260 BC; Roman phase, as preserved, ca. mid-second century AD; Roman phase, partially restored; Roman phase, fully restored, with hypothetical peristyle. (Fikret Yegül)

  • Fig. 4.2

    Ionic dipteroi of Asia Minor. Temple of Artemis (“second temple”), Ephesus, ca. 334–300 BC; Temple of Apollo, Didyma, ca. 300 BC and later; Temple of Hera, Samos (phase IV/Polycrates); Temple of Athena, Priene, ca. 334 BC. (FKY and DF redrawn after Gruben 1966, fig 260; Coulton,’ Greek Architects’, 27a)

  • Fig. 4.11

    South pteroma, western end; south wall, junction with middle crosswall, west and east cellas separated by east crosswall (looking east). (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. 2.65

    West wall, interior face, covering column foundations 80 and 79 (with column base 78 in foreground) (looking northwest). (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. 2.67

    West wall, connection to north wall (looking southeast). (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. 2.181

    East peristyle and east porch. (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. 3.64

    Hypothetical sequence of approach to the cella with cult images from the east porch. Top: View from east peristyle. Center: View from east door stairs, with open porch roofing system visible. Bottom: View from the entrance to the east cella. (Digital modeling R. Pellegrini, UCLA Experimental Technologies Center)

  • Fig. 2.157

    Column foundation 53. (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. 1.27

    Clarkson Frederick Stanfield, R.A. (1793–1867), Sardis. One of the Seven Churches, ca. 1835. Water- and bodycolor over pencil on paper, 23.3 × 35.5 cm. (London, Victoria and Albert Museum, SD.1000.)

  • Fig. 2.199

    Column 18, from south peristyle. (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. 4.12

    BEST store, Sacramento, mid-1970s. (Courtesy Richard Longstreth.)

  • Fig. 4.13

    Restored perspective by the author (ideal state), Roman period (east and south sides). See plate VIII for larger version. (Fikret Yegül)

  • Fig. 4.14

    Restored cutaway perspective by the author, Roman period (east porch, looking west into the east cella). See plate VIII for larger version. (Fikret Yegül)

  • Fig. 4.15

    Temple of Demeter, Pergamon, elevation (with Roman porch addition) and plan. (Fikret Yegül and Diane Favro after AvP XIII, pl. 52.)

  • Fig. 4.16

    Temple of Artemis, Magnesia on the Meander, west facade, computer-aided simulations of light conditions, approximately on a mid-afternoon in late autumn. Top: Light effects on the temple as a pseudodipteros. Center: Light effects on the temple as a dipteros. Bottom: Light effects on the temple as a peripteros. (Haselburger and Holman 2015, fig. 10; courtesy of the authors.)

  • Fig. 4.17

    Temple of Artemis, Sardis, author’s reconstruction of hypothetical east pronaos porch (open to the sky). (Fikret Yegül)

  • Fig. 3.63

    East cella, hypothetical arrangements of imperial cult statues. Left: A. Pius and Faustina occupy the central podium. Right: Hadrian and Sabina occupy the central podium. (Fikret Yegül)

Architectural Analysis and Comparisons

The application of the pseudodipteral plan at Sardis occurred not as a natural development in the wake of Hermogenes’s new and prestigious temple in Magnesia, nor its many prestigious followers in Asia Minor, but rather as a deliberate and historicizing choice centuries after the master’s architectural landmark and its canonization by Vitruvius (3.2.6; 3.3). In Magnesia, Hermogenes created (or planned) an ambulatory of two uniform, interaxial widths all around the cella; all columns are equally spaced, with the exception of the wider-spaced pair in the middle (Fig. 4.20). At Sardis, not only does spacing between the columns vary progressively following an Archaic convention, but also the ambulatories at the ends of the temple are wider than the sides (three-intercolumniation ends versus two-intercolumniation sides). As pointed out by Howe, the Roman era pseudodipteros at Sardis is not a pseudodipteros at all, technically speaking.56 Furthermore, by creating spacious, projecting six-column pronaos porches within the columnar enclosure of the temple’s ends, and pushing them forward into the normal intercolumnar space against the frontal rows (in effect, occupying this pseudodipteral corridor), the Sardis temple violates the basic rule behind pseudodipteral design—or, in terms of scenographic and visual drama, improves upon it.

The Creation of Space

A normal pseudodipteral arrangement (and one in line with other Roman era pseudodipteroi in Asia Minor, such as the temples at Ankara and Aezane; Figs. 3.75, 3.76) would have favored a simple tetrastyle-prostyle internal porch enveloped by a continuous, uniform double-intercolumniation-wide ambulatory. This “logical” and canonical solution, for which there were clear and available models, was obviously never the case for Sardis, where unusual, triple-intercolumniation-deep pronaos porches enclose and capture space like magnificent halls in a way that a straight-front tetrastyle or hexastyle porch arrangement does not and cannot.57 These large, immensely tall, “cubical halls” defined by the six columns of the porch and the projecting arms of the anta walls might have even been open to the sky (Fig. 4.17). Even if we accept that the original columns in antis were retained into the Hadrianic phase for simple structural expedience, and that this space was covered by a high, timber-trussed roof (which might have contained a more modest sky opening) by carving a void in the heart of the pronaos—as opposed to the forest of even-spaced columns that occupy the same areas in the contemporary Artemision of Ephesus or the Temple of Apollo at Didyma—the Sardian Artemision would have still projected a bold and expanding sense of space, exploited through its unorthodox plan. Space and visual drama, the play of light and shadow and of material and void induced by the great, soaring hole framed by tall, slender Ionic columns of the “pseudo-pseudodipteros”—the “dazzling effect” of propter asperitatem intercolumniorum—would have been overpowering. Consider first the ambulatories that are 9 m wide and 92 m long, defined on one side by the smooth, sheer, stark, unornamented nineteen-meter rise of a marble wall, and on the other a tightly spaced, overlapping rhythm of monumental columns of the same height (Fig. 4.18)—a lofty, hard-edged corridor of rhythmic lights and shadows terminated at each end by single, tall, anthropomorphic, iconic columns merged into the expanding volume of a pronaos, rising to about nineteen meters like a tower of light shot through a mass slashed by shadows. This was an experience of surface and space not commonly found in the orderly, canonical temple architecture of Hermogenes and his followers.58 In its monumental impact and “dazzling effect,” it went beyond anything that a Hermogenean pseudodipteros could create. I believe the sources and inspirations were hybrid, and although they were anchored at home and in its predecessor, they also lay further afield.

Mixed Sources and Traditions behind the Sardis Artemision

In assessing the nature of Roman architecture in Asia Minor over fifty years ago, J. B. Ward-Perkins identified two broad currents: buildings that followed local, Hellenistic traditions and those that took Italy and the West for inspiration due to a lack of suitable models at home. He found that religious architecture belonged “decisively to the former category.”59 Undoubtedly there is truth and simplification in this view. Such binary positions, more fashionable in the 1970s than now, have both strengths and limitations. While searching for cultural sources behind traditions is a valid form of analysis, one should be prepared to accept that the end product is often more than the sum of its parts. This is especially true in Asia Minor, where a rich patrimony of cultural resources and vigorous borrowings blur the hybrid origins of architecture, as well as its unusual and unique results. This is the background against which we should view the Temple of Artemis at Sardis, whose design does not represent the simple following of an idea or a recognizable type, but lots of ideas and types shaped by fortuitous circumstances and deliberate choices—a situation true not only for this one temple, but for Roman architecture in Anatolia in general.60 There is precedence but also serendipity here—the desire to follow tradition, to re-create memory, but also novelty and the deliberate rejection of memory. We have already emphasized the role of the great Ionian temples in shaping the cella of the first phase of the Sardian temple (Fig. 4.2; see in this chapter pp. 229-231). One must remember the daunting choice faced by the Roman architect(s) who inherited this long, narrow, austere, archaizing cella squeezed tight between a venerable altar at one end and the rising ground hugging the steep Acropolis at the other; the challenge to fashion a mantle of columns around a naked cella that must have been as formidable as it was unique (Figs. 4.1, 4.10). The resulting design, starting from an honored, extant structure and reshaping it into an idiosyncratic variant of a broadly conceived pseudodipteros, benefited from sources rooted in local as well as foreign traditions. The Roman re-creation of the Artemis Temple at Sardis is as remarkable as it is problematic because it has little to do with the familiar and regular Hermogenean tradition of pseudodipteros type that was established in Asia Minor by the middle or end of the second century BC.

The question lingers as to who was responsible for the exceptional design of Artemis’s temple at Sardis through its historical trajectory from Hermogenes to Hadrian. Who inherited history and reshaped it into a creative variant of a “broadly conceived pseudodipteros?” A vibrant, wealthy, busily building regional center like Sardis must have had many architects and master builders at work during the Imperial period. Yet given the paucity of information on the names and careers of ancient architects, especially for those working outside Rome, the task of finding who is daunting, achieved more through luck than design. Compared to the many questions we are confronted with regarding this great temple, the lack of information about its architect(s) is the one causing the least worry. Still, we may be visited by such luck in a very modest, tentative way.

An Architect from the Vine-Rich Tmolos

Among the various Greek inscriptions Captain Beaufort saw and transcribed in 1811–12 at Patara, in coastal Lycia, was the statue base of an architect identified as Dionysius of Sardis (rather, “Dionysius from vine-rich Tmolos”); it might have stood in the town’s odeion, as it celebrated the architect in reference to this building.61 It is a posthumous inscription that has been placed in a position of honor by the excavators at the entrance porch of the restored odeion (Fig. 4.19). Dionysius describes himself as “skilled in all works of Athena” and hopes for abiding fame for his engineering feat of putting a great roof over the odeion. Then, he wistfully informs the reader that he died in Patara far from home (“the foreign earth of Patara received and holds me”). The words the Patarans put into Dionysius’s mouth, and the sentiments in his heart in honoring him with a statue and an epitaph, seem to have been fulfilled in a modest way: we are, after all, talking about this master architect from Sardis. We have no idea who Dionysius the architect was—if he ever worked at Sardis, or whether his engineering skills inspired by Athena, for which he was justly proud, were ever put to test for Artemis in his home city. We do not know whether the architect was ever involved in the design or construction of the Temple of Artemis in any capacity.62 Since the Patara project can be dated to the Hadrianic–Antonine period, it is a possibility in chronological terms—and a tempting one. Indeed it would be unlikely for an architect whose fame and skill had traveled well beyond his Tmolos home, not to have had a hand in the most important and prestigious, ongoing architectural project of his city at some point in his career.63 Is it possible that Dionysius left Sardis as a young man, never to return? Or if he did visit, the prestigious architectural project of his hometown might have been already assigned to professional rivals with connections to the sources of power and prestige—it often happens that way.

Asia Minor and the Legacy of Hermogenes

Second- and first-century BC pseudodipteroi in Asia Minor (Fig. 4.20) are generally characterized by front and back porches with columns in antis, such as the Temple of Artemis at Magnesia, the Temple of Apollo Smintheus at Chryse-Gülpınar, the Temple of Apollo at Alabanda, the Temple of Hecate at Lagina (a variant with no opisthodomos), and as was probably the Temple of Aphrodite at Aphrodisias. They share conservative plans that go back to the Archaic and Classical Doric temples in Greece, but these temples follow the Anatolian preference for emphasizing the front; their front porches are invariably deeper than their back porches (or they have straight-backed cellas).64 In the Temple of Aphrodite at Mesa, Lesbos, an early pseudodipteros (or proto-pseudodipteros), which might predate the Hermogenes’s Artemision by as much as a century, the pronaos and the opisthodomos are exactly equal, with a pair of columns in antis, as in a mainland Doric temple. Three of the well-known pseudodipteroi from the Roman period—the Temple of Augustus and Roma at Ankara, ca. 25 BC (Fig. 3.75); the Temple of Zeus at Aezane, ca. AD 80–90 (Fig. 3.76); and the Temple of Domitian at Ephesus, ca. AD 80 (Fig. 4.20)—are routinely described in scholarly literature as having conservative plans that more or less follow the traditional, Hellenistic temple model of Asia Minor established by Hermogenes at Magnesia (Fig. 4.20).65 To this small group we might add the Wadi B temple at Sardis, an early Imperial pseudodipteros (Fig. 3.72 and pp. 217-220).66 What scholarly literature does not indicate is that all four of these Roman pseudodipteroi also deviate from the standard Hellenistic tradition by incorporating a prostyle tetrastyle porch within the peristyle instead of the usual columns in antis arrangement. This characteristic is significant not because such projecting porches inside the peristyle are unknown in the larger context of Greek temple architecture (consider its locus classicus to be the Parthenon in Athens), but because they are rare in Greece and Anatolia and simply do not exist among the known Hellenistic pseudodipteroi (Fig. 4.20).67 One wonders if this group of pseudodipteroi with pronaos porches created during a period of increasing Roman presence in Asia Minor represents a preferred Roman architectural type?

One should ask what is unique about the Sardis Artemision. I have dwelled on the use of the “pronaos porch” in the Roman pseudodipteroi of Asia Minor in order to highlight an architecturally related but far more unusual and interesting design motif of the Sardis temple: a six-column prostyle porch that projects out from the anta piers, all the way up to the front colonnade (see Fig. 4.22). Extending beyond the anta walls, the spatial potential of such lofty enclosures, especially in a huge temple like Sardis’s, is obvious. Already in 2004, Ü. Serdaroğlu identified them as “deep pronaos” temples uncommon in Anatolian usage. These columnar porches create internalized pockets of space deep within the peristyle and are not characteristic of temple architecture in the Greek East.68 Notable exceptions, all from the Roman Imperial era, have distinct links to Italy and include the side-by-side temples of Jupiter Heliopolis and the Temple of Bacchus at Baalbek, Syria (see in this chapter, below), and to a certain extent the Temple of Artemis at Gerasa in Jordan (the six-column “porch” is not within the peristyle but a part of it) and the Severan Temple at Lepcis Magna, in Libya—are all imperial projects (Fig. 4.20; see also Fig. 4.22). When such porches—with prostyle projections and columns on the returns—do occur they are invariably found in small, simple prostyle temples, such as the so-called Corinthian temple in Termessus (the largest of the group, a hexastyle prostyle; the rest are all tetrastyle); a very small, unidentified temple in Alakapı in southwest Anatolia (Fig. 4.20); and the early second-century BC Temple of Dionysus (later the Temple of Caracalla, though construction and ornament belongs to the late Hadrianic era) scenically located on the theater terrace in Pergamon (Figs. 4.20, 4.21).69 Among the larger Roman era temples of Asia Minor, the Temple of Artemis at Sardis is the only one known to us that has projecting porches placed within a peristyle.70

The Italian Connection

For the genesis, development, and articulation of temples which feature deep, spacious, and internalized pronaos porches, and privilege such porches as distinctive plan organizers, we turn to Italy.71 We start with Vitruvius and his description of the Etruscan temple, where he makes clear that the depth of the porch is to be made equal to the depth of the cella(s), with special attention paid to the “forecourt” or the pronaos: “Let the space before the cella be planned so that the corner columns are aligned with the anta walls” (Vitruvius 4.7.2). Vitruvius’s technical prescription—his emphasis of the “space before the cella” or its reasonable variants—finds ample confirmation on the ground.

The twin temples of Fortuna and Mater Matuta in the Sanctuary of Sant’Omobono (ca. mid-fifth century BC) are characterized by lateral walls that project from the back like long corridors and enclose space around the cellas (Fig. 4.22).72 The front porches, deeper than half the length of the temple and each with four wooden interior columns to hold the roof, open onto a very high platform and entirely dominate the design.73 The venerable Capitolium in Rome, dedicated in 509 BC and many times rebuilt, however, conforms to the Vitruvian scheme only in a generalized way. It had a triple cella and an exceptionally spacious porch with six frontal columns three rows deep, and the outer rows ran back to the projecting wings (alae) back wall of the cella.74 While the front of the Capitolium was well emphasized, the forest-of-columns effect, however austere and monumental, must have detracted from the sense of open space. In other words, it is not a model we seek to associate with future deep porch/pronaos schemes, nor with Sardis. Even relatively small and late temples, such as the third-century BC temple at Fiesole, or the second century BC Temple B at Ordona (Herdonia), represent the continuation of the peculiar arrangement of Sant’Omobono, where the lateral corridors envelop the cella and expand into the open “forecourt”—curiously, like the disposition of peripheral spaces at the ends of the Sardis temple. In this connection the “forecourt” design of the second-century BC Temple B at Pietrabbondante, which crowns the great, scenically laid Italic sanctuary, is of particular interest (Fig. 4.22).75 The wide prostyle tetrastyle with two columns along the sides projects before the anta walls of the triple cella—as if the temples shed their side walls to bring their deep, internalized porches out into the open. The grand porch thus created (9 × 16 m), like the pronaos porches of Sardis, defines a spacious, distinctive, hall-like space, still capable of being roofed. Whether these broad porches were roofed or not, the technical challenge to roof them seems to have been a conscious and positive aspect of the program of every “deep porch.”

Late Republican temples of central Italy experimented with endless variants of the Etrusco-Italic models. Two fundamental approaches are recognizable: the first is characterized by frontal emphasis with spacious prostyle porches (no side columns); the second, the periptero sine postico, which incorporates columnar side corridors with a clearly defined open porch, or “deep porch,” as the distinguishing element of the temple front.76 The latter tends to be more typical of the Late Republic, perhaps because of its affinity to Greek peripteral temple design.77 Among the well-known examples in Rome, which combine the Italic plan with the Greek side colonnades, are the Temple of Jupiter Stator (Fig. 4.22) and the temple in the Forum Holitorium, both dated to the late second century BC.78 Both are hexastyle, the front colonnade fully integrated into the structure with the six-column porch, and provide the Italian models for the Temple of Artemis at Gerasa (see Fig. 4.20). The octastyle temple that crowns the Sanctuary of Hercules Victor at Tivoli, ca. 50 BC, is a more elaborate version of the same design; its deep front porch projects not the usual two, but rather three intercolumniations, thus enclosing a very large space, ca. 15 × 11 m, defined by twelve columns (Fig. 4.22).79 For all its grandiosity, Tivoli was a provincial sanctuary. The utilization of a conservative sine postico plan a half-century later for the Temple of Mars Ultor in the Forum of Augustus in Rome, which is almost an enlarged carbon copy of the Temple of Hercules Victor, is a testimony to the acceptance of the type as the standard for prestigious monuments (Fig. 4.22).80

The most distinctive feature of this important temple that dominates perhaps the most politically charged and artistically accomplished public space in Rome is its majestic, pedimented front: octastyle, three intercolumniations deep before short anta walls, defining a ten-column pronaos porch. As observed by J. Senseney, “The architect could have filled the deep pronaos with a second row of columns, or in the case of the [latter two large temples] an additional two rows for a forest effect, yet has instead treated the central porch as an independent space[italics mine].”81 Could have but did not: this is what architecture is all about, a historicizing, aesthetic, and informed choice that must have dictated the architect’s decision to exploit and privilege this area as an open independent space, without columns. Whether the exceptionally wide cellas and deep prostyle-pronaos porches were a native Italian habit and choice that was governed by aesthetic, functional, or religious considerations may be beside the point. The choice of the pseudodipteral option at Sardis, however, was all but necessary and, in a sense, already decided for our architect.82

Let us look at the plan and consider: if the architect at Sardis had opted for a normal, peripteral scheme, the resulting temple would have had a six-column facade, incongruous and ill-matched to its enormous cella. Unlike the Italian octastyle temples with six columns across their wide cellas (the Temples of Venus Genetrix, Castor and Pollux, and Mars Ultor, all in Rome, and others), the Sardis cella could accommodate only four prostyle columns across its cella. In other words, Roman or Roman-style temples could be octastyle, even decastyle, and still accommodate peripteral schemes because they typically had wide cellas, hence the need to make the Artemis Temple at Sardis a “big temple,” commensurate with its extant narrow but long cella that mandated the use of the pseudodipteral (or dipteral) system with an octastyle front. Likewise, the monumental height of the existing cella—insofar as the Hellenistic cella of the Sardis temple matched the height of the Roman columns, or close to it—determined the height of the Roman peristyle. In a sense, the Hellenistic temple at Sardis (as an unfinished cella or as an intended dipteros) was the parent of the Roman one.

Even when Roman temple design followed thoroughly hellenized peripteral arrangements, as in the Augustan Temple of Castor in the Forum Romanum—and almost every other variant on Italian soil we care to consider—the native Italian porch dominates. Seen in this light, it is remarkable that the three relevant Asian pseudodipteroi of the Imperial era at Aphrodisias, Ankara, and Aezane (the first two started under Augustus) resisted the formation of deep pronaos porches (see Figs. 3.75, 3.76).83 In Syria and the East, where native versions of Hellenistic temples had neither been as strong nor as persistent as they were in Asia Minor, there was no such resistance, and well-established “Romanisms” could be embraced. The plan of the Temple of Jupiter-Heliopolis at Baalbek, a gigantic decastyle structure which took shape during the first half of the first century AD, displays a hexastyle prostyle projecting porch of suitably gigantic proportions (18 × 25 m), enveloped by a uniformly spaced, conventional pseudodipteral ambulatory; there is no opisthodomos (Fig. 4.22). The second-century Temple of Bacchus next to it is octastyle and closely follows the design of its larger neighbor, including an internalized, deep prostyle porch (with six columns across and one in the return), but this is placed within the typically tighter confines of a peripteros (Fig. 4.22).

How relevant are these Syrian developments to Sardis? In so far as creating spatially distinct, deep porches contained within the columnar boundaries of an external peristyle—the shaping of a “space within a space”—the Syrian and Sardian schemes are quite similar. Yet they are also substantially different. Both Syrian temples are podium temples, an obvious sign of their Italian origin, while the Sardis temple, almost certainly raised on a crepidoma of several steps over its platform of mortared rubble, was not what one could classify as a podium temple. The wide, continuous, and uniform-width corridor that circumambulates the deep porches of the Temple of Jupiter effectively isolates it, too. The porch and the corridor exist side by side as separate entities. At Sardis, the lateral north and south pteromas do not enfold the deep porches; in effect, they dead-end against them, leaving the tall volumes captured by deep pronaos porches as separate architectural entities. The easy continuity of space in one scheme is replaced by the blocked pathways and edgy discontinuities in the other. Hence at Sardis, like two conflicting but overlapping arguments in stone and space, the porch and the corridor share each other’s space and engender an ambiguous, juxtaposed, dynamic, and hybrid relationship—a wide pteroma and soaring marble wall set against tall Ionic columns capped by beautiful capitals, achieving a sense of unruly elegance.

Hadrian’s Temples

Surprisingly, one of the closest resemblances to the unique solution displayed by the Sardis porches is none other than the celebrated porch of Hadrian’s Pantheon in Rome (Fig. 4.23).84 Unlike the other Italian temples, which are mainly variants of peripheral schemes, the columnar arrangement of Pantheon’s octastyle porch mimics a pseudodipteral composition. The middle four columns with rows of three in return, define an area of ca. 15 × 15 m and constitute the equivalent of the “deep porch” motif (even though in this case the area was roofed daringly by trusses; the pronaos porch of the Sardis building measures ca. 18 × 13.40 m). The side colonnades, two intercolumniations wide, are comparable to the ambulatories that flank the deep porches at Sardis. A comparison of the plans of the two porches demonstrates that the basic morphology of the two columnar systems is almost identical.85 There are also important differences. The pseudodipterality of the Pantheon porch is only an illusion; the Hermogenean definition of pseudodipterality describes a particular form of columnar arrangement around a cella, not a frontal porch applied onto a building. Another difference is the creation of a true, independent, six-column porch within the larger porch we have at Sardis: this has been achieved by adding two columns (the “pedestal columns”) behind the two central columns of the front colonnade. Without these columns, the Pantheon’s deep porch appears even deeper, but also less differentiated from the lateral ambulatories. In fact, Pantheon’s porch resembles more a basilical arrangement with three covalent aisles rather than a hierarchical order of architectural entities, a spatial core distinctly drawn from its supporting envelope.

At the level of spatial experience, I admit that the similarity between the two designs ceases to be so close and striking. However, the connection was never claimed to be a direct one. Still, it would be hard to imagine that the Greco- Roman architect of the Artemision of Sardis would have been unaware of the great rotunda in Rome—the building was celebrated enough. Furthermore, a circular temple closely and deliberately modeled after it was built on Asian soil not that far from Sardis: the Temple of Asclepius at Pergamon. If my preferred Hadrianic dating of the Temple of Artemis is correct, as construction started at Sardis the great rotunda in the capital would have been the hottest thing on the architectural horizon (or so we might think); it is highly unlikely that the architect(s) of our temple would have been unaware of it.

What we need then is more than a formal similarity of porch plans, but an argument that could furnish credibility to such a comparison at a different level: we need to demonstrate the plausibility of a cogent and particular association between Rome and Sardis. The great decastyle Temple of Venus and Roma comes to mind, another Hadrianic project in the capital whose rectangular cella was divided into two equal, back-to-back chambers, much like the double cellas of the redesigned Roman phase of the Temple of Artemis (Fig. 4.24).86 One must point out, however, that the similarity between the two temples applies to this rare cella configuration, and not to their overall appearance. Recent investigations in the Temple of Venus and Roma indicate that the gigantic temple was not a pseudodipteros as was once believed, but a regular 10 × 22 dipteros with tripteral “forests-of-columns” at the ends, a total column count of 124.87 An academic exercise in Greek architecture, Hadrian’s temple in Rome undoubtedly looked back self-consciously to the great Ionian temples of Asia Minor, especially in its ornament, which had supplied the inspiration for Sardis in the first place.88 In fact, the “un-Roman” nature of the design allegedly occasioned the criticism of Apollodoros, Trajan’s aging architect, with an unfortunate result if the anecdote by Dio is to be believed: Hadrian, stung by this wise criticism, put the architect to death—unlikely.89 So, baldly stated, the architectural gaze here can be said to be not from Sardis to Rome but from Rome to Sardis, at least to Asia Minor in general.

The critical issue with respect to comparisons is not the overall temple scheme; it is the remarkable parallelism between their rarely seen, back-to-back cella designs.90 The problem of the double-temple or double-cella temple (ίερόν διπλοΰν, as Pausanias [2.25.1] described the Temple of Aphrodite and Ares outside Argos)—and as acknowledged by U. Fusco in a preliminary study of the type—is complex and cannot be tackled here.91 Whether the Temple of Aphrodite and Ares actually had back-to-back cellas and acted as a source, either immediate or as one of inspiration, during Hadrian’s visit to Greece and Argos in 123/24 is impossible to say.92 We simply lack solid information on the Argos temple, its date or any physical remains, besides what Pausanias tells us. Suffice it to say, however, that the back-to-back cella arrangement as we have at Sardis is far rarer than the traditional, Etrusco-Italic side-by-side cellas (or the vaulted subterranean cult crypts), which often pass as a “double-cella” type. The sources of the back-to-back cella arrangement anywhere else in the eastern Mediterranean are obscure. A potential earlier parallel to the Sardis arrangement might be the Temple of Apollo at Corinth, whose back-to-back, double-cella division has recently been proposed to belong to the early Roman empire (when the cella was emptied out of its columns), as opposed to the long-held belief that it was an original feature of the Archaic building. This intriguing theory, which has gained support, has as much evidence against it as for it. More importantly, Apollo’s temple was never a two-cult temple; it lacks any programmatic or conceptual links to the Sardis temple to make it a viable and meaningful model at any time (see pp. 201-204).

Temple of Venus and Roma in Rome is possibly the earliest example of a back-to-back double-cella arrangement housing double cults and is the locus classicus. If this is true, the Temple of Artemis at Sardis becomes its closest direct follower. Especially important is the unusual but comparable mode of their programs and ideologies: the imperial mandate to accommodate dual cults of a similar nature. While honoring an ancestral goddess, Venus of the Romans or Artemis of the Sardians (which incorporated the imperial cult), these temples dignify and celebrate the state—in the former through Roma and in the latter through the imperial cult and the deified imperial masters. But in Rome, Hadrian’s renewed emphasis on Roman state religion was presented within the imported context of Hellenic/Ionian architecture; at Sardis, the architecture of a genuine Hellenic/Ionian temple set in a genuine Anatolian/Greek cultural milieu was Romanized. The irony of this inversion is validated by the general tenor of Hadrian’s policies and personality. This may be our outside example—our control, as it were—for the meaningful association we seek to establish between the Pantheon in Rome and the Temple of Artemis at Sardis, or at a broader stretch, for parallel political and historical themes that play out in Hadrianic Italy and Asian Sardis, as co-opted by Hadrian.

We can further our arguments for such a connection by underlining the experimental and inventive nature of Hadrianic architecture as a fitting source of inspiration for the unusual planning of the Sardis pseudodipteros. We can also garner some historical support for our position by recalling Hadrian’s unmatched patronage of Asian cities and personal connection to Sardis: his visit to Lydia and almost certainly to Sardis in AD 123/24, and his awarding of the second neokorate to Sardis, also almost certainly, and consequently, the major rebuilding of the Temple of Artemis—all of which makes a conscious and contemporary architectural kinship between the Pantheon porch and the Artemision porch ponderable, as does the divided cellas and cults of the Temple of Venus and Roma and the Sardis Artemision.

We are confident that the architectural character of the Roman temple, as well as its admittedly limited but important ornament, points strongly to a Hadrianic date and design; it is possible, even likely, that much of the continuing Roman era construction of a building of such colossal scale might have occupied the Antonine period. In fact, while the exact dates of Sardis’s successive neokorate privileges do matter and could be elucidated further by future generations, whether the character of the temple’s remarkable architecture is Hadrianic or early Antonine, a period which famously followed Hadrianic initiatives in art and architecture, hardly does.

  • Fig. 4.20

    Comparative plans of the Temple of Artemis, Sardis, with second- and first-century pseudodipteroi in Asia Minor. Temple of Artemis, Sardis; Temple of Artemis, Magnesia on the Meander; Temple of Apollo Smintheus, Chryse (Gülpınar); Temple of Apollo, Alabanda; Temple of Hekate, Lagina; Temple of Domitian, Ephesus; Temple of Artemis, Gerasa; Temple, Alakapı (southwest Turkey); Temple of Dionysus (later Hadrian-Caracalla), Pergamon. (FKY and DF after Kraeling, Gerasa 1938; Lytellton, ‘Baroque Architecture,’ fig 37, p 217)

  • Fig. 3.75

    Temple of Augustus and Roma, Ankara, plan. (after Hoepfner, Hermogenes 1990, fig 43; Krencker and Schede, ‘Der Tempel in Ankara, Berlin 1936)

  • Fig. 3.76

    Temple of Zeus, Aezane, plan. (FKY/DF after Krencker 1936, in Schede and Naumann 1979, fig ??. OR, Hoepfner, Hermogenes 1990, fig 43)

  • Fig. 4.17

    Temple of Artemis, Sardis, author’s reconstruction of hypothetical east pronaos porch (open to the sky). (Fikret Yegül)

  • Fig. 4.18

    Temple of Artemis, Sardis, author’s reconstruction of the south pteroma. (Fikret Yegül)

  • Fig. 4.2

    Ionic dipteroi of Asia Minor. Temple of Artemis (“second temple”), Ephesus, ca. 334–300 BC; Temple of Apollo, Didyma, ca. 300 BC and later; Temple of Hera, Samos (phase IV/Polycrates); Temple of Athena, Priene, ca. 334 BC. (FKY and DF redrawn after Gruben 1966, fig 260; Coulton,’ Greek Architects’, 27a)

  • Fig. 4.1

    Temple of Artemis, Sardis, plans of temple phases, 1:200. Hellenistic phase, ca. 280–260 BC; Roman phase, as preserved, ca. mid-second century AD; Roman phase, partially restored; Roman phase, fully restored, with hypothetical peristyle. (Fikret Yegül)

  • Fig. 4.10

    Massing comparison of Hellenistic- and Roman-phase temple structures. (Fikret Yegül)

  • Fig. 4.19

    Inscribed statue base of architect Dionysius, Patara. (Courtesy H. İ. Işık, Patara Kazıları.)

  • Fig. 3.72

    Wadi B temple and terrace. (©Archaeological Exploration of Sardis/President and Fellows of Harvard College)

  • Fig. 4.22

    Roman temples with prostyle porches in Italy and Syria; Twin temples of Mater Matuta (left) and Fortuna (right), Sant’Omobono, Rome; Temple B, Pietrabbondante; Temple of Jupiter Stator, Rome; Temple of Hercules Victor, Tivoli; Temple of Mars Ultor, Rome; Temple of Jupiter-Heliopolis, Baalbek; Temple of Bacchus, Baalbek. (after Giuliani 2004, fig74)

  • Fig. 4.21

    Temple of Dionysus, theater terrace, Pergamon. (AvP IV, pl. 31.)

  • Fig. 4.23

    Temple of Artemis, Sardis, east porch (left) as compared to the porch of the Pantheon, Rome (right). (from Yegul, “Temple of Artemis” Byzas 2012, fig 10)

  • Fig. 4.24

    Temple of Venus and Roma, Rome. (FKY and DF redrawn from Stamper 2005, fig 155, based on A. Cassatella and S. Pensabene, in ArchLaz 10, 1990, fig 2, p 54)

Conclusions

Let us evoke the predicament of our unknown Roman architect and the portion of the community involved in the process of design, such as the members of the city council, sacerdotes of the imperial cult and other cults, etc., facing the problem of reshaping, restructuring, and finishing the city’s and the region’s most important and spectacular temple, which had been standing unfinished for three to four hundred years. The program for incorporating the imperial cult into Artemis’s temple, occasioned by the city’s newly attained second neokorate, must have been as exciting as it was economically daunting, technically challenging, and politically and socially demanding, perhaps even confusing.

Certain design decisions had already been made for the architect (the extant single-cella temple, earlier altars, honorific and religious monuments; the limiting conditions of the sanctuary; the limiting conditions of topography and geology, etc.). Since the cella was already established, the height of the Roman peripteral columns, which naturally are related to cella height, must have been pre-determined.93 The architect probably knew or inferred that the unfinished temple had been intended as a dipteros. But a dipteros, requiring some one hundred or one hundred two columns, might have been judged as ruinously expensive and too lengthy to build—and probably not so fashionable then. A hexastyle peripteros like the Temple of Athena at Priene, while technically possible, would have been architecturally illogical; this cella was too large and too long for the hexastyle arrangement (see Fig. 4.2). The resulting design would not have been sufficiently grand, as the original builders clearly wished their temple to be. A grand decastyle pseudodipteros (like the Temple of Jupiter at Baalbek or the Temple of Apollo at Didyma) would have been impossible; the cella proper was too narrow to fit the necessary six columns across (Figs. 4.2, 4.22). So an octastyle pseudodipteros scheme must have emerged as the natural choice.

Moreover, Hermogenes’s architectural heritage, though distant then, was still valued in its own right and used for many significant temples of the Imperial era in Asia Minor. It could be adopted in a general way. To adopt a specific, elaborate version of complex interaxial contraction in the end colonnades (at least on the new east front), on the other hand, was a deliberate, sophisticated, and idiosyncratic touch—self-consciously looking back. None of the Roman pseudodipteroi, not even their Hellenistic predecessors, used the system. Recalling the Archaic (possibly also the Classical) Artemision of Ephesus, the use of complex interaxial contraction was, indeed, a sophisticated and anachronistic footnote in local history: the city was tipping its hat to its Archaic and Lydian past in general, and to Ephesus in particular, as it appears to have been an inspiring mother sanctuary for Sardis. In the archaistic, Second Sophistic language of temple’s talking column, the power of logos and the mystery of mythos actually referred to the era when rich Croesus and his golden city had so lavishly patronized Ephesus, Ionia’s most prestigious sanctuary.

Another allusion to the past was the deliberate stylistic copying of the original Ionic capitals of the first temple in the second, Roman one. Given the eminence and the truly exceptional nature of the models before them, this imitation appears not only logical but inevitable and admirable. The old temple was a beloved and important landmark. Its capitals were perhaps its most distinctive part; they must have been admired for their beauty then as they are now. Roman craftsmen and artists, steeped in traditional and local styles, were willing to and capable of achieving that imitation.94 The results for the most part—as witnessed in the few Roman period capitals that we have, mainly capitals A and B on columns 6 and 7—were so successful that generations of scholars have taken the copies for the real thing (see pp. 121-125). It is highly likely, as Cahill has suggested and as supported by evidence, that some of the best originals, such as capital C, were never put to structural use during the Roman period; instead they were set up for proud display as a kind of model, somewhere near or in the temple, probably protected from the elements, given the superb condition of capital C. Memory lived and memory displayed (occasionally, memory rejected), as it historically does, endowed legitimacy and power to the city and its temple.

Work on the cella—dividing the chamber, moving the walls, building new ones, opening doors, changing some of the interior supports, roof beams, and partial roof, engaging in the production of new decorative sculpture—must have been good but routine work. Laying their foundations and raising the exterior columns, among the tallest and biggest of their kind, must have been more challenging—dramatically illustrated by one of the east-end columns stating with undisguised pride that its base was a massive and heavy monolith paid for with the temple’s own funds, and that it was the first to be finished. We posit that this project was conceived and begun within the second quarter of the second century (as indicated by epigraphic and literary-cultural grounds), and occasioned by the second neokorate honors granted to the city under Hadrian and associated with his almost-certain visit to Sardis in AD 124—with his approval, and possibly his help.

The technical undertaking celebrated by column 4 has received sufficient attention in our discussion here and elsewhere. Now, the financial undertaking that is indirectly celebrated by the same column, by extension, might deserve some emphasis in our concluding discussion. A consideration of the facts and figures related to the cost of building and raising the peripteral columns of the Temple of Artemis might offer a factual explanation as to why this temple (and other Hellenistic behemoths like it) was never finished, and also why the initiation of such a massive pseudodipteral scheme might have been an economic undertaking more in keeping with the relatively wealthier second century AD.

Let us consider some hard facts about the technical and economic aspects of raising these columns. With a total height of 17.87 m (including base and capital), the total volume of marble removed from the local Mağara Deresi quarries (in the mountains ca. three kilometers south of the temple) for one finished column shaft, not counting the base and the capital, would have been about eighty to ninety cubic meters. This represents an aggregate work of cutting, removing, transporting, shaping, and raising roughly 200–240 tons of marble per column shaft. Since fourteen to sixteen (or even eighteen) columns appear to have been finished, this means some 3,500–3,800 tons of stone for this first attempt! In addition to the work and expense required for the rough shaping and cutting of a column shaft, the shaping and removal of marble for the flutes of each column (although in the Roman case this was never done) would have been startling in its immensity. Based on an average flute (ca. 17–18 cm wide and half as deep), the volume of marble to be removed for just one flute is ca. 0.176 cubic meters, which is almost half a ton; for a single column with the standard twenty-four flutes, about eleven metric tons! The total material to be removed just from the flutes of the fifty-two peripteral columns of the Roman pseudodipteros (not counting any of the pronaos porch columns) would have been a staggering 572 tons of stone—hard to imagine unless you do the math—and goes a long way toward explaining why none of the Roman columns were fluted except for their very tops.95

Based on the records of payment kept during the building of the Temple of Apollo at Didyma, the production of a single, unfluted column in material and labor costs was estimated at ca. 40,000 drachmae (ca. $2.3 million today), a figure that would have been close for the Sardis columns (which are less than two meters shorter). This financial outlay is conservatively estimated to be equal to the temple’s construction budget for one whole year.96 With some sixty-four columns to build, it would have taken more than a half-century to complete just the peristyle, not counting other significant building costs, provided that the construction of a column was undertaken each year, and provided that all years were economically good ones.97

The situation might have been barely tenable in the best building era for the cities of Asia, the famously wealthy and ambitious decades of the second century AD, especially under Hadrian, and decidedly less so under the parsimonious Antonines. I believe it would have been an unlikely undertaking to even conceive of during the Julio-Claudian period, when the city was reeling under the devastation of the AD 17 earthquake and must have been forced to use its funds for its primary, numerous civic needs—not wholly restricted to building projects (notwithstanding generous imperial help under Tiberius, and accepting that the situation might have improved substantially by the second half of the century). If Tacitus’s detailed description of the scope of the devastation, physical and financial, upon the empire’s budget caused by the great fire of Rome in AD 64 could be used as a measure, the “great earthquake” of Sardis must have placed an even more devastating burden upon the city. The task of removing the construction debris must have been little known and little appreciated; it would have been taken to receptacles outside the city and probably used at some later date(s) as convenient fill for other architectural or landscaping projects; in Rome, Nero’s efforts on this subject are specifically mentioned by Tacitus.98 The city of Sardis and the temple must have made some good efforts towards repairing any damage the temple might have suffered, such as the roof, but probably did not start a new project of known, mammoth economic proportions.

The Sardis temple in its Roman re-creation followed an unusual, unorthodox, and unique variation of the pseudodipteral scheme. The foremost design motif that takes the Temple of Artemis outside the regular Hermogenean tradition of Hellenistic pseudodipteroi and makes it exceptional is the six-column pronaos porches placed within the peristyle enclosure, described as the creation of a space within a space. If these spaces were open to the sky (though the spatial analysis would hold even if they were not, or if they were roofed with large skylights—a plausible alternative), the dynamic sense of interior space formed by these tall, boxy volumes juxtaposed against the lofty lateral corridors would have been entirely alien to the conservative Hellenistic canon (see Fig. 4.17). With its soaring pteroma, set off by the simple whiteness of an austere, undecorated expanse of marble walls rising some nineteen meters (“shining walls,” as Cyriacus said) and with minimal use of ornament—in contrast to the ornamented articulation of cella walls admired in the work of Hermogenes—the Sardis temple must have had some sense of timelessness in architecture, of severity and permanence enhanced by its natural setting.

We know of no major temple in Asia Minor that parallels such an arrangement and projects such a sense of magnificent, even terrible purity and austerity. What is surprising is that the unknown architect of our temple probably admired but rejected not only the distant original Hermogenean models (then some three centuries old), but also its contemporary or nearly contemporary popular Roman era followers—canonical types with modular plans, uniform column spacing, and academic and correct use of orders that had become the common manner and the serviceable, successful, and standard Roman issue from Ankara to Aezane. If followed, this quintessentially orthodox Roman pseudodipteral model would have allowed for a simple, prostyle tetrastyle porch encased by a uniform and continuous ambulatory, without the awkwardness of the Sardis solution. Indeed, as admired in Hermogenes’s Magnesian masterpiece, though not by its numerous and somewhat mechanical imitations, this ambulatory could have been articulated through decorative enhancement. Recently, concepts of the “visuality of the pteron,” or the “visual feast [provided by] the pteron,” applied to the Magnesian temple were underlined by H. Drerup and O. Bingöl, and to an extent by L. Haselberger (see Fig. 4.16).99 Following the less-articulate but more-available models of orthodox Asian pseudodipteroi, with “four prostyle columns in front of the anta porches, and pairs in antis front and back,” could have been the easier thing to do, avoiding all anxiety of measuring up to distant, articulate, and challenging models.100 Instead, the Roman rebuilding of the Sardian Artemision, while retaining its old, ultimate anchor, must have turned to the West for new design inspiration in Rome and Italy; it is there that we see these spatially articulated, deep porches as integral to the native Italian tradition. We are even able to relate the design particulars as well as programmatic themes of our recreated, double-cella temple to specific Hadrianic projects in Rome. And what emperor and period would have been more attuned to creative experiments and unorthodox distortions in architecture than Hadrian and his times?

Thus the Roman rebuilding of Artemis’s temple at Sardis can be seen as a credible and logical consequence of Hadrian’s almost-certain visit to the site in AD 123/24 attendant on the awarding of its second neokorate (in addition to the Wadi B temple, strongly presumed to be the seat of the first neokorate). The major Roman construction (or reconstruction) of the temple to accommodate the imperial cult must have subsumed much of the second century AD, with little evidence of major work beyond that period except, perhaps, for adding more column foundations along the long north and south peristyles, as attested by their incomplete, roughed-out states. The decision to rebuild and set aside a full half of Artemis’s temple for the imperial cult might not have pleased some, but others, obviously, responded with practical and realistic considerations. In a world mostly tolerant of religious syncretism, it is possible to imagine that Artemis shared her temple with the new gods and old cults and allowed peaceful worship in her sanctuary, either separately or together, in ways and upon altars we may not know.

Architecturally, the Temple of Artemis at Sardis is an experimental, mannered, and eclectic building that is not necessarily the “end” of a popular pseudodipteral system that had been established in Asia Minor by Hermogenes and soon achieved a remarkable prescriptive authority—it had little to do with the academic version of that tradition, as previously indicated. Instead, it was one that refused to follow contemporary and orthodox manners and went beyond, either because of informed and creative design decisions, or sheer historical chance—or both.101 Set in the magical landscape of the Tmolos, there were neither equals to nor clear followers of our temple. Perhaps one should not expect progeny from an exceptional monument shaped by the uniqueness of circumstance and talent—genius does not bargain with its powers.

  • Fig. 4.2

    Ionic dipteroi of Asia Minor. Temple of Artemis (“second temple”), Ephesus, ca. 334–300 BC; Temple of Apollo, Didyma, ca. 300 BC and later; Temple of Hera, Samos (phase IV/Polycrates); Temple of Athena, Priene, ca. 334 BC. (FKY and DF redrawn after Gruben 1966, fig 260; Coulton,’ Greek Architects’, 27a)

  • Fig. 4.22

    Roman temples with prostyle porches in Italy and Syria; Twin temples of Mater Matuta (left) and Fortuna (right), Sant’Omobono, Rome; Temple B, Pietrabbondante; Temple of Jupiter Stator, Rome; Temple of Hercules Victor, Tivoli; Temple of Mars Ultor, Rome; Temple of Jupiter-Heliopolis, Baalbek; Temple of Bacchus, Baalbek. (after Giuliani 2004, fig74)

  • Fig. 4.17

    Temple of Artemis, Sardis, author’s reconstruction of hypothetical east pronaos porch (open to the sky). (Fikret Yegül)

  • Fig. 4.16

    Temple of Artemis, Magnesia on the Meander, west facade, computer-aided simulations of light conditions, approximately on a mid-afternoon in late autumn. Top: Light effects on the temple as a pseudodipteros. Center: Light effects on the temple as a dipteros. Bottom: Light effects on the temple as a peripteros. (Haselburger and Holman 2015, fig. 10; courtesy of the authors.)

Notes

  • 1Butler 1911, pp. 450–51; Sardis I.1, p. 65; Sardis II.1, p. 29; Sardis R1, pp. 78–79.
  • 2Cahill and Greenewalt 2016, p. 498.
  • 3Pre-temple architectural and topographic features naturally would have played an important role in shaping the design, as they do in any architectural project. However, the view that the close proximity of the altar and the temple, and the irregularity of the resulting design, “confounded the architects for centuries and was never satisfactorily resolved” (as stated by Cahill and Greenewalt [2016, p. 498]) may allow some variance. Some Greek temples of the Classical and Hellenistic periods were not wholly orthodox; a tight distance between a temple and its altar, even an immediate physical connection, was not uncommon. For a comparable example, see the Temple of Zeus at Euromos, where there is no space between the temple east front steps and the altar—resulting from a later enlargement of the temple (the altar is Hellenistic), not unlike our situation at Sardis. The relationship between the enlarged Temple of Zeus and its altar at Labraunda in Caria is also similar, providing us with a fourth-century BC example with a front row of columns that came up against an altar belonging to an Archaic temple and its smaller altar. Therefore, the close relationship between the altar and west colonnade of the temple at Sardis, especially in its Roman phase, may be more “awkward or confounding” in our eyes in than the eyes of its builders.
  • 4The frontality of Asian (or Anatolian) Ionic temples, achieved through making the pronaos deeper than the opisthodomos (or simply omitting the opisthodomos altogether, especially common in small temples), has been noted as a general design characteristic. See Akurgal 1978, pp. 20–21, 29–31, figs. 1–4; Şahin 2002, pp. 77–78.
  • 5For Gruben’s idealized and theoretical dipteros with a 20-foot interaxial spacing and 8 × 15 column arrangement on an idealized 300 × 150-foot stylobate (and with the wholesome proportions of 1 : 2, for which he took liberties with his measurements), see pp. 155-157 and Gruben 1961, pp. 184–91.
  • 6The Didymaion is raised on a 3.5 m crepidoma, ascended by seven large steps built over it; indeed, this is a very shabby block-and-rubble platform. The Early Classical Artemision of Ephesus is also reconstructed as raised on a platform (probably a measure against flooding). The Roman period Temple of Augustus and Roma in Ankara and the Temple of Zeus in Aezane both stand on many-stepped podiums. Bammer 1972; Akurgal 1978, pp. 222–30, 268–69, 284–87.
  • 7If the Hellenistic temple at Sardis had achieved its dipteral scheme as intended, its resemblance in plan to the Second Artemision at Ephesus, its putative contemporary, would have been striking. The three-intercolumniation front and back and two-intercolumniation side scheme of Ephesus was later echoed in the Roman era pseudodipteros of Sardis. A major difference would have been the sekos arrangement of Ephesus—a gigantic, open hypaereal center (as in the Temple of Apollo of Didyma), not necessary or desired at Sardis, which had an unusually narrow cella that was easily spanned by a double row of columns. One might even say that the open cella of Ephesus provided a sort of stimulus—at least on technical grounds—to the hypothetical, open pronaos porches of Sardis. The relationship of the pteroma to the cella interior, however, would have been reversed: At Ephesus one descended ca. 1.60–1.80 m into the sekos by steps; at Sardis, one ascended ca. 1.70 m from the east and west ends into the cella. It is intriguing and ironic that the design of the Hellenistic temple at Sardis appears more orthodox than its Ephesian model; the situation is reversed in its Roman era resurrection.
  • 8For a dramatic view of these terraces as they appeared at the end of the 1913 and 1914 seasons, see Sardis I.1, p. 133, figs. 149, 152, and 171.
  • 9Sardis R1, pp. 54–55; Sardis I.1, pp. 130–34, figs. 143–50. See also G. W. Olson, “Field Report on Soils of Sardis, Turkey” (1970, Archaeological Exploration of Sardis).
  • 10B. Marsh, “Toward a Landscape History at Sardis, 2007” (field report, Archaeological Exploration of Sardis).
  • 11Cahill and Greenewalt 2016, p. 494 (for the north pteroma trench of 2010, see pp. 475–79, fig. 7).
  • 12Sardis VII.1, p. 57, nos. 38–39; Cahill and Greenewalt 2016, p. 488; Sardis M14, p. 173, no. 574, translated as “[tile given] by the phyle of Kai[sareios?].”
  • 13Not counting those inscribed directly on the structure, 26 out of 101 of these inscriptions are included in the first relevant categories in Sardis VII.1, pp. 1–102, nos. 3–4, 6–8, 22, 24, 27, 37, 49, 50–54, 58, 72, 85–93. Of these, seventeen are datable to the Hellenistic and late Republican periods and nine to the Imperial period. In addition, there are some sixty or so stele bases and/or other bases for votive monuments and statuary preserved around the altar. Hanfmann assumed that these votive monuments, placed against the walls of the altar or leading up to it in two rows, probably represented a late Roman reorganization; see Sardis R1, pp. 53–73, fig. 181. For the contrary view to the “late Roman reorganization” idea, based on strong new evidence, see Cahill and Greenewalt 2016, pp. 499–500.
  • 14Sardis VII.1, p. 91, no. 85 and Sardis VI.1, pp. 38–39; Sardis I.1, pp. 125–27; Shear 1931, p. 129, fig. 8.See also pp. 208-209 above.
  • 15Sardis VII.1, pp. 9–12, 93–94, nos. 3–4, 89.
  • 16Sardis VII.1, p. 65, no. 49, fig. 40.
  • 17The first king is probably Attalid Eumenes II (Soter), whose dedication is from the spoils of the Battle of Magnesia (ca. 190 BC); the plaque was “probably fastened on a wall in the Temple of Artemis” (Sardis VII.1, pp. 92–93, no. 88). The “second king” refers to a record of Sardians honoring the deceased Antoninus Pius (after AD 161, but it commemorates the dead emperor as he was in 139, at the beginning of his reign); the text was probably recorded on a pedestal, now lost. As suggested by W. H. Buckler and D. M. Robinson in Sardis VII.1, pp. 71–72, no. 58, “the reminiscence presumably refers to some boon then conferred, and this may well have been the granting by him of the second neocorate.”
  • 18Sardis VII.1, pp. 91–92, no. 86; pp. 94–96, nos. 90–93.
  • 19Sardis VII.1, pp. 66–69, nos. 51–54, figs. 42–44.
  • 20Sardis VII.1, pp. 65–66, no. 50, fig. 41.
  • 21Anonymous, “Review of draft text of the Temple of Artemis at Sardis, June 2017,” 2.
  • 22For a scheme portraying such a thoroughly integrated relationship of stairs and columns, see Butler’s fanciful reconstruction proposal. Sardis II.1, p. 86, fig. 97, pl.A (here, Fig. 3.1); Sardis R1, p. 101, fig. 180.
  • 23Cahill and Greenewalt 2016, pp. 480–82, 499–500.
  • 24Among possible parallels for the use of high stairs in association with Hellenistic temples are the peripheral steps surrounding the Temple of Apollo at Didyma, rising an impressive 3.50 m, yet this is not a case of stairs leading up directly to the temple front, but rather a step system surrounding the temple peristyle, a crepidoma, which is not the same thing. The Hellenistic sanctuaries at Cos and Labraunda utilize many stair-and-wall combinations similar in appearance to our proposal, but those are structures linking terraces of different levels; the temples within the terraces are traditionally designed. One might remember the Temple of Dionysus located on the steeply rising hill on the north end of the theater terrace at Pergamon (see Fig. 4.21). Dated to the second century BC, it is an Ionic prostyle temple rising on a podium more than 4 m high, approached frontally by a stair of twenty-five steps. Yet all of the visible structure had been rebuilt in toto during the Roman Imperial period.
  • 25For an alternate solution with symmetrical stairs from the south and north sides leading up to a more formal, ceremonial plaza, see Gruben 1961, pp. 175–78, fig. 3. For this older argument see also Sardis R1, pp. 100–101, figs. 124–25. For a recent cogent view, see Cahill and Greenewalt 2016, pp. 499–501.
  • 26See also the discussion in Sardis II.1, p. 16.
  • 27It may be unlikely that Butler moved stele 15, although he excavated the area quite thoroughly and did remove some mortared-rubble retaining walls and mortared-rubble foundations, and reorganized many other stelae found in this area. The photograph of the west end of the temple and LA 1/LA 2 (looking south) in Sardis I.1, ill. 100, taken at the end of the 1912 season, shows stele base 15 with the lower part of a stele still in situ behind LA 2; it also shows a great many other stele bases, pedestals, and blocks in the area—as if the blocks were in the process of being moved and the area was being tidied (see also Fig. 4.5). The detailed plan of this area as it was freshly dug at the end of the 1910 season (“Plan of Lydian Building and Foundations at the West End of the Temple,” Sardis I.1, ill. 35) does not include this stele, even though this plan seems to be accurate in including almost every stele base placed against the north wall of LA 2 (and that are there now). It is also curious that Butler, who described the excavation of this area as it emerged through the 1910–12 seasons, and carefully included almost every stele and stele base in his description, made no mention of stele 15 as a feature in its present, presumably original position (Sardis I.1, pp. 41–42). Grounds for doubt remain.
  • 28“Misaligned and non-joining clamp cuttings show that these stair blocks were moved from another location… that they were already reused in that prior staircase.” This is the Hellenistic staircase leading from level *96.80 behind LA 2 to *100.0 of the west pronaos porch and putative pteroma. Cahill and Greenewalt 2016, pp. 480–81.
  • 29Sardis I.1, p. 42; Cahill and Greenewalt 2016, p. 500.
  • 30A minor excavation in 2010 disclosed the full extent and construction details of this wall that connected the east end of the northwest stairs to the north pteroma column foundations. Cahill and Greenewalt 2016, p. 501.
  • 31The heights of the risers of the preserved five courses, from bottom up, are: 23, 23.5, 23.5, 22, and 23 cm; the widths of the treads for the same are 33.5, 36.5, 36, 37, and 37 cm.
  • 32This unusual situation baffled Hanfmann, as it did many others: “There is no blinking the fact that the interpretation of the stucco on the eastern wall of LA 2 as stucco suitable for an external wall brings about a head-on collision with the evidence of the northwest staircase. If we accept both assertions, we obtain the peculiar result that a person going up to the western end of these steps would only do so to precipitate himself some five feet into the empty corridor separating structure LA 2 from the present western edge of the temple platform!” (Sardis R1, pp. 101–102).
  • 33It has been suggested, however, that this might have been the location for some second-century AD inscribed marble blocks found at the bottom of the northwest steps in 1911 that honor priestesses of Artemis. While they could have been built, there is no external support for this theory. See Sardis VII.1, pp. 66–68, nos. 51–52 and possibly no. 53.
  • 34Cahill and Greenewalt 2016, p. 481.
  • 35Following Gruben’s lead, those who subscribe to a late Hellenistic second phase are Akurgal (1978, pp. 127–31), who reproduces Gruben’s plans for the temple’s phases unchanged, Hoepfner (1990, p. 7), and more recently Şahin (2002, pp. 70–71).
  • 36Among the scholars who have contributed to the “early versus late” Hermogenes controversy: von Gerkan 1929, pp. 32–33 (150–130 BC); Yaylalı 1976, pp. 106–8 (ca. 150 BC); Şahin 2002, pp. 56–58, 77–85, 87–89 (180–130 BC); Akurgal 1990 (190–150 BC) (in his earlier work Akurgal followed von Gerkan’s mid-second century BC assessment: Akurgal 1978, pp. 20–31, esp. 25); Dinsmoor 1973, p. 274 (200–160 BC); Gros 1978, pp. 687–89, 695 (late third to mid-second century BC). Peter Herrmann had opted for an earlier date, ca. 220–190 BC, supported by an inscription of Antiochus III (223–187 BC) found at Teos in 1965; see Herrmann 1965, pp. 29–32. Others who also favor this date include Hoepfner 1990, pp. 29–30 (end of third to early second century BC); Kreeb 1990, pp. 103–13 (end of third to early second century BC); Stampolides 1990, n. 2; Kadıoğlu and Özbil 2015, p. 6. For a recent summary of the dating question of Hermogenes’s floruit see Haselberger (who also favors a date ca. 200 BC) in Haselberger and Holzman 2015, pp. 376–77, n. 13. For a recent discussion and reassessment of Hermogenes’s involvement in the Temple of Artemis in Magnesia in light of a new donor inscription (of the peristasis columns) datable to the first quarter of the first century AD, see Bingöl and Dreyer 2018, pp. 61–72.The early date for Hermogenes (as supported by Herrmann) was recently confirmed by M. Kadıoğlu on the basis of ceramic evidence found in 2019 in the foundations of the north crepidoma of theTemple of Dionysus atTeos that is datable at the latest to ca. 230–200 BC. Further work is planned. I am grateful to Kadıoğlu for sharing this important information (pers. comm., Nov. 7, 2019).
  • 37I suggest that a few inscribed marble tiles indicating tribal donations dating from the Julio-Claudian era provides evidence for the post-earthquake repair and restoration of the damaged roof (see p. 231, n. 12 above). See also Ambraseys 2009, pp. 105–7; Sardis R1, pp. 55–56.
  • 38Rumscheid 1999.
  • 39I would like to remember fondly many site visits and discussions on this subject with Dr. Bingöl, where we studied the peristasis elements and ornament of the Magnesia temple, judged it to be later than Hellenistic ornament, with a comparative view toward the Hellenistic- and Roman-phase ornament from the Temple of Artemis at Sardis—our judgment at the time independent of the strong evidence provided by the donor inscription mentioned above. See also Bingöl 2012, pp. 113–20.
  • 40In some of the Late Hellenistic temples of Asia Minor, starting with Hermogenes’s Temple of Artemis in Magnesia and later with the Roman pseudodipteroi at Ankara and Aezane, only the center intercolumniation is emphasized by wider spacing; the front, back, and sides are uniform.
  • 41New studies indicate that the plans of the Archaic and Late Classical temples were identical. Based on the state of preservation, it could be ascertained that the locations of one of the two northern rows of columns, both of the southern rows of columns, and the walls of the sekos were matching in both temples. The west front of both the Archaic and Classical temples had eight columns; the east has been traditionally restored with nine columns. As indicated on the foundation plan by A. E. Henderson (1915–16) and recent observations of their traces on the stylobate, the spacings of the outer two columns on the front are 6.08 m and 6.16 m (or 6.13 m); the inner two are interpolated to be 7.25 m and 8.60 m. I am grateful to Aenne Ohnesorg for her suggestions; see Ohnesorg 2012, pp. 23–24; Ohnesorg 2007, pp. 98–103; Ohnesorg 2008. See also Weissl 2002; Henderson 1915–16. For earlier studies of the Ephesian Artemision, where the complex interaxial contractions were thought to affect only three spatial gradations, from the widest at the center at 8.75 m, flanked by a pair of lesser widths at 7.20 m, down to the two pairs at the ends at 6.16 m, see Bammer 1972; Bammer 1984; Bammer and Muss 1996; and, more recently, Bammer 2008.
  • 42Gruben 1963, pp. 81ff.; Gruben 1966, pp. 339–54; Fehr 1972, pp. 14–59; Dirschedl 2012, pp. 46–48, fig. 5.
  • 43If foundations 20 and 22 on the south side had columns, then the likelihood that the empty positions of their counterparts on the north side (19 and 21) had foundations and columns above them gains some strength; a clear west-end alignment across the temple east wall would provide for a technically logical and aesthetically pleasing roof structure for the entire east end. The question was approached by the anonymous reader of this manuscript, who suggested that the bases for these empty positions, such as 19 and 21, “were built and then dismantled, e.g., to allow easier access for the construction of the east end columns” (Anonymous, “Review, June 2017,” 7).
  • 44For an elaborate and detailed list of over forty donors for thirty of the peristasis columns of the Temple of Artemis in Magnesia (all dating from the early Imperial period) studied by B. Dreyer, see Bingöl and Dreyer 2018, pp. 73–75, 83–84.
  • 45In considering the Sardis temple and similar instances of unfinished major projects in Asia Minor, R. R. R. Smith seeks an organic and culturally meaningful explanation for why it would have been “satisfying and effective” to leave the temple in its unfinished state. The financial explanation for this, to which we, too, mainly subscribe, is fine, but it is only “one possible interpretation.” Smith 2015, pp. 478–81, esp. 479.
  • 46Piranesi 1743, frontispiece. On the artistic and intellectual reception of ruins, and the differences among the concepts of “restoration,” “reconstruction,” and “restitution” (and the superiority of the latter over the two former), with special reference to ruins and eighteenth-century architects, see Pinto 2014.
  • 47Harbison 1992, pp. 99–130, esp. 99, 105. See also Jencks 2011, pp. 55–56; Macaulay 1967, pp. xv–xvii, 1–39.
  • 48Besides suggesting similar mortared rubble–backed steps for the Wadi B temple, excavators Ratté, Howe, and Foss aptly refer to the “disappearance of the rubble underpinning for the steps [of the Temple of Zeus] at Aizanoi” (Ratté, Howe, and Foss 1986, pp. 52–53 n. 20); see also Naumann, Krencker, and Schede 1979, pl. 10.
  • 49For studies exploring the different roof constructions of Greek temples, some only partially and awkwardly finished, see Ohnesorg 1993; Schulz 2012b; Schulz 2010.
  • 50The late third–early second century BC Temple of Demeter in the Sanctuary of Demeter at Pergamon, with its simple, narrow facade of two columns in antis, was enlarged by a tetrastyle prostyle deep porch during the mid-second century AD. While the details of the connection between the higher roofline of the original temple and its slightly lower, wider Roman pediment are not known, connecting two rooflines at different heights must have posed problems similar to those of the east porch at Sardis. The new, deep, and spacious Roman porch of the Demeter temple (composed of six columns: four in front and two in the returns) also offers a general comparison to the deep pronaos porch of the Sardis temple, albeit at a much smaller scale—an illustration of the Roman predilection for “space” achieved through grafting deep porches onto existing Hellenistic structures. I hasten to add, however, that the similarities here are superficial: at Pergamon, the tetrastyle facade was a simple frontal addition—a facade that served little purpose other than to project “imperial” power at the expense of (some might say) coarsening the fine Hellenistic facade. At Sardis, the unfinished temple was finished; the deep pronaos porch was not a graft but an integrated element within the peribolos. I am grateful to G. Kökdemir and E. Erten for bringing this interesting example to my attention and for their insightful conversation on the subject. Radt 1999, pp. 180–86, esp. 184, figs. 79, 126, 128; Altertümer von Pergamon IV, pl. 52.
  • 51In an evocative think piece, K. Moe (2017) explores architecture’s broader modalities in changing and moving references and visualizes magnificence as a dynamic set of “flow fields” that define a building in human and material contexts. The unfinished Temple of Artemis, in its exceptional setting and with its rich multiplicity of references, must have appeared virtuous and certainly “magnificent.”
  • 52Haselberger and Holzman 2015. See also Gros 1990, pp. 111–16; Gros 1990.
  • 53As Haselberger (Haselberger and Holzman 2015, p. 379, fig. 6) underlines, the full reconstruction drawing by Thomas Howe of the Temple of Artemis at Sardis is a “painterly” pen-and-ink sketch of the front that emphasizes the deep contrasts of light, shadow, and spatial depth (Howe 1999, p. 206, fig. 11.5; see p. iii of this book). This may be so, however a trained architect has no difficulty in “seeing” space and depth (even shadow and light) in well-executed line drawings and perspectives—witness German Neoclassical architect Karl Friedrick Schinkel’s superb, precise, linear drawings.
  • 54The Sardis pteroma, ca. 9 m wide, 92–93 m long, and rising to some 19 m in height, encloses nearly three times the spatial volume that the pteroma of the Magnesia temple does.
  • 55Favro 2011; Favro, “Reading Augustan Rome: Materiality as Rhetoric” (unpublished lecture, American Society for the History of Rhetoric, May 26, 2016, Atlanta, GA). Inveighing against the too-clean and lifeless appearances of typical paper restorations of Roman environments, F. S. Dunn, in an insightful gem of an article penned over a century ago (with the memorable title “Rome, the Unfinished and Unkempt”), underscored the messy life of the almost-endless construction in real Roman cities, and observed that no city, then or now, was without ruins. Yet these shabby, dilapidated urban ruins that always need rebuilding (vetustate corruptum restituit) appear pale and insignificant in view of the impure vitality of a city constantly in the process of building. It may be viewed as the energy of process against the repose of arrival. Dunn 1915, pp. 318–20.Butcher (2017) discusses the “incomplete, emergent nature” of ancient temples.
  • 56Howe 1999, p. 204.
  • 57See Howe 1999 for an expansion on this description and further discussion.
  • 58For most of us, the visual effect that such a wall could produce is only an abstract notion: marble walls of such dimensions existed in a few colossal temples, but none have been well enough preserved to be experienced. As I write these lines in view of the six-story-high sheer marble wall of the library atrium of the National Gallery of Art’s East Building in Washington, DC (which, at seventy-two feet high, is barely two meters higher than the cella wall of the Sardis temple but has only one-fifth of its length and is without columns), I marvel at the pure visceral effect, the awe (cannot think of a better word) such a physical presence would have induced. There is nothing like it.
  • 59Boëthius and Ward-Perkins 1970, pp. 391–92.
  • 60Yegül 1991a, pp. 345–55.
  • 61The inscription was first recorded in the pages of Captain Beaufort’s travel journal (or log) in 1812. He showed some interest in its “elegiac verse” and transcribed it but offered no translation or comments. It was published in his Karamania in 1817; Beaufort 1817, pp. 4–5. The inscription was fully reproduced again, in reference to Beaufort’s field notes and diary, in the Journal des Savants (May 1819: 261). This was followed by a short reference to it three years later in the Rev. Robert Walpole’s travelogue (Walpole 1820, no. 42). In modern epigraphic publications: Tituli Asiae Minoris II.417; Corpus inscriptionum graecarum III, 4286 and addendum, 1126; Robert 1960, p. 276; Merkelbach and Stauber 2002, p. 40. See also Thomas 2007, p. 91; Foss 1982, p. 180 n. 3; Hellmann 1994; Augusta-Boularot and Seigne 2005, p. 300. More recently, there are simple references to and acknowledgments of the inscription but no attempt to understand it or to connect it even at the most basic level to Sardis and the great Temple of Artemis; see Korkut 2006, pp. 93–97, n. 11; Tacoma and Tybout 2016, pp. 345–89. The inscription is also reproduced in Kaibel, Epigr. Gr., no. 412, where the text of the Dionysius inscription is compared to another one from Patara that uses the same epigraphic formula, datable to the Hadrianic period (no. 356), hence helping to date the former inscription. In interpreting Dionysius’s interesting epitaph, my thanks go to Angelos Chaniotis and Eleni Hasaki.
  • 62Considering Dionysius’s fame in engineering, one could imagine that his involvement at the temple, if at all, would have been centered on the great building’s technical challenges, while others could have been responsible for design. He might have managed the erection of gigantic columns, transportation of heavy stones, and indeed, putting a massive roof over the east end and possibly the new west cella, whose 18.40 m width comes close to the 21–22 m clear span the architect mastered at Patara. Although, the short span of the odeion is 30.60 m, when the geometry of the building—its stage, bowing back, and wall thicknesses—is taken into consideration, and the clear span can be reduced significantly. See Patara II.1, for roof reconstruction, pp. 55–56, 89, 93–96, fig. 123.
  • 63Epigraphic and archaeological evidence indicate a mid-second century date for the second Roman phase of the Patara odeion; excavators suggest a date soon after the earthquake of AD 141; for the renovation of the damaged building with its new, famous roof, see Işık, İşkan, and Aktaş 2011, pp. 62–64; Korkut and Grosche, Patara II.1, p. 73.
  • 64For underlining this development from “Polycrates to Hermogenes,” see Bingöl 2013.
  • 65As described by Dinsmoor in reference to the Temples of Aphrodite at Aphrodisias and of Zeus at Aezane, they were “executed by Greek artists still working in the ancient tradition, and so retain much greater purity of style than that found in most Roman work” (Dinsmoor 1973, p. 277). See also Boëthius and Ward-Perkins 1970, p. 392; Gros 1996, pp. 182–83. For the comprehensive study of the Temple of Zeus at Aezane (Rudolf Naumann’s publication is based on the precise fieldwork and reconstruction studies of D. Krencker and M. Schede in the 1930s) see Naumann, Krencker, and Schede 1979; for the late Flavian dating of this temple, see Jes, Posamentir, and Wörrle 2010. For the Temple of Augustus in Ankara, see Krencker and Schede 1936; Kadıoğlu, Görkay, and Mitchell 2011, pp. 79–108; Görkay 2012.
  • 66Ratté, Howe, and Foss 1986.
  • 67A rare example is the small hexastyle peristyle temple with a tetrastyle pronaos and deep porch located inside the temenos above the theater in Aegae. See plan in Bohn 1889, pp. 36–38, figs. 37, 40; Serdaroğlu 2004, pp. 142, 198, pl. 58.
  • 68Although he left the idea undeveloped, the late Ümit Serdaroğlu had alluded precisely and cogently to the typological affinity between some of the “deep pronaos” temples of Asia Minor and the Etruscan and central Italic temples of Italy, and he perceptively noted the rarity of the “deep pronaos” [“derin pronaos”] type in Anatolia and clearly enunciated this specialty: “Gerçekte Anadolu dışında tripteros ve tetrapteros olarak da uygulamalar vardır; fakat bunlar bir Ege gele-neğinden çok Etrüsk geleneğine bağli düzenlerdir ve karakteristik örnekleri… (Roma ve İtalya’dadır)… Doğrudan doğruya İtalya usulu [“the Italian manner”] olarak nitelemek istediğimiz bu şeklin Ana-dolu’da örneklerine pek raslanmaz [In reality, there are tripteros and tetrapteros applications outside Anatolia; however, these arrangements follow Etruscan traditions rather than the Aegean and their characteristic examples [are in Rome and Italy]… In Anatolia, we have scant representations of this type, which we would like to designate as [temples] in the Italian manner]” (Serdaroğlu 2004, pp. 155–56).
  • 69Serdaroğlu 2004, pp. 155–56, pls. 39, no. 3, 42, no. 1; see prior note. For Termessos see Lanckoronski 1892, pp. 121–26, esp. 79–91; Akurgal 1978, pp. 84–85, figs. 33, 95; Schwandner 1990, p. 93; Radt 1999, pp. 188–93, fig. 132.
  • 70An interesting demonstration of Anatolia’s staunch preference for conservative prostyle or in antis schemes, even when fashioning Roman-style sanctuaries, is the comparison of a group of three parallel podium temples located inside a temenos at Lydai (coastal Lycia, ca. second century AD), which display two-column in antis frontages, to the almost-identical disposition of three, small Capitoline temples in Sbeitla (Tunisia) with prostyle porches (six-column porches with four frontal columns and two on returns), or to the similarly planned three Capitoline temples in Baelo Claudia (Belo) in southern Spain. All of these strikingly similar temple ensembles from distant lands make common reference to the late Republican and early Imperial models of Italy, but they adapt freely from Italian sources. Serdaroğlu 2004, pp. 55–61, pls. 14–15; Gros 1996, p. 154, fig. 174; Mierse 1990, pp. 189–95; MacDonald 1986, pp. 119–20; Duval and Baratte 1973, pp. 23–28; Kirsten 1961, pp. 77–78.
  • 71I gratefully acknowledge John Senseney’s excellent research and ideas on this subject as a part of a seminar he took with me in 1996.
  • 72For Etruscan and Roman Republican temples, see Boëthius and Ward-Perkins 1970, pp. 29–42, 51–56, 108–14, 132–33; Boëthius 1960, pp. 58–63; Coarelli 1974, pp. 279, 281–84; Gros 1996, p. 125, fig. 130; Stamper 2005. For comparative plans of Etrusco-Italic temples of central Italy, see also Gros 1996, pp. 127, 131, figs. 134, 140.
  • 73Even the Archaic Temple II at Sant’Omobono (ca. 540 BC) represents frontal emphasis with a deep, double-column porch. See Hopkins 2016, pp. 66–70, figs. 37–38; pp. 146–52, fig. 118.
  • 74Gjerstad 1960, pp. 168–89; Gjerstad 1962, pp. 151–55; Stamper 2005, pp. 19–33. See also Hopkins 2010; Hopkins 2012, fig. 6.3; Hopkins 2016, pp. 97–122, figs. 86–87, 97. The columnar arrangement of the Capitolium is actually a variant of the Tuscan temple type, described by Vitruvius as peripteros sine postico (3.2.5), or a temple whose side colonnades fade against its back wall.
  • 75For Temple B at Pietrabbondante: Cianfarani 1960, pp. 23–25, pl. 1; Boëthius and Ward-Perkins 1970, pp. 168, 170; Gros 1996, p. 125, figs. 128, 132; Strazzulla 1972; La Regina 1976.
  • 76We simplify our typology in order to identify the lines that are more relevant to the morphology of the prostyle porches of Sardis. One important and popular type, which derives from the Vitruvian-Etruscan model (which we largely ignore), is the small- to medium-size prostyle temple type with a deep and soaring frontal porch and two, three, or even four columns in the returns. Widely diffused in the West, the type includes such well-known examples as the Temple of Portunus in the Forum Boarium, Rome; the pseudodipteral Temple of Apollo on the Palatine, Rome; the Maison Carrée in Nimes, France; the Temple of Augustus at Pula, Croatia; the Doric temple at Cori, Italy; the Antonine Capitolium in Dougga, Tunisia; and the previously mentioned triple temples at Baelo, Spain. In Asia Minor this type is represented, with a few exceptions, in its Greek form, i.e., a tetrastyle prostyle porch, or simply a pronaos with columns in antis. For the great variety that exists among fundamental types, see comparative plans of Republican temples in Pensabene 1991; Coarelli 1988, pp. 205–63; Castagnoli 1966–67; Boëthius and Ward-Perkins 1970, pp. 108–12; Stamper 2005, pp. 34–67, fig. 95. See also useful charts in Gros 1996, pp. 123–33, figs. 134, 140, and extensive bibliography. I leave out of my discussion the Archaic and Classical West Greek temples of southern Italy and Sicily, with their curious proto-pseudodipteral plans and deep, frontal porches set within rambling peristyles. However, we accept the view that promotes the primacy of Italian influence in the formation of experimental schemes, such as the gigantic Temple G at Selinunte or the Temple of Athena (Ceres) at Paestum. Lawrence 1967, pp. 119–29; Barletta 1983; Mertens 1996, pp. 320–22. For a divergent view, see Dinsmoor 1973, pp. 73–105, esp. 75. I would also like to include with thanks N. Cahill’s observation that south Italian and Sicilian temples with deep prostyle porches within a peristyle, as at Sardis, are “set on stairs surrounding the temple in a properly Greek fashion (while) the ‘Italianizing’ examples… are podium temples accessible only from the front,… an important distinction” (correspondence with the author, May 10, 2017).
  • 77For Greek and Hellenistic influences on Late Republican temple design and the Hellenization of the well-known Italian types, see Boëthius 1960, pp. 58–62; Boëthius and Ward-Perkins 1970, pp. 132–40; Gros 1996, pp. 123–33; Giuliani 1984; Lauter 1979; Stamper 2005, pp. 49–67; for temples in Late Republican sanctuaries in Latium, see Coarelli 1987.
  • 78For the temple of Jupiter Stator, see Boëthius and Ward-Perkins 1970, pp. 132–33, fig. 73; Stamper 2005, pp. 53–55, fig. 36. For Italian podium temples and the architectonic advantages of podia: MacDonald 1986, pp. 136–40. See also Castagnoli 1966–69; Colonna 1984. See previous note.
  • 79Giuliani 1970, p. 7; Giuliani 2004, pp. 74–77; Boëthius and Ward-Perkins 1970, p. 166, figs. 157, 202, 206, 232; Coarelli 1982, pp. 77–82; Coarelli 1987, pp. 85–103; Gros 1996, pp. 139–40.
  • 80Boëthius and Ward-Perkins 1970, pp. 190–91; Coarelli 1974, pp. 107–11; Zanker 1968; Gros 1976; Ganzert and Kockel 1988; Ganzert and Herz 1996, p. 47.
  • 81J. Senseney, “The Temple of Artemis at Sardis: Contextual Evidence,” seminar report, University of California, Santa Barbara, 1996, p. 7.
  • 82The irony that the Hermogenean pseudodipteros type, despite Vitruvius’s obviously admiring representation, never gained a foothold in Italy, should be noted. Vitruvius 3.3, esp. 3.3.8; Stamper 2005, pp. 51–53.
  • 83Two other colossal temples in Asia Minor should be mentioned in connection to the Roman phase of the Temple of Artemis at Sardis. The first, the Temple of Zeus and Hadrian in Cyzicus, a close contemporary of the Sardis structure and regarded as one of the Seven Wonders of the World in late antiquity, was an all-marble Corinthian behemoth with 8 × 15 columns, measuring some 92 × 47 m at stylobate level but set on a much larger high platform. Although some suggest that it might have been a pseudodipteros, sketches and a description provided by Cyriacus of Ancona, who visited Cyzicus in 1430 and 1444, do not allow for a clear understanding of its plan. A. Baratollo shows a dipteros with a three-intercolumniation front and back and front and back columns in antis (in an alternative arrangement, the opisthodomos is blank; only the front has a pair of columns in antis). Hypothetical reconstructions include four rows of columns inside the front porch between anta walls, comparable to the arrangement at Didyma. The other colossal building is a decastyle temple known as Dönük-taş in Tarsus, probably also dedicated to Hadrian/Zeus. Recent investigations indicate that the temple was approached by a massive frontal ramp in concrete, as were the podium, the foundations, and the ashlar-faced walls. The cella, too large to be roofed, was raised on a podium of ca. 107 × 50 m; the columns, probably carrying Corinthian capitals, are reconstructed with a diameter of ca. 2.10 m and a height of 19.95 m. It was probably unfinished. Comparisons to the Temple of Apollo in Didyma or the Temple of Jupiter-Heliopolis at Baalbek (whose columns stand at 19.90 m) are apt. Neither of these colossal Roman temples, Cyzicus or Tarsus, however, seems to utilize a plan type at all comparable to the Sardis temple or to feature a deep projecting porch. For Cyzicus, see Dinsmoor 1973, p. 283; Boëthius and Ward-Perkins 1970, p. 392; Ashmole 1956. For reconstructed plan studies of the Cyzicus temple, see Ashmole 1956, pp. 182–83; Ashmole 1959; Bodnar 2003, letter 14.3, diary I.28–43, 73–81; Bodnar and Mitchell 1976, pp. 27–31, figs. 1–6. See also DeLaine 2002; Gülbay 2009, pp. 22, 72–83; Barattolo 1995, figs. 2–3; Perrot and Guillaume 1894; Koçhan 2011, pp. 77–81. For Donüktaş see Koldewey 1890, pp. 178–85; Dinsmoor 1973, p. 283; Baydur 2010; Seçkin and Baydur 2001; Seçkin n.d.
  • 84Boëthius and Ward-Perkins 1970, pp. 256–60; Ward-Perkins 1977, pp. 133–42; MacDonald 1982a, pp. 94–118; MacDonald 1976; De Fine Licht 1968; Wilson Jones 2000, pp. 177–212; Stamper 2005, pp. 184–205, esp. fig. 151; Ziolkowski 1999; Opper 2008, pp. 110–25; Yegül and Favro 2019, pp. 355–70. For a recent but controversial hypothesis attributing the Pantheon to Trajan’s master architect Apollodorus and hence dating the building to the late years of Trajan’s reign, see Hetland 2007.
  • 85Yegül 2012, pp. 107–9, fig. 10; Yegül 2010c, p. 381, fig. 11.
  • 86Boëthius and Ward-Perkins 1970, pp. 265–66; MacDonald 1982a, pp. 129–36; Stamper 2005, pp. 206–12, fig. 155; Barattolo 1978; Opper 2008, pp. 126–27; Howe 1999, p. 204. On the possibility of the existence of architectural awareness and information exchange between Sardis and Italy, consider also that the shaping of the base of column 4 (with the “talking inscription”) as a victory wreath complete with lemnisci was a direct reference to the wreath-base of the Column of Trajan in Rome, the earliest and most prestigious specific example of the type.
  • 87Cassatella and Panella 1990, fig. 2; Gros 1996, pp. 178–80. Another comparable, very large dipteros with triple columnar rows at ends is the Olympieion in Athens. This long-lasting project (started in the sixth century BC) was continued ca. 175 BC under the Roman architect Cossutius and was finally finished by Hadrian. See comparative plan in Stamper 2005, p. 208, fig. 153.
  • 88Amply noted is the style of the ornament of the Temple of Venus and Roma and its use of Proconnesian marble, harkening back to models from Asia Minor. Judging by the similarity of its decorative moldings and profiles to those used in the Trajaneum at Pergamon, an Anatolian architect (and even workmen) may be imagined. The decorative relationship is fully and convincingly discussed by Strong (1953, p. 133). See also Boëthius and Ward-Perkins 1970, p. 266; Sear 1982, pp. 182–83.
  • 89Dio Cassius 69.4.1–5.
  • 90The architectural specificity of a cella physically divided by a wall compared to an undivided cella shared by two deities (as in the Temple of Zeus-Trajan/Hadrian in Pergamon) should, again, be emphasized. A good example of two deities or cults sharing a single cella (“temple sharing”) is the Temple of Roma and Augustus in Ankara, whose handsome yet fantastical reconstruction drawing by Perrot and Guillaume shows the seated statues of Roma and Augustus, back-to-back, facing opposite directions (Fig. 4.25). Perrot and Guillaume 1862, pl. 22; Güven 1998, pp. 36–37 n. 44. Early suggestions that this temple originated as just a cella of Hellenistic date housing the Anatolian cult of Meter Theon, and was later expanded with a peristasis to incorporate the cults of Roma and Augustus, are unsupportable. See Kadıoğlu, Görkay, and Mitchell 2010, pp. 79–91.
  • 91Fusco 2015–16, pp. 7–11. At Letoon (in Lycia), the easternmost of the three Hellenistic temples is assigned to Apollo and Artemis, on the iconographical evidence of a cella mosaic. Its cella appears to have been divided in two during the second century BC (or else this modification simply represents the creation of an adyton, based on the shaky evidence of an internal wall). This is then a possible “double-cult/double-cella” parallel for our temple; further fieldwork is indicated. I express my thanks for this information to Baki Demirtaş, formerly on the Letoon research team (personal communication, Oct. 2019); see paper by B. Demirtaş, “New Observations on the Letoon Temple in the Light of Hermogenean Architecture,” presented at the Karia ve Karialılar Sempozyumu, Milas, August 2018. See also Courtils 2003, pp. 142–45.
  • 92Even if Hadrian had seen the Temple of Aphrodite and Ares in Argos in 124, the Temple of Venus and Roma in Rome (begun in AD 121) might have already advanced beyond its broad platform, establishing its divided-cella plan and its position as the locus classicus of the unusual type by that date. For an early (and mostly outdated) note on the “origins of the plan” of the Temple of Venus and Roma, see Picard 1945. For a loosely defined consideration of temples with multiple cellas in western Asia Minor, see Anabolu 1992.
  • 93Considering that the Roman construction features on top of the Hellenistic anta capitals indicate some overhauling (similar to some of the Hellenistic anta capitals whose top surfaces were recarved)—hence at least the upper parts of the southeast and northeast antae during Roman rebuilding—one could ruminate over whether the Roman design altered the height of the original cella walls—perhaps unlikely, certainly unknowable (see also pp. 136–138).
  • 94I have elsewhere argued that certain decorative motifs of the Ionic capitals of the Severan Marble Court in Sardis were also modeled after the temple’s capitals; see Sardis R3, p. 137 n. 13, fig. 172.
  • 95I owe this idea about the effort expanded in just fluting the columns and these related statistics to Amanda Claridge, who some time ago calculated the figures for the columns of the Temple of Hadrian in Rome, a project contemporary with that of Sardis whose columns are about three-quarters the size of ours; fluting the marble of one column shaft of the Hadrianeum would have been at just under seven tons to our eleven tons; Claridge 1983, p. 119. The “ton” we use here is the weight equivalent of 1,000 kg, or 2,240 lbs, sometimes called the “metric ton” or the “long ton.”
  • 96Converting the figure for Didyma to its near-equivalent in modern Germany for the year 2000 (at 100 euros per day per worker; the euro officially went into use on Jan. 1, 2002, but it was officially adopted on Jan. 1, 1999). G. Gruben estimated the cost of each Didyma column to be around 2 million euros (more now); Gruben 2001, pp. 5, 406. See also Voigtlander 1975, pp. 74–82, 92–102; Bingöl 2004, pp. 150–62, esp. 161–62; Yegül 2014, pp. 219–20.
  • 97The study of costs undertaken above is limited to materials, labor, and transportation costs of the Artemis temple’s peripheral columns. Other expenses—such as the massive foundations for these columns and the skilled labor that carved the ornament (bases and capitals)—have not been considered. A thorough economic study of the temple’s construction must be reserved for a future study, one that could benefit from the excellent research on similar subjects collected in a volume focused on Hierapolis and southwestern Anatolia (Ismaelli and Scardozzi 2016). The cost study of the “Marble Stoa” at Hierapolis by D. Maschek could be singled out in this collection as a relevant model (Maschek 2016).
  • 98Tacitus, Annals 15.38–43. See also MacDonald 1982a, pp. 25–31.
  • 99For the light manipulation of the Temple of Artemis at Magnesia, see Drerup 1964; contra Wesenberg 2012. For an analysis and a development of the notion of Hermogenes’s decorative, or visual, pteron, see Bingöl 2013, pp. 107–13. For a sophisticated argument on the dramatic lighting and spatial qualities of the same temple, see Haselberger and Holzman 2015. One would also like to point out the conspicuously decorative treatment of the cella walls of the pseudodipteral Temple of Zeus at Aezane.
  • 100I owe the inspiration for the concept of “anxiety in measuring up in the arts” from the literary parallels of the process of imitation and influence from Harold Bloom’s critical early book, The Anxiety of Influence: A Theory of Poetry (1973; 2nd ed. 1997). See also C. Benfrey’s review (2015) of Bloom’s The Daemon Knows: Literary Greatness and the American Sublime (2014).
  • 101I am in agreement with Howe, who described the architecture of the Sardis temple as “not so much a late continuation of the Hellenistic tradition… inherited from Hermogenes, as the end of that tradition” (Howe 1999, p. 210), but by maintaining the view that our temple was not an “end” of that tradition because it never was a part of it, I prefer to place this finer point on it. It is instructive and ironic that over half a century ago, H. Drerup summed up Hermogenes’s masterpiece pseudodipteros in Magnesia in similar terms: “Hermogenes and his pseudodipteros was not the beginning of a development [Archaic and Classical] but its conclusion” (Drerup 1964, p. 19).
  • Fig. 3.1

    Hypothetical restoration by Butler of the west end of the temple, incorporating the “Lydian Building” (LA 1 and LA 2). (Sardis II.1, ill. 97.)

  • Fig. 4.21

    Temple of Dionysus, theater terrace, Pergamon. (AvP IV, pl. 31.)

  • Fig. 4.5

    Area between west porch and LA 2, with steles and stones on the northwest steps in the process of being reorganized by Butler, 1913. (Princeton University, Butler Archaeological Archive, research photograph A.54.)

  • Fig. 4.25

    Georges Perrot (1832–1914), Temple de Rome et d’Auguste, Coupe longitudinale restaurée. (From Georges Perrot and Edmond Guillaume, Exploration archéologique de la Galatie et de la Bithynie (Paris, 1862), pl. 22.)