The Arab author Ibn Shaddad Al-Halabi mentioned a single tower, Burj-Al-Yatim, existing on the site of Zefat before the Frankish conquest. This same source reported that the Franks initially fortified the site in 1101/2 CE (AH 495). King Amaury of Jerusalem handed the site in 1168 over to the Templars, who were still in charge of the castle, when Saladin besieged it in 1188, a year and half after the Frankish army was routed out from the battle of the Horns of Hattin. Following the Frankish surrender, the fortress remained in Muslim hands until the beginning of the thirteenth century. Sometimes between 1218 and 1220, the Muslims undertook to dismantle it, fearing that its falling into Frankish hands will enable them to regain control over the Galilee. The Franks did gain some of their former possessions, including Zefat, after an agreement with the Emir of Damascus in 1240. This episode is the most famous and best known event of the site's history, due to the testimony of the bishop of Marseilles, Benoît d'Alignan, to whom a Latin source from 1264, De Constructione Castri Saphet, which provides a detailed and well-informed description of the site, is attributed. The Mamluks under the command of Baybars besieged the fortress in 1266, which surrendered after six weeks. Soon after the conquest, in 1266–1267, the Mamluk sultan carried out construction works, in which he himself participated. The site was preserved as a military post, first under Ottoman rule from 1516 onward, until the beginning of the nineteenth century. Then, it served as a base for the revolt against the Turks during the eighteenth and nineteenth centuries; it was even occupied for a while by a contingent of Bonaparte’s troops in 1799. The fortress was partially ruined in several earthquakes throughout its history, but that of January 1st, 1837 laid it to waste. As of this date on, the stones of the fortress were used for rebuilding the city, causing its nearly complete destruction.
The excavation was concentrated on the southwestern sector of the fortress, although the entire southern side of the site was cleared by mechanical equipment.
The Frankish Remains
The Inner Rampart. Two units of walls that could be interpreted as part of a curtain wall were uncovered. The first unit was a rectilinear wall (W200; width 2 m), oriented north-northwest–south-southeast and extending over 25 m (Fig. 1). The two faces of W200 were composed of carefully cut, average-sized stretchers, attached to a masonry core. Five niches (I–V; width 1.8 m, depth 1.2 m) were integrated into W200, each containing an arrow loop (slit width 9 cm). Niches I–IV were completely cleared and Niche V was identified by the slit of the arrow loop. Their preserved height did not allow for the restoration of the roof. The threshold of the arrow loops was a parapet, set at a level slightly higher than the floor of the niche. It always had a deep and strong tilt, providing a firing angle directed at the base of the curtain wall. The absolute elevation of the floor in Arrow Loop I was somewhat lower than that of Niches II–IV, wherein the loop holes had an even elevation. The slit axes of the first and second arrow loops were 9.7 m distant from each other, whereas the four others in Niches II to V were set at a regular interval of 4.5 m. Niche III, which had been the best preserved one (Fig. 2), was subsequently sealed.
The second unit (W109) was situated immediately south of W200 and had no direct architectural connection to it. Wall 109 was imposing, preserved over a distance of 6 m. It was built in the same technique as that of W200; one of the stretchers in its facing bore an axe-shaped mason’s mark. However, W109 differed from W200 in two respects: First, it was 3.2 m thick, unlike the thirteenth-century sources that designated a width of 1.5 canna for the top of the inner rampart (Cypriot canna equals 2.2 m; Provençal canna equals 1.95 m); the evidence here may point to the use of the Cypriot canna. Second, the layout of the wall was definitely curved. Despite the systematic stone robbing in the southern part of the site, the masonry core of W109 survived rather well and enabled the satisfactory restoration of its curve. Running parallel to W109 across its inner face was another curved wall (W210). The continuation of this wall eastward may have been another curved wall that was supported by small buttresses (W300; Fig. 1). It was excavated in 1951 by M. Dothan and had a paved passageway with a drainage water channel alongside its southern facing. The space between W109 and W210 formed an enclosed thoroughfare (S9) that accessed a niche (Fig.1, A) within W109, which had an arrow loop with a tilting threshold; the niche was subsequently sealed. The staircase at the northern end of S9 ascended into a room to the north (S2) that was partially excavated, although its original plan was not identified. At a later stage, Room S2 functioned as an entrance into the Mamluk tower-gate (see below), the facade of which clearly evidenced two different phases of construction (Fig. 3). The first construction phase comprised W109, which formed the foundation course of the entrance, up to the level of its threshold. The wall slightly sloped northward in this section, disappearing into the masonry of the tower-gate; here was the only excavated place where the glacis nature of W109 could be observed. Remains of plastered ribbon-like joints (width c. 2 cm) were preserved on its facing and a mason’s mark was engraved on one of the stretchers, showing two small triangles joined at their top, much like an hourglass. During the second construction phase, W109 had been partially dismantled when the tower-gate was adhered to its western side and was cut through to accommodate the entrance (W122) from the tower-gate into the enceinte.
Exploring the area near the outer facing of the western curtain wall (W200) partly revealed the remains of a hall with an ogive-groined vault (S8; Fig. 1). Its architecture showed fine ribs that compared to those of the eastern entrance to the enceinte of Caesarea or, geographically closer, the castle of Montfort. The location of Hall S8 in the space between the two sections of the inner rampart (W109, W200), as well as its architectural features, such as the plastered ribbon-like joints on its facing, testify to a Frankish construction between the years 1240–1266.
The preliminary mechanical removal of debris from the area uncovered the tops of walls of several structures within the boundaries of the curtain wall. These consisted of a succession of rooms, positioned at a fairly straight alignment from west to east. The most complete and interesting quadrangular room (S12) was originally roofed with an ogive-groined vault, attested by the springs of the arch ribs that rested on consoles in each corner of the room, in various states of preservation. The best-preserved example in the southwestern corner (Fig. 4) was composed of ribs that were set on a conical base decorated with a lily and surmounted with a leaf-shaped decoration. This decoration recalls the architecture of the Templar castle at Tortosa (Tarţūs), located on the Syrian coast and dated to the thirteenth century. The northern wall of Room S12 had an opening toward the western curtain wall (W200). To the east of Room S12 was another room, apparently roofed with a barrel vault and oriented somewhat differently (S15). The inner facings of these two rooms were covered with plastered ribbon-like joints. While mechanically clearing the space of Room S15, a carved head was discovered (Fig. 5). On the basis of style, and comparisons with similar heads from Belvoir and Nazareth, this sculpture can be dated to the twelfth century. However, a protrusion in the nape indicates it was used as an architectural element, to be compared with sculpted heads that served as wall decorations at ‘Akko and Tortosa, or used as consoles in ‘Athlit (Château Pélerin), and are dated to the thirteenth century.
Finally, a cistern (S14) to the south of Room S15 was characterized by the thick hydraulic plaster coating on its northern and eastern walls (W204, W211).
The Mamluk Period
The Tower-Gate. The massive quadrangular tower (15 x 20 m) was exposed to most of its height, although its foundations were not reached. Its western wall (W107) was never completely buried and appears on a drawing in the Survey of Western Palestine (Fig. 6; SWP I, 1881. Sheet IV, between pp. 248–249). The tower’s walls (thickness 7 m) were faced with large blocks, laid as stretchers against a thick masonry core. The facing blocks displayed a distinct embossing that slightly protruded from the chiseled frame. The general profile of the facing was that of a glacis. A deep probe dug into the fill of the southwestern corner in Room S1 revealed an inner stepped facing below the level of the floor. The fill consisted of a single homogeneous layer of earth and gravel (height 2 m) and appears to have been contemporary with the construction of the tower-gate. The finds in this fill were dated to the Mamluk period. The remains of two broad openings were visible in W107 and W108. They were undoubtedly used as windows, providing light to the inner room of the tower (S1); however, their shutting system was completely missing. Arrow loops were most probably set in these windows, although their preserved height does not furnish the necessary evidence. Latrines were located in the southeastern corner, at the joint with the facing of the curtain wall (W109) from the previous period. The latrines were connected by a rectangular conduit that reached the upper floor, as indicated by its preservation. At the northwestern side of the tower, the threshold and holes for the lower hinges and the casings of an entrance were preserved, denoting the use of a double door. A quadrangular deep and plastered hole was visible next to the eastern casing. It was most likely intended to accommodate a sliding wooden beam that would have locked the door. However, no evidence of a similar installation existed in the opposite wall. The upper parts of this entrance, including the lintel, were destroyed; yet, the partial preservation of the lower sections of the intrados suggested that a ribbed vault covered it.
At the end of the Mamluk period or at the beginning of the Ottoman occupation, installations for craft activities (furnaces, hearths) were built on top of the tower-gate floors; one of the installations sealed the gate entrance in W109. Italian Majolica of the Renaissance era appeared among the finds from this phase (sixteenth century).
The access ramp. Visitors approached the tower-gate through a ramp (S4; length 24 m; width 7 m to the north to c. 8 m to the south; Fig. 7) that was accessed via an entry at its northwestern corner. The presence of casings, hinge holes and holes for a barrier evidenced the closing system that used a double door. The ramp sloped down to the northwest and consisted of stairs that were separated from each other by paved or earthen floors. Pillars were set at regular intervals at the inner face of the ramp's western wall (W100). Between them were oblique arrow loops within niches (1, 2). The northernmost and best-preserved Arrow Loop 1 was oriented toward the gate of the ramp, to defend it (Fig. 8). Arrow Loop 2 was inside a niche, partly sealed by a later phase, which covered a trench that served for water drainage. Only the southern side of the slit in Arrow Loop 2 was partially preserved. The length of W100 enabled to visualize two more arrow loops inside niches, although the state of preservation was too poor and precluded reconstruction. The ramp was equipped with water drainage on both sides of the thoroughfare. A corridor of unclear function was set along its northeastern half (S4bis). It had a beaten-earth floor in its original state, which sloped up from south to north, contrary to the north–south slant of the access ramp, reaching a podium at its northernmost end. At a later phase, still within the Mamluk period, Corridor S4bis probably served as a latrine.
Additions and modifications were noted in various places of the ramp. These included a reduction of the original entrance to the ramp by the addition of a new wall at the northern side (W132), against the inner facing of W159, the doubling of the third pillar (W127), the thickening of the outer facing of the corridor's wall (W150) and the building of a supporting low wall (W151). These changes imply the installation of roofing over the area. The floor of the ramp slightly sank 0.2–0.3 m at the joint with the threshold of the tower-gate. These observations, as well as several cracks that were clearly visible in the structure of the tower-gate are evidence of damages caused by seismic activities. Therefore, the alterations to the ramp's construction may be understood as repairs after an earthquake.
During the preliminary mechanical clearing of the area, a large square limestone block was uncovered at a destruction level, opposite the gate of the ramp (Fig. 9). A prancing lion is carved on one of its faces, with the right foreleg raised and part of the head missing. The block's position and its comparison with other remains indicate that it may be interpreted as part of a lintel, decorating the entrance to the access ramp. Originally, this lintel bore two heraldic lions, which were the symbol of the Mamluk ruler Baybars.
The fills that almost entirely sealed the preserved architecture of the ramp yielded finds dating to the fourteenth and fifteenth centuries, down to the end of the Mamluk period. The most interesting finds were retrieved from the fill inside Corridor S4bis. The large quantity of pottery and glass finds that consisted of many restorable pieces provides a reference group due to its diversity and quality. The red fabric of the plain ware, which is characteristic of the Mamluk period, is associated with large footed dishes decorated with underglazed incised motives (green and yellow Gouged Ware). Various imports from Syria, Egypt, Italy (Venetian), China (Celadon and Ming) and Spain are present together with local productions. The quality of the assemblage and the absence of kitchen ware denote a rich table set that belonged to a high-ranking person, perhaps the Mamluk governor of the fortress. A small hoard of 37 silver coins that was found together with this pottery deposit included 19 Mamluk coins and 18 Venetian coins. The minting dates suggest the hoard was hidden in the second quarter of the fifteenth century.
The partial excavation of the ramp floors and that of the corridor also yielded finds from the same period. Some potsherds that derived from the fill below the corridor's floor were clearly of an earlier, thirteenth century date.
The Circular Tower. A big tower was partly uncovered during the preliminary mechanical clearing of the site (W400; Figs. 1, 10); its excavation was completed manually. The stratigraphy of the tower could not be studied due to the excavation carried out in 1961–1962 by A. Druks. The latter’s work revealed an entrance and a portion of a corridor. A circular cistern (diam. 10 m; height 10.5 m) was located in the center of the big tower. The tower was most probably built by the Mamluk ruler Baybars. No Frankish sources mention a big tower that would have certainly constituted the largest of its kind in the military architecture of the Crusaders. The tower remains match relatively well the ‘splendid tower’ described by Al-Ottmani at the end of the fourteenth century and attributed to Baybars. In addition to the measured diameter, which agreed with the 70 cubits mentioned by this writer, he also described a ring-shaped inner corridor that is also present in the tower. The negative traces of the tower's original floor are reminiscent of the spiral structure described in the text, as do the stone courses that are slightly slanted and match the structural details in the text. Finally, the cutting of the blocks that compared with that of the tower-gate and its access ramp, are well identified as a Mamluk achievement of the second half of the thirteenth century.
The Ottoman Period
Remains were attributed to the Ottoman period on the basis of their relative association with medieval remains, such as the big circular tower, as well as their building techniques; they were revealed on the upper level of the site, clustered around the big circular tower. The most evident were Walls 500–503 (Fig. 1) that partly abutted the southern side of the tower. The walls were part of a large building, set against and above the ruins of the big circular tower.
Architectural remains of the Frankish fortification from the twelfth century were not identified in this sector of the site; the piece of sculpture is, for the time being, the only artifact predating the year 1188. Two sections of the inner curtain wall were clearly identified, although their architectural relations are not fully understood at present. Their layout fits well the plan of a large concentric castle, as described in the sources of the thirteenth century. Notwithstanding the difference in the thickness of the walls, both show the same building technique and the same type of arrow loops inside niches. The association of the niches with the arrow loops appears to be neither an invention of the thirteenth century, nor of the Frankish period. Examples already occurred in Byzantine constructions of the tenth century in Syria. An enclosed thoroughfare would have been built, at least in the southern section of the inner rampart. From the twelfth century onward, it became a regular feature of Frankish architecture. Early samples of enclosed thoroughfares also existed, as in the castle of Ukhaydir in Iraq, dating to the nineth century, or in the Templars dungeon of Tortosa, built in 1169. However, these two samples included simple arrow loops set in the outer wall of the gallery. The curtain wall of the Cairo enceinte, built by Saladin in the second half of the twelfth century, is characterized by the association of the enclosed thoroughfare with the arrow loops inside niches; it could be regarded as a precursor of the Zefat Fortress. This system seems to have been taken over by Frankish architecture of the Holy Land only during the thirteenth century. The lack of historical data or archaeological evidence, attesting to an Arab building campaign from 1188 to 1240, indicates that the two curtain wall sections of the Zefat castle should be dated to the rebuilding of 1240, as their construction technique is in good agreement with Frankish architecture of the middle of the thirteenth century. Plastered ribbon-like joints, used frequently in Frankish constructions, are known from Mezudat Gadin (Qal‘at Jiddin) in the Galilee, dating to 1220–1230 and from the enceinte of Caesarea, as well as from the Ayyubid citadel of Damascus, dating to 1210. The large vaulted room (S5) is difficult to assign to any given phase of the fort, as it is integrated between the external western facing of the curtain wall (W200) and the access ramp of the tower-gate complex. At any rate, it cancelled the arrow loops inside the niches of the inner rampart for a length of at least 42 m.
One of the most convincing results of the excavation was the evidence showing a significant Mamluk building campaign, immediately following the conquest of the castle by sultan Baybars. The dating of the tower-gate complex and its access ramp is clear, yet the choice of its location probably depended on the position of the Frankish remains. The hall with the ogive-groin vault (S8), which was inserted in the space between the two sections of the inner curtain wall, could be interpreted as the remains of a gate from the Frankish phase. In this case, the Mamluk architects would only have to modify an existing plan. The additional constructions could have been set mainly on the inner moat, which, in this sector at least, would have been filled. These Mamluk constructions were very soon destroyed by an earthquake, probably that of the year 1303. Despite attempts to consolidate them, the maintenance of this complex appears to be increasingly neglected throughout the fourteenth and fifteenth centuries. Its original functions were abandoned at the end of the fifteenth or the beginning of the sixteenth centuries, replaced by some craft activity in the Ottoman period.