Almost every building on earth is assembled. Stone is quarried, dressed, lifted and set; if a block is wrong, it is replaced. The caves at Ajanta and Ellora were not assembled. They were subtracted. The building was already there, inside the rock, and the work consisted of taking away everything that was not the building.

This sounds like a technicality. It changes every decision in the design, because subtraction has no undo. A misplaced block in a temple can be pulled out and recut. A misplaced cut in basalt is permanent, and the only remedy is to keep cutting until the error looks intentional.

Why the proportions are conservative

Look at the span-to-height ratio in any rock-cut hall and compare it with a structural temple of the same period built of blocks. The excavated hall is almost always stockier. That is not a difference in taste. It is what happens when the builder cannot add a buttress, cannot thicken a wall afterwards, and cannot test a beam before committing to it.

In a built temple the mason can be adventurous because failure is local: a cracked lintel is replaced. In an excavated one, a collapsed roof is the end of the project, and the rock above it is now a hazard rather than a roof. So the columns come closer together, the ceilings sit lower, and the whole interior is designed with a margin that a free-standing structure would not need. The conservatism is a signature of the method, and once you know it you can date an excavated hall by how confident it is.

Ajanta: painting as the point

Ajanta is a monastery, not a temple complex, and its thirty caves are mostly viharas — residential halls with cells around a central square — plus a few chaityas, the barrel-vaulted assembly halls with a stupa at the far end. The earliest work is second century BCE; the bulk of what survives is fifth century, the product of a single intense period under the Vakataka dynasty.

What makes Ajanta matter is not the architecture but what is on it. The walls carry the largest surviving body of ancient Indian painting, executed on a prepared ground of clay and lime and worked while still damp enough to bond. The subjects are narrative — the previous lives of the Buddha, court scenes, processions — and the technique is closer to fresco than to panel painting, which is why the colour has survived in a way that pigments on dry plaster never do.

The architectural consequence is easy to miss. These interiors are dark by design. There is no window; the only light comes from the doorway. Painting a wall that will only ever be seen by a visitor whose eyes have adjusted, at a distance of two metres, changes the painter's palette: fewer mid-tones, stronger outlines, larger figures. The cave is a viewing chamber with a controlled light level, and the painting is calibrated to it. Remove the painting and the architecture is a fairly ordinary vihara. The building exists to be a container for the image.

Ellora: three religions in one escarpment

Ellora is a different proposition, and it is the more interesting site for anyone thinking about architecture rather than art. Two kilometres of basalt cliff contain thirty-four caves in use, cut between roughly 600 and 1000 CE: twelve Buddhist, seventeen Hindu, five Jain, in that order, north to south.

The sequence is the argument. The Buddhist caves come first and are the simplest — viharas and chaityas, close to Ajanta's model. The Hindu caves in the middle are the most ambitious. The Jain caves at the end are small, late, and extremely refined, as if the technique had outlived the budget. Nobody planned this as a statement of religious harmony; the caves were cut over four centuries by different patrons with different money. But the effect is a stratigraphy of belief in a single rock face, and it is the most legible thing at the site.

Cave 16: building from the ceiling down

Kailasa, Cave 16, is the one that breaks the category. It is not a room cut into a cliff. It is a complete temple — courtyard, gatehouse, pillared halls, subsidiary shrines, a tower over the sanctum — carved as a monolith out of the hillside, with no part of it assembled.

To do that, the masons had to work from the top down. They began at the highest point of what would become the tower and cut downward, and outward, removing roughly 200,000 tonnes of rock in the process. For most of the project the finished building did not exist above them; it existed below them, still encased. The first day on site, a mason stood at the level of the final roof. That is the opposite of how every other building in the world is made, and it means the whole design had to be held in the head, or in a model, before the first cut.

It also explains the plan. A temple built up from the ground can be extended by adding a mandapa later. A temple cut down from the top cannot: the courtyard is fixed by the first cut, because everything else has to fit inside it. Kailasa looks like a designed composition because it had to be one.

Why the rock stays cool

The Deccan basalt around these caves is a poor conductor of heat and a large mass. A cave interior sits at roughly 25 to 27 degrees when the plateau outside is at 40, and the lag is not a trick of shade — it is the rock acting as a reservoir. During the day the surface absorbs heat and the interior barely moves; at night the surface radiates and the interior still barely moves. The amplitude is damped by the mass, and the deeper the room, the flatter the curve.

This is the third time the same principle has appeared in this pillar. The stepwell used the ground as a cool volume. The courtyard used night air as a cold store. The cave uses rock as a thermal reservoir. In all three cases the ornament ends up where the structure is heaviest, and in all three the modern visitor notices the ornament first.

How to read one

At Ajanta, take a torch and go to Cave 1 and Cave 26. Stand in the doorway for two full minutes without moving — this is not reverence, it is your eyes adapting, and it is how the room was designed to be entered. Then look at where the painting is best preserved: almost always the ceiling and the upper wall, because that is where hands and lamps reached least.

At Ellora, walk the escarpment north to south in order rather than jumping to Cave 16. The point is the sequence, and the sequence is lost if you start with the masterpiece. Then at Cave 16 climb to the south gallery and look down: that is the view the masons had on day one, and it is the only place where you can see the method rather than the result.

Sources and checks

  1. UNESCO World Heritage Centre Ajanta Caves, ref. 242 — inscribed 1983, criteria (i), (ii), (iii), (vi)
  2. UNESCO World Heritage Centre Ellora Caves, ref. 243 — inscribed 1983, criteria (i), (iii), (vi)
  3. Spink, Walter M. Ajanta: History and Development, 7 vols. Brill, 2005–2017
  4. Michell, George The Penguin Guide to the Monuments of India, vol. 1. Viking, 1989
  5. Archaeological Survey of India Aurangabad Circle — conservation reports, Ajanta and Ellora

Last checked 28 October 2026 Yellowsea Huang Corrections