Along the Indus River and on the plateau and escarpments of the Salt Range in upper Pakistan, a sequence of venerable stone temples is preserved (Fig. 1). These temples have been little studied, but are significant for understanding the evolution of north Indian temple architecture. By our chronology, presented previously in Expedition (Meister 1996), the earlier of these temples, built in the 6th-7th centuries AD, follow many architectural and constructional conventions of Buddhist architecture in the Gandhara region prior to the 5th century AD. They also, however, display a local adaptation of the unique curvilinear tower used for Hindu temples across north India—providing a missing link in the evidence for its evolution. This distinctive form of superstructure is called Nagara by scholars (Meister et al. 1988, 1991).A second group of temples. with upper chambers in the superstructure (Meister 1996), extends this regional school in the 9th and l0th centuries in the period of the Hindu Sahi. kings. This evolving regional school adapts conventions of Nagara architecture to the local craft traditions of the Indus region.
Our recent excavations at one of the more significant of these sacred sites, Kafirkot, demonstrate that existing temples at the site were appropriated in the 9th century by Hindu Sahis, reconfigured into more complex structures, and reformulated in their pattern of worship. These excavations also recovered a previously undocumented temple of the earlier period that was likewise extended and enlarged in the Hindu Sahi period (Fig. 2).
Background
Scholarship has often relegated all these Salt Range temples solely to the period of the Hindu Sahib kings, who ruled the region from the 9th to the early 11th century Al) (Rehman 1979), and they were often associated with the politics of neighboring Kashmir (Hark i986). Based on our new research we now know that one group of temples was built before the time of the Hindu Sahi kings and that in only the one instance of the l0th century temple at Malot did architects of the Punjab mimic the distinctive pyramidal form of Kashmiri architecture (Meister nod. b). The school as a whole shows a continuing ex presentation with the curvilinear Niagara formula of north India, as occasionally noted by previous scholars (Lohuizen-De Leeuw 1959; Hark 1986), providing otherwise missing evidence for the evolution of this form.
For this development of Nagara architecture in the Punjah region and along the Indus, we have proposed a chronology (Meister n.d. a) that begins in the 6th century AD and ends in the early 11th century, when the Islamic forces of Mahmud of Ghazni captured the fort at Bandana, bringing the region under Islamic control. Recent excavations at the hill-fort of north Kafirkot (Fig. I), where some of the earliest of these structures survive, carried out under the auspices of the Pakistan Heritage Society in Peshawar and the University of Pennsylvania, have provided archaeological evidence that has reinforced and refined our chronological frame. The most revelatory discovery is the previously unknown 6th century temple (Fig. 2).
The First Season’s Excavations
In our first season at Kafrkot (1996) we cleared the mound of rubble covering the eastern face of the structure labeled Temple C on the Archaeological Survey’s site plan (Archaeological Survey of India t921-22), exposing the moldings of the high platform on which the temple stood (Figs. 3 and 4). The long stairway forming the approach to the platform resembles those leading up to Buddhist stupas (shrines) in Gandhara. Two walls framed a chamber in front of the sanctum, fronted by a pair of large round pillars made of kanjur (a porous form of sedimentary stone used in Gandhara). This form of antechamber suggests the formula found much earlier at the Parthianperiod Prandial temple at Taxiia, ca. 1st century AD.
We discovered as well, however, that limestone masonry (cut as square blocks called ashlar) had been used to extend this platform on either side the entry stairway, mimicking the earlier moldings. This extension made a larger pillared hall possible. We found rectilinear limestone pillar bases had been placed in front of the two round kanjur pillars and had also been inserted along the moldings of the walls that framed the temple entry. Similar limestone pillar bases were also found jumbled where they had fallen from the edge of the platform—in the trench below. Wooden beams were used to frame a sill for a new doorway to the sanctum, which was later filled in, perhaps to prevent the temple from collapsing (Fig. 5). The stones of the entry hall show severe discoloration, suggesting an ancient fire, and layers of ash were found elsewhere at the site. We are not ready yet to write of “Flames over Kafirkot,” to mimic Mortimer Wheeler’s famous title (1968), but it is apparent that a major conflagration brought down many structures in this fortress.
That archaeological evidence can confirm two phases for this structure, corresponding to pre–Hindu Sahib and Hindu Sahi periods, was significantly reinforced by the results of a second season.
The Second Season’s Excavations
In a second season (t997) we first continued exploration of Temple C at Kafirkot. We exposed part of a second-phase limestone wall that formed a compound connecting the temple to a ruined two-story kanjur structure on axis to the east, locally called the Mari (“ancient ruin”). The Mari once served as a gateway to the east, and perhaps as a resthouse for pilgrims. We located entry stairs to the compound on the north and a pilastered cell in the wall on the south. In the vicinity of the Mari, limestone walls had been added above kanjur ones, and a thick layer of ash suggested a serious fire had once burned out the wooden beams and floor of the second story of this structure. We also determined two construction phases—one of kanjur and one of limestone—in a square pedestal set between the temple and the Mari that would once have supported a free-standing standard-bearing pillar (Fig. 6).
Most importantly, when we made a trench to separate the kanjur masonry of the temple’s first- phase base platform from that of its limestone addition, we revealed to the north of the entry stair a completely preserved “Indo-Corinthian” pilaster. Its volutes and acanthus leaves (Fig. 7a) are in a detailed way reminiscent of Indo-Corinthian capitals dating many centuries earlier (Fig. 7b) such as those excavated at the Bactrian Greek city of Ai-Khanum (Wheeler 1968). At the floor level of the expanded compound, in front of the temple stairs, we also recovered a coin of the Hindu Sahi ruler Samantha of the late 9th/10th century AD that suggests the period of the temple’s second phase.
Discovering a New Temple
In this second season, we also began a new excavation in a depression to the south of the building called Temple A. We were led to this site by the plan of the fort made by A. Hargreaves early in this century (see Fig. 3). Some physical evidence for a buried structure remained, as well as of a trench that had been cut in from the north, then filled with rubble, suggesting the spot had once been plundered. After clearing the ground, we began a new trench along the south edge of what seemed to be a buried platform, quickly exposing a low pilastered wall made of kanjur, with fallen pillar bases, shafts, and capitals jumbled in the trench. We followed this pilastered wall to its southeast corner, and again we found that limestone masonry had been used to extend the platform, mimicking the kanjur moldings.
We exposed the full east face of this expanded platform and the elegant white limestone stairway at its center. Here, as at Temple C, we recovered coins of the early Hindu Sahi period at floor level in front of these stairs. This newly excavated temple at Kafirkot we have called Temple E (Fig. 8).
Our lead archaeologist, Abdur Rehman, appropriately questioned whether we had a case of two phases or simply a change in plan, and sug gested that we needed to find out whether a second stairway was hidden below the one we had uncovered. As with Temple C, we began to separate the kanjur platform from the rubble remains of the limestone extension along its eastern face. We did find a second stairway in kanjur preserved below the limestone one.
More startlingly, however, we found two sub-shrine cells filled with rubble set into the platform to either side of the older stairway (see Fig. 2). These have a fronting vestibule, like a narthex, set between the trefoil entrance and the interior cella. A square and a circular pillaret placed to either side support the trefoil entrance-arch and foliated half-arc patterns (candrasalas) typical of Nagara decoration. Beyond this entry hall is the plain doorway to the small sanctum.
The rather grand entrances to these small subshrines may perhaps reproduce the entrance used for a larger hall enclosing the temple above that doesn’t survive (Fig. 9). A much simpler trefoil doorway does survive as the entry to the masonry hall of an 8th century temple at Mari-Indus, a site in the area, and there are cinquefoil entries for two 9th/l0th century temples at Amb that demonstrate the continuity of this local tradition (Mumtaz and Siddiq-a-Akbar 1989; Meister 1996).
The central tower of Temple E above this platform must once in many respects have resembled Temple A still standing to its north (Figs. to and 15). Temple A preserves an early experiment with the curvilinear tower typical of Nagara architecture. This was set within an ambulatory hall, of which only sockets for the beams of its roof remain. The subshrines embedded in the front face of the platform of Temple E. however, suggest a ritual function not typical of Hindu structures elsewhere. In our excavation of Temple C we did find two deep unornamented niches on the east face of the kanjur platform (see Fig. 5). At the site of Bilot (Meister t996: fig. i6), two east-facing sancta sunk into the platform for the main early temple there were replaced at a later period by the addition of curvilinear subshrines facing each other to the north and south on the platform above. None of these, however, has the architectural presence of those found on Temple E at Kafirkot.
Further excavation exposed an extensive compound built during the limestone phase for Temple E. There is a flight of steps into the compound on the east and a colonnade around its interior that intrudes into the ancient platform for Temple A to the north (Fig. 10).
Recovering an Image
In this northwest region, no Hindu image had until now been recovered in situ from its architectural context. From Temple E, excavations recovered only an enigmatic furl of a carved ribbon (Fig. II), suggestive of Sasanian” ribbons found sometimes in the northwest and in Central Asia (Rowland t974) and in Shi-period images on the market today. However, in clearing steps in front of the slightly earlier Temple B, north of Temple A, the team discovered a unique image, ca. 53 cm high, broken in two pieces (Fig. 12). The figure is that of a corpulent male seated on a broad-petaled lotus. He wears a short loin cloth, sacred thread, and a double string of beads, and his upper left hand holds a lotus. Breaks in the stone suggest that he had three other arms. He has three heads, not just faces, each set on a sturdy neck. These are coifed with elegant twisted strands of hair tied by pearl chains and a diadem. Abdur Rehman has identified the image as Siva Mahesvara (Rehman 1996). Temple B’s sanctum is only ca. 1.5 m wide, and this free-standing image would have fit comfortably within it (Fig. 13).
This quality of “emerging” form, not yet consolidated into the cliché of multiple faces, is characteristic of Kushan-period experiments with the use of multiple body parts to express divine manifestation (Maxwell 1996). This characteristic, the style of the image, and the form of the temple (Fig. 14) reinforce Temple B’s likely 6th century date, supported also by its experimental proto-Nagara superstructure.
Two Phases
The two phases of construction we have uncovered at Kafirkot Temples C and E suggest that there was a pre—Hindu Sahi period when Temples A, B, and C were built in kanjur stone (Fig. t3), and a later Hindu Sahi phase when these temples were reclaimed for worship and expanded, in part perhaps as jirnodhar (sacred rebuilding) of an an-dent site and in part as an expression of Hindu Sahib political self-promotion and re-legitimization in the region. The reconfiguration of these temples goes beyond simple expansion: there is a conscious attempt to mimic the forms of the earlier phase, although building with a different technology and material; and at the same time there is an attempt to reorganize ritual space, suggesting some discontinuity with the religious life of the earlier phase. We would tentatively date the first phase to the 6th and 7th centuries AD by considering both its links to the region’s Gandharan past and the tentative nature of its Nagara experimentation. We would date the second phase to the first century of Hindu- Sahi hegemony in ca. AD 825-925. We know that the now-destroyed pre-7th century temple of the Sun at Multan remained of very great significance in the pre-Islamic life of this region (Hasan ibn Yasid 1733). Perhaps these excavations at Kafirkot bring us back in contact with that world.
Acknowledgments
Support for Penn’s participation in the excavations at Kafirkot has come from the University Research Foundation, Department of South Asia Regional Studies, and the American Institute of Pakistan Studies, with a license to the Pakistan Heritage Society from the Department of Archaeology and Museums, Government of Pakistan, who kindly have assigned staff to attend the excavation.