Temple-Monasteries, Buddhist Monks, and Architectural Exchange Between India, Java, and Tibet in the Late 8th Century

Abstract

The once-dominant view that architectural developments in mediaeval Southeast Asia closely followed Indian ‘influence’ is now largely rejected. Recent scholarship has shifted its focus onto the agency of local artists and architects in driving architectural innovations across the region. However, specific cases of transregional exchanges in architectural ideas and practices remain underexplored. This study examines three geographically distant Buddhist sites—Paharpur in northern Bangladesh, Candi Sewu in Central Java, Indonesia, and Samye Monastery in central Tibet—active in the late 8th century. I consider the significance of specific similarities and their temporal correlations within a broader range of styles, materials, and technologies. I argue that the activity at these sites reveals a shared architectural agenda transmitted over vast distances by religious experts, including Buddhist monks, in the last decades of the 8th century. Central to the network of three temple-monasteries proposed is the role that a specific architectural type was understood to play in protecting the kingdom and extending a king’s sovereignty while manifesting his spiritual aspirations. By distinguishing between architectural forms, architectural agendas, and modes of production, this study clarifies the complex nature of transregional architectural exchange in the premodern world.

1. Introduction

Following the excavation of Paharpur in northern Bengal, the Indian archaeologist K. N. Dikshit suggested that parallels for this Buddhist monastery and temple complex might be found in ‘Further India, especially Burma, Java, and Cambodia’ (Dikshit 1938, p. 7). Specifically, he noted that ‘the nearest approximation to the plan and superstructure of the Paharpur temple […] is afforded by the temples known as Chandi Loro Jongrang and Chandi Sewu close to Prambanan in Central Java’ (Dikshit 1931, pp. 38–45). This comparison between buildings separated by more than 4000 km across land, sea, and several political and cultural boundaries reflects a sense of pan-Asian cultural geography prevalent during Dikshit’s time. The ‘Further India’ he invoked aligns with the ‘Greater India’ most closely associated with the historian Ramesh Chandra Majumdar. This idea transplanted some of the characteristics and extent of European empires in Asia into the ancient and mediaeval past. In brief, Majumdar argued that similarities in the culture, religion, and society of mediaeval India and what is now Southeast Asia1 were the legacy of ancient colonisation: Indian kings, he proposed, had supplanted indigenous rulers across the region centuries earlier, bringing with them new religions, languages, art, and architecture (Majumdar 1927, 1937, 1944).

The study of ‘Indianisation’, as this process became known, has centred on two key questions: What influence did India exert on Southeast Asia in the past? And how was this influence transmitted? When these questions are accepted (though often they are not), responses generally focus on three main issues (Ali 2011; Acri 2017). The first problem is one of agency—who was responsible for this cultural transfer? Three hypothetical agents have been proposed to challenge Majumdar’s theory of large-scale colonisation: small groups of adventuring warriors (the kṣatriya theory), religious experts (the brāhmaṇa theory), or merchants (the vaiśya theory) have all been linked to this process (Bosch 1961, pp. 1–22; Cœdès 1968, pp. 15–16). Driving these potential carriers of cultural knowledge and practices is the implicit allure of ‘legitimation.’ Drawing on Max Weber’s ideas, scholars have suggested that the Indian mode of charismatic kingship, religion, and social organisation appealed to Southeast Asian rulers who were looking for ways to extend and consolidate their authority (Wheatley 1982; Pollock 2006, p. 516).

Some historians have questioned the scale and extent of this cultural transfer, however, arguing that the influence of India on Southeast Asia has been overstated, while the creative potential of localisation as a process has been overlooked (Wolters 1982; de Casparis 1983). Of particular interest to historians recently is the possibility of multi-directional transfer and the effort to reframe a relationship in which India is often assumed to have been the creative centre and Southeast Asia a passive recipient periphery (Acri and Sharrock 2022). Hermann Kulke has further argued that the dichotomy underpinning this discussion is misleading, if not entirely false (Kulke 1990, 2014). He argues that the similarities and simultaneity in cultural, religious, and social practices observed across mediaeval southern Asia reflect such a high degree of near-constant interaction that it is more accurate to view these regions as ‘converging’ in this period. Running parallel to these often-theoretical discussions is the material issue of chronology. As Ian Mabbett has argued, the scholarly study of Indianization has often flattened periods of activity that, when considered diachronically, reveal distinct and characterful differences (Mabbett 1977a, 1977b).

The study of architecture has been a constant feature of this scholarly discussion. R. C. Majumdar, F. D. K. Bosch, and George Cœdès all compared sacred buildings—mostly temples—to support and articulate their arguments about the relationship between India and Southeast Asia (Bosch 1924; Majumdar 1937, pp. 347–54; Cœdès 1968, pp. 31–32). Architecture has been central to this discussion for several reasons: First, large-scale architectural production in a pre-industrial society required substantial social and economic resources, making it a valuable indicator of broader cultural trends. More importantly, architecture—unlike most other forms of cultural production—does not move. Buildings are typically fixed in place, so the transmission of an architectural form or idea necessitates a mediator: a text, a physical mode, or an expert. In this way, evidence of transregional architectural exchange has provided a relatively stable peg on which to hang broader hypotheses about cultural transfer.

In this context, I examine a specific moment in the exchange of architectural knowledge between eastern India, the island of Java, and Tibet in the last decades of the 8th century CE. By comparing three Buddhist monastery complexes—Paharpur in northern Bangladesh, Candi Sewu in Central Java, Indonesia, and Samye Monastery in central Tibet—I consider a set of shared architectural ideas that encompass both general and specific formal features and reveal a broader ideological agenda of temple buildings that emerged within Buddhist communities during this period.

2. Paharpur Naogaon District, Bangladesh

2.1. A Temple-Monastery in Northern Bengal

The complex at Paharpur comprises a large quadrangle with a perimeter of around 1 km enclosing a monumental central temple and numerous smaller structures of various kinds (Figure 1). The quadrangle that encloses this complex will be familiar as a greatly expanded four-sided monastic dormitory (catūḥśāla-vihāra) of a type that became standard among Buddhist communities in India around the middle of the first millennium CE (Copplestone 2024, pp. 24–18, 88–90). Almost all of the 200 ‘cells’ in this structure at Paharpur contain large brick and stone pedestals, however; these once supported sculptures, perhaps of brick and clay, or wood, if not metal—none of which survive in place (Figure 2). Further clusters of three ‘shrine cells’ were discovered by excavators in the middle of three sides of the quadrangle, where stone drains in the form of mythical leonine vyālas indicate that these structures were used for the ritualised worship of images (pratimā-pūjā) (Figure 3 and Figure 4). The presence of pedestals and drains here challenges the established typology of Buddhist architecture, ruling out the possibility that this quadrangle was principally residential and revealing that the vast quadrangle at Paharpur was a monumental multi-shrine temple in its own right, as I have argued elsewhere (ibid., p. 91).

The temple at the centre of the Paharpur complex continues to tower over the site today. It is built on two terraces with a basement that projects in the cardinal directions reflecting the orientation of the four shrines at its centre, producing a characteristic cruciform configuration (Figure 5). Several hundred earthenware plaques are fixed in rows around the outside of this temple between alternating bands of narrow bricks made in moulds with geometric and vegetal patterns (Figure 6). These plaques depict an enigmatic array of imagery suggesting a divine assembly around the temple, and a similarly obscure programme is found in the irregular arrangement of sixty-three stone sculptures fixed into the basement of the temple, most of which are buried today (Kozicz 2022; Copplestone 2024, pp. 42–46). A low wall around the basement of the temple forms an enclosed circumambulatory path which once led past these stone sculptures and below several stone drains that lead from the four inner shrines of the upper temple. A large stone sculpture of the bodhisattva Avalokiteśvara found on the southern side of the temple with a brick-built basin directly below it suggests that this image had priority in the ritual activity that took place in this area.

The wall around the central temple at Paharpur projects on the northern side to create an inner courtyard between the temple’s staircase and the main entrance of the surrounding quadrangle. None of the surely monumental sculptures that once occupied the inner shrines of the central temple at Paharpur have survived, nor has anything of the superstructure that once rose above them (Figure 7). At the very core of the structure, however, excavators discovered a brick-lined chamber with two clay cups buried beneath layers of brick dust, ash, and clay (Dikshit 1938, pp. 14–15). An important feature of the architectural remains at Paharpur is its complicated physical history, which has been gradually revealed over generations of excavation and analysis (Sen et al. 2014, Table 1; Copplestone 2024, pp. 47–48). And the stratigraphy of this site reveals that the complex underwent several phases of renovation and reconstruction, notably in the construction of pillared antechambers attached to each of the four shrines of the central temple (Figure 8) (Dikshit 1938, pp. 12–14; Alam 2004, p. 52). I have dated this phase of architectural activity at Paharpur to c. 850 based on inscriptions found on the pillars that were erected in this period, several decades after the initial establishment of this temple-monastery complex (Copplestone 2024, pp. 112–13). And in another phase, the ground level of the entire courtyard was raised, either in response to a flood or to alter an existing hydrological system (Sen et al. 2014, pp. 60–64).

2.2. The King Dharmapāla Great Monastery at Somapura

The discovery of a large number of earthenware sealings and several stone inscriptions during the excavation of Paharpur has identified this temple-monastery as the King Dharmapāla Great Monastery at Somapura and dated its establishment to the late 8th century CE (Dikshit 1938, p. 91 and Pl. LIXh). This Great Monastery (mahāvihāra) was built at the heart of the historical-cultural region of Varendra, a region that would be eulogised several centuries later as the ‘fatherland’ (janaka-bhūmi) of the Pāla dynasty (Brocquet 2010, pp. 185–86). Although this institution does not appear to have been suburban to any major settlement besides an otherwise unknown ‘Somapura’, it is 40 km away from the historical city at Mahasthangarh, near Bogra (Figure 9). The architectural remains at Paharpur are in this way closely associated with the ruling Pāla dynasty (r. c. 750–1161) and particularly its second king, Dharmapāla (r. c. 775–807), with whom it had a ‘namesake’ relationship.

Most of what is known about the social history of the King Dharmapāla Great Monastery has been derived from a copper plate charter at the Indian Museum in Kolkata, which has been read and discussed at length by Ryosuke Furui (Furui 2011). This charter was issued in the 26th year in the reign of King Dharmapāla (r. c. 775–807; i.e., c. 801) and it may be understood to have acted as something like the institution’s ‘founding document’. It principally records a petition made to the king by a subordinate (mahāsāmanta) called Bhadraṇāga and his queen (rājñikā) Saṇhāyika so that they might grant land in favour of several monastic communities connected to the monastery at Somapura and in their own domain; this land would then become exempt from the usual obligations (Furui 2011, p. 154).

As Annette Schmiedchen and Furui have observed, an important distinction in charters such as this lies between the two participants in the establishment of a religious institution: the founder (here, the king) and donors (here, a mahāsāmanta and a rājñikā) may be understood to have performed consecutive pious acts, although only the latter is usually recorded in charters (Schmiedchen et al. 2023). The establishment of the monastery at Somapura thus was underwritten by a system of composite or shared patronage reflecting the emergent political structure of the Pāla realm—the relationship between this king and his subordinates was built into the Great Monastery at Somapura as an institution. The Indian Museum copper plate charter specifically connects both donors, Bhadraṇāga and Saṇhāyika, with particular buildings within the greater complex—an image shrine (gaṇḍhakūṭī) and a small monastery [?] (vihārikā) respectively—each of which appears to have had a distinct monastic community (saṁgha).2

The relationship between the Pāla King Dharmapāla and the Great Monastery at Somapura may have been more than nominal. The system of shared patronage that underwrote the establishment of this institution implicates architectural activity in the power dynamics of this early mediaeval polity, as Furui has discussed (Furui 2017). It is also likely that the connection between King Dharmapāla reveals a public and spectacular effort to cultivate a close and personal relationship with a specified Buddhist community. This is consistent with other sources for the early Pāla kings, too, which tell us that King Dharmapāla was close with two named Buddhist monks: Haribhadra (fl. 775) and Buddhajñānapāda (fl. 800) (Sanderson 2009, p. 90 and note 162). And according to the 16th-century Tibetan historian Kunga Nyingpo (Tāranātha), who was writing centuries after the fact, Dharmapāla tasked the monk Buddhajñānapāda with performing a regular fire sacrifice (homā) to ensure the longevity of the Pāla dynasty (Sanderson 2009, p. 93 and note 172).

Although it is not made explicit in the Indian Museum copper plate charter, the relationship between King Dharmapāla and the monastery that bore his name at Somapura may be understood in terms of the traditional role of the ‘monastery-owner’ (vihāra-svāmin) (Furui 2011, p. 151; Datta 2017a, p. 468). That is, the associated monastic community or communities (naivāsika-bhikṣu-saṃgha) may have understood themselves within an ongoing and obligatory relationship with this king and vice versa. As Gregory Schopen has shown, the Mūlasarvāstivāda Vinaya—the monastic rule that was probably followed in eastern India at this time—prescribes that monks recite verses for the monastery’s ‘owner’, who was usually the head donor (dānapati) or founder, on a daily basis, even after their death (Schopen 1996).

The Indian Museum charter is more explicit in associating the institutional and architectural activity it records with the path to liberation (apavarga-mārga) traversed by Bhadraṇāga and Saṇhāyika, their parents (māta-pitroḥ), all beings (sakalasya), and, indeed, their king (bhaṭṭāraka) (Furui 2011, p. 154). The transaction described here is a normative feature of the bodhisattva path (bodhisattvayāna) and it was made in perpetuity—‘for as long as the sun and the moon’ (ācandra-ārka-kṣiti-samakāla) (Furui 2011, p. 154). It also provides us with a glimpse of the intersection of worldly and supramundane aspirations at which the architectural activity at Paharpur was undertaken at the end of the 8th century.

Paharpur was not the only temple-monastery of its kind built in eastern India. Although unknown to K. N. Dikshit at the time of his excavation at Paharpur, Antichak in eastern Bihar and at Mainamati in southeastern Bangladesh would later reveal vast temple-monastery complexes that broadly conformed to the ‘Paharpur type’ architecturally and in their close association with specific named kings (Copplestone 2024, pp. 84–133). The remains at Antichak appear to date to the time of the Pāla King Devapāla (r. c. 812–850), about a generation after Paharpur (Verma 2011). And it is likely, though not confirmed, that Antichak is the famous Vikramaśīla Great Monastery, which was named after its founder, King Devapāla (Copplestone 2024, pp. 66–67).

As at Paharpur, the four shrines of the central temple at Antichak are empty, and it is impossible to determine with certainty which sculptures were once worshipped there. The situation at Mainamati is slightly different. Several temple-monastery complexes at Mainamati, such as Shalban Vihara—the King Bhavadeva Great Monastery named for the Early Deva king Bhavadeva (r. c. 750)—may have been built as much as a generation before the King Dharmapāla Great Monastery at Somapura.3 Several copper-alloy sculptures found during excavations at Mainamati, moreover, may indicate that the four shrines of the central temple at Shalban Vihara once housed larger-than-life images of bodhisattvas.4 And the discovery of a small copper-alloy caitya during the excavation of this site presents an iconographic arrangement of four bodhisattvas—Mañjuśrī, Avalokiteśvara, Vajrasattva, and Cuṇḍa—that may have corresponded to the lost sculptures of its central temple (Haque and Gail 2008, p. 287).

3. Candi Sewu Central Java, Indonesia

3.1. A Buddhist Temple Complex

Candi Sewu is one of several Buddhist and Śaiva temple complexes that were built between the late 8th and late 9th centuries around the Opak River on the Kewu Plain (also known as the Prambanan Plain) in what is today an eastern suburb of Yogyakarta in Central Java, Indonesia (Williams 1981, pp. 25–46; Dumarçay 1993, p. 289; Klokke 2006, p. 57). Indeed, Candi Sewu might have been designed as part of a larger architectural zone comprising the royal centre at Ratu Boko and several other sites such as Candi Bubrah, Candi Gana, and Candi Lumbung (Figure 10) (Degroot 2009, pp. 65–70, 151–76; Sundberg 2003, p. 163). Candi Prambanan, which is only a kilometre away from Candi Sewu, is the largest temple on the Kewu Plain and it remains the most popular with visitors today.

The Candi Sewu complex comprises two hundred and fifty structures arranged in four concentric quadrangles around a central temple (Figure 11). It was built using a jigsaw-like system of interlocking blocks cut from fine-grained volcanic stone, and this method of construction has enabled archaeologists to reassemble many of the collapsed structures in the Sewu complex, resulting in architectural remains that stand in striking contrast to the crumbling brick excavations at Paharpur. The temple at the centre of the Sewu complex comprises a single east-facing shrine surrounded by four antechambers on a platform that projects in staircases in each of the cardinal directions, producing a broadly cruciform configuration. The four staircases lead onto a narrow passage around the central temple and provide access to its single main shrine. The superstructure of this temple has been reassembled from its collapsed parts (Figure 12). And it relates to an architectural system that is often associated with South India: the ‘Drāviḍa’ architectural ‘language’ within which a single unit (the aedicule or kūṭa) repeats at multiple scales and altitudes to create structural elevation and an ornamental surface, affecting a sense of movement and expansion—in this case, the repeated unit takes the form of a stūpa (see Hardy 2007, pp. 206–10).

As at Paharpur, the principal image in the main shrine of the temple at Candi Sewu does not survive in place—although the throne-back that does survive indicates that this image was monumental (Figure 13). During the reassembly of this temple in the 1980s, several deposits were discovered in its foundations, including a small mould-made clay image of a stringed instrument and a ceramic bowl made in the Pearl River delta of Guangdong Province in southern China, which contained fragments of charcoal (Dumarçay 1987, p. 290; Dupoziat 2006, pp. 105–6). Well-carved stone drains in the form of mythical crocodilian makaras are found at intervals around the basement of the central temple at Sewu, and these appear to have carried water from the terrace or main shrine into the surrounding enclosure (Figure 14). Several more drains of the same type are found in the enclosure wall that immediately surrounds this temple, suggesting the possibility of a hydro-architectural feature, as has been proposed at Candi Prambanan (Jordaan 1996; Sundberg 2022) (Figure 15).

Four concentrically arranged quadrangles of small shrines and six mid-sized shrines surround the central temple at Candi Sewu in a regular and symmetrical arrangement. Based on the recovery of large numbers of monolithic buddha images of a uniform size in the complex, it has been suggested that each of these shrines contained an image of a seated buddha making one of several gestures (mudrās)—bhūmisparśa, varada, and dhyāna—although these do not appear to have been grouped iconographically (Degroot 2009, pp. 136–37). The exterior surfaces of the shrines and temples at Candi Sewu are carved in deep relief with a regular pattern of imagery: kīrttimukhas and gavākṣas appear over doorways, and bodhisattvas, floral patterns, and overflowing vessels (pūrṇaghaṭa) provide vertical planes of ornament. An animal-and-roundel pattern found on the exterior of the central temple has been compared to a kind of textile that was produced in Sogdia (today Uzbekistan) in the 8th century, as a comparison with a jacket at the Cleveland Museum of Art reveals (Figure 16 and Figure 17) (Woodward 1978). The composition of the architectural ornament at Candi Sewu does not correspond in a precise way to any known textile fragment, however (Sardjono 2017, pp. 22–25).

Jacques Dumarçay found four distinct phases of architectural activity in the central temple at Candi Sewu, including two major phases that, he argued, correspond to the two inscriptions that I discuss below (Dumarçay 1981, pp. 23–24 and Figure 29). Based on this, Dumarçay argued that the Candi Sewu complex was built initially in circa 782 CE and then renovated a decade later when the gateways in the galleries were lowered and narrowed, niches were added in flanking walls of the subsidiary chambers, and new walls were constructed to connect these chambers to the exterior wall of the main shrine. Marijke Klokke has similarly found two phases in the central temple based on her stylistic analysis, and she dates the renovation of the central temple to between 828 and 855, several decades later than Dumarçay and closer in time to a similar renovation of the central temple at Paharpur (Klokke 2006, p. 57).

3.2. A Royal Temple?

It is a measure of the little that is known about Candi Sewu that we do not know what the monumental temple complex was called by those who built it. Candi Sewu is conventionally associated with two stone inscriptions based on their find-spots, content, and dates relative to the architecture itself, however. And if the connection between this Buddhist temple complex and these inscriptions is allowed, we learn a little more. The Kĕlurak Inscription was found 300 m south of Candi Sewu near Candi Lumbung and consists of twenty verses in Sanskrit carved in a northern Indian Siddhamātṛkā script (Krom 1923, p. 272; Sarkar 1971, pp. 41–47). This inscription is dated to the 7th month of Śaka year 704 (782 CE) and it glorifies the bodhisattva Mañjuśrī while recording the deeds of King Saṅgrāmadhanañjaya (r. c. 775–800), who describes it as an ‘ornament of the Śailendra dynasty’ (śailendra-vaṁśa-tilaka) and a ‘destroyer of the best brave enemies’ (vairi-vara-vīra-mardana). Specifically, it remembers the Śailendra king Saṅgrāmadhanañjaya’s victory over the ‘kings of all quarters’ (dikcakra-rāja-vijaya-arjita-vikrameṇa) and his establishment of an image of the bodhisattva Mañjuśrī (mañjuśrī-pratimā-kṛtiḥ).

This image (pratimā) of the bodhisattva Mañjuśrī was likely a sculpture, and the inscription identified it as a ‘pillar of glory’ (kīrtta-stambha)—an ‘unparalleled bridge of religion’ (atula-dharmma-setuḥ). More specifically, it associates this image of Mañjuśrī with the protection of all beings (rakṣa-arthaṁ sarva-satvānāṁ) and the country from its enemies (deśasya-tasya-para-ārkṣāṁ). Interestingly, the Kĕlurak Inscription also describes the bodhisattva Mañjuśrī (or Mañjuvāc, ‘Gentle-Voiced’) as ‘possessing the glory of Brahmā-Viṣṇu-Maheśvara, the Lord containing all the gods’ (śrīmān brahmā viṣṇurmmaheśvaraḥ।sarva-deva-mayaḥ svāmī). This may reflect the understanding that this bodhisattva is capable of appropriating the teachings and capacities of these Brahmanical gods, as he does in several passages of the circa 8th-century Buddhist text the Mañjuśrīyamūlakalpa (Granoff 2000, pp. 404–12). It might also reveal a more particular conception of the bodhisattva Mañjuśrī that was developing in this period, as Mark Long has discussed in reference to a tantric Buddhist text, the Nāmasaṅgīti (Long 2015, pp. 7–9). The Kĕlurak Inscription is perhaps most famous, however, for its reference to a preceptor (guru) from Bengal (Gauḍī-dvīpa) called Kumāraghoṣa, who likely composed the inscription and is singled out and eulogised for his role in the activities recorded in connection to the Śailendra King Saṅgrāmadhanañjaya.

Another stone inscription was found during the clearance of the outer row of shrines on the southwestern side of the Candi Sewu complex itself (Dumarçay 1981, p. 3; Griffiths 2020).5 This is the so-called Mañjuśrīgṛha Inscription, composed in Old Malay, inscribed in the Old Javanese script, and dated to the 8th month of Śaka year 715 (792 CE)—a decade after the Kĕlurak Inscription. The Mañjuśrīgṛha Inscription eulogises a temple (prāsāda) called the ‘House of Mañjuśrī’ (mañjuśrī-gṛha) and identifies the merit accrued through its construction (?) with a bodhisattva vow (praṇidhāna) to attain awakening (Griffiths 2020, pp. 230–36). It moreover associates this temple with a vision seen by a Javanese nobleman (Jav. daṅ nāyaka) called Lūravaṅ, who saw the House of Mañjuśrī ‘while at the Vajrāsana’. As Arlo Griffiths has discussed, much of this inscription remains unclear and this particular passage may mean that Lūravaṅ visited the Vajrāsana (i.e., Bodhgaya) in India, that he saw this place internally as a vision, or that he saw an image or a temple that he associated with Vajrāsana iconographically, or in some other way (ibid., p. 245, note 86).

Marijke Klokke is correct to observe that the relationship between these inscriptions and Candi Sewu is not definite, not least because both inscriptions could have been moved (Klokke 2006, p. 57). I believe that the similar dates of these inscriptions, their proximity in time to the activity at Candi Sewu, and their find-spots support their value as sources for this site (van Lohuizen-de Leeuw 1981, pp. 19–20). Depending on how one understands the connection between these two inscriptions and Candi Sewu, one may learn that a temple or shrine within the Candi Sewu complex was called the House of Mañjuśrī and that this structure was associated with a named Javanese nobleman and his vision of the Vajrāsana, perhaps in a patronage relationship. It may never be known whether the image of Mañjuśrī referred to in connection to King Saṅgrāmadhanañjaya and his eastern Indian preceptor Kumāraghoṣa is the same as that in Lūravaṅ’s House of Mañjuśrī. Even if no direct connection between these two inscriptions and Candi Sewu is allowed, they still provide a window into the social and religious situation on the Kewu Plain in the late 8th century, where noblemen, kings, and their Buddhist preceptors used temples and images to commemorate their worldly achievements and satisfy their spiritual aspirations.

4. Samye Monastery Central Tibet

4.1. The First Monastery in Tibet

Samye Monastery (Tib. Bsam yas mi ’gyur lhun grub gtsug lag khang; ‘The Spontaneous Unchanging Inconceivable Temple’) sits near the northern bank of the Yarlung Tsangpo River in the Chimpu Valley, central Tibet (Figure 18). It is one of very few surviving Buddhist institutions that are known to have been established in the Imperial Period (c. 618–848) during the ‘first transmission’ of Buddhism to Tibet (Vitali 1990, pp. 5–25). Samye is perhaps most famous as the venue of the legendary debate between the Indian monk Śāntarakṣita (or his disciple Kamalaśīla) and his Chinese counterpart, Heshan Moheyan (fl. 8th century) held by King Khri Srong lde brtsan (r. c. 755–794) to determine the direction of Buddhism in Tibet (Dalton 2004). The architectural history of Samye presents a significant challenge, however. As it appears today, the Samye complex is the product of several phases of destruction and reconstruction. We know, for example, that the monastery was rebuilt at least partly on three occasions between the 17th and 19th centuries, and again in the 1980s following damage sustained in the 1960s (Lokesh Chandra 1961, pp. 10–11; Tenpa 2021). Nevertheless, the layout of this temple complex today and several buildings within it appear broadly consistent with their appearance in earlier photographs, paintings, and descriptions and can be discussed historically, if only in broad terms (Chayet 1988, pp. 19–29; Mémet 1988, pp. 30–32).

The Samye complex comprises numerous shrines and buildings of various kinds arranged around a central temple (Tib. dbu rtse) enclosed by a roughly circular wall (Figure 19). Due to Samye’s history of reconstruction, we cannot determine with certainty how the complex was originally built, but it was likely constructed by filling wooden frames with sun-dried brick masonry and rammed earth, then coating the entire structure with a thick layer of plaster, followed by white-washing and painting in broad planes of red and black. While the central temple appears to be a single structure, it is made up of three separate parts: an east-facing main shrine with a pillared assembly hall (Tib. du khang) at the centre is surrounded by two separate galleries, both of which project in the middle of each side to give the outward impression of a cruciform structure. The main shrine building has three storeys, and these are popularly thought to reflect different styles—Tibetan, Indian, and Chinese—leading to its traditional name: the ‘[temple of] three kinds’ (Tib. rigs gsum).

The low superstructure of the central temple today comprises several tiered roofs with projecting eaves covered in gold, at the very middle of which are four small cupolas or turrets surrounding a larger finial (Figure 20). The programme of the sculptures enshrined within the central temple at Samye has inevitably changed over time (Tenpa 2021, pp. 10–11). And today, the three shrines on its three floors contain images corresponding to the ‘three bodies’ of the Buddha (trikāya; Tib. sku gsum): an image of Śakyamuni on the ground floor, Vairocana on the first floor, and Sarvavid Vairocana on the upper floor. Several of the smaller shrines in the surrounding complex appear to communicate with projections in the plan of the central temple itself—four mid-sized shrines face towards the cardinal projections and four coloured stūpas (Tib. chod rten) sit at intervals between them.

4.2. A State-Supporting Temple

The events surrounding the construction of Samye Monastery may be pieced together from several inscriptions and historical accounts. Both the Samye Pillar Inscription, which is found nearby at Drakmar, and the Samye Bell Inscription are understood to be contemporary to the establishment of Samye (Walter and Beckwith 2010, pp. 303–4). The Pillar Inscription was likely composed during the reign of King Khri Srong lde brtsan shortly after the establishment of the Samye Monastery, and it principally records a vow (Tib. yi dam) made by the king, his heirs, lords, and ministers that the monastery (Tib. gtsug lag khang) at Samye and those at Ra sa (Lhasa) and Drakmar would be supported in perpetuity without reduction or diminishment (Tib. de las myi dbri myi) over successive generations of kings (Tib. bstan po yab sras) (Richardson 1985, pp. 26–31; van Schaik and Iwao 2009, pp. 11–12). This inscription also refers to an ‘explanatory edict’ (Tib. bka mchid) that, as Lewis Doney has argued, further situates the construction of Samye Monastery within a project to associate the propagation of Buddhism with the rule of the Tibetan kings and cultivates the sense that King Khri Srong lde brtsan specifically was a dharma king (Tib. chos rgyal) on the path to awakening (Tib. byang chub) (Coblin 1990, pp. 165–85; Doney 2014, p. 72). The shorter Samye Bell Inscription is found on a bronze bell commissioned by Khri Srong lde brtsan’s queen, Rgyal mo brtsan, and son, and it is composed in the manner of an ‘aspirational prayer’ (Tib. smon lam); it supports the acquisition of merit for the sake of the Three Jewels and the ten directions in favour of King Khri Srong lde brtsan and his path to awakening (Li and Coblin 1987, pp. 332–39).

Both inscriptions tell us something about the social and ideological context in which the Samye Monastery was built, particularly the relationship between the establishment of this monastery and the wider project of its founding king, revealing the intersection between architectural activity and general worldly and spiritual aspirations. Doney has argued that the establishment of Samye should be understood within a broader effort to present temple-building as a traditional responsibility of a Buddhist king and, specifically, to legitimise the reign of Khri Srong lde brstan; a feature that becomes more apparent in epitaphs composed after the death of this king identifying him as a mahābodhisattva (Tib. byang chub chen po) (Doney 2011, p. 109). A 9th-century prayer found at Dunhuang further cultivates this idea by situating this Tibetan king and his post-mortem awakening in relation to three historical Indian kings: ‘Dharmāśoka, Kaniṣkā, and Śīlāditya’ (ibid., p. 116).

Our understanding of Samye and its construction is significantly improved by a narrative account of the event contained in The Testament of Ba (Tib. dba’ bzhed), a lengthy chronicle of the aristocratic Ba Family (Stein 1961; Wangdu and Diemberger 2000). The use of this text as a historical source was for a long time disputed on the understanding that it was composed many centuries after the events it describes (van der Kuijp 1984, p. 180). However, the recent discovery of fragments of this text in the British Library reveals that significant parts of this narrative were circulating widely by the 9th or 10th century (van Schaik and Iwao 2009; Doney 2011, p. 34).6 Among these early fragments are sections narrating the arrival of the eastern Indian monk Śāntarakṣita at the court of Khri Srong lde brtsan immediately before the establishment of Samye Monastery (Doney 2011, p. 32). And this section is crucial to my understanding of the transregional architectural history of Samye. I summarise this narrative here based on a manuscript of The Testament of Ba dated to the 14th century published by Pasang Wangdu and Hildegard Diemberger (Wangdu and Diemberger 2000, p. 34).

The Tibetan aristocrat Ba Salnang travelled to eastern India in the reign of Khri Srong lde brtsan, where he worshipped at the Mahābodhi temple at Bodhgaya and visited the Great Monastery at Nalanda before helping to persuade the renowned monk Śāntarakṣita to visit Tibet and support the transmission of Buddhism (Wangdu and Diemberger 2000, pp. 39–73). Śāntarakṣita—who is referred to as Upādhyāya Bodhisattva (Tib. mkhan po bo dhi sa twa)—arrives in Tibet and declares that King Khri Srong lde brtsan had reached ‘the right age to build a temple’, and he offers his services in its establishment. It is unclear how closely to read this passage, but it would suggest that it was conventionally understood that a Buddhist king was expected to build a temple at a particular stage in their life. The Indian monk proceeds to survey the land for imperfections and identifies the ideal place for this new temple-monastery. The Indian monk then requests that the king send for the tantric ritual officiant (ācārya) Padmasambhava, who lived in Oḍḍiyāna (the Swat Valley in Pakistan) and a mason (Tib. phywa mkhan) from Nepal (Tib. bal yul) (Wangdu and Diemberger 2000, note 141). Both Padmasambhava and the unnamed architect from Nepal were presumably understood to have participated in the construction of Samye Monastery according to their expertise: the legendary Padmasambhava performing kīla (Tib. phur ba) rituals to subdue the ground, and the architect guiding construction, perhaps by laying out the complex.

The Testament of Ba dates the construction of Samye Monastery to four years between the Hare Year (775 CE) and the Sheep Year (779 CE), when Śāntarakṣita is said to have performed the foundation ritual (Wangdu and Diemberger 2000, note 201). Interestingly, this narrative identifies specific shrines within the complex with named ministers of the king, suggesting that the Tibetan polity was architecturally reflected in Samye Monastery (Tucci 1956, p. 280).7 On completion, the monk Śāntarakṣita is said to have smeared earth onto the king’s forehead and pronounced that the construction was ‘excellently accomplished’ (Tib. si ti pha la) (Wangdu and Diemberger 2000, p. 73). Śāntarakṣita then performed the ritual consecration (Tib. pra ti), at which time a light issued from the top of the central temple that he identified as ‘the light of Amitābha’s coming’ and ordered a shrine of Amitābha be built on the upper storey (Tib. yang thog) of the temple (Wangdu and Diemberger 2000, pp. 64–65).

The narrative contained in the Testament of Ba points to several general and specific ways in which Samye Monastery was understood. It identifies the establishment of this monastery with state support and the worldly ambitions and spiritual aspirations of its founder king. It tells us that the boundary wall of the complex demarcated a sacred, purified space regulated by the washing of hands and feet and a prohibition on animals (Wangdu and Diemberger 2000, p. 68). It also associates the spatial layout and structure of the Samye Monastery complex to several ‘continents’ (Tib. gling) and ‘subcontinents’ (Tib. gling phan), relating its architecture to ideas about geography and cosmology (Vitali 1990, pp. 256–57, note 374; Wangdu and Diemberger 2000, p. 67; Huntington 2019, p. 173). In one version, this text also records Śāntarakṣita’s decision to model Samye Monastery on the Great Monastery of Uddaṇḍapura, an institution that is more or less lost archaeologically but is understood to have been established by the first Pāla King Gopāla (r. c. 750–775)—the progenitor of the Pāla dynasty and father of Dharmapāla (Wangdu and Diemberger 2000, note 213; Beglar 1878, pp. 75–77). Some of these details may not date to the earliest recension of this text, and it is beyond the scope of this study to interrogate this further. But this discussion of the establishment of the Samye Monastery in the late 770s prompts several points of comparison with the activity at Paharpur in eastern India and Candi Sewu in central Java in the subsequent decades.

5. Comparison

5.1. Temple-Monasteries and Mandalas

When comparing the three Buddhist temple-monastery complexes mentioned above, several similarities emerge, not only in their architectural features but also in the social and ideological contexts surrounding their construction. I argue that these features reveal a particular set of architectural ideas and practices that was carried between eastern India, Tibet, and central Java at the end of the 8th century. Paharpur, Candi Sewu, and Samye Monastery share a broad spatial arrangement: a central temple enclosed by a systematic series of subsidiary structures and a clearly defined boundary (Figure 21). While this may seem too broad a comparison to be revealing, particularly when these built environments differ in the materials and technology of construction and the style of their ornament, it is important to note that this feature is distinctive; it is not found at most other major Buddhist centres in this period.8

The layout of these temple-monasteries is indicative of their use as built environments and is suggestive of their conceptual significance within a cosmic and ritual imagination. The arrangement of these temple-monasteries encourages movement along two axes. One leads from the periphery to the centre—east to west at Candi Sewu and Samye, and north to south at Paharpur. The other axis flows centripetally, guiding movement around a series of shrines towards the central temple. This latter axis of movement is, of course, ancient in a Buddhist context: the circumambulation of a sacred structure with one’s right side facing inwards (pradakṣiṇam) is an act of devotion and respect apparent in the ancient Buddhist stūpas and Brahmanical temples too.

The systematic and largely symmetrical arrangement of the three temple-monasteries discussed above, and the hypothetical identification of the lost sculpture within them, has led scholars to independently identify each with either a generalised or specific tantric Buddhist mandala, such as those described in the 8th-century Sarva-tathāgata-tattva-saṅgraha. Following this hypothesis, the iconographic arrangement of the images installed within the central temple and surrounding shrines is understood to correspond to the assembly of deities described in this text centred on the buddhas Vairocana, Ratnasambhava, Amitābha, and Amoghasiddhi (Figure 22).

The relationship between Candi Sewu and the Vajradhātu-maṇḍala of the Sarva-tathāgata-tattva-saṅgraha was first suggested by F. D. K. Bosch, and this idea has been developed by others more recently (Bosch 1961, pp. 124–30; Lokesh Chandra 1979; Long 2015). Jacques Dumarçay, for example, supported this identification with the caveats that Candi Sewu was built in two phases, and the association between this temple complex and this mandala was only possible after the second phase (Dumarçay 1981, pp. 29–33). And, secondly, he argued that the temple was dedicated to the bodhisattva Mañjuśrī based on his interpretation of the two inscriptions I have discussed above and could not, therefore, correspond exactly with the Vajradhātu-maṇḍala, which does not include this bodhisattva.

In response to Dumarçay, Mark Long has argued that Candi Sewu may be associated with a particular variation of the Vajradhātu-maṇḍala known to the 8th-century Indian monk Vilāsavajra, with whom the Śailendra-guru Kumāraghoṣa might reasonably have been familiar (Long 2015, pp. 7–8 and note 37). Vilāsavajra’s Vajradhātu-maṇḍala includes Vajrasattva as a sixth Buddha alongside the standard five Buddhas and, as Long has argued, this mandala may be hypothetically associated with the number and position of shrines in the central temple at Candi Sewu. In Vilāsavajra’s Vajradhātu-maṇḍala, the central deity Vajrasattva may be understood to have had a subtle and multivalent relationship with Mañjuśrī (Mañjughoṣa) (ibid., pp. 13–17 and Table 4).

Although K. N. Dikshit did not associate the architecture at Paharpur with a tantric Buddhist mandala, instead viewing it within a multi-religious history of temple-building, it is increasingly common to interpret the vast complex in these terms (Gail 1999; Samuel 2002; Kozicz 2022). And the four empty shrines of Paharpur’s central temple have been identified according to the Vajradhātu-maṇḍala described in the Sarva-tathāgata-tattva-saṅgraha: Vairocana at the centre (or in the missing superstructure), surrounded by Ratnasambhava, Amitābha, and Amoghasiddhi in their respective directional shrines. However, the number and placement of shrines in the central temple and in the surrounding complex do not appear to support the close correspondence that has been proposed at Candi Sewu, most recently by Mark Long.

Despite the little that survives of its original architecture and sculpture, textual sources provide us with a significant discourse about the historical conception of the Samye Monastery. While it is not uncommon to read that Samye was designed as a representation of the Vajradhātu-maṇḍala in architectural form in scholarly literature (e.g., Kapstein 2001, p. 43), the reality seems more complicated, Anne Chayet and Eric Huntington have both observed (Chayet 1988, pp. 19–29; Huntington 2019, pp. 172–75). The conception of Samye Monastery appears to have transformed over time: while this temple-monastery was initially related to a general Indic cosmology, increasingly explicit associations with a taantric Buddhist mandala of deities emerged over time. The Testament of Ba describes Samye Monastery in a way that is suggestive of Mount Sumeru surrounded by the four continents—that is, as a representation of the Indic cosmos (Figure 23). In his History of Buddhism in India and Tibet Butön Rinchen Drub (1290–1364) provides a similar cosmological interpretation, as does Gö Lotsawa Zhönnu-pel (1392–1481) in the Blue Annals (Roerich 1949, pp. 69–70; Obermiller 1986, pp. 188–90).

The relationship between these three late 8th-century temple-monasteries and a tantric Buddhist mandala, such as the Vajradhātu-maṇḍala, presents two problems requiring different approaches. The first problem relates to their physical history, which, in each case, remains obscure: there is little material evidence to confirm or refute the hypothesis that the principal shrines in any of these complexes initially contained images corresponding to this or any other mandala. The second problem concerns the relationship between architectural production and tantric Buddhist mandalas described in texts generally. This is both a problem of meaning and architectural form, and a question of mandalas in the history of Buddhist art. While mandalas do appear in Sanskrit texts in the 8th century, if not before, and they are made as drawings and paintings on cloth by Buddhists in East Asia, there remains little and only ambiguous evidence for their widespread use among mainstream Buddhist communities in Indian and Southeast Asian contexts at this time (Luczanits 2022; Linrothe 2022).

Without discounting the possibility that these three temple-monastery complexes were designed to have a formal relationship with a tantric Buddhist mandala, it must be remembered that the mandalas described in texts did not emerge from nothing; they have historical and conceptual relationships to a range of other pre-existing ideas and practices in rituals and visual cultures. In the Sarva-tathāgata-tattva-saṅgraha itself, the Vajradhātu-maṇḍala is introduced through a narrative retelling of Buddhist awakening: the initial antagonist, Sarvārthasiddhi (here, not, Siddhārtha), is consecrated (abhiṣeka) and receives the vajra-name ‘Vajradhātu’ in an assembly of buddhas in Akaniṣṭha Heaven. He then ascends to the peak of Mount Sumeru, where he is enthroned on a lion throne (siṃhāsana) within a vajra-jewelled peaked temple (vajra-maṇi-ratna-śikhara-kuṭāgāra) in the presence of the buddhas Akṣobhya, Ratnasambhava, Lokeśvarāja (Amitābha), and Amoghasiddhi forming the Vajradhātu-maṇḍala (Lokesh Chandra 1987, pp. 4–5).

This narrative and the image of a buddha enthroned in an assembly of buddhas on the peak of Mount Sumeru interacts with a range of Buddhist and multi-religious ideas about the heavens and the Indic cosmos. It also depends on architectural imagery: the peaked temple (kūṭāgāra) in this text was a recurrent feature both in tantric mandalas and Buddhist texts over a longer period (Bollée 1983; Copplestone 2024, pp. 143–49). The internal visualisation (and description) of this architecture constitutes the ‘cultivation stage’ (utpattikrama) of mandalas described in Buddhist Tantras, in which a practitioner imagines this temple atop Mount Sumeru, often with a four-part internal structure of shrines. The broader relationship between architecture, real and imagined, and tantric Buddhist mandalas is beyond the scope of this paper, as is the specific question of the possible relationship between the design of the temple-monasteries at Paharpur, Candi Sewu, and Samye Monastery and the tantric Buddhist mandala. I want to emphasise the complexity, contingency, and fluidity inherent to any such relationship between a building and an abstruse textual prescription or ritual device.

The art and architecture of Buddhist monasteries in India and their framing in inscriptions have engaged with a range of cosmological ideas from at least the middle of the first millennium CE when this architectural tradition began to play with the difference between mundane monastic dormitories and supramundane assemblies in heavenly realms (Tournier 2020, pp. 206–10). This development is consistent with ideas that were emerging across the region and beyond Buddhist communities. As Phyllis Granoff has convincingly argued with respect to temples dedicated to Hindu gods in mediaeval India, modern scholars often attempt to connect architectural forms to specific and abstruse symbolic systems when the texts closely associated with them tend to cultivate broader relationships with a deity’s immanent presence in a heavenly palatial residence (Granoff 1997).

Architecture rarely embodies singular meanings, and this is true of temples and monasteries built across Buddhist contexts in mediaeval Asia. The design and construction of the temple-monasteries at Paharpur, Candi Sewu, and Samye Monastery, I argue, should be understood on a continuum of ideas without necessarily adhering to any single fixed meaning—for which, in any case, the arguments must remain hypothetical pending further discoveries. What stands out in the history of Buddhist architecture is the scholarly distance between monumental Buddhist temple projects and the ideas and practices linked to royal temple-building across various religious communities towards the end of the first millennium CE.

5.2. Buddhism’s Imperial Temples

While it may be assumed that kings had always built Buddhist temples to protect the polity, commemorate their achievements, and fulfil both spiritual and worldly ambitions, it is only around the turn of the 9th century that these ideas began to be clearly expressed in inscriptions and, I argue, reflected in the design of these Buddhist temple-monasteries. The establishment of the King Dharmapāla Great Monastery at Sompura, as I have previously mentioned, is closely tied to the internal dynamics of the emerging Pāla polity. Similarly, under King Khri Srong lde brtsan and with the guidance of the eastern Indian monk Śāntarakṣita, the Samye Monastery was founded through a comparable system, whereby the structures and institutions of this temple-monastery reflected the organisation of the Tibetan Empire.

Inscriptions related to the establishment of Samye Monastery locate this architectural project at the intersection of several worldly and supramundane aspirations: the project to cultivate Khri Srong lde brtsan as a dharma king on the path to awakening, and to support the longevity of his dynasty. As Christopher Beckwith has observed, the Samye Monastery instantiated the king’s authority over the universe in new and distinctive ways (Beckwith 1987, pp. 173–96). The construction of the King Dharmapāla Great Monastery, mobilised significant resources to similar objectives. Although no sources explicitly detail King Dharmapāla’s political project, many of the defining features of Pāla rule, which would persist for four centuries, began to take shape during his reign, such as the standardisation of the dynasty’s copper plate charters (Datta 2017b). The late 8th century also marked a critical phase in the protracted conflict between the Pālas and their subcontinental rivals—the Rāṣṭrakūṭas and the Gurjara-Pratihāras—with the military victories of this period remembered in copper plates issued by Pāla kings for centuries (Sircar 1985, pp. 43–47; Furui 2024).

The Mañjuśrīgṛha Inscription offers a glimpse into late 8th-century society on the Kewu Plain, where artistic production, and possibly tantric ritual activity associated with the bodhisattva Mañjuśrī, were harnessed to support the Śailendra King Saṅgrāmadhanañjaya, under the guidance of the eastern Indian monk Kumāraghoṣa. If this inscription is considered in relation to Candi Sewu, the temple-monastery may be identified as a temple of Mañjuśrī, established to commemorate the king’s martial and worldly achievements while also serving to protect the polity. The Kĕlurak Inscription may further implicate this temple-complex within a system of shared patronage reflecting the political organisation of the Śailendra polity. In this context, one might ask whether the nobleman Lūravaṅ (Jav. daṅ nāyaka) established a shrine within the larger royal temple complex on the Kewu Plain in the 790s, in a manner similar to his contemporaries Bhadraṇāga, Saṇhāyikayā, and the ministers of Khri Srong lde brtsan, who did so in northern Bengal and central Tibet around the same time.

In this respect, these three temple-monasteries may be understood as Buddhist inflexions of the ‘Imperial Temple’ as defined by Ronald Inden and Hermann Kulke (Kulke 2001a, 2001b; Inden 1985, 2000, p. 54). Both Inden and Kulke have proposed the concept of the Imperial Temple as a category to explore the connections between architectural monumentality, the act of temple-building, and the social, political, and cosmological aspirations that emerged in early mediaeval India. Kulke defines the Imperial Temple in practical and ritualistic terms, viewing it as the central institution of early mediaeval society through which rulers consolidated their power and unified a diverse population under a singular religious and political order. Inden, on the other hand, interprets the Imperial Temple as fundamental to the ideology of kingship during this period, seeing it as the culminating event in the narrative of royal cosmo-political supremacy, embodying the world system established by the founding king and his chosen religious community.

According to Inden, the paradigmatic Imperial Temple is described in the Viṣṇudharmottara-purāṇa—an encyclopaedic treatise composed or compiled in a Pāñcarātra Vaiṣṇava context, likely at the Karkoṭa court in Kashmir during the 8th century. Specifically, Inden locates the paradigmatic Imperial Temple in the Sarvatobhadra Temple (sarvatobhadra-prāsāda) described as the final temple in a list of 101 temples in this text. This temple is familiar to art historians through Stella Kramrisch’s discussion in The Hindu Temple (1946), which included a plan of the structure (Kramrisch 1946, p. 418). The Sarvatobhadra Temple described in this text is, in fact, a complex. At its centre stands a temple consisting of a four-sided platform with a staircase on each side, supporting four shrines, each housing one of the four aspects of Viṣṇu (Figure 24) (Shah 1958, pp. 243–48, vv. 87.1–63).

The central temple described in this text is situated within a vast courtyard filled with celestial waters and enclosed by a compound wall lined with ‘cave shrines’ (garbhaguhā), which house additional images of Viṣṇu. The description of the Sarvatobhadra Temple in the Viṣṇudharmottara-purāṇa is addressed to an unnamed king and spoken by the sage Markandeya. It concludes by enumerating the various benefits of the temple: the king for whom it is built will be reborn as a universal ruler (cakravartin), erase his sins, be reborn in heaven, and eventually attain union (sāyujya) with Viṣṇu. Additionally, one who merely sees the temple will be freed from sin and acquire merit (puṇya), and anyone who enters the temple will be protected from disease, calamity, and unnatural death.

No temple complex in India has been found that corresponds in every detail to the textual prescriptions of the Sarvatobhadra Temple described in the Viṣṇudharmottara-purāṇa, nor should we expect to find one. This textual prescription has, however, been compared to several mediaeval temple projects. For example, Michael Meister has argued that the unfinished 8th-century rock-cut Śaiva temple at Masrur in Himachal Pradesh, India, corresponds to many of the ideas outlined in the text. While it may not precisely match the Sarvatobhadra Temple, it reflects similar concepts to such an extent that this Śaiva temple can be understood in comparable cosmo-political terms (Meister 2006). Specifically, Meister has argued that the Masrur temple complex was an unfinished intervention in the cosmo-political rivalry between two kings—Yaśovarman (r. c. 725–754) and Muktāpīḍa Lalitāditya (r. c. 724–760).

Ronald Inden, S. K. Saraswati, and others have identified formal correspondences between the Sarvatobhadra Temple described in the Viṣṇudharmottara-purāṇa and the architecture of the King Dharmapāla Great Monastery (Inden 2000, p. 54; Saraswati 1976, p. 85). This comparison requires careful historical and methodological framing, especially when extended to the two other temple-monasteries discussed in this study. The relationship between a temple complex described in a text of this nature and an actual built temple complex is not absolute. The gap between a brief textual description and the realities of architectural activity is considerable, and it is not useful to imagine that architects and builders were directly referring to such texts on mediaeval building sites. Nor should the explicitly Vaiṣṇava orientation of the Viṣṇudharmottara-purāṇa be taken to imply that this type of temple complex was inherently Vaiṣṇava or exclusive to any particular religious community. Instead, the compelling similarities between this textual prescription and the architectural features and ideologies I have outlined suggest, in my view, that those who designed and built the three temple-monasteries at Paharpur, Candi Sewu, and Samye Monastery engaged with a system of ideas about temple-building that circulated widely across regions and religious communities during the second half of the 8th century.

5.3. Material and Technology

The formal similarities between Paharpur, Candi Sewu, and Samye, while compelling at the level of layout and broad outward appearance, must be reconciled with the significant differences that emerge upon closer comparison. Architecture is profoundly influenced by the material and topographical conditions of a site—whether a material is scarce or abundant plays a crucial role in determining construction choices, and the availability of clay for bricks, quality stone, or wood for construction, for example, varies widely across and between regions. The materials and construction technologies used at each of these sites are entirely distinct, and these differences are reflected in the buildings’ appearances and specific features. Paharpur, situated in a clay-rich, stone-poor region on the eastern edge of the Ganges River basin, was constructed primarily from fired bricks made in moulds and earthenware plaques. This relatively lightweight material facilitated the creation of large-scale structures but was less suited for deep ornamentation or figural carving. As a result, architectural decoration at Paharpur is largely confined to bands of patterned bricks and hundreds of small plaques featuring figural forms.

Candi Sewu, by contrast, was constructed from blocks of volcanic rock, abundant in Central Java. Once assembled, these blocks allowed sculptors to carve deep reliefs into the stone, creating an ornamental surface that flows seamlessly across much of the temple’s exterior. Seams between the blocks were concealed by a thin layer of white plaster applied to the entire structure. In central Tibet during the 8th century, the materials and technologies available to architects were different yet again. While the exact construction methods remain uncertain, the absence of high-quality clay and fuel for kilns, and the scarcity of abundant stone would have constrained efforts to achieve significant verticality or detail in architectural ornamentation, apart from the use of polychromy and roof fixtures.

The scale of the differences between the architecture of Paharpur, Samye, and Candi Sewu, when examined closely, might seem to challenge the historical connections between these three building projects that I argue for in this paper. However, I contend that material and stylistic variations of this kind are exactly what we should expect in certain forms of transregional exchange in the premodern world. When an architectural mode—the conceptual framework and purpose of a building—is transmitted across vast geographic and cultural distances, it is reasonable to assume that materials, technologies, and styles might not be transferred with it. These differences do not simply indicate partial transmission; rather, they offer valuable insights into the nature of architectural knowledge and the individuals who carried it across mediaeval Asia.

6. Architectural Exchange

6.1. Type, Style, and Mode

It is a central premise of architectural history that the degree of formal similarity between two buildings often reflects the depth of the historical relationship between them. Such similarities may suggest that the buildings share an architect, are linked to the same patron, or were constructed around the same time. However, exceptions exist: formal similarities may sometimes not indicate direct historical relationships, while, conversely, buildings that appear very different have significant historical connections. The latter is often the case when the transregional exchange of architectural ideas occurs, when forms are transmitted in parts, transposed into different materials, translated into distinct styles, or transformed with new or altered meanings and associations. As a result, in this case, the historical relationships between two or more buildings may be easily overlooked.

The practice of formal comparison in architectural history is often organised around three categories of analysis—type, style, and mode—which correspond to distinct agents involved in the design and construction of buildings, as Tamara Sears has argued (Sears 2015). Even in cases in which sources detailing the social history of temple-building are absent, it can be assumed that architectural production in mediaeval Asia was a socially complex process. Labourers, artisans, architects, patrons, and religious experts each contributed to specific parts of a temple, operating within their own networks and guided by their individual or communal agendas. Temples, when viewed as a whole, represent a composite of these various interacting agents and the distinct dispensations they brought to the process of construction.

As Sears has shown, this framework can be used to extrapolate a hypothetical social history from formal analysis: the plan and elevation of a temple, which often determine its type, may correspond to the expectations or aspirations of a patron; the ornamental surface, which typically defines its style, may reflect the work of a particular group of artisans; and the main ritual image and the iconography of the temple’s sculpture generally align with the ritual agenda of its religious community or a particular religious expert. Architectural mode, as a historical category, overlaps several of these aspects but is more likely shaped by the collaboration between the patron and their religious expert, rather than by the artisan or architect.

Analogies with translation are, I think, useful in understanding transcultural architectural exchange. Linguists recognise that the relationship between form and content is inherently unstable, and that the two can be transmitted either together or separately. Similarly, in architectural exchange, the underlying ideas or agenda (content) may travel independently from the form it takes, resulting in transformations of various kinds: material, technological, or ideological. Specific architectural ideas or forms, much like a word or phrase translated into a new language, undergo a transformation in which both loss and gain can occur (Koehler and Saletnik 2018, p. 3). When the temple-monastery—and indeed the Buddhist ‘Imperial Temple’ as a mode that I have identified with Paharpur, Samye Monastery, and Candi Sewu was exchanged, we may expect it to undergo both material and conceptual transformation in this manner.

This understanding of transregional architectural production further highlights the inadequacy of common cultural geography categories such as ‘Indian’, ‘Javanese’, ‘Tibetan’, or the more general terms ‘local’ and ‘foreign’. When applied to cases of transregional exchange in architectural ideas—where buildings or built environments are recognized as composites, and where formal and conceptual translation occurs—these categories cease to be useful. Although often applied heuristically, such categories not only fail to accurately reflect mediaeval identities, as far as we understand them, but they also flatten the composite and historically contingent character of transregional architectural activity.

6.2. Mechanisms of Exchange

F. D. K. Bosch argued that architects from Java, rather than India, were responsible for determining the design of temples built in Central Java during the 8th and 9th centuries (Bosch 1924, pp. 30–35). He based this argument on a comparison of architectural ornamentation across a corpus that included major temples in Central Java as well as South and Northwest India, notably overlooking parallels in eastern India (see Jordaan 1999, p. 224). A generation later, Reginald le May conducted a similar study, arriving at the same conclusions (Le May 1954, pp. 28–29). Most recently, Marijke Klokke compared motifs in temple ornamentation between Central Java and South India, finding a broader and less coherent field of correspondence, along with a pronounced degree of ‘localization’ (Klokke 2008, pp. 154–66). This suggests that the artisans who carved relief ornament onto temple surfaces in Central Java and South India were not the same and were likely unfamiliar with each other’s motifs and practices.

Short of artisans travelling across the seas from India to Central Java—a scenario neither suggested by historical sources nor supported by formal analyses of architectural ornament—how else might transregional exchange have occurred? Heather Grossman and Alicia Walker have outlined several ways in which architectural ideas circulated in the mediaeval world, emphasizing the particular roles played by plans and drawings; texts; mediatory objects such as tools, furniture and textiles; and memories (Grossman and Walker 2013). They argue that each mechanism carries a particular kind of information and circulates in its own distinct way, with none being mutually exclusive. The transregional exchange of architectural knowledge might occur through one or more of these mechanisms, each of which informs architectural production in its own way.

Nearly a century ago, Bosch argued that architectural knowledge was exchanged between India and Central Java through architectural treatises—texts that were transported by Javanese pilgrims and studied by Javanese architects (Bosch 1924, pp. 15–17). Roy Jordaan associated this exchange with the Śailendra kings themselves, who, he argued, had their origins in eastern India and were familiar with the cultural practices of that region (Jordaan 1999, 2003).

The complexity of the comparisons undertaken in this study, along with its findings, reveals several distinct mechanisms of exchange. At this point, I might choose to trace the trajectories of one or more of these mechanisms—such as the appearance of textile patterns originating in Sogdia and ceramics from southern China at Candi Sewu, for example. I am arguing that a particular type of temple-monastery, and more broadly a mode of the Buddhist Imperial Temple, was transmitted between eastern India, Central Java, and Tibet at the end of the 8th century. This is not to suggest that other networks of exchange, or indeed local invention, were not also active at these sites during this period and across these regions. However, my focus is on exploring how this particular set of ideas and practices was transmitted.

The Testament of Ba, as introduced above, associates three key individuals with the architectural activity at Samye Monastery in Tibet during the second half of the 770s: Śāntarakṣita, a monk from eastern India; Padmasambhava, a ritual expert from the Swat Valley in Pakistan; and an unnamed architect from Nepal. This text does not mention the labourers—masons, diggers, hod carriers, and others—who were surely also involved in the construction and likely came to the site from the local area. This division of labour may apply more generally—trans-local or even cosmopolitan mobile experts such as the Śailendra-guru Kumāraghoṣa working alongside less-mobile artisans and labourers.

It can be assumed that the role played by these monks in the architectural activity in Tibet and Central Java principally concerned the religious, ideological, and ritual aspects of the building process. Of course, they would have carried with them specific knowledge related to ritual practices, but also the memory of the temples and monasteries in which they had lived or practised in eastern India—such as the Uddaṇḍapura Great Monastery, which has been mentioned in connection with Śāntarakṣita. Memory of this kind constitutes a distinct and highly portable form of architectural knowledge, one that focuses on the religious, political, and social significance of a building rather than the technical aspects of its fabrication (Grossman 2012).

These three temple-monasteries, I argue, form part of a late 8th-century ‘architectural network’ operating within the broader social phenomenon that Tansen Sen has referred to as the ‘Buddhist Cosmopolis’ (Sen 2018, p. 346). Śāntarakṣita, Padmasambhava, and Kumāraghoṣa are each, moreover, associated with the ‘first wave’ in the transmission of Buddhist tantra proposed by Andrea Acri (Acri 2016, pp. 1–26, Map 1.1). The spatial and chronological limits of this architectural network, as well as its relationship with other such networks from the same period, remain to be explored. The copper plate charter issued in the 35th year of Devapāla’s reign (c. 850 CE), found at Nalanda Monastery in Bihar, reveals the existence of other kinds of architectural networks. In this context, it is possible that the Śailendra King Saṅgrāmadhanañjaya, mentioned in the Kĕlurak Inscription, may have served as a patron for a monastery in eastern India (Shastri 1924).9 Not only does this reveal a multi-directional exchange in itself, but the date of this charter also aligns with a pivotal period in temple-building in Central Java, when the ‘new cultural impulse’ described by Jacques Dumarçay took hold (Klokke 2006, p. 51). It was also around this time that the central temples at Paharpur and Candi Sewu appear to have been renovated.

7. Conclusions

This study has examined a specific case of architectural exchange between India, Java, and Tibet that, I argue, took place in the final decades of the 8th century. It has focused on three Buddhist temple-monastery complexes: Paharpur, Candi Sewu, and Samye Monastery. While the architecture at each site differs significantly in terms of materials, construction technologies, and style, similarities in layout, and particularly in the broader ideological context, suggest that an architectural mode—the Buddhist Imperial Temple—was transmitted alongside a specific type of temple-monastery complex across vast geographical and cultural distances during this period.

These Buddhist temple-monasteries share features that suggest an effort to align the built environment with a conception of the Indic cosmos, and possibly more specifically with emerging ideas related to tantric Buddhist mandalas, if not directly with the Vajradhātu-maṇḍala described in the Sarva-tathāgata-tattva-saṅgraha. I have argued that the layout and spatial arrangement of these three Buddhist temple-monasteries also invites comparisons with a broader set of architectural ideas that became current across India and religious communities in the 8th century. The Imperial Temples, such as the Sarvatobhadra Temple described in the Viṣṇudharmottara-purāṇa, suggest a wider base of multi-religious architectural concepts within which, I propose, these three temple-monasteries were designed and built.

The material and stylistic differences between these three temple-monasteries offer valuable insights into the nature of architectural exchange during this period. I have argued that this exchange reflects a transmission of architectural knowledge facilitated by Buddhist monks and other ritual experts, rather than artists or artisans, who appear to have been sourced more locally. However, these artisans may have interacted along other networks of artistic exchange in textiles, ceramics, and other media that are beyond the scope of this study. The involvement of Buddhist monks and ritual experts—and the mode of Buddhist temple-building they promoted—enabled kings across this architectural network to express their sovereignty and spiritual ambitions in a manner that carried transregional legitimacy.

Scholarship has often divided southern Asia into distinct spheres of cultural exchange, typically differentiating between networks across the Indian Ocean and those over the Himalayas. This distinction, particularly between maritime and trans-Himalayan land routes, is especially prominent in the historiography of Buddhism, where the movement of art and ideas between India, Southeast Asia, and Tibet has rarely been considered together. However, this study identifies an architectural network that reveals the rapid exchange of architectural ideas between eastern India, the island of Java, and central Tibet in the final decades of the 8th century.