High above the mighty Indus, on hills streaked red with salt, forts with citadels, habitation sites, and temples were built from the sixth to the eleventh centuries A.D. Largely ignored by scholarship in this century, and orphaned since partition, these structures form an important missing link in the history of architecture in South Asia (Mumtaz 1989).  A new integrated archaeological study of these sites, undertaken by this author with colleagues in Peshawar, has begun to recover new aspects of this important period of South Asia's antiquity.

The far northwest in ancient India what now is the Panjab and Northwest Frontier provinces of Pakistan, Swat, and parts of Afghanistan is better known for the presence of eclectic cross currents over many centuries, at important archaeological sites such as the city of Taxila (Marshall 1951), and for the massive numbers of Buddhist sculptural and structural remains we associate with the region of Gandhara (Ingholt 1957).  These Gandharan remains already show visually a local vocabulary that lets architectural traditions from India, Central Asia, and the classical world stand together, as in many Gandharan Buddhist narrative steles or as ornament on the famous shrine of the double-headed eagle or the Dharmarajika stupa at Taxila.  The Chinese pilgrim, Hsüan Tsang, visiting Gandhara in the seventh century A.D., noted hundreds of Hindu structures as well as many declining Buddhist sites in the region (Watters 1904-5).

If there is a Gandharan legacy in Hindu temple architecture, however, it takes two paths:  one, a unique tradition of pyramidal pent-roof temples built in Kashmir from before the reign of Lalit ditya in the eighth century A.D. (Kak 1933; Meister, et. al., 1988:351-93), and a separated tradition in Gandhara itself, along the upper Indus, and onto the high plateau and escarpments of the Salt Range in the Panjab (Lohuizen-De Leeuw 1959). 

As an example, I might contrast the eighth-century (or  earlier) temple at Loduv in Kashmir (Meister, et. al., 1988:361-3) and one of the surviving masonry sub-shrines at the Hindu pilgrimage site of Katas in the Panjab Salt Range.  The square Kashmir shrine's circular interior space and hemispherical dome, for which a Gandharan prototype, a masonry structure at Guniar in Swat, is sometimes cited (Kak 1933:55-6; Meister, et. al., 1988:362), once was covered by a simple peeked pyramidal pent roof, as indicated by the pent-roofed frame decoratively surrounding its doorway. 

The temple at Katas, on the other hand, while sharing the formula of a simple square plan with plain masonry walls and cantoned corner pilasters, struggled to give height to the temple by quite different means. The Katas sub-shrine's elevation can be reconstructed as a series of cornices with intermediate tiny rows of pillars and a crowning ribbed stone (amalaka), a type of "pre-Nagara" tower I have labeled "bhumi-prasada" after its use of many multiple stories.This early type of simply storied structure can best be paralleled at Sarnath, in Saurashtra, and elsewhere across northern India and the Deccan in the sixth century A.D. (Meister 1986; Meister, et. al., 1988: passim).

By contrast, the distinctive mode of monument that became the signature for Lalitaditya's powerful dynasty in Kashmir used a gabled  pent roof, as is well preserved on temples at Narastan, Pandrethan, or Payar  from the eighth to tenth centuries (ibid.: 351-93), a form already  marked on the doorway of the earlier temple at Laduv. Antecedents for this gabled roof-type already can be seen in the architectural vocabulary of Gandhara, as can be suggested by the "classical" niche-pediments represented on the shrine of the double-headed eagle at Taxila as early  as the first century B.C. (Harle 1986:74), or by the split pyramidal  pediments in sculpture and on stupas from Gandhara .

At the northern Kafirkot ("foreigners' fortress") in the North West Frontier Province near the Chashma barrage north of Dera Ismail Khan, however, a true local experiment with Nagara architecture – the curved temple form of northern India – had already begun.  The two earliest temples in this fort can most closely be related to early Garulaka or Maitraka dynasty temples in Saurashtra (coastal western India), at sites like Bhanasar and Dhank, from the sixth and early seventh centuries A.D., and Saindhava-dynasty temples from the same region in the eighth century (Nanavati and Dhaky 1966; Meister, et. al. ,1988:167-206).  Even the name of the little understood "Saindhava" dynasty seems to indicate their link with the Indus.

Scholarship in the past century including that of Aurel Stein (1937), Alexander Cunningham (1872-3:87-8), and Ananda Coomaraswamy 1927:108, 143) focused primarily on one particular temple in the Salt Range, however, that from the tenth century at Malot and on its links to the architecture of Kashmir.  This whole group of temples in the northwest scholars has tended to date "post Islamic contact" because of the use of mortar, rubble-fill between masonry walls, arches, and squinched interior domes (Archaeological Survey of India Annual Report, 1920-1:6-7).

Percy Brown (1942), in his history of Indian architecture, placed these temples as a branch of Kashmir architecture, as also did James Harle (1986:197-8) in his Pelican History of Art volume.  Harle, however, takes notice briefly, but only in his chapter on Kashmir temples, that "another group in Dera Ismail Khan … forms an extension of post-Gupta Madhyadesa" (ibid.198) rather than of Saurashtra as I suggest.

The early tenth-century temple at Malot does indeed mimic peak-roofed temples in Kashmir at a time of marital alliance between the Utpalas of Kashmir and the Hindu Shahi kings of Hund in Gandhara (Rehman 1979).  It signals its difference on its walls, however, by placing curvilinear Nagara shrine models that mimic local Gandhara-Nagara temples at other contemporary Hindu Shahi sites in the tenth  century, such as a pair of temples in a second important fortress, Bilot  (south Kafirkot), to the west of the Indus near Dera Ismail Khan

Sources for this Indus group of temples Malot excepted can much better be found in the Gandharan substrata and in the ferment of Nagara formation in other areas of north and western India (Meister 1981) than in Kashmir.  Whether in the domed Buddhist compounds at Takht-i-bahi or the fifth-century moldings facing the Dharmarajika stupa at Taxila, antecedents are close at hand.  Certainly the basic molding sequence of Gandhara-Nagara temples begins as early as Taxila, compared to those from the later temples at Bilot and Kafirkot.  The typical slender pseudo-corinthian pilasters at Kafirkot as well as true arches can be seen also on the second/fourth-century Buddhist stupa at Guldhara in Afghanistan (Harle 1986:73).  The characteristic sloping batter of niches and doorways (and sometimes walls) on these temples has its clear antecedents in Gandharan conventions.  Even the use of interior squinches and masonry domes is not new to the region, nor is much of the architectural ornament in these temples unfamiliar in the Gandh ra region.

What is new is the Nagara modality of superstructure as it had developed in North India for the first time in the fifth and sixth centuries A.D. (Meister 1986, 1988).  One might compare the shrine model on the wall of temple D at Bilot with the much better known proto-Nagara shrine model represented on the early sixth-century doorway to the "Gupta" temple at Deogarh in Central India, for example, or one on a brick stupa-base at Nalanda in Eastern India (Meister 1986:46-7).

Remarkably, this region preserves an almost continuous record of temples that define the evolution of distinctive school of Gandhara-Nagara architecture.  In this preliminary report and stylistic analysis of these monuments, let me give a brief review of these remains to frame this local and continuous craft tradition.

At northern Kafirkot, temples B and A represent the earliest experiments in this region with the developing Nagara formula.  The much larger temple D at Bilot awkwardly formulates a Nagara tower on a square base, much like the pre-Nagara temple at Bile vara in Saurashtra in the seventh century (Meister, et. al., 1986:181-4), then marks its walls with a model of a proto-Nagara shrine. 

Temple C at north K firkot and temple A at Bilot, with damaged Nagara towers, project one central offset on each wall and modulate ornamental elements of their superstructures in a more integrated way compared to Bilot temple D, marking new confidence and knowledge of Nagara formulas late in the seventh century.   Temple C at north Kafirkot for the first time tentatively introduces a version of north India's common vase-and-foliage capital for its corner pilasters, while retaining the local neo-Corinthian type for the central offset.

Two striking temples, located on hills east of the Indus opposite  Kalabagh at Mari – which I would date in the eighth century – continue but  refine this local N gara tradition, but still with only a single central  offset on their walls.  Temple A places thin pilasters on the corners of each offset, while temple B pairs pilasters for the first time on its corner buttresses.  Temples in this sequence in turn seem to provide a central shrine-model on each wall to represent a slightly preceding local experiment with the formula for a Nagara temple.  Each also seems to carry some architectural element forward, as in the trefoil arched niches at Bilot, trefoil doorway at Mari-Indus, and five-cusped entry to the smaller temple at Amb in the ninth century.

The first temple in this tradition that can have its date confirmed by any evidence other than style and decorative context is the elegant fired brick structure at Kallar .  It's layering of five offsets making up its walls, and its developed ornamentation with  vase-and-foliage pilasters and other distinctive details, place it  parallel to temples in central and western India from late in the eighth  and early in the ninth century A.D. (Meister, et. al., 1991), a date  suggested also by a single coin found near the foundations.  This comes from the reign of the first Hindu Shahi ruler, the beginning of whose dynasty can now fairly firmly be dated to ca. 821 A.D. (Rehman 1993).  Only further archaeological explorations, however, and perhaps carbon-14 dating of wood beams used to support the interior domes of some of these temples, can fix the dates and historical frame suggested here more firmly.

Sub-shrines that were added above the eastern corners of the  platform supporting temple D at Bilot perhaps early in the eighth century  echo but re-orient two domed cells sunk into the front corners of the  temple's platform .  The small temple D at Kafirkot, built near the north gateway to that fort late in the ninth century, mimics some distinctive details of these sub-shrines.

In the spectacular fortress at Amb, on the southern edge of the Salt Range; at Bilot in the tenth century; and at Nandana, larger temples began to be built under the patronage apparently of the Hindu Shahi kings.  These still were Latina temples, with curvilinear single spires, but they had a stairway within their walls leading to an interior ambulatory corridor surrounding an embedded upper chamber, in this respect unlike any other Nagara temples elsewhere in India. 

This remarkable regional experiment with multiple levels folded within a Latina tower (s. 21-4) came to an end when the great fortress at Nandana on the eastern flank of the Salt Range fell to Mahmud of Ghazni, who sought to control the significant routes across the Panjab toward Multan and Delhi early in the eleventh century.  The Hindu Shahi kings then took refuge with their cousins in Kashmir.  In this sequence, only this last temple built at Nandana suggests corner turrets on its  single-spired tower.  These suggest, however, the multi-spired shrine-models placed on the walls of the tenth-century temple at Malot, even as they reflect a multi-spired convention common in central and western India by the tenth century (Meister, et.  al., 1991). That these forts and temples survive along the Indus must be a reminder to us of how untouched many of South Asia's traditions are; of how insular scholarship can become; and of our task as scholars to weave a comprehensive image of the past, even as we reproblematize colonial scholarship and its assumptions.

I might end this preliminary report with a footnote to demonstrate the mighty weight of finding a new monument in the field.  At the site of Mari, in addition to the two eighth-century temples already discussed, there also are two mounds higher up the hill to the west, badly ravaged by treasure hunters, that past reports have labeled primarily as places of residence (Cunningham 1879:25-26; Mumtaz  1989:32). These in fact are ruins of two large temples placed on high platforms. One still preserves remains of an inner sanctum and an enclosing ambulatory wall.  On the north side, this wall preserves a central niche with a distinctive "Kashmiri-style" pent roof, but the shattered remains of the temple's superstructure suggest instead a complex multi-spired tower with curvilinear Latina spirelets.  This temple seems, in fact, to have been almost a reverse response to the unique local experiment with Kashmiri style found at Malot, and an answer to it.  Let scholars beware!

Author: Michael W. Meister