Ellora Caves Comparisons to Other Sites

About Ellora Caves Comparisons to Other Sites

The comparison frame for the Ellora Caves is set by a single procedural fact: every monument here was made by removing rock, not by stacking it. The 34 caves cut into the basalt scarp of the Charanandri Hills (c. 600-1000 CE) — twelve Buddhist (caves 1-12), seventeen Hindu (caves 13-29), and five Jain (caves 30-34) — are the most concentrated record of subtractive monumental architecture in the ancient world, and the Kailasa Temple (Cave 16) is its outer limit. The Baroda copper-plate inscription of the Rashtrakuta feudatory Karka II, dated Saka 734 (= 812-813 CE), names Krishna I (r. c. 756-773 CE) as the patron who "built a temple on a hill at Elapura, of a wonderful structure" so impressive that "the immortals were astonished, exclaiming, this is not made by human hand." Hermann Goetz, in Artibus Asiae 15 (1952), pp. 84-107, parsed ten distinct sculptural styles within the temple and argued for a long, multi-reign campaign; M.K. Dhavalikar, working from a tighter reading of inscriptional and stratigraphic evidence in 1982 and consolidated in his Oxford Monumental Legacy volume Ellora (2003), defended the Krishna I attribution as the principal building phase. The dispute is the productive one for any cross-site comparison: how a single dynastic command could produce a structure on this scale, and how Ellora's procedural choice — carve down, never build up — sets it against the rest of the ancient world.

The peer corpus for that comparison is wide. Within rock-cut technique alone, Ellora has near contemporaries (Petra's late Nabataean phases), older predecessors (Abu Simbel under Ramesses II), and a sister site 100 km northeast (Ajanta, almost entirely Buddhist, c. 200 BCE-480 CE per Walter M. Spink's "short chronology" published in his Ajanta: History and Development series (Brill, 7 vols. published 2005-2016)). Among cosmic-mountain temples, Kailasa is the earliest of a tightly clustered group of medieval Indianised giants — Borobudur, Khajuraho, Angkor Wat, Konark — each encoding the same Mount Meru / Mount Kailasa cosmology in radically different materials and methods. And among multi-traditional sites, Ellora's adjacency of Buddhist, Hindu, and Jain caves has no real equivalent in the ancient corpus; the closest analogues (Knossos, Karnak) document sequential reuse rather than simultaneous coexistence. The five comparison axes that follow are the ones that defend on the evidence.

The cleanest comparison Ellora invites is the chronologically near-simultaneous one with Borobudur in central Java, built by the Sailendra dynasty c. 780-825 CE — a generation after Krishna I commissioned Kailasa. Both are three-dimensional cosmological models. Kailasa renders Mount Kailasa, Shiva's Himalayan abode; Borobudur renders the Mahayana Buddhist mandala as a stepped pyramid culminating in a central stupa. Both encode the axis-mundi convention shared across early-medieval Indianised Asia. The contrast is in the verb of construction.

Borobudur was assembled from approximately 55,000 cubic metres of andesite (a figure consistently given by UNESCO and the post-1973 restoration team), corresponding to roughly 150,000-165,000 tonnes at andesite densities of 2.7-3.0 t/m³. Workers cut the blocks at neighbouring quarries and along the Progo riverbed, transported them to the site, and laid them dry — interlocking by knobs, indentations, and dovetails rather than mortar. The 1973-1983 UNESCO restoration removed and individually catalogued more than one million stones to install drainage. Every block was a discrete decision; misplaced stones could be repositioned. The monument's nine stacked platforms (six square, three circular) and its 2,672 relief panels and 504 Buddha statues were laid out on a 118-123 m square footprint roughly 32 m high; the constructed plain on which it sits could be expanded, and the project's five identifiable building stages — Sailendra c. 780, c. 792, an unfinished fourth stage c. 824, and a fifth Sanjaya phase c. 833 — show that the pyramid grew incrementally outward and upward.

Kailasa was made by erasing approximately the same order of magnitude of stone, but the stone was already in place. Three vertical trenches were cut from the scarp top to isolate a U-shaped block; conventional figures of the temple-block give roughly 60 m long × 33 m wide × 30 m high (about 200 × 100 × 100 ft), set within a courtyard about 82 m × 46 m. Estimates of basalt removed cluster in the 200,000-400,000 tonne range across popular and academic sources; Dhavalikar's labour-rate calculation (a workforce of roughly 250 over about 5.5 years) is broadly consistent with the lower end of that range and with the comparable mass of Borobudur's superstructure. The structural difference is irreversibility. Borobudur could be rebuilt, and was. Kailasa, like every rock-cut work, allows no edit: a chisel stroke that goes too deep cannot be filled. This forced the architects to design the entire structure — exterior tower, mandapa, garbhagriha, elephant balustrades, drainage, stair runs, the bridge from the gopuram — as a complete three-dimensional vision before the first cut.

Adam Hardy in The Temple Architecture of India (Wiley, 2007, 256 pp.) reads Kailasa as a built quotation of the Pattadakal Virupaksha temple, completed c. 740-745 CE under Vikramaditya II's queen Lokamahadevi (the inscriptional name is Shri Lokeshvara Mahasila Prasada). The Virupaksha is itself modelled on the earlier Pallava Kailasanatha at Kanchipuram, so Ellora's Kailasa stands at the third remove in a transmission chain — Pallava Kailasanatha at Kanchipuram (built c. 700-728 CE under Narasimhavarman II Rajasimha) → Chalukya masonry replication at Pattadakal → Rashtrakuta rock-cut amplification at Ellora. The Rashtrakutas had recently defeated the last Chalukya, Kirtivarman II; reproducing the Virupaksha at Ellora reads as both architectural homage and dynastic appropriation, with the medium itself — carved monolith versus assembled blocks — as the upgrade. Borobudur, by contrast, has no such direct masonry predecessor; its mandala-pyramid form appears to be a Sailendra synthesis rather than a quotation.

One narrow caveat is worth keeping. Borobudur's incremental construction left visible course-lines and interlock joints, which the 1907-1911 van Erp and 1973-1983 UNESCO restorations could trace. Kailasa's erasure left no such record of sequence — there is no equivalent to Borobudur's stage-by-stage stratigraphy because the chisel marks are continuous across the temple's surfaces. What dating evidence Ellora offers comes from the inscriptional and stylistic seams Goetz (1952) and Dhavalikar (2003) read; both methods are interpretive in ways that block-counting at Borobudur is not.

Within the smaller world of subtractive monumental architecture, three sites carry the case for what the technique can do: Ellora's Kailasa, Petra's Al-Khazneh and the Royal Tombs (1st century BCE - 2nd century CE), and Abu Simbel (Ramesses II, c. 1264-1244 BCE). All three were carved from cliff faces. The differences are instructive about what each tradition was willing to attempt.

Petra's Al-Khazneh — façade 24.9 m wide × 38.77 m high, cut into Cambrian Umm Ishrin sandstone — is essentially a relief: an ornate two-storey Hellenistic-Nabataean front behind which the interior is a comparatively plain set of chambers. The Nabataean carvers used a top-down sequence (so as not to damage finished work below them), and the technical achievement is in the precision of the façade's classical orders, not in volumetric depth. Abu Simbel takes the rock-cut concept inward: Ramesses II's Great Temple extends roughly 56 m (185 ft) into the Nubian sandstone cliff, with four 20 m colossi facing east on the exterior and a hypostyle hall, vestibule, and sanctuary cut as connected interior volumes — but the exterior, beyond the four seated kings, is essentially the cliff itself.

Kailasa is the only one of the three that is fully three-dimensional on every face. It has an exterior visible on all sides because the surrounding rock has been removed, an interior volume that you can walk into, a roof shaped from above, sculptural programs on every wall, and a base at the level of the courtyard floor. The temple's most celebrated panel — Ravana shaking Mount Kailasa, on the lower southern face — is itself a comment on the structure's representational program: the demon-king Ravana attempts to lift the mountain on which Shiva and Parvati sit, and the panel depicts the cosmic mountain that the building is. Dhavalikar (2003) dates this panel to roughly three to four decades after the principal building phase, on stylistic grounds linked to similar work in the nearby Lankeshvar cave; the temple's interior wall surfaces continued to receive sculptural treatment well into the late 8th century even as the structural carving may have wrapped up under Krishna I. Carmel Berkson's photographic survey Ellora: Concept and Style (Abhinav Publications in association with the Indira Gandhi National Centre for the Arts, 1992; 392 pp., ~270 black-and-white plates) documents the all-around treatment of the structure in a way text cannot — the same panel appears in plates from four exterior angles plus one interior, showing that no face was treated as secondary.

Across the three sites, the choice of stone matters. Petra's sandstone (Cambrian Umm Ishrin, Mohs hardness ~3-4) carves easily but weathers visibly within a single century; Abu Simbel's Nubian sandstone is similar. Ellora's Deccan basalt is much harder (Mohs 6-7) and homogenous — formed by the Deccan Traps flood basalts that erupted across the Cretaceous-Paleogene boundary ~66 million years ago, with no fracture planes that would allow blocks to spall away. The basalt's hardness made the carving labour-intensive (iron and steel chisels were the only tool available) but produced surfaces that have survived twelve centuries with only modest erosion. The trade-off is exactly what each tradition was buying: Petra got rapid execution and ornamental subtlety on a relatively soft façade; Ellora got permanence at the cost of every cubic centimetre being chiselled by hand.

Kailasa belongs to a family of medieval temples that translate cosmic-mountain symbolism into architectural form: Angkor Wat (Suryavarman II, c. 1113-1150 CE) as Mount Meru in Khmer sandstone; Khajuraho's Chandela temples (c. 950-1050 CE) as Vastu-Purusha-Mandala compositions; the later Konark Sun Temple (c. 1250 CE) as a chariot of Surya. Stella Kramrisch's The Hindu Temple (Calcutta University, 1946; two volumes) remains the foundational synthesis of how each of these monuments encodes cosmological diagrams in plan and elevation: the temple as a built body, the garbhagriha as the cave-womb of the mountain, the vimana as the peak.

The chronological ordering matters. Kailasa (c. 756-773 CE for the principal phase) precedes Borobudur (c. 780-825 CE), which precedes Khajuraho (c. 950-1050 CE) and Angkor Wat (c. 1113-1150 CE) by long stretches. Adam Hardy (Temple Architecture of India, 2007) treats Kailasa as a moment when the Dravidian temple type — already developed in masonry at Pattadakal under the Chalukyas — was imported into the rock-cut tradition at full scale. Krishna I, who had defeated the last Chalukya Kirtivarman II, may have been quoting their architecture in a medium that erased their authorship: a stone victory column rendered as a temple.

Comparison with Angkor Wat shows the limit of the rock-cut approach. Angkor's plan is roughly 162 hectares (galleries, moats, libraries, the central quincunx of towers); the central tower rises 65 m. No subtractive method could produce this footprint — Angkor required a constructed plain, hauled sandstone, and an irrigation system. Kailasa's freestanding monolith, by contrast, sits in a courtyard the size of a small temple precinct. The cosmic-mountain symbolism is the same; the territorial ambition is not. Khmer architects could keep adding; Rashtrakuta architects had to decide where the cliff ended.

Khajuraho extends the comparison sideways. The Chandela temples of c. 930-1050 CE — Lakshmana (c. 930-950), Vishvanatha, Kandariya Mahadeva (c. 1025-1050) — render the Vastu-Purusha-Mandala as a mountain of ascending shikharas, with the central tower clustered with subsidiary turrets that telescope upward. Hardy (2007) argues that Khajuraho takes the Nagara temple type into a baroque elaboration where every surface, including the famously erotic Mithuna couples, becomes part of the temple-as-mountain reading. Kailasa is structurally Dravidian (southern), Khajuraho structurally Nagara (northern); they bracket the medieval Indian temple lexicon. Konark, two centuries later again under the Eastern Ganga king Narasimhadeva I (c. 1250 CE), pushes the cosmological program further by rendering the temple as Surya's chariot — wheels, horses, and all carved at architectural scale. Read in sequence — Kailasa, Khajuraho, Angkor Wat, Konark — the medieval cosmic-mountain temple develops from a rock-cut quotation of an existing masonry type into ever more explicit architectural cosmology.

Ellora's coexistence of Buddhist, Hindu, and Jain caves at one site over four centuries is rare among ancient monumental complexes. The closest analogues are sites of religious continuity, not pluralism: Knossos (Minoan c. 1950-1450 BCE, then Mycenaean reuse, then Greek and Roman cult layers); Karnak Temple at Thebes (c. 2055 BCE - 100 CE, with the Amun-Ra cult evolving across two millennia of Egyptian dynasties through Ptolemaic and Roman reuse). Both Knossos and Karnak preserve sequential religious phases — one tradition succeeded by another, often appropriating prior structures.

Ellora is structurally different. The Buddhist caves (1-12, c. 600-800 CE) and the Hindu caves (13-29, including Kailasa, c. 600-900 CE) overlap chronologically. Both communities were carving the same cliff face during the same decades. The Jain caves (30-34, c. 800-1000 CE) follow but do not displace; the earlier caves were neither destroyed nor systematically modified. Dhavalikar (2003, p. 12) emphasises the absence of iconoclastic damage: "There is no record at Ellora, archaeological or epigraphic, of one community destroying the work of another. The cliff was shared." This is unusual at this scale. The Rashtrakutas, who patronised Kailasa, are documented as patrons of Buddhist and Jain institutions as well — a cross-religious patronage pattern that Hardy (2007, ch. 5) treats as characteristic of medieval Deccan polity rather than as exceptional tolerance.

The contrast with Knossos sharpens the point. At Knossos, layered religious occupation produced a palimpsest where one phase obscures another; archaeologists must dig through Linear A material to reach earlier Minoan strata, and Mycenaean Linear B sits over both. At Ellora, you walk laterally along the cliff and pass between traditions without descending in time: cave 10 (Buddhist Visvakarma chaitya) and cave 16 (Hindu Kailasa) are spatial neighbours, not stratigraphic layers. The site preserves religious diversity as adjacency rather than as overlay. Whether this represents genuine inter-traditional respect or simply concurrent patronage by traditions that did not see themselves as competing for territory is debated; Walter M. Spink, in his Ajanta-Ellora studies, pointed out that Buddhist sponsorship at Ellora declines exactly as Hindu sponsorship rises, suggesting economic and dynastic substitution more than spiritual rapprochement.

The Karnak comparison is in some ways closer. Karnak's two-millennium accretion (c. 2055 BCE - 100 CE) records the Amun-Ra cult's evolution through Middle Kingdom, New Kingdom, Third Intermediate, Late Period, Ptolemaic, and Roman phases — but the cult itself stays continuous. Pharaohs added pylons, courts, and chapels; later rulers reused or recut earlier blocks; the same deity received offerings across a span longer than recorded Indian history. Ellora's coexistence is the opposite kind of pluralism: three different religious systems, with three different soteriologies, sharing a single sacred geography across four centuries without one absorbing the others. There is, in this respect, no real precedent in the ancient corpus for what Ellora records. The closest later parallel is the medieval Indian institutional norm Hardy (2007, ch. 5) describes — kings patronising multiple traditions simultaneously to broaden their legitimacy — but Ellora is the most concentrated material expression of that pattern.

One comparison that gets made and should be refused is between Ellora and decorated natural caves like Lascaux (c. 17,000-15,000 BCE, Magdalenian). The shared word "cave" misleads. Lascaux is a natural limestone cavity whose makers added pigment to surfaces they did not create; the cave existed before the painters arrived. Ellora's caves did not exist before the carvers arrived. Each chamber was an act of architecture — the wall, ceiling, floor, columns, and figures all produced from the same operation that hollowed the rock. There are no found surfaces at Ellora, only made ones.

The technical comparison breaks down at the level of intent. A Magdalenian painter at Lascaux worked with a void already given; the engineering question — how do you make a chamber large enough to stand in — was already answered by geology. A Rashtrakuta sculptor at Ellora worked with no void at all; the entire space he was decorating was the same operation as the space's existence. This is why the comparison with Petra and Abu Simbel is the productive one: those are also made caves, where the cavity and its decoration emerged together.

What this set of comparisons reveals is Ellora's distinctive position. Among rock-cut sites, Kailasa is the only monument to be carved fully on every face, in three dimensions, with an integrated interior — the technique pushed past the relief of Petra and the embedded interior of Abu Simbel into a freestanding form. Among cosmic-mountain temples, it is the earliest of the medieval Indian and Indianised-Asian giants, and the only one made by erasure rather than by accumulation. Among multi-religious sites, it is the rare instance where the traditions were spatial neighbours during overlapping centuries rather than chronological successors. The procedural fact that begins this article — the temple was revealed, not built — is what holds the comparison set together. Borobudur was stacked. Angkor was paved. Karnak was added to. Ellora was uncovered. The Rashtrakuta workforce, several hundred strong over fifteen to twenty concentrated years per Dhavalikar's reading (or several reigns per Goetz's longer chronology), removed the rock that was not the temple and left the temple standing. The conceptual demand — to see the finished structure in the unworked cliff before the first chisel stroke — is the demand the site keeps making of every visitor who tries to understand it.

One last comparative note belongs in the synthesis. The Kailasa Temple's location on the medieval Deccan plateau trade-route network is not incidental to its survival or its scale. The Rashtrakuta capital at Manyakheta (modern Malkhed) lay roughly 400 km south-southeast, and the dynasty's mid-eighth-century military and economic dominance of the Deccan supplied the patronage budget that a project of this complexity required. Hardy (2007) and Dhavalikar (2003) both note that the absence of a securely identified architect-name at Kailasa — no equivalent to Divakarapandita, Suryavarman II's royal preceptor whom Khmer sources name as the conceptual force behind Angkor Wat, or to the Sailendra patrons named on contemporary inscriptions like the Karangtengah plate at Borobudur — reflects the Indian convention that monumental work is attributed to its dynastic patron rather than its design author. What we have is the Baroda inscription's astonished register: the immortals exclaim that the building is not made by human hand. Whether the panel of Ravana shaking the temple's mountain is the Rashtrakuta architects' own self-aware comment on what they have produced — that the temple is the mountain — or simply a standard Hindu iconographic scene rendered at unusual scale, is not finally decidable. But the panel works either way: the demon trying to dislodge the cosmic mountain, the temple that succeeds where the demon failed, both rendered from one piece of basalt. That is the level at which Ellora's comparisons close.

Significance

Ellora's distinctive place in the ancient world rests on three intersecting facts that no other site holds together. It is the only monumental rock-cut work treated fully in three dimensions on every face, with an integrated interior — pushing the technique past the relief of Petra and the embedded interior of Abu Simbel. Its Kailasa Temple, securely attributed by Hermann Goetz (Artibus Asiae 15, 1952) and re-confirmed by M.K. Dhavalikar (Ellora, Oxford 2003) to Krishna I c. 756-773 CE, is the earliest of the great medieval cosmic-mountain temples and the only one made by erasure rather than by accumulation. And its Buddhist, Hindu, and Jain caves were spatial neighbours during overlapping centuries rather than chronological successors — an arrangement Knossos and Karnak document as sequential overlay, but which Ellora preserves as adjacency. The procedural verb is what unites these claims: Borobudur was stacked, Angkor paved, Karnak added to. Ellora was uncovered.

Connections

Ellora Caves — the parent entity. This sub-page focuses on cross-site comparisons; the parent covers Ellora's 34 caves, the Kailasa Temple's construction, and the site's religious history in standalone depth.

Borobudur — the chronological twin: c. 780-825 CE Sailendra Buddhist mandala assembled from ~55,000 m³ of andesite, where Kailasa was carved by removing a comparable mass of basalt. Same scale, opposite verb.

Petra — Nabataean rock-cut peer (1st century BCE - 2nd century CE). Al-Khazneh's 24.9 m × 38.77 m sandstone façade is essentially a relief; Kailasa's monolithic block is fully volumetric. The comparison maps the range of subtractive technique.

Abu Simbel — Ramesses II's c. 1264-1244 BCE rock-cut temple extends 56 m (185 ft) into the Nubian cliff. With Kailasa and Petra, completes the family of stone-carved monumental architecture.

Angkor Wat — c. 1113-1150 CE Khmer cosmic-mountain temple to Vishnu, later Buddhist. Renders Mount Meru in additive sandstone where Kailasa renders Mount Kailasa in subtracted basalt.

Khajuraho — c. 950-1050 CE Chandela Vastu-Purusha-Mandala compositions. Together with Kailasa and Konark, traces the medieval Indian temple as cosmological diagram.

Konark Sun Temple — c. 1250 CE solar chariot in Odisha; later phase of the same Indian temple-as-sculpture tradition Kailasa belongs to.

Knossos — Minoan-Mycenaean-Greek-Roman religious continuity (c. 1950 BCE forward) preserved as stratigraphic overlay. Contrasts with Ellora's lateral adjacency of three traditions.

Karnak Temple — c. 2055 BCE - 100 CE additive Amun-Ra precinct at Thebes. Two-millennium accretion site against Ellora's four-century plural-tradition site.

Lascaux — Magdalenian decorated natural cave (c. 17,000-15,000 BCE). The anti-comparison: Lascaux's cavity preceded the painters; Ellora's cavity was the same act as its decoration.

Further Reading